&EPA
United States
EnviroimnU Protection
Agency
Quantitative Risk and Exposure Assessment
for Carbon Monoxide - Amended
-------
-------
EPA-452/R-10-009
July 2010
Quantitative Risk and Exposure Assessment
for Carbon Monoxide - Amended
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Health and Environmental Impacts Division
Research Triangle Park, North Carolina
-------
-------
DISCLAIMER
This document has been reviewed by the Office of Air Quality Planning and Standards,
U.S. Environmental Protection Agency (EPA), and approved for publication. This document has
been prepared by staff from the Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency. Any opinions, findings, conclusions, or recommendations are
those of the authors and do not necessarily reflect the views of the EPA Mention of trade names
or commercial products is not intended to constitute endorsement or recommendation for use.
Any questions or comments concerning this document should be addressed to Dr. Stephen
Graham, U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards,
C504-06, Research Triangle Park, North Carolina 27711 (email: graham.stephen@epa.gov ).
Elements of this report have been provided to the U.S. Environmental Protection Agency (EPA)
by Abt Associates, Inc. and TRJ Environmental, Inc. in partial fulfillment of Contract No. EP-D-
08-100, Work Assignments 0-08 and 1-15.
-------
-------
PREFACE
This amended Quantitative Risk and Exposure Assessment for Carbon Monoxide (CO
REA) contains revised results reflecting corrections to one of the input variables used in the dose
model to generate estimates of carboxyhemoglobin (COHb). Modeling results presented in the
May 2010 CO REA were generated with erroneous values for altitude for both the Los Angeles
and Denver study areas. This error has been corrected and the model simulations repeated
resulting in increases to the COHb estimates in both study areas. These increases, in terms of the
ambient contribution to COHb levels, were small for the Denver study area and negligible for the
Los Angeles study area, with somewhat larger increases to the total COHb estimates. Because
altitude was not used in calculating exposure estimates, none of the exposure results previously
presented were affected by this error. The sections of the REA revised to reflect the corrected
dose estimates include several tables and associated text in chapters 5, 6, 7 and 8 and Appendix
B. The specific areas of the report containing revisions to address this error include section
5.10.4 (Table 5-27), section 6.2 (Tables 6-15 through 6-23, Figures 6-5 and 6-6), section 6.3
(Tables 6-24 and 6-25), section 7.2.1.2 (Table 7-4), section 7.2.2 (Tables 7-5 through 7-8, 7-10,
7-11, and 7-14), chapter 8 and Appendix B.6 (Tables B-4 through B-6, Figures B-2 through B-5).
In addition, the presentation of exposure estimates has been modified to reflect a correction to
the rounding convention for percentages of the population less than 0.1%. Previously, values
between 0.05 and 0.09% were rounded upwards to 0.1%. This correction is reflected in the
presentation of exposure estimates in Tables 6-4, 6-7, 6-10, 6-13, and 6-14.
-------
-------
Table of Contents
List of Figures v
List of Tables vi
I INTRODUCTION 1-1
1.1 BACKGROUND 1-1
1.2 PREVIOUS REVIEWS AND ASSESSMENTS 1-3
1.3 CURRENT REVIEW, CAS AC ADVICE AND PUBLIC COMMENT 1-6
1.4 REFERENCES 1-9
2 CONCEPTUAL OVERVIEW: ASSESSING AMBIENT CARBON
MONOXIDE EXPOSURE AND RISK 2-1
2.1 SOURCES OF CARBON MONOXIDE 2-1
2.2 EXPOSURE PATHWAYS AND IMPORTANT
MICROENVIRONMENTS 2-2
2.3 EXPOSURE AND DOSE METRICS 2-6
2.4 AT-RISK POPULATIONS 2-7
2.5 HEALTH ENDPOINTS 2-11
2.5.1 Cardiovascular Disease-related Effects 2-12
2.5.2 Other Effects 2-14
2.6 RISK CHARACTERIZATION APPROACH 2-15
2.7 KEY OBSERVATIONS 2-19
2.8 REFERENCES 2-21
3 AIR QUALITY CONSIDERATIONS 3-1
3.1 AMBIENT CO MONITORING 3-1
3.1.1 Monitoring Network 3-1
3.1.2 Analytical Sensitivity 3-2
3.1.3 General Patterns of CO Concentrations 3-4
3.1.4 Policy-Relevant Background Concentrations 3-11
3.1.5 Within-Monitor CO Concentration Trends 3-11
3.2 STUDY AREAS SELECTED FOR CURRENT ASSESSMENT 3-20
3.3 KEY OBSERVATIONS 3-20
3.4 REFERENCES 3-22
-------
OVERVIEW OF APEX MODELING SYSTEM FOR ESTIMATING CO
EXPOSURE AND COHB DOSE LEVELS 4-1
4.1 PURPOSE 4-1
4.2 MODEL OVERVIEW 4-1
4.3 MODEL HISTORY AND EVOLUTION 4-2
4.4 MODEL SIMULATION PROCESS 4-3
4.4.1 Characterize Study Area 4-6
4.4.2 Generate Simulated Individuals 4-6
4.4.3 Construct Activity Sequences 4-8
4.4.4 Calculate Microenvironmental Concentrations 4-13
4.4.5 Estimate Energy Expenditure and Ventilation Rates 4-31
4.4.6 Calculate Exposure 4-33
4.4.7 Calculate Dose 4-34
4.4.8 Model Output 4-37
4.5 Key Observations 4-37
4.6 REFERENCES 4-38
5 APPLICATION OF APEX4.3 IN THIS ASSESSMENT 5-1
5.1 PURPOSE 5-1
5.2 OVERVIEW 5-1
5.3 STUDY AREAS 5-2
5.4 EXPOSURE PERIODS 5-3
5.5 STUDY POPULATION 5-7
5.5.1 Simulated at-Risk Subpopulations 5-7
5.5.2 Time-Location-Activity Patterns 5-12
5.5.3 Construction of Longitudinal Diaries 5-12
5.6 EXPOSURE SCENARIOS 5-13
5.7 AMBIENT AIR QUALITY DATA 5-16
5.7.1 Unadjusted 1-Hour Ambient Concentrations 5-16
5.7.2 Method for Estimating Missing 1-Hour Ambient Concentrations 5-17
5.7.3 Adjusted 1-Hour Ambient Concentrations 5-21
5.8 METEOROLOGICAL DATA 5-25
5.8.1 Method for Estimating Missing 1-Hour Temperature Data 5-26
11
-------
5.9 MICROENVIRONMENTS MODELED 5-27
5.9.1 The Micronenvironmental Model as Implemented by APEX4.3 5-27
5.9.2 Microenvironmental Mapping 5-30
5.9.3 Selection of Microenvironmental Method Used 5-31
5.9.4 Air Exchange Rates and Air Conditioning Prevalence 5-31
5.10 ADDITIONAL EXPOSURE AND DOSE OUTPUT GENERATED
USING REDUCED APEX SIMULATIONS 5-33
5.10.1 Estimate of Microenvironmnet Contribution to At-Risk Population
Exposure Levels 5-34
5.10.2 Estimate of Microenvironment-to-Ambient Concentration Ratios 5-34
5.10.3 Estimate of Ambient Exposure Contribution to Total COHb Level 5-35
5.10.4 Comparison of the Exposure and Dose Results Genererated Using
50,000 Persons Versus 5, 000 Persons Simulation 5-36
5.11 KEY OBSERVATIONS 5-38
5.12 REFERENCES 5-40
6 SIMULATED EXPOSURE AND COHB RESULTS 6-1
6.1 ESTIMATED EXPOSURES 6-2
6.1.1 Air quality "As Is" 6-2
6.1.2 Air quality adjusted to just meet the current 8-hour standard 6-11
6.1.3 Air quality adjusted to just meet alternative air quality scenarios 6-19
6.2 ESTIMATED COHB DOSE LEVELS 6-22
6.2.1 Air quality "As Is" 6-22
6.2.2 Air quality adjusted to just meet the current 8-hour standard 6-27
6.2.3 Air quality adjusted to just meet alternative air quality scenarios 6-31
6.3 COMPARISON OF COHB ESTIMATES OBTAINED FROM THE 2000
PNEM/CO AND 2010 APEX/CO ASSESSMENTS 6-35
6.4 KEY OBSERVATIONS 6-40
6.5 REFERENCES 6-43
7 VARIABILITY ANALYSIS AND UNCERTAINTY
CHARACTERIZATION 7-1
7.1 ANALYSIS OF VARIABILITY 7-1
7.2 CHARACTERIZATION OF UNCERTAINTY 7-3
7.2.1 Considerations in Characterizing Sources and Uncertainty 7-9
7.2.2 Sensitivity Analysis 7-22
7.3 KEY OBSERVATIONS 7-32
7.4 REFERENCES 7-34
8 SUMMARY OF KEY OBSERVATIONS 8-1
in
-------
APPENDICES
A. Technical Memorandum on Updates to APEX Physiology.Txt File A-l
B. COHb Module for APEX4.3 B-l
C. Isaacs et al. (2009) Reference Used in Developing D and A Statistics Input to APEX
Model C-l
D. Microenvironmental Mapping D-l
E. Analysis of CHAD Diaries for Time Spent in Vehicles E-l
F. Differences in Human Activity Patterns Between Individuals With and Without
Cardiovascular Disease F-l
IV
-------
List of Figures
Figure 3-1 Spatial and temporal trends in the 2nd highest 1-hour (top) and 8-hour average
(bottom) CO ambient monitoring concentrations in Denver, Colorado(left) and
Los Angeles, California (right), years 1993 - 2008 3-6
Figure 3-2 Diurnal distribution of 1-hour CO concentrations in Denver
(monitor 080310002) by day-type (weekdays-left; weekends-right),
years 1995 (top) and 2006 (bottom) 3-9
Figure 3-3 Diurnal distribution of 1-hour CO concentrations in Los Angeles (monitor
060371301) by day-type (weekdays-left; weekends-right), years 1997 (top)
and 2006 (bottom) 3-10
Figure 3-4 Comparison of high concentration year (1997) versus a low concentration year
(2006) at four ambient monitors in Denver 3-13
Figure 3-5 Comparison of a high concentration year (1997) versus a low concentration year
(2006) at four ambient monitors in Los Angeles 3-14
Figure 4-1 Conceptual model and simplified data flow for estimating population exposure
and dose using APEX4.3 4-5
Figure 5-1 Ambient monitor locations, air districts (black circles), meteorological zones
(blue circles), and study area (red circle) for the Denver exposure
modeling domain 5-5
Figure 5-2 Ambient monitor locations, air districts (black circles), meteorological zones
(blue and pink circles), and study area (red circle) for the Los Angeles exposure
modeling domain 5-6
Figure 5-3 Relationship between microenvironment-to-ambient concentration ratios using
estimated indoor-residential concentrations (left panel) and inside-vehicle
concentrations (right panel) in Denver -as is air quality 5-35
Figure 6-1 Estimated microenvironmental contributions to time spent at or above selected
exposure concentrations using the Denver CHD population - as is air quality 6-5
Figure 6-2 Estimated microenvironmental contributions to time spent at or above selected
exposure concentrations using the Los Angeles CHD population
-as is air quality 6-9
Figure 6-3 Estimated microenvironmental contributions to time spent at or above selected
exposure concentrations using the Denver CHD population - air quality just
meeting the current 8-hour standard 6-13
Figure 6-4 Estimated microenvironmental contributions to time spent at or above selected
exposure concentrations using the Los Angeles CHD population- air quality just
meeting the current 8-hour standard 6-17
Figure 6-5 Estimated percent of the CHD population in Denver (top) and Los Angeles
(bottom) experiencing repeated COHb levels - as is air quality 6-26
Figure 6-6 Estimated percent of the CHD population in Denver (top) and Los Angeles
(bottom) experiencing repeated COHb levels - air quality just meeting the
current 8-hour standard 6-30
-------
Figure 7-1. Comparison of commute and travel times for persons residing in Denver
and Los Angeles counties to those persons surveyed in CHAD 7-17
List of Tables
Table 3-1 Within monitor temporal variability in Denver using historical (1995-97) and
recent (2005-07) air quality data - 2nd highest 8-hour average 3-17
Table 3-2 Within monitor temporal variability in Los Angeles using historical (1995-97)
and recent (2005-07) air quality data - 2nd highest 8-hour average 3-17
Table 3-3 Within monitor temporal variability in Denver using historical (1995-97) and
recent (2005-07) air quality data - 99th percentile 1-hour daily maximum 3-18
Table 3-4 Within monitor temporal variability in Los Angeles using historical (1995-97)
and recent (2005-07) air quality data - 99th percentile 1-hour daily maximum 3-18
Table 3-5 Within monitor temporal variability in Denver using historical (1995-97) and
recent (2005-07) air quality data - 99th percentile 8-hour daily maximum 3-19
Table 3-6 Within monitor temporal variability in Los Angeles using historical (1995-97) and
recent (2005-07) air quality data - 99th percentile 8-hour daily maximum 3-19
Table 4-1 Summary of activity pattern studies comprising the recent version of CHAD 4-10
Table 4-2 Variables used by APEX4.3 in the mass balance model 4-16
Table 4-3 Variables used by APEX4.3 in the factors model 4-19
Table 4-4 Estimated values of distribution parameters and variables in equation 4-11 as
implemented in the application of pNEM/CO to Denver and Los Angeles
(Johnson et al., 2000) 4-22
Table 4-5. Parameters of bounded lognormal distributions defined for proximity factors used
in applications of APEX3.1 to Los Angeles (Johnson and Capel, 2003) 4-30
Table 5-1. Attributes of fixed-site monitors selected for the Denver study area 5-3
Table 5-2. Attributes of fixed-site monitors selected for the Los Angeles study area 5-4
Table 5-3. National prevalence rates for diagnosed coronary heart disease by age range 5-8
Table 5-4 National prevalence rates for diagnosed coronary heart disease by gender 5-9
Table 5-5 Estimated national prevalence rates for coronary heart disease stratified by age
range and gender 5-9
Table 5-6 Estimated national prevalence rates for coronary heart disease, including
diagnosed and undiagnosed cases, stratified by age and gender 5-10
Table 5-7 National prevalence rates for all types of diagnosed heart disease by age range. ..5-11
Table 5-8 National prevalence rates for all types of diagnosed heart disease by gender 5-11
Table 5-9 Estimated national prevalence rates for all types of diagnosed heart disease,
stratified by age and gender 5-11
Table 5-10 Estimated national prevalence rates for all types of diagnosed heart disease plus
undiagnosed coronary heart disease, stratified by age and gender 5-12
Table 5-11 Array of alternative standard forms and levels informed by modeled exposure
scenarios in Denver 5-15
VI
-------
Table 5-12 Array of alternative standard forms and levels informed by modeled exposure
scenarios in Los Angeles 5-16
Table 5-13 Descriptive statistics for hourly carbon monoxide concentrations before and
after estimation of missing values - Denver 1995 5-18
Table 5-14 Descriptive statistics for hourly carbon monoxide concentrations before and
after estimation of missing values - Denver 2006 5-18
Table 5-15 Descriptive statistics for hourly carbon monoxide concentrations before and
after estimation of missing values - Los Angeles 1997 5-19
Table 5-16 Descriptive statistics for hourly carbon monoxide concentrations before and
after estimation of missing values - Los Angeles 2006 5-20
Table 5-17 Design values and adjustment factors used to represent air quality just meeting
the current and potential alternative standards 5-22
Table 5-18 Descriptive statistics for hourly carbon monoxide concentrations after adjusting
to just meet the current 8-hour standard - Denver (adjusted 1995 data) 5-24
Table 5-19 Descriptive statistics for hourly carbon monoxide concentrations after adjusting
to just meet the current 8-hour standard - Los Angeles (adjusted 1997 data) 5-24
Table 5-20 Locations of meteorological stations selected for Denver 5-25
Table 5-21 Locations of meteorological stations selected for Los Angeles 5-26
Table 5-22 Parameters of bounded lognormal distributions defined for proximity factors used
in the application of APEX4.3 to Los Angeles and Denver 5-29
Table 5-23 List of microenvironments modeled and calculation methods used 5-31
Table 5-24 Lognormal distributions of indoor air exchange rates used in Denver 5-32
Table 5-25 Lognormal distributions of indoor air exchange rates used in Los Angeles 5-33
Table 5-26 Comparison of exposure summary output generated when simulating 50,000
persons versus that of simulating 5,000 persons - as is air quality 5-37
Table 5-27 Comparison of dose summary output generated when simulating 50,000 persons
versus that of simulating 5,000 persons -just meeting the current standard 5-38
Table 6-1. Estimated daily maximum 1-hour or 8-hour exposure for simulated at-risk
populations in the Denver study area-as is air quality 6-3
Table 6-2 Estimated distribution of microenvironment-to-ambient concentration ratios using
the Denver CHD population - as is air quality 6-6
Table 6-3 Estimated distribution of microenvironmental concentrations using the Denver
CHD population -as is air quality 6-6
Table 6-4 Estimated daily maximum 1-hour or 8-hour exposure for simulated at-risk
populations in the Los Angeles study area - as is air quality 6-8
Table 6-5 Estimated distribution of microenvironment-to-ambient concentration ratios using
the Los Angeles CHD population - as is air quality 6-10
Table 6-6 Estimated distribution of microenvironmental concentrations using the
Los Angeles CHD population - as is air quality 6-10
Table 6-7 Estimated daily maximum 1-hour or 8-hour exposure for simulated at-risk
populations in the Denver study area - air quality just meeting the current 8-hour
standard 6-12
Table 6-8 Estimated distribution of microenvironment-to-ambient concentration ratios
using the Denver CHD population - air quality just meeting the current 8-hour
standard 6-14
vn
-------
Table 6-9 Estimated distribution of microenvironmental concentrations using the Denver
CHD population - air quality just meeting the current 8-hour standard 6-14
Table 6-10 Estimated daily maximum 1-hour or 8-hour exposure for simulated at-risk
populations in the Los Angeles study area - air quality just meeting the current
8-hour standard 6-16
Table 6-11 Estimated distribution of microenvironment-to-ambient concentration ratios using
the Los Angeles CHD population - air quality just meeting the current 8-hour
standard 6-18
Table 6-12 Estimated distribution of microenvironmental concentrations using the Los Angeles
CHD population- air quality just meeting the current 8-hour standard 6-18
Table 6-13 Estimated daily maximum 1-hour and 8-hour exposures for simulated at-risk
populations in the Denver study area - alternative air quality scenarios 6-20
Table 6-14 Estimated daily maximum 1-hour and 8-hour exposures for simulated at-risk
populations in the Los Angeles study area - alternative air quality scenarios 6-21
Table 6-15 Portion of the simulated at-risk populations in the Denver study area estimated
to experience a daily maximum end-of-hour COHb at or above specified levels
-as is air quality 6-23
Table 6-16 Portion of the simulated at-risk populations in the Los Angeles study area
Estimated to experience a daily maximum end-of-hour COHb at or above
specified levels - as is air quality 6-23
Table 6-17 Percentages of the simulated CHD populations in the Denver and Los Angeles
study areas estimated to experience a daily maximum end-of-hour COHb
contribution from ambient exposure alone at or above specified levels
-as is air quality 6-27
Table 6-18 Portion of the simulated at-risk populations in the Denver study area estimated
to experience a daily maximum end-of-hour COHb at or above specified levels
- air quality just meeting the current 8-hour standard 6-29
Table 6-19 Portion of the simulated at-risk populations in the Los Angeles study area
Estimated to experience a daily maximum end-of-hour COHb at or above
specified levels - air quality just meeting the current 8-hour standard 6-29
Table 6-20 Percentage of simulated CHD populations in the Denver and Los Angeles study
areas estimated to experience daily maximum end-of-hour COHb contribution
from ambient exposure alone at or above specified levels - air quality just
meeting the current 8-hour standard 6-31
Table 6-21 Portion of the simulated at-risk populations in the Denver study area estimated
to experience a daily maximum end-of-hour COHb at or above specified levels
-air quality just meeting potential alternative standards 6-33
Table 6-22 Portion of the simulated at-risk populations in the Los Angeles study area estimated
to experience a daily maximum end-of-hour COHb at or above specified levels
-air quality just meeting potential alternative standards 6-34
Table 6-23 Percentage of simulated CHD populations in the Denver and Los Angeles study
areas estimated to experience daily maximum end-of-hour COHb contribution from
ambient exposure alone at or above specified levels - air quality just meeting a
99th percentile daily maximum 8-hour average concentration of 5.0 ppm 6-35
Vlll
-------
Table 6-24 Percentage of Denver adults with coronary heart disease (CHD) estimated to
experience a daily maximum end-of-hour COHb level - air quality just meeting the
current 8-hour standard 6-39
Table 6-25 Percentage of Los Angeles adults with coronary heart disease (CHD) estimated to
experience a daily maximum end-of-hour COHb level - air quality just meeting the
current 8-hour standard 6-39
Table 1-1. Summary of how variability was incorporated into the assessment 7-2
Table 7-2. Characterization of key uncertainties in the assessment 7-8
Table 7-3. Frequency of CO concentrations reported as zero in Denver and Los Angeles
ambient monitoring data 7-11
Table 7-4. Percentage of simulated at-risk CHD population in Los Angeles with highest
daily maximum end-of-hour COHb levels at or above indicated COHb level
considering potential alternative standards 7-15
Table 7-5. Comparison of highest estimated daily maximum end-of-hour COHb levels for
Denver HD population for two model simulations - all monitor concentrations
versus the design monitor concentrations - as is air quality 7-24
Table 7-6. Comparison of highest estimated daily maximum end-of-hour COHb levels for
Los Angeles HD population for two model simulations - all monitor
concentrations versus the design monitor concentrations - as is air quality 7-24
Table 7-7. Comparison of highest estimated daily maximum end-of-hour COHb levels for
Denver HD population for two model simulations - all monitor concentrations
versus the design monitor concentrations - air quality just meeting the current
8-hour standard 7-25
Table 7-8. Comparison of highest estimated daily maximum end-of-hour COHb levels for
Los Angeles HD population for two model simulations - all monitor concentrations
versus the design monitor concentrations - air quality just meeting the current
8-hour standard 7-25
Table 7-9. Estimated alternative prevalence rates for CHD, stratified by age and gender 7-26
Table 7-10. Comparison of the portion of the simulated CHD population in the Denver study
area estimated to experience a daily maximum end-of-hour COHb at or above
specified levels using base and alternative undiagnosed CHD prevalence rates
-as is air quality 7-27
Table 7-11. Comparison of the portion of the simulated CHD population in the Denver study
area estimated to experience a daily maximum end-of-hour COHb at or above
specified levels using base and alternative undiagnosed CHD prevalence rates
-air quality just meeting the current standard 7-28
Table 7-12.Hemoglobin levels below which anemia is present in a population (from
WHO etal., 2001) 7-29
Table 7-13.Descriptive statistics of blood hemoglobin content measured in various groups
(fromNHANES 1999-2008) 7-30
Table 7-14. Comparison of the portion of the simulated CHD population in the Denver study
area estimated to experience a daily maximum end-of-hour COHb at or above
specified levels when sampling from the base and anemic hemoglobin content
distributions - air quality just meeting the current standard 7-31
IX
-------
-------
1 INTRODUCTION
This document, Quantitative Risk and Exposure Assessment for Carbon Monoxide,
describes the quantitative human exposure assessment and risk characterization being conducted
to inform the U.S. Environmental Protection Agency's (EPA's) current review of the National
Ambient Air Quality Standards (NAAQS) for carbon monoxide (CO). Given the significant time
constraints of this review,1 results of the analyses are provided in this document without
substantial interpretation. Rather, interpretative discussion of these results is provided in the
Policy Assessment document for the review (US EPA, 2010a).
1.1 BACKGROUND
The EPA is presently conducting a review of the national ambient air quality standards
for CO. Sections 108 and 109 of the Clean Air Act (Act) govern the establishment and periodic
review of the NAAQS. These standards are established for certain pollutants that may
reasonably be anticipated to endanger public health and welfare, and whose presence in the
ambient air results from numerous or diverse mobile or stationary sources. The NAAQS are to
be based on air quality criteria, which are to accurately reflect the latest scientific knowledge
useful in indicating the kind and extent of identifiable effects on public health or welfare that
may be expected from the presence of the pollutant in ambient air. Based on periodic reviews of
the air quality criteria and standards, the Administrator is to make revisions in the criteria and
standards, and promulgate any new standards, as may be appropriate. The Act also requires that
an independent scientific review committee advise the Administrator as part of this NAAQS
review process, a function performed by the Clean Air Scientific Advisory Committee
(CASAC).
The current NAAQS for CO includes two primary standards to provide protection for
exposures to carbon monoxide. In 1994, EPA retained the primary standards at 9 parts per
million (ppm), 8-hour average and 35 ppm, 1-hour average, neither to be exceeded more than
once per year (59 FR 38906). These standards were based primarily on the clinical evidence
relating carboxyhemoglobin (COHb) levels to various adverse health endpoints and exposure
modeling relating CO exposures to COHb levels. With the 1994 decision, EPA also reaffirmed
an earlier decision that the evidence did not support the need for a secondary standard for CO (59
FR38906).
A subsequent review of the CO NAAQS was initiated in 1997, which led to the
completion of the 2000 Air Quality Criteria Document for Carbon Monoxide (US EPA, 2000;
1 As noted below, the schedule for this review is governed by the terms of a court order.
1-1
-------
henceforth referred to as the 2000 AQCD) and a draft exposure analysis methodology document
(US EPA, 1999). EPA put on hold the NAAQS review when Congress requested that the
National Research Council (NRC) review the impact of meteorology and topography on ambient
CO concentrations in high altitude and extreme cold regions of the U.S. In response, the NRC
convened the Committee on Carbon Monoxide Episodes in Meteorological and Topographical
Problem Areas, which focused on Fairbanks, Alaska as a case-study. A final report, "Managing
Carbon Monoxide Pollution in Meteorological and Topographical Problem Areas" (NRC, 2003),
offered a wide range of recommendations regarding management of CO air pollution, cold start
emissions standards, oxygenated fuels, and CO monitoring. Following completion of this NRC
report, EPA did not conduct rulemaking to complete the review.
EPA initiated the current review of the NAAQS for CO on September 13, 2007, with a
call for information from the public (72 FR 52369) requesting the submission of recent scientific
information on specified topics. A workshop was held on January 28-29, 2008 (73 FR 2490) to
discuss policy-relevant scientific and technical information to inform EPA's planning for the CO
NAAQS review. Following the workshop, EPA outlined the science-policy questions that would
frame this review, outlined the process and schedule that the review would follow, and provided
more complete descriptions of the purpose, contents, and approach for developing the key
documents for the review in a draft Plan for Review of the National Ambient Air Quality
Standards for Carbon Monoxide (US EPA, 2008a). After CASAC and public input on the draft
plan, EPA made the final plan available in August 2008 (US EPA, 2008b). In January, 2010,
EPA completed the process of assessing the latest available policy-relevant scientific information
to inform the review of the CO standards. This assessment, the Integrated Science Assessment
for Carbon Monoxide (hereafter, "ISA") (US EPA, 201 Ob), includes an evaluation of the
scientific evidence on the health effects of CO, including information on exposure, physiological
mechanisms by which CO might adversely impact human health, an evaluation of the clinical
evidence for CO-related morbidity, and an evaluation of the epidemiological evidence for CO-
related morbidity and mortality associations.2
EPA's Office of Air Quality Planning and Standards (OAQPS) has developed this Risk
and Exposure Assessment (REA) describing the quantitative assessment conducted by the
Agency to support the review of the primary CO standards. This document is a concise
presentation of the methods, key results, observations, and related uncertainties associated with
the quantitative analyses performed. The REA builds upon the health effects evidence presented
in the ISA, as well as CASAC advice (Brain, 2009; Brain and Samet, 2009; Brain and Samet,
2 The ISA also evaluates scientific evidence for the effects of CO on public welfare which EPA will
consider in its review of the need for a secondary standard. EPA has not developed a quantitative risk assessment
for the secondary standard review.
1-2
-------
2010a; Brain and Samet, 201 Ob) and public comments on a scope and methods planning
document for the REA (hereafter, "Scope and Methods Plan") (US EPA, 2009a) and on the first
and second draft REA documents (US EPA, 2009b; US EPA, 2010c). This final REA was
completed by May 28, 2010, consistent with the court order governing the schedule for
completion of this review. The court order also specified that EPA sign for publication notices
of proposed and final rulemaking concerning its review of the CO NAAQS no later than October
28, 2010 and May 13, 2011, respectively.
The ISA and REA are used to inform the policy assessment and rulemaking steps that
lead to final decisions on the CO NAAQS. The policy assessment is described in a Policy
Assessment (hereafter, "PA") document, which include staff analyses of the scientific basis for
alternative policy options for consideration by the Administrator prior to rulemaking (US EPA,
2010a). The PA integrates and interprets information from the ISA and the REA to frame policy
options for consideration by the Administrator. The PA is intended to link the Agency's
scientific and technical assessments, presented in the ISA and REA, to judgments required of the
Administrator in determining whether it is appropriate to retain or revise the existing standards.
Development of the PA is also intended to facilitate elicitation of CAS AC's advice to the
Administrator on the adequacy of existing standards, and any new standards or revisions to
existing standards as may be appropriate.
1.2 PREVIOUS REVIEWS AND ASSESSMENTS
Reviews of the CO NAAQS completed in 1985 and 1994 included analyses of exposure
to ambient CO and associated internal dose, in terms of COHb levels, which were used to
characterize risks for populations of interest (50 FR 37484; 59 FR 38906). These prior risk
characterizations compared the numbers and percent of the modeled population that exceeded
several potential health effect benchmarks, expressed in terms of COHb levels. The COHb
levels of interest in these reviews were drawn from the evidence of COHb levels associated with
reduction in time to exercise-induced angina and other indicators of myocardial ischemia in
controlled human exposure studies involving short-term (shorter than 8 hours) exposures of
patients with diagnosed ischemic heart disease (IHD)3 to elevated CO concentrations (US EPA,
1979; US EPA, 1984; US EPA, 1991).
3 Ischemic heart disease is a category of cardiovascular disease associated with narrowed heart arteries; it is
often also called coronary artery disease (CAD) and coronary heart disease (CHD). Individuals with CHD have
myocardial ischemia, which occurs when the heart muscle receives insufficient oxygen delivered by the blood.
Exercise-induced angina pectoris (chest pain) occurs in many of them. Among all patients with diagnosed CAD, the
predominant type of ischemia, such as that indicated by ST segment depression, is asymptomatic (i.e., silent). Also,
patients who experience angina typically have additional ischemic episodes that are asymptomatic (2000 AQCD,
section 7.7.2.1).
1-3
-------
In the review completed in 1994, this characterization was performed for the population
of interest in the city of Denver, Colorado (US EPA, 1992; Johnson et al., 1992). That analysis
indicated that if the current 8-hour standard were just met, the proportion of the nonsmoking
population with cardiovascular disease4 experiencing exposures to ambient CO at or above 9
ppm for 8 hours decreased by an order of magnitude or more as compared to the proportion
under then-existing ambient CO levels, down to less than 1 percent of the total person-days in
that population. Likewise, just meeting the current 8-hour standard reduced the proportion of the
nonsmoking cardiovascular-disease population person days at or above COHb levels of concern
by an order of magnitude or more relative to then-existing ambient CO levels. More specifically,
upon just meeting the 8-hour standard, EPA estimated that less than 0.1% of the nonsmoking
cardiovascular-disease population would experience a COHb level of about 2.1% as a result of
exposure to ambient CO.5 A smaller percentage of the at-risk population was estimated to
exceed higher COHb levels. The analysis also considered additional exposure scenarios that
included certain indoor sources (e.g., passive smoking, gas stove usage). However, the indoor
sources were shown to contribute to total CO exposure to a much greater extent than ambient air
CO sources, leading to a conclusion that inclusion of indoor sources was of limited utility in
considering risk related to CO in ambient air. Further, it was noted that these indoor source
emissions would not be effectively mitigated by setting more stringent ambient air quality
standards (59 FR 38914).
In the review initiated in 1997, EPA consulted with CAS AC (Mauderly, 1999) on a draft
exposure analysis methodology document, Estimation of Carbon Monoxide Exposures and
Associated Carboxyhemoglobin Levels in Denver Residents (Johnson et al., 1999), using the
Probabilistic NAAQS Exposure Model (pNEM/CO, Version 2.0). Although the EPA did not
complete the review initiated in 1997, OAQPS continued work on the CO exposure assessment
to further develop the exposure assessment modeling component of EPA's Total Risk Integrated
Methodology (TRIM). A subsequent draft technical report (Johnson et al., 2000) was produced
documenting the application of the CO exposure and dose modeling methodology for two study
4 In characterizing the population of interest with regard to demographics (age and sex), the assessment for
the review completed in 1994 drew from estimates of the prevalence of ischemic heart disease (IHD) provided by
the National Health Interview Survey and corresponding estimates of undiagnosed ischemia developed by EPA.
Estimates of undiagnosed IHD were based on two assumptions: (1) there are 3.5 million persons in U.S. with
undiagnosed IHD (drawn from estimate by American Heart Association) and (2) persons with undiagnosed IHD are
distributed within the population in the same manner as persons with diagnosed IHD (US EPA, 1992).
5 In the 1992 assessment, the person-days (number of persons multiplied by the number of days per year
exposed) and person-hours (number of persons multiplied by the number of hours per year exposed) were the
reported exposure metrics. Upon meeting the 8-hour standard, it was estimated that less than 0.1% of the total
person-days simulated for the nonsmoking cardiovascular-disease population were associated with a maximum
COHb level greater than or equal to 2.1% (US EPA, 1992; Johnson et al., 1992).
1-4
-------
areas (Denver and Los Angeles). The exposure and dose estimates were obtained by applying
pNEM/CO version 2.1, a predecessor to the currently used Air Pollutants Exposure Model
(APEX), to adults with IHD residing within each urban area.6 This report was subjected to an
external peer review by three exposure modeling experts and convened by Science Applications
International Corporation (SAIC, 2001).
In the 2000 pNEM/CO assessment, the Denver study area was defined as all census tracts
located within 10 km of each of six fixed-site monitors within the Denver metropolitan area. Air
quality data for 1995 reported by these monitors were used to represent existing conditions in the
study area. Because the second highest non-overlapping 8-hour average CO concentration
equaled 9.5 ppm, the existing conditions in Denver for 1995 were considered to approximate just
meeting the 8-hour average CO standard.7 In a similar manner, the Los Angeles study area was
defined as all census tracts within 10 km often fixed-site monitors in the Los Angeles area,
though air quality data for 1997 were adjusted downwards so that the concentrations associated
with the design monitor just met the 8-hour NAAQS. A total of 15 distinct microenvironments
were modeled using a mass balance model accounting for the infiltration of outdoor (ambient)
CO, air exchange rates, as well as CO emissions from two indoor sources (residential gas stoves
and passive cigarette smoke).
In the 2000 pNEM/CO assessment, approximately 0.5% of the non-smoking IHD
population in both urban areas was estimated to experience a maximum end-of-hour COHb level
of about 2.0% as a result of exposure to ambient CO under air quality conditions just meeting the
current 8-hour standard.8 A smaller percentage of the at-risk population was estimated to exceed
higher COHb levels (e.g., <0.1% of persons were estimated to have COHb levels at or above
3.0% in either location). Indoor CO sources were a much greater contributor to COHb levels,
with their inclusion impacting a much larger portion of the simulated population at the higher
COHb levels (i.e., those persons with >1% COHb). For example, in Denver with indoor sources
included, nearly 20% of persons with IHD were estimated to have a maximum end-of-hour
COHb level of about 2.0%. In Los Angeles with indoor sources included, the estimated percent
of persons having a COHb level at or above 2.0% was lower (i.e., about 17%), though still a
6 This is consistent with the demographic group modeled in the 1992 assessment described above (Johnson
et al., 1992; US EPA, 1992), and drew from updated information with regard to prevalence demographics (Johnson
etal., 2000, section 2.5.2).
7 A rounding convention allows the second highest 8-hour average CO concentration (i.e., the design value
(DV)) to be as high as 9.4 ppm for the 8-hour CO NAAQS of 9 ppm (Laxton, 1990).
8 Note that the contemporaneous design value for Denver was 0.1 ppm above just meeting the current 8-
hour standard (9.5 versus 9.4 ppm).
1-5
-------
much greater percentage than that estimated in the absence of indoor sources (i.e., <1% of the
simulated at-risk population).
1.3 CURRENT REVIEW, CASAC ADVICE AND PUBLIC COMMENT
In preparing the draft Scope and Methods Plan for the REA (US EPA, 2009a), we
considered the scientific evidence presented in the ISA and the key science policy issues raised
in the IRP (US EPA, 2008b). EPA held a consultation with CASAC to solicit comments on the
draft Scope and Methods Plan during a May 2009 CASAC meeting. Public comments were also
requested (74 FR 15265). Those CASAC and public comments were considered in developing
the first draft REA (US EPA, 2009b) which implemented a simplified, screening-level approach
to assess population exposure and dose in two urban study areas, Denver and Los Angeles. The
current version of EPA's exposure model for CO (APEX/CO) was used to estimate exposure and
dose for a simulated at-risk population within 20 km of a single fixed-site monitor9 in each
location. Only two microenviroments were simulated; one in-vehicle and the second comprising
all other locations persons might visit.10 In using this simplified approach, the results were
considered by staff as likely more representative of upper level exposure and doses experienced
by a portion of the simulated at-risk population rather than the simulated at-risk population as a
whole.
Following the review of the first draft REA by CASAC (Brain and Samet, 2010a) and by
public commenters, we made a number of modifications to our initial approach and developed
the second draft REA (US EPA, 2010c) to better estimate population exposure and dose
distributions in each location modeled.11 Specifically in the second draft assessment, we 1)
expanded each of the two original modeling domains to include a greater number of ambient
monitors used as input to APEX, 2) increased the number of microenvironments modeled from
two to eight, 3) improved the representation of variability in estimated microenvironmental
concentrations, including in-vehicles, 4) included an algorithm that adjusts for spatial
heterogeneity in estimated outdoor concentrations across each model domain, 5) implemented
the mass-balance model for estimating concentrations in all indoor microenvironments, 6)
implemented the algorithm that allows commuters to experience home-tract and work-tract
9 The single monitor used in each location was the design monitor, that is, the monitor used to evaluate
concordance with the NAAQS. This monitor would measure the highest CO concentrations pertaining to the
NAAQS (i.e, the greatest 2nd highest 8-hour (or 1-hour) average CO concentration).
10 In the first draft REA, in-vehicle concentrations were estimated by applying a factor of 2.0 to ambient
CO concentrations. All other microenvironmental concentrations (i.e., both outdoor and indoor) were assumed to be
the same as measured at the single ambient monitor.
11 Public Comments on the first draft REA were submitted to the docket for this review and also presented
in March, 2010 at the CASAC second draft review meeting.
1-6
-------
ambient concentrations, and 7) expanded the at-risk population to also include undiagnosed
persons with CHD.
This final REA was produced in consideration of comments received on the second draft
REA from CAS AC (Brain and Samet, 201 Ob) and the public. The approach used to estimate
population CO exposure and COHb levels in this final REA has remained largely the same as
that used in the second draft REA, with the following adjustments or additions:
• Inclusion of a further expanded simulated at-risk population based on
prevalence rates for all types of heart disease (as well as including the previous
estimates of persons with undiagnosed CHD);
• Evaluation of endogenous CO production and ambient CO exposure separately
and their contributions to individual and population COHb levels in a larger
and more representative population subset;
• Identification of the specific microenvironments that contribute to low- and
high-level exposures;
• Inclusion of estimates of persons experiencing multiple occurrences per year at
or above selected COHb levels;
• Evaluation of the distribution of microenvironmental factors used to estimate
exposure concentrations in response to concerns regarding the application of
the microenvironmental algorithm; and
• Performance of additional sensitivity analyses including
o An evaluation of the impact additional monitors had on estimated
COHb dose levels experienced by the at-risk populations;
o An evaluation of the potential impact to estimated COHb dose levels
experienced by the at-risk populations by varying undiagnosed
prevalence rates by gender; and
o An evaluation of the potential impact to estimated COHb dose levels
experienced by a hypothetical anemic CHD population by using
alternative hemoglobin content distributions.
The chapters and appendices that follow describe the technical details in the exposure and
dose modeling approach used for this assessment, as well as the data analysis results. More
specifically,
• Chapter 2 provides a conceptual overview of the assessment of CO exposure
and risk with particular focus on aspects pertinent to this REA;
1-7
-------
Chapter 3 summarizes some of the general trends in CO ambient monitoring
across the U.S. and presents additional air quality analyses relevant to the two
urban areas of focus in this REA;
Chapter 4 provides a technical overview of EPA's APEX model including
model algorithms and databases common to most pollutant applications as well
as the description of approaches used specifically for estimating CO exposure
and dose;
Chapter 5 details the site- and pollutant-specific data used for the application of
APEX to the two study areas assessed in this REA;
Chapter 6 provides the exposure and dose results;
Chapter 7 presents an analysis of how variability was addressed in this
assessment and qualitatively characterizes how uncertainties in input data and
model algorithms might affect exposure and dose results; and
Chapter 8 summarizes the key observations associated with each chapter.
-------
1.4 REFERENCES
Brain JD. (2009). Letter from Dr. Joseph Brain to Administrator Lisa Jackson. Re: Consultation on EPA's Carbon
Monoxide National Ambient Air Quality Standards: Scope and Methods Plan for Health Risk and Exposure
Assessment. CASAC-09-012. July 14, 2009.
Brain JD and Samet JM. (2009). Letter from Drs. J.D. Brain and J.M. Samet to Administrator Lisa Jackson. Re:
Review of the EPA's Integrated Science Assessment for Carbon Monoxide (First External Review Draft).
EPA-CASAC-09-011. June 24, 2009.
Brain JD and Samet JM. (2010a). Letter from Drs. J.D. Brain and J.M. Samet to Administrator Lisa Jackson. Re:
Review of the Risk and Exposure Assessment to Support the Review of the Carbon Monoxide (CO) Primary
National Ambient Air Quality Standards: First External Review Draft. EPA-CASAC-10-006. February
12, 2010.
Brain JD and Samet JM. (2010b). Letter from Drs. J.D. Brain and J.M. Samet to Administrator Lisa Jackson. Re:
Review of the Risk and Exposure Assessment to Support the Review of the Carbon Monoxide (CO) Primary
National Ambient Air Quality Standards: Second External Review Draft. EPA-CASAC-10-012. May 19,
2010.
Johnson T, Capel J, Paul R, Wijnberg L. (1992). Estimation of Carbon Monoxide Exposure and Associated
Carboxyhemoglobin Levels in Denver Residents Using a Probabilistic version of NEM. Report prepared
by International Technology for U.S. EPA, Office of Air Quality Planning and Standards, Durham, NC,
Contract No. 68-DO-0062.
Johnson T, Mihlan G, LaPointe J, Fletcher K, Capel J. (1999). Estimation of Carbon Monoxide Exposures and
Associated Carboxyhemoglobin Levels in Denver Residents Using pNEM/CO (Version 2.0). Report
prepared by ICF Kaiser Consulting Group for U.S. EPA, Office of Air Quality Planning and Standards,
under Contract No. 68-D6-0064, WA Nos. 1-19 and 2-24. Available at:
http://www.epa.gov/ttn/fera/human_related.html.
Johnson T, Mihlan G, LaPointe J, Fletcher K, Capel J. (2000). Estimation of Carbon Monoxide Exposures and
Associated Carboxyhemoglobin Levels for Residents of Denver and Los Angeles Using pNEM/CO
(Version 2.1). Report prepared by ICF Consulting and TRJ Environmental, Inc., under EPA Contract No.
68-D6-0064. U.S. Environmental Protection Agency, Research Triangle Park, NC. Available at:
http://www.epa.gov/ttn/fera/human_related.html.
Laxton, WG. (1990). Memorandum on "Ozone and Carbon Monoxide Design Value Calculations". June 18, 1990.
Available at: http://www.epa.gov/ttn/naaqs/ozone/ozonetech/laxton.htm.
Mauderly J. (1999). Letter from Dr. Joe Mauderly, Chair, Clean Air Scientific Advisory Committee, to
Administrator Carole M. Browner. Re: Notification of a Consultation on the Estimation of Carbon
Monoxide Exposures and Associated Carboxyhemoglobin Levels in Denver Residents using pNEM/CO
(Ver. 2.0). July 12, 1999.
National Research Council. (2003). Managing Carbon Monoxide Pollution in Meteorological and Topographical
Problem Areas. Washington, DC, The National Academies Press.
SAIC. (2001). Memo to Harvey Richmmond, EPA, Technical Peer Review (including reviewers comments) of
"Estimation of Carbon Monoxide Exposures and Associated Carboxyhemoglobin Levels for Residents of
Denver and Los Angeles Using pNEM/CO (version 2.1)", Docket EPA-HQ-OAR-2008-0015. Available
at: http://www.epa.gov/ttn/fera/human_related.html.
US EPA. (1979). Air Quality Criteria for Carbon Monoxide. Research Triangle Park, NC: Office of Health and
Environmental Assessment, Environmental Criteria and Assessment Office, report no. EPA/600/8-79-022.
1-9
-------
US EPA. (1984). Review of the NAAQS for Carbon Monoxide: Reassessment of Scientific and Technical
Information. Office of Air Quality Planning and Standards, report no. EPA-450/584-904. Research
Triangle Park, NC.
US EPA. (1991). Air Quality Criteria for Carbon Monoxide. Research Triangle Park, NC: Office of Health and
Environmental Assessment, Environmental Criteria and Assessment Office, report no. EPA/600/8-90/045F.
US EPA. (1992). Review of the National Ambient Air Quality Standards for Carbon Monoxide: Assessment of
Scientific and Technical Information. Office of Air Quality Planning and Standards Staff Paper, report no.
EPA/452/R-92-004.
US EPA. (1999). Total Risk Integrated Methodology - TRIM.Expo Technical Support Document, External Review
Draft, November 1999. Office of Air Quality Planning and Standards, U.S. Environmental Protection
Agency, Research Triangle Park, NC, report no. EPA-453/D-99-001. Available at:
http://www.epa.gov/ttn/fera/trim_fate.html#1999historical.
US EPA. (2000). Air Quality Criteria for Carbon Monoxide. National Center for Environmental Assessment,
Office of Research and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC,
report no. EPA/600/P-99/001F. June 2000. Available at:
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=18163.
US EPA. (2008a). Draft Plan for Review of the National Ambient Air Quality Standards for Carbon Monoxide.
U.S. Environmental Protection Agency, Research Triangle Park, NC, report no. EPA/452D-08-001. Also
known as Draft Integrated Review Plan. Available at:
http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_pd.html.
US EPA. (2008b). Plan for Review of the National Ambient Air Quality Standards for Carbon Monoxide. U.S.
Environmental Protection Agency, Research Triangle Park, NC, report no. EPA/452R-08-005. Also known
as Integrated Review Plan. Available at: http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_pd.html.
US EPA. (2009a). Carbon Monoxide National Ambient Air Quality Standards: Scope and Methods Plan for Health
Risk and Exposure Assessment. U.S. Environmental Protection Agency, Research Triangle Park, NC,
report no. EPA-452/R-09-004. Available at: http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_pd.html.
US EPA. (2009b). Risk and Exposure Assessment to Support the Review of the Carbon Monoxide Primary
National Ambient Air Quality Standards: First External Review Draft. U.S. Environmental Protection
Agency, Research Triangle Park, NC, report no. EPA-452/P-09-008. Available at:
http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_rea.html.
US EPA. (2010a). Policy Assessment to Support the Review of the Carbon Monoxide Primary National Ambient
Air Quality Standards. U.S. Environmental Protection Agency, Research Triangle Park, NC, report no.
EPA-452/R-10-007. Available at: http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_pa.html.
US EPA. (2010b). Integrated Science Assessment for Carbon Monoxide. U.S. Environmental Protection Agency,
Washington, DC, report no. EPA/600/R-09/019F. Available at:
http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_isa.html.
US EPA. (2010c). Risk and Exposure Assessment to Support the Review of the Carbon Monoxide Primary
National Ambient Air Quality Standards: Second External Review Draft. U.S. Environmental Protection
Agency, Research Triangle Park, NC, report no. EPA-452/P-19-004. Available at:
http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_rea.html.
1-10
-------
2 CONCEPTUAL OVERVIEW: ASSESSING AMBIENT CARBON
MONOXIDE EXPOSURE AND RISK
In this chapter, we have summarized the conceptual model for assessing exposure to
ambient CO and associated health risk. Subsections focus on different components of the model
including identification of the key emission sources to ambient concentrations (section 2.1),
exposure pathways and key microenvironments (section 2.2), exposure and dose metrics (section
2.3), at-risk populations (section 2.4), health endpoints (section 2.5), and the risk characterization
approach (section 2.6). Section 2.7 presents the key observations for this chapter.
2.1 SOURCES OF CARBON MONOXIDE
Carbon monoxide in ambient air is formed primarily by the incomplete combustion of
carbon-containing fuels and photochemical reactions in the atmosphere. The amount of CO
emitted from these reactions, relative to the amount of carbon dioxide (CO2) generated, is
sensitive to conditions in the combustion zone. CO production relative to CO2 generally
decreases with any increase in fuel oxygen (O2) content, burn temperature, or mixing time in the
combustion zone (ISA, section 3.2). As a result, CO emissions from large fossil-fueled power
plants are typically very low because optimized fuel consumption conditions make boiler
combustion highly efficient. In contrast, internal combustion engines commonly used to power
mobile sources have widely varying operating conditions. Therefore, higher and more variable
CO emission levels result from the operation of these mobile sources (ISA, section 3.2). In
2002, CO emissions from on-road vehicles accounted for a substantial majority of total
emissions by individual source sectors in the U.S. (ISA, Figure 3-1).l As in previous NAAQS
reviews, mobile sources continue to be a significant emission source of CO to ambient air,
although in some areas, local stationary sources may be important contributors to ambient CO
concentrations.
Sources of CO inside buildings include infiltration of ambient air indoors, as well as,
where present, indoor (nonambient) sources such as gas stoves and tobacco smoke (ISA, section
3.6.5.2). In addition to infiltration of ambient air, CO inside motor vehicles may also receive
contributions from nonambient sources in the cabin, which can be substantial under air
ventilation modes that limit inflow from outside the vehicle (ISA, p. 3-89). In past CO
assessments, nonambient sources were estimated to have a substantially greater impact on the
highest total exposures experienced by the simulated population than have ambient sources (as
1 The 2002 National Emissions Inventory (NEI; US EPA, 2006) was the most recently available NEI
meeting data quality objectives for the ISA (US EPA, 2010a). The NEI includes data from various sources such as
industries and state, tribal, and local air agencies (ISA, p. 3-1).
2-1
-------
summarized in section 6.3 below). However, the focus of this REA, conducted to inform the
current review of the CO NAAQS, is on sources of ambient CO. We provide quantitative
estimates of population exposure and dose originating from ambient CO in two urban areas
(details on site selection are provided in chapter 3 below). The exposure modeling in this
assessment does not quantitatively estimate the contribution of indoor sources to estimated
population exposure and dose. In section 2.2 below, however, we qualitatively draw upon
available information regarding potential indoor source contributions to estimated population
exposure and dose.
2.2 EXPOSURE PATHWAYS AND IMPORTANT MICROENVIRONMENTS
Human exposure to CO involves the contact (via inhalation) between a person and the
pollutant in the various locations (or microenvironments) in which people spend their time.
Studies of personal exposure have generally found that the largest portion of the day is generally
spent indoors and the largest percentage of the time in which an individual is exposed to ambient
CO occurs indoors (ISA, sections 2.3 and 3.6). As a result, CO concentrations in indoor
microenvironments are an important determinant of an individual's total CO exposure. Recent
population exposure studies conducted in Milan, Italy support this conclusion (Bruinen de Bruin
et al., 2004), indicating that over 80% of the population exposure to CO can occur in indoor
microenvironments (ISA, Table 3-13). Taking into account the infiltration of ambient CO
indoors, indoor CO concentrations are similarly an important determinant in an individual's
exposure to ambient CO.
Microenvironments that may influence CO exposures typically include residential indoor
environments and other indoor locations, near-traffic outdoor microenvironments and other
outdoor locations, and inside vehicles. Consideration of microenvironmental exposures
illustrates the variability in the relationship between personal exposure and ambient
concentrations. For example, one study summarized the relationship between personal CO
exposure concentrations in five broadly defined microenvironments (i.e., indoor residence,
indoor other, outdoor near road, outdoor other, and in-vehicle) and ambient CO concentrations2
in Baltimore, MD (ISA, section 3.6.5.2; Chang et al., 2000). For most of the microenvironments,
the mean indoor-to-ambient and outdoor-to-ambient concentration ratios were about one, though
most of the individual ratios observed across this set of indoor and outdoor microenvironments
were less than one. With the exception of ratios for the in-vehicle microenvironments, which as
a group had most of the ratio distribution (as well as the mean ratio) above one, few ratios were
2 The ambient CO concentrations were those measured at a fixed site monitor (winter) or reflected average
concentrations across three fixed-site monitors (summer) (Chang et al., 2000).
2-2
-------
above unity (ISA, p. 3-85, Figure 3-46). Given the expected stability of CO as it infiltrates
indoor microenvironments from outdoor air and the lack of significant removal mechanisms of
CO in outdoor microenvironments, it is likely that the variability in personal- or
microenvironmental-to-ambient monitor and outdoor-to-ambient monitor concentration ratios is
the result of variability in outdoor concentrations that are not correlated with simultaneously
measured ambient concentrations at fixed-site monitors. This lack of correlation is a function of
the presence of local ambient and nonambient source emissions as well as local meteorology
Typically the highest CO exposure concentrations are experienced while inside vehicles.
Because motor vehicle emissions continue to be important contributors to ambient CO
concentrations, both the time spent in motor vehicles and the elevated CO concentrations
occurring on and near heavily trafficked roads continue to be important contributors to personal
exposures. For example, in the study summarized above on personal exposures occurring within
particular microenvironments (i.e., Chang et al., 2000), most in-vehicle CO exposure-to-ambient
concentration ratios were greater than one, with the median being approximately 2.5. The
average ratio was approximately 2.5 in summer, but a few somewhat higher in-vehicle
measurements in the winter period, contributed to a winter average of approximately 4 (ISA,
section 3.6.5.2, Figure 3-46; Chang et al., 2000 Figure 5).3 Given this relationship, it should not
be surprising that while about 8% of a person's time per day is spent in transit, approximately
13-17% of their total daily exposure occurs within an in-vehicle microenvironment (e.g., Bruinen
de Bruin et al., 2004; Scotto di Marco et al., 2005).
Similarly, the influence of mobile sources to microenvironmental concentrations and
personal exposurs was observed in the CO population exposure studies conducted in Denver CO
and Washington, DC during the winter of 1982 and 1983 (Akland et al., 1985).4 In both cities,
when comparing the distribution of measured CO concentrations from the monitoring network to
measured personal exposures, two common phenomena were observed. At the lowest
percentiles of each distribution, ambient CO concentrations were consistently greater than the
personal exposures. At the highest percentiles of each distribution, ambient concentrations were
consistently lower than the personal exposures (US EPA, 2000). These studies determined that
the highest average CO concentrations occurred when subjects were in a mobile source-
influenced microenvironment (e.g., inside parking garages, in-vehicles). Commute time was also
a factor; those who commuted 6 hours or more per week had higher average exposures than
3 Information on the distance of the ambient monitors from highly trafficked roadways or potential for in-
vehicle (nonambient) sources was not provided.
4 Both studies collected measurements and activity pattern diaries from a random sample of the population,
defined as including non-institutionalized, non-smoking residents, 18 to 70 years of age, who lived in each
respective city's metropolitan area (Akland et al., 1985).
2-3
-------
those who commuted fewer hours per week. Furthermore, mean CO concentrations within in-
vehicle microenvironments (ranging from 7.0 to 9.8 ppm) were greater than common outdoor
locations (ranging from 1.4 to 3.2 ppm) (US EPA, 2000). In considering the results from the
Denver and Washington personal exposure studies it is important to recognize that CO emissions
from motor vehicle sources have declined dramatically since the early 1980's when these studies
were conducted. Consequently, both ambient fixed-site CO concentrations and in-vehicle CO
concentrations have also been reduced significantly since that time period.
Given their influence on ambient exposures, exposures to CO near roadways and in
vehicle microenvironments are of particular importance in this assessment. Data from several
studies that have compared concentrations inside vehicles to concentrations immediately outside
vehicles indicate that indoor/outdoor concentration (I/O) ratios on average range from just above
to just below unity (Chan et al., 1991; Rodes et al., 1998; Boulter and McCrae, 2005; Sharp and
Tight, 1997). These studies are supported by a review by Flachsbart (1999) regarding other
studies published between 1982 and 1992 that measured interior and exterior CO concentrations
simultaneously during motor vehicle trips and reported I/O ratios just below unity (Petersen and
Allen, 1982; Koushi et al., 1992). Some studies reported no effect of ventilation setting on I/O
ratios, while others reported an effect. For example, one study described in the ISA indicated
I/O ratios could exceed unity with the ventilation set to re-circulate vehicle air (Abi Esber and
El-Fadel, 2008). However, the study authors attributed this finding to unaccounted sources of
CO that caused increases in CO concentrations within the vehicle cabin under those conditions
(ISA, section 3.6.6.2; Abi Esber and El-Fadel, 2008).
In general, the above results suggest that the I/O ratio tends toward unity when there are
no interior sources of CO, the automobile engine does not contribute directly to its own interior
concentrations, and the measurement probes are properly installed on the vehicle. This
conclusion is consistent with theoretical expectations for a non-reactive pollutant. For example,
CO concentrations inside vehicles can be estimated as a function of outside CO concentration,
air exchange rate, a penetration factor, and the emission rates of indoor sources (e.g., exhaust
leaks, smoking). If one assumes that (1) steady-state ventilation conditions exist, (2) the indoor
removal rate (K) is zero (i.e., no loss of CO as it moves from outside to inside the vehicle), and
(3) there are zero emissions from interior sources, then the CO concentration inside a vehicle can
be simplified to a function of outside CO concentrations and the penetration rate (i.e., infiltration
is generally equivalent to penetration).5 Under these stated conditions, the I/O ratio would
ultimately converge to unity.
5 See section 3.6.2 of the ISA.
2-4
-------
There are a few studies that have measured both in-vehicle and fixed-site monitoring
concentrations. The data from these studies can also inform the development of
microenvironmental factors used for estimating in-vehicle CO exposures. The ISA notes that
studies summarized in the 2000 CO AQCD found that in-vehicle CO concentrations were
generally two to five times higher than ambient CO concentrations obtained at fixed-site
monitors within the cities studied. For example, Shikiya et al. (1989) reported such
concentrations measured as part of a southern California study. When using the reported in-
vehicle CO measurements, one could estimate concentration ratios ranging from 1.8 to 2.7, a
range of ratios dependent on the time-of-year measurements were collected. Note however that
there are several factors that could contribute to variability in reported or calculated
concentration ratios. For example, often times in these measurement studies, the averaging time
associated with the companion measurements differ, that is there may be a much shorter
sampling interval for the in-vehicle measurement when compared with that of the ambient
monitor. More specifically, Shikiya et al. (1989) measured in-vehicle CO concentrations during
commutes lasting, on average, 33 minutes, while fixed site monitoring values averaged over 4
hours. It is possible that the time-averaged concentrations are less than that of the true fixed-site
concentrations that occurred during the 33 minute commute, perhaps resulting in an
overestimation of the concentration ratios when using this data. Furthermore, Shikiya et al.
(1989) reported seasonal differences for the in-vehicle CO concentrations (winter averaged 10.1
ppm; summer averaged 6.5 ppm), but not for the fixed-site monitor (average for both seasons
was 3.7 ppm). Typically ambient concentrations are greater in winter (e.g., ISA Figure 3-22 for
Los Angeles). Therefore, when using the fixed-site seasonal average and in-vehicle seasonally
stratified measurements from Shikiya et al. (1989) to calculate the ratios as was done above, the
winter I/O ratio may be overestimated while the summer value could be underestimated. In
addition to the factors mentioned above, this relationship can vary based on several other factors
that may influence the fixed-site monitor concentration, such as the nearby roadway traffic
density, the monitor siting characteristics (e.g., proximity to the roadway), and local scale
meteorology (e.g., downwind), with each described in greater detail in chapter 3. Of the few
studies reporting in-vehicle and companion fixed-site measurements, most do not measure all of
the potentially influential factors or provide the data stratified by such factors. Thus, a general
range of two to five may be adequate to represent the total variability for this particular
relationship, recognizing that there are limitations in the available measurement data to better
define this relationship.
Although not the focus of this review, indoor sources such as gas stoves and
environmental tobacco smoke can, where present, be important contributors to total CO exposure
and may, consequently, be of particular concern for such at-risk populations as individuals with
2-5
-------
cardiovascular disease, among others (see section 2.4 below). For example, some assessments
performed for previous reviews have included modeling simulations both without and with
indoor sources (gas stoves and tobacco smoke) to provide context for the assessment of ambient
CO exposure and dose (e.g., US EPA, 1992; Johnson et al., 2000). The 2000 pNEM/CO
simulations with indoor sources indicated that the impact of such sources on the proportion of
the population experiencing higher exposures and COHb levels can be substantial (Johnson et
al., 2000), as summarized in section 1.2 above and in section 6.3 below.6
2.3 EXPOSURE AND DOSE METRICS
Exposure concentration over a time period of interest (e.g., one hour or eight hours) is a
common exposure metric which reflects the integration of exposures to pollutant concentrations
that occur in each microenvironment in which time is spent (see section 4.4.6 below). In the case
of CO, for which the common mechanism underlying biological response is binding to heme
proteins, COHb level in blood is well recognized as an important internal dose metric used in
evaluating CO exposure and the potential for health effects (ISA, p. 2-4, sections 4.1, 4.2, 5.1.1).
Accordingly, COHb levels are used in this assessment.
Carboxyhemoglobin occurs in the blood due to endogenous CO production from
biochemical reactions associated with normal breakdown of heme proteins, as well as in
response to inhaled (exogenous) CO exposures (ISA, section 4.5).7 Levels of COHb associated
with endogenous CO production in healthy individuals have been described to range down to
0.3% and generally be less than 1% (ISA, pp. 4-9, 4-23, 2-6). However, the production of
endogenous CO and levels of endogenous COHb vary with several physiological characteristics
(e.g., slower COHb elimination with increasing age), as well as some disease states, which can
lead to higher endogenous levels in some individuals (ISA, section 4.5). Other factors affecting
CO uptake and elimination include physical activity and altitude (ISA, section 4.4).
The amount of COHb formed in response to exogenous CO is dependent on the CO
concentration and duration of exposure, exercise (which increases the amount of air removed and
replaced per unit of time for gas exchange), the pulmonary diffusing capacity for CO, ambient
pressure, health status, and the specific metabolism of the exposed individual (ISA, chapter 4;
2000 AQCD, chapter 5). The formation of COHb is a reversible process, but the high affinity of
CO for Hb, which affects the elimination half-time for COHb, can lead to increased COHb levels
6 As has been recognized in previous CO NAAQS reviews, such sources cannot be effectively mitigated by
setting more stringent ambient air quality standards (59 FR 38914), and are therefore not a focus of this assessment.
7 The dosimetry and pharmacokinetics of CO are discussed in detail in chapter 4 of the ISA.
2-6
-------
in some circumstances.8 Exogenous CO, ambient and nonambient9, can contribute to CO uptake
and increased levels of COHb. As recognized in sections 2.1 and 2.2 above, nonambient
(indoor) sources of CO (ISA, section 3.6.5.2) can result in much greater CO exposures and
associated COHb levels than those associated with ambient sources.10 Further, baseline COHb
levels in active smokers have been estimated to range from 3 to 8% for one- to two-pack-per-day
smokers. As a result of their higher baseline COHb levels, smokers may exhale more CO into
the air than they inhale from the ambient environment when not smoking. Tobacco smoking can
also contribute to increased CO exposures and associated COHb levels in nonsmokers (2000
AQCD, p. 7-4). In order to focus on the impact of ambient CO sources on population COHb
levels, exposure modeling for this REA does not include indoor CO sources; the impact of
indoor sources has been evaluated in previous assessments (see section 6.3 below).
As described in section 4.4.7 and Appendix B, blood levels of COHb have been
estimated in this REA using a nonlinear solution of the Coburn-Forster-Kane (CFK) model
(Coburn et al., 1965), which remains "the most extensively validated and applied model for
COHb prediction (ISA, section 4.2.3).
2.4 AT-RISK POPULATIONS
The term 'susceptibility' (and the term "at-risk") has been used to recognize populations
that have a greater likelihood of experiencing effects related to ambient CO exposure (ISA,
section 5.7). This increased likelihood of response to CO can potentially result from many
factors, including pre-existing medical disorders or disease states, age, gender, lifestyle or
increased exposures (ISA, section 5.7). For example, medical disorders that limit the flow of
oxygenated blood to the tissues have the potential to make an individual more susceptible to the
potential adverse effects of low levels of CO, especially during exercise. Based on the available
evidence in the current review, coronary artery disease (CAD), also known as coronary heart
disease (CHD) is the "most important susceptibility characteristic for increased risk due to CO
exposure" (ISA, p. 2-11). While persons with a normal cardiovascular system can tolerate
8 Fortunately, mechanisms exist in normal, healthy individuals to compensate for the reduction in tissue
oxygen caused by increasing levels of COHb. Cardiac output increases and blood vessels dilate to carry more blood
so that the tissue can extract adequate amounts of oxygen from the blood (ISA, chapter 4). As discussed in sections
2.4 and 2.5 below, however, there are several medical disorders that can make an individual more susceptible to the
potential adverse effects of low levels of CO, especially during exercise.
9 A significant source of nonambient CO long recognized as contributing to elevated COHb levels is
tobacco smoking (e.g., ISA, Figure 4-12).
10 For example, in addition to COHb estimates from previous assessments discussed in sections 2.1 and 2.2,
indoor source-related exposures, such as faulty furnaces or other combustion appliances, have been estimated in the
past to lead to COHb levels on the order of twice as high as those short-term exposures to ambient CO considered
more likely to be encountered by the general public (2000 AQCD, p. 7-4).
2-7
-------
substantial concentrations of CO if they vasodilate or increase cardiac output in response to the
hypoxia produced by CO, those that are unable to vasodilate in response to CO exposure may
show evidence of ischemia at low concentrations of COHb (ISA, p. 2-10). There is strong
evidence for this in controlled human exposure studies of exercising individuals with CAD,
which is supported by results from recent epidemiologic studies reporting associations between
short-term CO exposure and increased risk of emergency department visits and hospital
admissions for individuals affected with ischemic heart disease (IHD)11 and related outcomes
(ISA, section 5.7). This combined evidence, briefly summarized in section 2.5.1 below and
described in more detail in the ISA, supports the conclusion that individuals with CAD represent
the population most susceptible to increased risk of CO-induced health effects (ISA, sections
5.7.1.1 and 5.7.8). The 2007 estimate of the size of the U.S. population with coronary heart
disease, inclusive of those with angina pectoris (cardiac chest pain) and those who have
experienced a heart attack (ISA, Table 5-26) is 13.7 million people, some fraction of whom have
IHD (ISA, pp.5-117). Further, there are estimated to be several million additional people with
silent ischemia or undiagnosed IHD (AHA, 2003). In combination this represents a large
population that is more susceptible to ambient CO exposure when compared to the general
population (ISA, section 5.7).
Other types of cardiovascular disease12 may also potentially contribute to increased
susceptibility to the adverse effects of low levels of CO, especially during exercise (ISA, section
5.7.1.1). For example, some evidence with regard to other types of cardiovascular disease such
as congestive heart failure, arrhythmia, and non-specific cardiovascular disease, although more
limited for peripheral vascular and cerebrovascular disease, indicates that "the continuous nature
of the progression of CAD and its close relationship with other forms of cardiovascular disease
suggest that a larger population than just those individuals with a prior diagnosis of CAD may be
susceptible to health effects from CO exposure" (ISA, p. 5-117).
11 Ischemic heart disease is a category of cardiovascular disease associated with narrowed heart arteries,
which is often also called CAD (coronary artery disease) and CHD (coronary heart disease). Individuals with CHD
have myocardial ischemia, which occurs when the heart muscle receives insufficient oxygen delivered by the blood.
Exercise-induced angina pectoris (chest pain) occurs in many of them. Among all patients with diagnosed CAD, the
predominant type of ischemia, as identified by ST-segment depression, is asymptomatic (i.e., silent). Also, patients
who experience angina typically have additional ischemic episodes that are asymptomatic (2000 AQCD, section
7.7.2.1). In addition to such chronic conditions, CHD can include myocardial infarction (ISA, p. 5-24).
12 Cardiovascular disease comprises many types of medical disorders, including heart disease,
cerebrovascular disease (e.g., stroke), hypertension (high blood pressure), and peripheral vascular diseases. Heart
disease, in turn, comprises several types of disorders, including ischemic heart disease (i.e., CHD, CAD, myocardial
infarction, and angina), congestive heart failure, and disturbances of cardiac rhythm (dysrhythmias and arryhthmias)
(2000 AQCD, p. 7-7).
2-8
-------
Beyond persons with cardiovascular diseases, other populations may be potentially
susceptible to CO-related health effects. These populations are listed in the paragraphs below.
However, little empirical evidence is available by which to specify health effects associated with
ambient or near-ambient CO exposures in these potentially at-risk groups.
Populations with other preexisting diseases, such as chronic obstructive pulmonary
disease, diabetes or anemia have been identified as potentially susceptible to CO-induced health
effects (ISA, p. 5-123). For example, although there are no controlled human exposure or
epidemiological studies examining potential CO-induced effects in people suffering from
hematologic diseases, such as anemia, that affect oxygen-carrying capacity or transport in the
blood, it is reasonable to assume that the potential combination of hypoxic effects of CO together
with reduced oxygen availability and/or elevated baseline COHb levels in people suffering with
anemia13 may make those with anemia susceptible to CO-induced effects (ISA, section 5.7.1.4).
Included in this category of anemia diseases is sickle cell anemia, which is documented at a
higher incidence in African-American populations (ISA, section 5.7.1.4). Asthma and COPD are
other oxygen-limiting diseases which may be exacerbated by CO-related oxygen limitation.
Another population that may be potentially susceptible to CO includes those persons that may
have increased endogenous production of CO and potentially higher endogenous COHb levels
such as diabetics, for which a few epidemiological studies provide suggestive evidence of
increased risk for cardiovascular emergency department visits and hospital admissions compared
to non-diabetics in response to short-term CO concentrations (ISA, section 5.7.1.3).
Additionally, older adults, especially those with compromised cardiovascular function, represent
a potentially susceptible population (ISA, section 5.7.2.1).
The developing young (e.g., gestational development and newborns) may also represent a
population potentially susceptible to CO-induced health effects (ISA section 5.7.2.2; 2000
AQCD, section 7.7.1). For example, although the effects of CO on maternal-fetal relationships
are not well understood, fetal circulation is likely to have a higher COHb level than the maternal
circulation because of differences in uptake and elimination of CO from fetal Hb, which may
contribute to an enhanced sensitivity to CO exposure during gestation (ISA, section 5.7.2.2).
The comparatively higher rate of oxygen consumption and lower oxygen-transport capacity for
Hb in newborn infants as compared to adults may make them susceptible to CO-induced hypoxic
effects (2000 AQCD, section 7.7.1). Data from laboratory animal studies on CO developmental
toxicity suggest that prolonged exposure to high CO levels (>60 ppm) during gestation may
produce reduction in birth weight, transient cardiomegaly and delayed behavioral development,
13 Individuals affected with anemias of different etiologies may have low hematocrit, reduced capacity of
the blood to carry oxygen, or increased COHb levels due to increased endogenous CO production, all of which
would decrease the oxygen available for organs and tissues (ISA, pp. 118-119).
2-9
-------
or may disrupt the normal physiological roles of endogenous CO in the body (ISA, section
5.4.2.2); multiple-day prenatal animal exposures to exposures at or above 12 ppm indicated
effects on the developing auditory system (ISA, pp. 5-75 to 5-76). Limited epidemiological
evidence suggests some association of short-term ambient CO exposure with pre-term birth and
birth defects, and weak evidence suggests an association with reduction in birth weight and fetal
growth, and infant mortality (ISA, section 5.7.2.2; 2000 AQCD, section 7.7.1), although a clear
understanding of the mechanisms by which CO may induce those effects and at what exposure
levels is lacking (ISA, section 5.4.3).
Other populations that may be potentially susceptible due to impacts on endogenous CO
production, uptake and elimination of CO, or increased exposure concentrations include visitors
to high-altitude locations, persons using medicinal or recreational drugs with central nervous
system depressant properties or that that increase endogenous formation of CO, and people that
spend a substantial amount of time on or near heavily traveled roads which may contribute to
higher CO exposures (ISA, section 5.7).
As discussed in section 2.5 below, the sensitive endpoint which is the focus of this
quantitative assessment is exacerbation of myocardial ischemia. Based on the current evidence
for this endpoint, two target populations have been identified for this REA: (1) adults with CHD
(also known as ischemic heart disease IHD or CAD), both diagnosed and undiagnosed;14 and (2)
adults with diagnosed heart disease (HD) which includes CHD as well as other conditions (e.g.,
arrhythmias), along with undiagnosed CHD.
As mentioned above, there is little empirical evidence currently available by which to
specify health effects associated with relevant CO exposures in the other, potentially at-risk
groups identified above. Such evidence characterizing the nature of specific health effects of CO
in these populations is extremely limited and does not include COHb levels related to a particular
health effect identified in these potentially susceptible populations. Quantitative evidence
relating exposure or an applied dose to an adverse health outcome is requisite to the conduct of a
quantitative exposure and risk assessment. As a result, while we continue to recognize the
potential susceptibility of the larger cardiovascular disease population to health effects of CO, as
has been recognized in past reviews, as well as the potential susceptibility of several other
populations identified above (ISA, section 5.7), the at-risk populations simulated in this
assessment are individuals with CAD (diagnosed and undiagnosed and inclusive of individuals
with angina pectoris and heart attacks), as well as the larger HD population. We additionally
note that the still broader cardiovascular disease population and the potential susceptibility of
14 As described in section 1.2 above, this is the same population group that was the focus of the
exposure/dose assessments conducted previously (e.g., US EPA, 1992; Johnson et al., 2000).
2-10
-------
other populations is further considered with regard to the review of the CO NAAQS in the Policy
Assessment document (US EPA, 201 Ob).
2.5 HEALTH ENDPOINTS
Carbon monoxide elicits various health effects by binding to heme proteins and altering
the function of a number of heme proteins (ISA, section 2.4.2). The level of CO bound to
hemoglobin as carboxyhemoglobin (COHb) in the blood is the best characterized dose metric for
evaluating CO exposure and the potential for associated health effects, as described in section 2.3
above.
The best characterized health effect associated with CO levels of concern is hypoxia
(reduced oxygen availability) induced by increased COHb levels in blood (ISA, section 5.1.2).
The formation of COHb reduces the oxygen carrying capacity of the blood and impairs the
release of oxygen from oxy-hemoglobin complexes to the tissues. Accordingly, CO is especially
harmful in individuals with impaired cardiovascular systems (as discussed in section 2.4 above)
and the clearest evidence of causal relationships with CO exists for cardiovascular effects. In
characterizing the combined evidence, the ISA concluded that cardiovascular effects are likely
causally related to short-term exposures to CO at relevant concentrations, with "relevant CO
concentrations" defined in the ISA as "generally within one or two orders of magnitude of
ambient CO concentrations" (ISA, p. 2-5). The "most compelling evidence of CO-induced
effects on the cardiovascular system comes from a series of controlled human exposure studies
among individuals with coronary heart disease (CHD) (ISA, sections 5.2.4 and 5.2.6).
Other potential effects of CO which are less well characterized at relevant exposure
concentrations are those on the central nervous system, reproduction and prenatal development,
and the respiratory system (ISA, section 2.5). These additional health endpoints, for which the
limited available evidence is suggestive of causal relationships (ISA, sections 5.3, 5.4 and 5.5),
are also considered in this review and are discussed in detail in the ISA and summarized briefly
in section 2.5.2 below. Across the health endpoints identified here, however, the focus of the
quantitative analysis described in this document is on cardiovascular disease-related effects that
have been observed in adults with CHD, most specifically decreased time to exercise-induced
angina and changes to the ST-segment of an electrocardiogram that are indicative of myocardial
ischemia. This focus is based on the strength of the evidence and availability of quantitative
information from human studies of controlled CO exposures in which the resulting COHb levels
were associated with these effects (as discussed in sections 2.5.1 and 2.6 below).
2-11
-------
2.5.1 Cardiovascular Disease-related Effects
The best characterized cardiovascular disease-related effects associated with CO are
markers of myocardial ischemia observed in studies of controlled CO exposures of CHD
patients15 and effects on exercise duration and maximal aerobic capacity observed in controlled
exposure studies of healthy adults.16 As noted in the ISA, the decreases in exercise duration
among healthy adults (associated with COHb levels from 3 up to 20%) were relatively small and
only likely to be noticed by competing athletes, although they are considered to provide
coherence with the exercise-induced cardiovascular effects of greater concern that have been
demonstrated in CHD patients. The controlled human exposure studies involving individuals
with preexisting CHD provide strong evidence for an association between short-term exposure to
CO and measures of ischemia (US EPA, 2000, section 6.2.2; ISA, section 5.2.4). Multiple
controlled human exposure studies have shown that short-term exposure to CO and subsequent
elevation of COHb to levels of approximately 2-6% reduces time to onset of exercise-induced
myocardial ischemia in individuals with preexisting CAD, with no evidence of a threshold at the
lowest levels tested (ISA, section 5.2.4).
The controlled exposure study of principal importance is a large multi-laboratory study
designed to evaluate myocardial ischemia, as documented by reductions in time to change in the
ST-segment of an electrocardiogram17 and in time to onset of angina, during a standard treadmill
test, at CO exposures targeted to result in mean subject COHb levels of 2% and 4%, as measured
by gas chromatographic technique18 (ISA, section 5.2.4, from Allred et al., 1989a, 1989b, 1991).
In this study, subjects on three separate occasions underwent an initial graded exercise treadmill
test, followed by 50- to 70-minute exposures under resting conditions to average CO
concentrations of 0.7 ppm (room air concentration range 0-2 ppm), 117 ppm (range 42-202 ppm)
15 Epidemiological studies have consistently shown associations between ambient CO measurements and
emergency department visits and hospital admissions for IHD, which is coherent with the effects observed in
controlled human exposure studies of CAD patients (ISA, p. 2-14, section 5.2.6.1). Additional studies have shown
associations between ambient CO and hospital admissions for congestive heart failure and cardiovascular disease as
a whole (which includes IHD), although this evidence is not as consistent among studies as the IHD evidence (ISA,
sections 5.2.3 and 5.2.6.1).
16 Human clinical studies of individuals without diagnosed heart disease that were conducted since the 2000
CO AQCD did not report an association between CO and ST-segment changes or arrhythmia (ISA, section 2.5.1).
17 The ST-segment is a portion of the electrocardiogram, depression of which is an indication of insufficient
oxygen supply to the heart muscle tissue
18 As stated in the ISA, the gas chromatographic technique for measuring COHb levels "is known to be
more accurate than spectrophotometric measurements, particularly for samples containing COHb concentrations <
5%" (ISA, p. 5-41). CO-oximetry is a spectrophotometric method commonly used to rapidly provide approximate
concentrations of COHb during controlled exposures (ISA, p. 5-41). At the low concentrations of COHb (<5%)
more relevant to exposures to ambient CO, co-oximeters are reported to overestimate COHb levels compared to GC
measurements, while at higher concentrations, this method is reported to produce underestimates (ISA, p.4-18).
2-12
-------
and 253 ppm (range 143-357 ppm). After the 50- to 70-minute exposures, subjects underwent a
second graded exercise treadmill test, and the percent change in time to onset of angina and time
to ST endpoint between the first and second exercise tests was determined. Relative to clean-air
exposure that resulted in a mean COHb level of 0.6% (post-exercise), exposures to CO resulting
in post-exercise mean COHb concentrations of 2.0% and 3.9%19 were shown to decrease the
time required to induce ST-segment changes by 5.1% (p=0.01) and 12.1% (p<0.001),
respectively. These changes were well correlated with the onset of exercise-induced angina the
time to which was shortened by 4.2% (p=0.027) and 7.1% (p=0.002), respectively, for the two
CO exposures (ISA, section 5.2.4; Allred et al., 1989a, 1989b, 1991).
No human clinical studies have been specifically designed to evaluate the effect of
controlled exposures to CO resulting in study mean COHb levels lower than 2% (ISA, section
5.2.6). However, an important finding of the multi -lab oratory study was the dose-response
relationship observed between COHb and ischemia without evidence of a measurable threshold
effect (Allred et al., 1989b, 1991). As reported by the authors, the results comparing "the effects
of increasing COHb from baseline levels (0.6%) to 2 and 3.9% COHb showed that each
produced further changes in objective ECG measures of ischemia" implying that "small
increments in COHb could adversely affect myocardial function and produce ischemia" (Allred
et al., 1989b, 1991). For each 1% increase in COHb resulting from the experimentally increased
CO exposure concentrations the dose-response analysis performed by the authors indicated
decreases of 1.9% in time to exercise-induced angina and 3.9% in time to exercised-induced ST-
segment change in persons with pre-existing CAD (ISA, section 5.2.4, from Allred et al., 1989a,
1989b, 1991).
Other controlled human exposure studies (Adams et al., 1988; Anderson et al., 1973;
Kleinman et al., 1989, 1998) involving individuals with stable angina have confirmed the Allred
et al. findings at COHb concentrations between 3 and 6% (as measured by CO-oximeter) (ISA,
section 5.2.4). Among the evidence is also a study of a small group of patients with CAD which
reported no change in time to onset of angina or maximal exercise time following a 1 hour
exposure targeted to result in 4% COHb. A subsequent study conducted by the same laboratory
reported a significant increase in number of ventricular arrhythmias during exercise relative to
room air among individuals with CAD following a 1-hr CO exposure targeted to yield 6%
COHb, but not following a 1-hr exposure targeted to yield a COHb level of 4% (ISA, p. 5-42;
Sheps et al., 1987, 1990). Although there was no clear pattern across the different studies with
respect to the magnitude of the decreased time to onset of angina versus dose level, differences
19 The corresponding co-oximeter measured post-exercise levels were 2.7% and 4.7%. The post-exposure,
pre-exercise COHb levels for the two CO exposures were 2.4% and 4.7% by GC and 3.2% and 5.6% by co-oximetry
(ISA, p. 5-41).
2-13
-------
in study protocols and analytical methods do not allow for an informative pooled or quantitative
meta-analysis of the dose-response relationship across studies (ISA, section 5.2.4). Although the
subjects evaluated in the controlled human exposure studies described above are not necessarily
representative of the most sensitive population, the level of disease in these individuals ranged
from moderate to severe, with the majority either having a history of myocardial infarction or
having > 70% occlusion of one or more of the coronary arteries (ISA, p. 5-43).
We also note that, in the current review, a number of epidemiological studies are now
available that investigate associations of cardiovascular morbidity with ambient measurements of
CO (ISA, sections 5.2.4 and 5.2.5). These studies have observed associations between ambient
monitor CO concentrations and increases in emergency department visits and hospital
admissions for cardiovascular disease (ISA, sections 5.2.1.9). In considering the epidemiological
evidence in the case of CO, we recognize that there is coherence between the available clinical
and much expanded epidemiological evidence since the prior review, with regard to the health
effects of CO in the cardiovascular system (primarily for ischemia-related events). As discussed
in the ISA, the epidemiological studies reported associations of CO concentrations at ambient
monitors with emergency department visits and/or hospital admissions for IHD and other
cardiovascular disease-related outcomes that are plausibly related to the effects on physiological
indicators of myocardial ischemia (e.g., ST-segment changes) demonstrated in the controlled
human exposure studies, providing coherence between the two sets of findings. Furthermore, in
consideration of the epidemiological studies for cardiovascular outcomes in light of the larger
body of evidence, the ISA notes that the "known role of CO in limiting O2 availability lends
biological plausibility to ischemia related health outcomes following CO exposure", providing
coherence between the two sets of findings.
2.5.2 Other Effects
Other health effects for which the evidence is suggestive of causal relationships with CO
exposures include some effects on the central nervous system, reproduction and prenatal
development, and the respiratory system (ISA, section 2.5).
High CO exposures have "long been known to adversely affect central nervous system
(CNS) function", although the evidence does not include such effects associated with exposures
close to ambient CO concentrations (ISA, p. 5-49). Further, the evidence indicates that healthy
adults may be protected against such CNS effects at ambient levels through compensatory
responses such as increased cardiac output and cerebral blood flow, although these compensatory
mechanisms may be impaired among certain groups, such as those with reduced cardiovascular
function (ISA, section 5.3.3).
2-14
-------
Epidemiological and toxicological studies provide limited evidence of CO effects on the
developing fetus and newborn infants, as summarized in section 2.4 above. For example, some
epidemiological studies have reported associations of CO exposure during early pregnancy with
pre-term births and cardiac birth defects, with animal toxicological studies providing some
support and coherence for these effects at prolonged exposure concentrations ranging from 60-
500 ppm (ISA, section 5.4.3, pp. 5-80 and 5-120). The ISA notes that overall, there is limited
though positive evidence for CO-induced effects on pre-term birth and birth defects, and weak
evidence for a decrease in birth weight and fetal growth, and infant mortality; with animal
toxicological studies providing support and coherence for those effects. A clear understanding of
the mechanisms by which CO may induce those effects is still lacking (ISA, section 2.5.3).
New epidemiologic studies report positive associations for CO-induced lung-related
health outcomes, although interpretation is affected by uncertainties including with regard to the
biological mechanism that could explain CO-induced respiratory outcomes (ISA, section 5.5.5).
While only briefly summarized here, the evidence for the health effects identified here is
further discussed and considered with regard to the review of the CO NAAQS in the Policy
Assessment.
2.6 RISK CHARACTERIZATION APPROACH
In identifying an approach to characterize the risk of cardiovascular effects of exposures
to ambient CO, we considered 1) approaches employed in previous assessments, 2) the currently
available evidence regarding associations between CO concentrations and cardiovascular
outcomes, and 3) advice from CAS AC (Brain, 2009; Brain and Samet, 2009, 2010a, 201 Ob). As
summarized in section 1.2, the last CO NAAQS review included analyses of exposure to ambient
CO and associated internal dose, in terms of COHb levels, which were used to characterize risks
for the population of interest (US EPA, 1992). The prior risk characterization considered the
percent of the modeled population that exceeded COHb levels of interest which were drawn from
the evidence of COHb levels associated with a decrease in time to exercise-induced angina in
controlled human exposure studies involving short-term (shorter than 8 hours) exposures of
patients with diagnosed CAD20 to elevated CO concentrations (US EPA, 1991).
In the current review, the controlled human exposure studies among individuals with
CAD continue to provide the clearest evidence of CO-induced effects on the cardiovascular
system as the most sensitive endpoint. In contrast to epidemiological studies, human exposure
studies also provide quantitative information linking CO exposures through COHb levels with
20 Study subjects met certain criteria with respect to evidence of coronary artery disease, often also called
CHD or IHD.
2-15
-------
these effects. Among these studies, the multilaboratory study of Allred et al. (1989a, 1989b,
1991) continues to be the principal study informing our understanding of the effects of CO on
individuals with pre-existing CAD at the low end of the range of COHb levels studied (US EPA,
1991, 2000, 2010a). The strength of the evidence more broadly continues to support the use of
COHb level as the internal dose metric for assessing exposure to ambient levels of CO and
characterizing associated potential for cardiovascular disease-related health risk. Thus, based on
the strength of the evidence and the availability of quantitative information from controlled
human exposure studies, this REA also focuses on estimates of the percent of the simulated at-
risk population expected to experience maximum end-of-hour COHb levels of interest based on
findings of those studies.
As noted in section 2.5.1 above, a number of epidemiological studies are now available in
the current review that have observed associations between ambient monitor CO concentrations
and increases in emergency department visits and hospital admissions for cardiovascular disease
(ISA, sections 5.2.1.9). These studies are coherent with the controlled human exposure studies
(ISA, section 5.2.6), however, a number of uncertainties complicate their use for our purposes in
a quantitative risk assessment (ISA, pp. 2-14 to 2-17, section 5.2.3). These uncertainties are
discussed and considered in greater detail in the Policy Assessment. Accordingly, in light of the
longstanding body of evidence that links exposures to effects through the internal dose metric,
COHb, we have characterized health risk of ambient CO exposures in this assessment using
estimates of associated COHb levels and a benchmark level approach, with benchmarks
identified in consideration of the controlled human exposure literature.21
In drawing from the results of the controlled human exposure studies to inform the
characterization of potential CO risk in this assessment, staff considered a number of factors,
listed below.
• Myocardial ischemic effects, as documented by reductions in times to exercise-induced
change in the ST-segment of an electrocardiogram and to exercise-induced onset of
angina, were observed in response to CO exposures involving subjects with pre-
existing CAD. Staff gives primary focus here to the multi-laboratory study in which
COHb was analyzed by the more accurate GC method (Allred et al., 1989a, 1989b,
1991).
• Relative to clean-air exposure that resulted in a mean level of 0.6% COHb (post-
exercise), exposures to CO resulting in post-exercise mean COHb levels of 2.0% and
21 While not used for the purposes of this quantitative assessment, EPA is considering all of the current
health evidence, including the epidemiological studies, in the Policy Assessment, along with considerations based on
the risk and exposure assessment findings.
2-16
-------
3.9%22 were shown to decrease the time required to induce ST-segment changes by
5.1% (p=0.01) and 12.1% (p<0.001), respectively. These changes were well correlated
with the onset of exercise-induced angina, the time to which was shortened by 4.2%
(p=0.027) and 7.1% (p=0.002), respectively, for the two CO exposures (Allred et al.,
1989a, 1989b, 1991).
• There is no evidence of a threshold for the measures assessed at the lowest levels
tested, with incremental additions of COHb from baseline mean levels of 0.6% to 2 and
3.9% COHb showing changes in the monitored measures of ischemia (Allred et al.,
1989b, 1991). The average of the regressions of the individual study subject data for
these measures at baseline COHb and the two COHb levels resulting from the two
controlled CO exposures was summarized by the authors as indicating decreases of
roughly 1.9% in time to exercise-induced angina and 3.9% in time to exercise-induced
ST-segment change per 1% increase in COHb concentration in persons with pre-
existing CAD (ISA, section 5.2.4; Allred et al., 1989a, 1989b, 1991).
• Studies have not been designed to evaluate similar effects of exposures to increased
CO concentrations eliciting average COHb levels below the 2% target level of Allred
et al. (1989a, 1989b, 1991). In addition, these studies do not address the fraction of the
population experiencing a specified health effect at various dose levels. These aspects
of the evidence contributed to EPA's conclusion that at this time there are insufficient
controlled human exposure data to support the development of quantitative dose-
response relationships which would be required in order to conduct a quantitative risk
assessment for this health endpoint, rather than the benchmark level approach.
In drawing on this information, staff recognize the uncertainty associated with
interpretation of COHb levels estimated to result from CO exposure concentrations in this
assessment that are much lower than the CO exposure concentrations used in the clinical studies
to elicit increases in participant's COHb levels to target levels for the study.
We have reviewed the daily maximum end-of-hour COHb estimates developed in this
REA with attention to both the total COHb levels, which represent the combined influence of
ambient CO exposures and endogenous CO production, and the ambient CO contribution to
COHb levels, derived by subtracting the COHb produced in the absence of any CO exposure
from the total COHb level (see section 6.2 below). Results from the model simulations are
reported in terms of percent of population expected to experience daily maximum end-of-hour
COHb levels (or ambient CO contribution to daily maximum end-of-hour COHb levels) at or
above a series of levels that range as low as 1%. These results are interpreted in the Policy
Assessment document in light of potential health effects benchmarks.
22 Subjects were exposed to two levels of CO exposure, resulting in COHb levels in the range of 2.0 to
2.4% and 3.9 to 4.7%, respectively. The upper end of each range is the average COHb level obtained post-exposure
and the lower end is the average COHb level obtained after the subsequent exercise test (Allred et al., 1989a, 1989b,
1991).
2-17
-------
With regard to total COHb, staff identified benchmark levels of 1.5%, 2.0%, 2.5% and
3% COHb based on consideration of the evidence from controlled human studies of CHD
patients discussed above, and is inclusive of the range of levels considered in the review
completed in 1994 (US EPA, 1992). This range extends below the lowest mean COHb level
(e.g., 2.0% post-exercise in Allred et al., 1989b) resulting from controlled exposure to increased
CO concentration in the clinical evidence. This extension reflects comments from the CASAC
CO panel on the draft Analysis Plan (Brain and Samet, 2009) and consideration of the
uncertainties regarding the actual COHb levels experienced in the controlled human exposure
studies; that these studies did not include individuals with most severe cardiovascular disease;
the lack of studies evaluating effects of controlled short-term CO exposures resulting in COHb
levels below study mean 2.0-2.4% and the lack of evidence of an effect threshold at these levels.
We note that CASAC comments on the first draft REA recommended the addition of a
benchmark at 1% COHb and staff has presented results for this COHb level in this REA. In
considering this advice, we recognize, however, that a level of 1% COHb overlaps with the
upper part of the range of endogenous levels in health individuals as characterized in the ISA
(ISA, p. 2-6) and with the upper part of the range of baseline COHb levels in the study by Allred
et al. (1989b, Appendix B). As a result, while noting population dose estimates in relation to this
level, we have not placed weight on this level as a potential health effects benchmark in
discussions of the results below and in the Policy Assessment document.
We additionally consider the assessment results in light of the multi-laboratory clinical
study conclusions regarding response to specific increases in COHb level over the subjects' pre-
exposure or air exposure, with the increased COHb resulting from short-term controlled CO
exposure exposures of persons with pre-existing CAD (ISA, section 5.2.4; Allred et al., 1989a,
1989b, 1991).23 For this, we present the percentage of the simulated populations estimated to
experience ambient CO contribution or increment to daily maximum end-of hour COHb levels
greater than a series of levels that range as low as 1%. These results are interpreted in the Policy
Assessment document in light of potential health effects benchmarks, which for this ambient
contribution (or increment) to daily maximum end-of-hour COHb levels include the range from
23 Relative to clean-air exposure that resulted in a mean COHb level of 0.6% (post-exercise), exposures to
CO resulting in post-exercise mean COHb concentrations of 2.0% and 3.9% were shown to decrease the time
required to induce ST-segment changes by 5.1% (p=0.01)and 12.1%(p<0.001), respectively. These changes were
well correlated with the onset of exercise-induced angina the time to which was shortened by 4.2% (p=0.027) and
7.1% (p=0.002), respectively, for the two CO exposures. A dose-response analysis in which the individual
regressions of study subject responses at baseline COHb and at the two increased COHb levels were averaged was
summarized as indicating decreases of roughly 1.9% in time to exercise-induced angina and 3.9% in time to
exercised-induced ST-segment change per 1% increase in COHb concentration in persons with pre-existing CAD
(ISA, section 5.2.4; Allred etal., 1989a, 1989b, 1991).
2-18
-------
1.4% up to 2.4%, COHb increments associated with reduced time to exercise-induced angina and
ST-segment change in those studies.
The benchmark levels identified are used to interpret COHb levels estimated to occur in
the modeled population in response to exposures to ambient CO in different air quality scenarios
in light of the evidence discussed above for cardiovascular effects observed in individuals with
CHD when exposed to CO. More specifically, we have estimated the number of persons and
percent of the simulated at-risk population expected to experience COHb levels below each of
these potential health effect benchmark levels as a result of ambient CO exposures associated
with a set of air quality scenarios employed to inform the current review of the CO NAAQS (see
chapter 5 below). As noted in chapter 1 above, given the significant time constraints of this
review, results are provided in this document without substantial interpretation. Rather,
discussion of health risk and public health implications of these results in the context of the
NAAQS review is provided in the Policy Assessment.
2.7 KEY OBSERVATIONS
Presented below are key observations for this conceptual overview of the assessment of
ambient CO exposure and health risk.
• Carbon monoxide in ambient air is formed primarily by the incomplete combustion of
carbon-containing fuels and photochemical reactions in the atmosphere, with on-road
mobile sources representing significant sources of CO to ambient air.
• Microenvironments influenced by on-road mobile sources are important contributors to
ambient CO exposures, particularly in urban areas. Where present, other (nonambient)
CO sources can also be important influences on total CO exposure and on the impact of
ambient CO exposure on COHb levels.
• The formation of COHb is a key step in the elicitation of various health effects by CO.
Further, COHb level is commonly used in exposure assessment and is considered the
best biomarker for evaluating CO exposure and potential for health effects of concern.
• An individual's COHb levels reflect their endogenous CO production, as well as CO
taken into the body during exposure to ambient and nonambient CO sources. CO
uptake into the bloodstream during exposure is influenced by a number of variables
including internal levels of CO and COHb, such that net uptake may be lower or
negligible in instances where a preceding exposure has been substantially higher than
the current one. Thus, the magnitude of the change in COHb level in response to
ambient CO exposure may decrease with the presence of concurrent or preceding
nonambient CO exposure.
• Individuals with CHD are the population with greatest susceptibility to short-term
exposure to CO, and the population for which the current evidence indicates health
effects occurring at the lowest exposures. The evidence further indicates a potential for
other underlying cardiovascular conditions, particularly other types of heart disease, to
2-19
-------
contribute susceptibility to CO effects. Other populations potentially at risk include
those with diseases such as chronic obstructive pulmonary disease (COPD), anemia, or
diabetes, and those in prenatal or elderly life stages.
• Cardiovascular effects are the category of health effects for which the evidence is
strongest and indicative of a likely causal relationship with relevant short-term CO
exposures, particularly for people with CHD. Other endpoints for which the evidence
is suggestive of causal relationships include effects on the central nervous system,
reproduction and prenatal development, and the respiratory system.
• The specific cardiovascular effects occurring at the lowest COHb levels studied in
CHD patients are reduced time to exercise-induced angina and other markers of
myocardial ischemia, in particular, specific changes to the ST-segment of an
el ectrocardi ogram.
• Risk is characterized in this REA through evaluation of COHb estimated in simulations
involving ambient CO exposures experienced by two target populations: (1)
individuals with CHD (including undiagnosed CHD persons) and (2) individuals with
HD, including CHD (diagnosed and undiagnosed).
• Two types of COHb estimates are considered for the two target populations: (1) daily
maximum end-of-hour COHb levels and (2) ambient contribution to daily maximum
end-of-hour COHb levels (i.e., the change in COHb associated with ambient CO
exposure alone).
• Results from simulations are reported in terms of percent of the simulated at-risk
population expected to experience daily maximum end-of-hour COHb levels (or
ambient CO contribution to daily maximum end-of-hour COHb levels) at or above a
series of levels that range as low as 1%. These results are interpreted in the Policy
Assessment document in light of potential health effects benchmarks.
• For daily maximum end-of-hour COHb levels (absolute), these benchmarks
range from 1.5%, which is below the lowest study mean COHb level resulting
from experimental CO exposure in controlled human exposures of subjects
with CAD, up to 3.0%, a level within the range associated with effects in those
studies. For ambient contribution to daily maximum end-of-hour COHb
levels, the comparison benchmarks include the range from 1.4% up to 2.4%,
which are the COHb increments associated with effects in those studies.
• Beyond the at-risk populations and myocardial ischemia-related effects that are the
focus of this quantitative REA, the current evidence regarding other potentially
susceptible populations and other health effects associated with CO exposures is
discussed and considered with regard to the review of the CO NAAQS in the Policy
Assessment.
2-20
-------
2.8 REFERENCES
Abi-Esber L and El-Fadel M. (2008). In-vehicle CO ingression: validation through field measurements and mass
balance simulations. Sci Total Environ. 394:75-89.
Adams KF, Koch G, Chatterjee B, Goldstein GM, O'Neil JJ, Bromberg PA, Sheps DS, McAllister S, Price CJ,
Bissette J. (1988). Acute elevation of blood carboxyhemoglobin to 6% impairs exercise performance and
aggravates symptoms in patients with ischemic heart disease. J Am Coll Cardiol. 12:900-909
Akland GG, Hartwell TD, Johnson TR, Whitmore RW. (1985). Measuring human exposure to carbon monoxide in
Washington, DC, and Denver, Colorado, during the winter of 1982-1983. Environ Sci Technol. 19:911-
918.
Allred EN, Bleecker ER, Chaitman BR, Dahms TE, Gottlieb SO, Hackney JD, Pagano M, Selvester RH, Walden
SM, Warren J. (1989a). Short-term effects of carbon monoxide exposure on the exercise performance of
subjects with coronary artery disease. NEnglJMed. 321:1426-1432.
Allred EN, Bleecker ER, Chaitman BR, Dahms TE, Gottlieb SO, Hackney JD, Hayes D, Pagano M, Selvester RH,
Walden SM, Warren J. (1989b). Acute effects of carbon monoxide exposure on individuals with coronary
artery disease. Cambridge, MA: Health Effects Institute; research report no. 25.
Allred EN, Bleecker ER, Chaitman BR, Dahms TE, Gottlieb SO, Hackney JD, Pagano M, Selvester RH, Walden
SM, Warren J. (1991). Effects of carbon monoxide on myocardial ischemia. Environ Health Perspect.
91:89-132.
AHA. (2003). Heart and Stroke Facts. American Heart Association, Dallas, TX. Available at:
http://www.americanheart.org/downloadable/heart/1056719919740HSFacts2003text.pdf.
Anderson EW, Andelman RJ, Strauch JM, Fortuin NJ, Knelson JH. (1973). Effect of low level carbon monoxide
exposure on onset and duration of angina pectoris. Annals of Internal Medicine. 79:46-50.
Boulter P and McCrae I. (2005). Carbon Monoxide Inside Vehicles: Implications for Road Tunnel Ventilation. In:
Annual Research Review 2005. TRL Academy.
Brain JD. (2009). Letter from Dr. Joseph Brain to Administrator Lisa Jackson. Re: Consultation on EPA's Carbon
Monoxide National Ambient Air Quality Standards: Scope and Methods Plan for Health Risk and Exposure
Assessment. CASAC-09-012. July 14, 2009.
Brain JD and Samet JM. (2009). Letter to EPA Administrator Lisa Jackson: Clean Air Scientific Advisory
Committee's (CASAC) Peer Review of the Agency's 1st Draft Carbon Monoxide Integrated Science
Assessment. EPA-CASAC-09-011. June 24, 2009.
Brain JD and Samet JM. (2010a). Letter from Drs. J.D. Brain and J.M. Samet to Administrator Lisa Jackson. Re:
Review of the Risk and Exposure Assessment to Support the Review of the Carbon Monoxide (CO)
Primary National Ambient Air Quality Standards: First External Review Draft. CASAC-10-006. February
12, 2010.
Brain JD and Samet JM. (2010b). Letter from Drs. J.D. Brain and J.M. Samet to Administrator Lisa Jackson. Re:
Review of the Risk and Exposure Assessment to Support the Review of the Carbon Monoxide (CO)
Primary National Ambient Air Quality Standards: Second External Review Draft. EPA-CASAC-10-012.
May 19,2010.
Bruinen de BruinY, Hanninen O, Carter P, Maroni M, Kephalopoulos S, Scotto De Marco G, Jantunen M. (2004).
Personal carbon monoxide exposure levels: contribution of local sources to exposures and
microenvironmental concentrations in Milan. J Expo Anal Environ Epidemiol. 14:312-322.
2-21
-------
Chan C, Ozkaynak H, Spengler J, Sheldon L. (1991). Driver exposure to volatile organic compounds, CO, ozone,
and NOa under different driving conditions. Environ Sci Technol. 25(5):964-972.
Chang LT, Koutrakis P, Catalano PJ, Suh HH. (2000). Hourly personal exposures to fine particles and gaseous
pollutants - results from Baltimore, Maryland. J Air Waste Manage Assoc. 50:1223-1235.
Coburn R, Forster R, Kane P. (1965). Consideration of the physiological variables that determine the blood
carboxyhemoglobin concentrations in man. J Clin Invest. 44:1899-1910.
FlachsbartP. (1999). Human exposure to carbon monoxide from mobile sources. Chemosphere - Global Change
Science. 1:301-329.
Petersen WB and Allen RA. (1982) Carbon monoxide exposures to Los Angeles area commuters. J Air Pollution
ControlAssoc. 32:826-833.
Johnson T, Mihlan G, LaPointe J, Fletcher K, Capel J. (2000). Estimation of Carbon Monoxide Exposures and
Associated Carboxyhemoglobin Levels in Denver Residents Using pNEM/CO (Version 2.1), prepared by
ICF Kaiser Consulting Group for U.S. EPA, Office of Air Quality Planning and Standards, under Contract
No. 68-D6-0064, WANos. 1-19, 2-24, 2-30, and 3-3, Docket EPA-HQ-OAR-2008-0015-0010.
Kleinman MT, Davidson DM, Vandagriff RB, Caiozzo VJ, Whittenberger JL. (1989). Effects of short-term
exposure to carbon monoxide in subjects with coronary artery disease. Arch Environ Health. 44:361-369.
Kleinman MT, Leaf DA, Kelly E, Caiozzo V, Osann K, O'Niell T. (1998). Urban angina in the mountains: effects
of carbon monoxide and mild hypoxemia on subjects with chronic stable angina. Arch Environ.Health.
53:388-397.
Koushi P, Al-Dhowalis K, Niazi S. (1992). Vehicle occupant exposure to carbon monoxide. J Air Waste Manag.
42:1603-1608.
Rodes C, Sheldon L, Whitaker D, Clayton A, Fitzgerald K, Flanagan J, DiGenova F, Hering S, Frazier C. (1998).
Measuring Concentrations of Selected Air Pollutants Inside California Vehicles. Final Report. Prepared
by Research Triangle Institute under Contract No. 95-339. California Air Resources Board. Sacramento,
California.
Scotto di Marco G, Kephalopoulos S, Ruuskanene J, Janrunene M. (2005). Personal carbon monoxide exposure in
Helsinki, Finland. Atmos Environ. 39Z:2697-2707.
Sharp D and Tight M. (1997). Vehicle occupant exposure to air pollution. In: Policy, Planning, and Sustainability:
Proceedings of Seminars C and D Held at PTRC European Transport Forum, Brunei University. Pages
481-492.
Sheps DS, Adams KF Jr, Bromberg PA, Goldstein GM, O'Neil JJ, Horstman D, Koch G. (1987). Lack of effect of
low levels of carboxyhemoglobin on cardiovascular function in patients with ischemic heart disease. Arch
Environ Health. 42:108-116.
Sheps DS, Herbst MC, Hinderliter AL, Adams KF, Ekelund LG, O'Neil JJ, Goldstein GM, Bromberg PA, Dalton
JL, Ballenger MN, Davis SM, Koch GG. (1990). Production of arrythmias by elevated
carboxyhemoglobin in patients with coronary artery disease. Ann Intern Med. 113:343-351.
Shikiya D, Liu C, Kahn M, Juarros J, Barcikowski W. (1989). In-Vehicle Air Toxics Characterization Study in the
South Coast Air Basin. Office of Planning and Rules, South Coast Air Quality Management District.
US EPA. (1991). Air Quality Criteria For Carbon Monoxide. Research Triangle Park, NC: Office of Health and
Environmental Assessment, Environmental Criteria and Assessment Office; report no. EPA/600/8-90/045F.
2-22
-------
US EPA. (1992). Review of the National Ambient Air Quality Standards for Carbon Monoxide: Assessment of
Scientific and Technical Information. Office of Air Quality Planning and Standards Staff Paper, report no.
EPA/452/R-92-004.
US EPA. (2000). Air Quality Criteria for Carbon Monoxide. National Center for Environmental Assessment,
Office of Research and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
27711; report no. EPA/600/P-99/001F. June 2000. Available at:
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=18163.
US EPA. (2006). 2002 National Emissions Inventory Data and Documentation. U.S. Environmental Protection
Agency, Office of Air and Radiation, Office of Air Quality Planning and Standards, Research Triangle
Park, NC. Available at: http://www.epa.gov/ttn/chief/net/2002inventory.html.
US EPA. (2010a). Integrated Science Assessment for Carbon Monoxide. U.S. Environmental Protection Agency,
Washington, DC, report no. EPA/600/R-09/019F. Available at:
http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_isa.html.
US EPA. (2010b). Policy Assessment for the Review of the Carbon Monoxide National Ambient Air Quality
Standards, External Review Draft. Office of Air Quality Planning and Standards Staff Paper, report no.
EPA-452/P-10-005.
2-23
-------
3 AIR QUALITY CONSIDERATIONS
Ambient air quality data can be used as an indicator of exposure or used in conjunction
with other information to estimate exposure concentrations. How well the ambient air quality is
represented in a particular location is dependent on a number of factors including the ambient
monitoring network design relative to the spatial and temporal characteristics of the pollutant as
well understanding the concentration contribution from important local source emissions. This
chapter summarizes findings about the current air quality conditions and their temporal and
spatial distribution, with particular focus on aspects informative to the design and conduct of this
assessment and including descriptions of CO measurement methods, monitor siting
requirements, and monitor locations (section 3.1). Section 3.2 then draws upon the information
presented in sections 3.1, among other data, to select ambient air quality/study locations most
useful in meeting the objectives of the REA. Finally, key observations of the chapter are
presented in section 3.3.
3.1 AMBIENT CO MONITORING
In this section, a broad overview of the monitoring network is provided (section 3.1.1)
and is followed by a summary of analytical detection issues (section 3.1.2). Ambient CO
concentrations and their spatial and temporal variability are characterized in section 3.1.3.
Estimates of policy-relevant background (PRB) concentrations which are defined as those
ambient concentrations that would occur in the US in the absence of anthropogenic emissions in
continental North America are presented in section 3.1.4. And finally, section 3.1.5 presents an
analysis of the specific CO concentration trends observed in individual monitors.
3.1.1 Monitoring Network
Ambient CO concentrations are measured by monitoring networks that are operated by
state and local monitoring agencies in the US, and are funded in part by the EPA. The main
network providing ambient data for use in comparison to the NAAQS is the State and Local Air
Monitoring Stations (SLAMS) network. The subsections below provide specific information
regarding the methods used for obtaining ambient CO measurements and the requirements that
apply to states in the design of the CO network.
Minimum monitoring requirements for CO were revoked in the 2006 revisions to ambient
monitoring requirements (see 71 FR 61236, October 17, 2006). This action was made to allow
for reductions in measurements of some criteria pollutants (CO, SO2, NO2, and Pb) where the
current measured levels were all well below the applicable NAAQS. CO monitoring activities
5-1
-------
have been maintained at some SLAMS and these measurements of CO at these monitoring sites
are required to continue until discontinuation is approved by the EPA Regional Administrator.
CO monitors are typically sited to represent one of the following spatial scales.1
• Microscale: Data represent concentrations within a 100 meter (m) radius of the
monitor. For CO, microscale monitors are sited 2 - 10 m from a roadway.
Measurements are intended to represent the near-road or street canyon environment.
• Middle scale: Data represent concentrations averaged over areas defined by 100 - 500
m radii. Measurements are intended to represent several city blocks.
• Neighborhood scale: Data represent concentrations averaged over areas defined by 0.5
- 4.0 km radii. Measurements are intended to represent extended portions of a city.
In addition to monitoring required for determining compliance with the NAAQS, the
EPA is currently in the process of implementing plans for a new network of multi-pollutant
stations, called NCore, that is intended to meet multiple monitoring objectives. A subset of the
SLAMS network, these NCore stations are intended to address integrated air quality
management needs to support long-term trends analysis, model evaluation, health and ecosystem
studies, as well as the more traditional objectives of NAAQS compliance and Air Quality Index
reporting.2 The complete NCore network, required to be fully implemented by January 1, 2011,
will consist of approximately 63 urban and 20 rural stations and will include some existing
SLAMS sites that have been modified to include additional pollutant and meteorological
measurements. Each state will contain at least one NCore station, and 46 of the states plus
Washington, D.C. will have at least one urban station. CO will be measured using high
sensitivity monitors (see section 3.1.2 below), as will SO2, NO, and NOy.3 The majority of
NCore stations will be sited to represent neighborhood, urban, and regional scales, consistent
with the NCore network design objective of representing exposure expected across urban and
rural areas in locations that are not dominated by local sources.
3.1.2 Analytical Sensitivity
To promote uniform enforcement of the air quality standards set forth under the C AA,
EPA has established provisions in the Code of Federal Regulations (CFR) under which analytical
methods can be designated as federal reference methods (FRMs) or federal equivalent methods
(FEMs). Measurements for determinations of NAAQS compliance must be made with FRMs or
1 A complete description of spatial scales is listed in 40 CFR Part 58 Appendix D, section 1.2. Ambient
monitoring of other NAAQS pollutants such as NO2 and SO2 follow the same general spatial scales.
2 (http://www.epa.gov/ttn/amtic/ncore/index.html).
3 NCore sites must measure, at a minimum, PM2 5 particle mass using continuous and integrated/filter-based
samplers, speciated PM25, PM10.25 particle mass, speciatedPM10_25, O3, SO2, CO, NO/NOY, wind speed, wind
direction, relative humidity, and ambient temperature (http://www.epa.gov/ttn/amtic/ncore/index.html).
O O
3-2
-------
FEMs.4 Specifications for CO monitoring are designed to help states utilize equipment that has
met performance criteria utilized in the FRM or FEM approval process; operational parameters
are documented in 40 CFR Part 53, Table B-l. Given the levels of the current CO NAAQS (35
ppm, 1-hour; 9 ppm, 8-hour average), the required 1.0 ppm lower detectable limit (LDL)5 is well
below the NAAQS levels and is therefore determined sufficient for demonstration of
compliance. However, with ambient CO levels now routinely near or below 1 ppm, there is
greater uncertainty in a larger portion of the distribution of monitoring data because a large
percentage of these measurements are below the LDL of conventional monitors. For this reason,
a new generation of ambient CO monitors has been designed that provides measurements with
improved sensitivity at or below the typical ambient CO levels measured in most urban and all
rural locations. Additionally, the higher sensitivity CO measurements are needed to support
additional objectives such as validating the inputs to chemical transport models and assessing the
role of transport between urban and rural areas because policy relevant background CO
concentrations on the order of 0.1 ppm are well below the LDL of conventional monitors.
Newer GFC instruments have been designed for automatic zeroing to minimize drift (US EPA,
2000).
Currently, a total of 13 approved FRMs are in use in the SLAMS network, based on a
retrieval of data reported between 2005 and 2009. Among these methods, nine are "legacy"
monitors with a federal method detection limit (MDL)6 given as 0.5 ppm according to records in
EPA's Air Quality System (AQS).7 As discussed in the ISA (US EPA, 2010), many of the
reported concentrations in recentyears are near or below these MDLs (ISA, section 3.5.1.2).
Four of these new methods are high sensitivity methods with a federal MDL of 0.02 ppm and
there is a growing body of ambient data from high sensitivity CO instruments becoming
available. Among newer gas filter correlation (GFC) high sensitivity instruments, manufacturer-
declared LDLs range from 0.02 - 0.04 ppm, with 24-hour zero drift varying between 0.5%
4 As of August 2009, twenty automated FRMs had been approved for CO measurement. All EPA FRMs
for CO operate on the principle of non-dispersive infrared (NDIR) detection and can include the gas filter correlation
(GFC) methodology. An extensive and comprehensive review of NDIR, GFC, and alternative, non-FRM techniques
for CO detection was included in the 2000 CO AQCD (US EPA, 2000).
5 Defined in 40 CFR Part 53.23 as the minimum pollutant concentration which produces a signal of twice
the noise level.
6 Defined in 40 CFR Part 136 as the minimum concentration of a substance that can be measured and
reported with 99% confidence that the analyte concentration is greater than zero and is determined from analysis of
a sample in a given matrix containing the analyte.
7 Among several of the older instruments (Federal Reference Method codes 008, 012, 018, 033, 041, 050,
051, and 054), performance testing has shown LDLs of 0.62- 1.05 ppm, with 24-hour drift ranging from 0.044 -
0.25 ppm and precision ranging from 0.022 - 0.067 ppm at 20% of the upper range limit of the instrument (Michie
etal., 1983).
5-3
-------
within 1 ppm and 0.1 ppm, and precision varying from 0.5% to 0.1 ppm. EPA performed MDL
testing on several high sensitivity CO monitors in 2005 and 2006 following the 40 CFRPart 136
procedures. Those tests demonstrated MDLs of approximately 0.017 - 0.018 ppm, slightly
below the stated LDL of 0.02 - 0.04 ppm.
Based on a retrieval of data reported to AQS for the time period between 2005 and 2009,
a total of 36 high sensitivity CO monitors have reported data with the majority of these monitors
currently active. Most of these active monitors are associated with the implementation of the
NCore network. The extent to which high sensitivity monitors become integrated into non-
NCore SLAMS stations, however, will depend on the availability of funding for states to replace
operating legacy CO monitors as well as the possibility that monitoring requirements for CO
might either encourage or require increased sensitivity.
3.1.3 General Patterns of CO Concentrations
As discussed in the ISA, the spatial and temporal patterns of ambient CO concentrations
are heavily influenced by the patterns associated with mobile source emissions (ISA, section
3.2.1). Based on the 2002 National Emissions Inventory (NEI; US EPA, 2006), on-road mobile
sources comprise about half of the total anthropogenic CO emissions, though in metropolitan
areas of the US the contribution is as high as 75% of all CO emissions due to greater motor
vehicle density. For example, emissions in Denver county originating from on-road mobile
sources is about 71% of total CO emissions (ISA, section 3.2). When considering all mobile
sources (non-road and on-road combined), the contribution to total CO emissions is roughly 80%
nationwide and can be higher in some metropolitan areas. Again using Denver County as an
example, all mobile sources combined contribute to about 98% of the total CO emissions in the
county. Temporally, the national-scale anthropogenic CO emissions have decreased 35%
between 1990 and 2002. Nearly all the national-level CO reductions since 1990 are the result of
emission reductions in on-road vehicles (ISA, Figure 3-2).
Nearly 400 ambient monitoring stations report continuous hourly averages of CO
concentrations across the US. Over the period 2005-2007, 291 out of 376 monitors met a 75%
completeness requirement, spread among 243 counties, cities, or municipalities (ISA, section
3.4.2.2). No violations of the NAAQS were reported at these monitoring sites during this time
period. For example, in 2007, none of the monitors reported a second-highest 1-hour CO
concentration above 35 ppm, the level of the current 1-hour NAAQS, while only two sites
reported 2nd highest 1-hour CO concentrations between 15.1 and 35.0 ppm (ISA section 3.5.1.1).
Only five counties reported a 2nd highest 8-hour CO concentration of 5.0 ppm or higher.
The current levels of ambient CO across the US reflect the steady declines in ambient
concentrations that have occurred over the past several years. On average across the US the
5-4
-------
decline has been on the order of 50% since the early 1990s (ISA, Figure 3-34). As an example,
Figures 3-1 illustrate the trends observed in Denver and Los Angeles ambient concentrations, for
several selected monitors within the urban core of each area during 1993 through 2008. Note
that there is a significant decrease in the 2nd highest 1-hour and 8-hour average CO
concentrations since the last review.
Ambient monitor siting characteristics can influence ambient CO concentrations.
Microscale and middle scale monitors are commonly used to measure significant local source
impacts, while neighborhood and urban scale monitors are designated for population-oriented
monitoring (40 CFR Part 58 Appendix D). As CO concentrations primarily originate from
vehicle emissions, the microscale and middle scale data can be a useful indicator of near-road air
quality. Such data analyzed in the ISA were concluded to be consistent with hourly
concentrations reported in the literature for the near-road environment in the US (ISA, p. 3-57).
Further, when considering monitoring scale across ambient monitors in the US, the median
hourly CO concentration measured at microscale monitors was about 25% higher than at middle
scale monitors and 67% higher than at neighborhood scale monitors (ISA, Table 3-12). In
general, similar patterns were present in the 1-hour daily max, 1-hour daily average, and 8-hour
daily max distributions (ISA, Table 3-12). These patterns are also consistent with findings
presented by other researchers regarding the relative decrease in concentration with increasing
distance from roadways, though the magnitude of the relationship can vary. Two studies
summarized in the ISA (Zhu et al., 2002; Baldauf et al., 2008) indicate that near-road CO
concentrations measured within 20 meters of an interstate highway can range from 2-10 times
greater than CO concentrations measured as far as 300 meters from a major road possibly
influenced by wind direction and on-road vehicle density (ISA, Figures 3-29 and 3-30).
5-5
-------
22
Denver
Los Angeles
? 20
Q.
S 18 H
O
O 16
V
S? 14
10 -
8
6
4
2
0
16
O 12
O n^
V
O)
5 10
?.
O
I
00
4-1
I
g>
±
060370113
060371002
060371103
060371201
060371301
060372005
060374002
060590001
060591003
— -060595001
6
2
060370113
060371002
060371103
060371201
060371301
060372005
060374002
- - - 060590001
060591003
— -060595001
Year
Year
Figure 3-1.
i ml
Spatial and temporal trends in the 2 highest 1-hour (top) and 8-hour average (bottom) CO ambient monitoring
concentrations in Denver, Colorado (left) and Los Angeles, California (right), Years 1993 - 2008.
3-6
-------
While recognizing that monitoring site attributes are not available for all monitors in the
current network and that data for some attributes may not reflect current conditions,8 the ISA
also evaluated the average annual daily traffic (AADT) data available for each ambient monitor.
The ISA noted that only two microscale monitors and two middle scale monitors in the existing
network are sited at roads with >100,000 AADT, although it is not uncommon for roadways
within Consolidated Statistical Areas (CSAs) to have several roads with AADT > 100,000. The
AADT ranged from 160,000-178,000 for the near-road monitors used in the aforementioned
study by Zhu et al. (2002) where CO concentrations were up to 10 times greater than monitors
sited at 300 m from a major road.9 Existing microscale sites near roads having only moderate
traffic count data (<100,000 AADT) may record concentrations that are not substantially
different from those obtained from neighborhood scale measurements (ISA, section 3.5.1.3).
Within a specific urban area, however, consideration of only monitor scale or other
attributes reported in AQS, such as AADT estimates, may be of limited use in efforts to
characterize the monitoring data as to its representation of local near-road CO concentrations.
For example, of the five monitors meeting a 75% completeness criterion in the Denver CSA,
three were microscale and two were neighborhood scale (ISA, section 3.5.1.2). While one of the
microscale monitors sited within downtown Denver measured the highest hourly ambient CO
concentrations (ID 080310002), another microscale monitor (ID 080130009) located outside the
urban core measured the lowest hourly ambient CO concentrations (ISA, Figure 3-19). Further,
the AADT estimate for a major road near the microscale monitor within the urban core (ID
080310002, AADT=17,200) was lower than that listed for the microscale monitor outside the
urban core (ID 080130009, AADT=20,000) (ISA, Table A-2). And, a third microscale monitor
located 1.3 km from monitor ID 080310002, within the urban core, and measuring somewhat
lower CO concentrations (but not lower than the monitor outside the urban core) had only 500
AADT listed for the nearest major road. It is likely that the higher CO concentrations measured
at the downtown monitor reflect influences of the denser roadway network surrounding that
monitor in the downtown Denver area (ISA, Figure 3-17).10
Thus, to better characterize the representation of near-road CO concentrations for many
of the existing ambient monitors, additional analyses would need to be performed that go beyond
8 Recorded AQS monitoring site attributes are not always available for each monitor or may not always
reflect potential source influences. For example, of 24 CO monitors in the Los Angeles CSA, AQS had no
information regarding the monitoring scale for 16 monitors (ISA, Figure 3-22).
9 Local-scale meteorology may have also contributed to the heightened concentrations, given that the Zhu
et al. (2000) study was designed to capture CO concentrations downwind of the roadway.
10 We also recognize there is uncertainty in how well the AQS estimated AADT reflects current conditions
at this monitor site.
3-7
-------
the AQS standard list of monitoring site attributes. Such analyses could include local-scale
meteorological data, using GIS to determine detailed monitor-to-roadway characteristics (e.g.,
monitor distance from roadways, the number and type of roads within close proximity of the
monitor), and obtaining current traffic count data for all roads.
Carbon monoxide also exhibits hourly variability within a day, with two distinct temporal
patterns noted for weekdays and weekends (ISA, section 3.5.2.2). The diurnal variation is
inherently linked to the typical commute times-of-day that occur within urban locations. In
general, in recent years observed mean and median concentrations for all hours of the day and
across all monitors within urban areas demonstrated limited variability, however 90th and 95th
percentile hourly concentrations generally exhibit early-morning and late afternoon peak CO
concentrations during weekdays (ISA, Figure 3-36). The weekend diurnal variation in ambient
CO concentrations was much lower than that occurring during weekdays as expected due to the
relative absence of commuter vehicle traffic during the morning and evening hours of the day.
Most urban areas have relatively stable concentrations throughout weekend days at each of the
selected percentiles, though a few locations (e.g., Phoenix, Los Angeles, Seattle) did have a more
pronounced late afternoon peak (ISA, Figure 3-37).
We investigated local hourly variation at two separate CO monitors located in Denver
and Los Angeles to illustrate similar trends. The monitor in Denver is a microscale monitor
located in downtown Denver and expected to reflect concentrations resulting from dense
downtown traffic in that city; it is the monitor measuring the highest ambient CO concentrations
in the Denver area. The monitor in Los Angeles is a middle scale monitor located in Lynwood;
it is also the monitor measuring the highest ambient CO concentrations in the Los Angeles area.
Figure 3-2 indicates that on average, peak ambient CO concentrations that occur during typical
commute times in Denver ranged from about 1 to 5 ppm during weekdays in 1995, while
currently, ambient CO concentrations during morning and afternoon commutes range from about
1 to 2 ppm. Weekends tend to exhibit less variability throughout the day. On average, CO
ambient concentrations generally ranged from 1 to 3 ppm throughout the day in 1995, while
current weekend concentrations are less than 1 ppm for most hours of the day. In Los Angeles,
both the concentration levels and variability are greater than when compared with similar years
and times of day in Denver (Figure 3-3). Peak ambient CO concentrations are more prominent
during morning commutes and generally ranged from 2 to 10 ppm in 1995, while when
considering more recent ambient monitor concentration (2006), most commuting times are
associated with hourly concentrations ranging from between 1 and 5 ppm. The weekend profile
exhibits some variation when considering either year, with maximum concentration levels and
variability commonly exhibited during the late-night/early morning hours.
3-8
-------
Weekdays in 1995
Weekends in 1995
o
o
0 o
-------
Weekdays in 1997
Weekends in 1997
o
o
o
\ I I I I I I I I I I I I I I I I I I I I I I I
0123456789 11 13 15 17 19 21 23
Clock Hour
o
o
o
\ I I I I I I I I I I I I I I I I I I I I I I I
0123456789 11 13 15 17 19 21 23
Clock Hour
o
o
Weekdays in 2006
Qiifl
I I I I I I I I I I
01 23456789
11
Clock Hour
\ I
13
15
I I
17
I I I
19 21
23
o
o
Weekends in 2006
i i i i i i i i i i
0123456789
\ I I I
11 13
Clock Hour
15 17
I I \
19 21
I i
23
Figure 3-3. Diurnal distribution of 1-hour CO concentrations in Los Angeles (Monitor 060371301) by day-type (weekdays-
left; weekends-right), years 1997 (top) and 2006 (bottom). The box encompasses concentrations from the 25th to 75th
percentiles or IQR, the line bisecting the box is the median, the solid dot within the box is the mean, the whiskers
represent 1.5 times the IQR, and concentrations outside the whiskers are indicated by open circles. Note there are
differences in the y-axis scale for the two monitoring years.
3-10
-------
3.1.4 Policy-Relevant Background Concentrations
EPA has generally conducted NAAQS risk assessments that focus on the risks associated
with ambient levels of a pollutant that are in excess of policy-relevant background (PRB).
Policy-relevant background levels are defined, for purposes of this document, as concentrations
of a pollutant that would occur in the US in the absence of anthropogenic emissions in the US,
Canada, and Mexico.
Over the continental US (CONUS), the 3-year (2005-2007) average CO PRB
concentration is estimated to range from 0.118 to 0.146 ppm (ISA, section 3.5.4). Outside the
CONUS, the 3-year average CO PRB in three Alaskan sites is estimated to range from 0.127 to
0.135 ppm, and from 0.095 to 0.103 ppm in two Hawaiian monitoring locations. The estimated
PRB concentrations exhibit significant within-location seasonal variation, with minimum
concentrations observed in the summer and fall and maximum concentrations occurring in the
winter and spring. For example, PRB in two California sites is estimated to range from about
0.085 to 0.170 ppm, and PRB in one site in Colorado ranged from about 0.080 to 0.140 ppm
(ISA, Figure 3-43).
Given that ambient concentrations of interest in this REA are well above the estimated
PRB levels discussed above and, thus the contribution of PRB to overall ambient CO
concentrations is very small, EPA is characterizing risks associated with ambient CO levels
without regard to estimated PRB levels.
3.1.5 Within-Monitor CO Concentration Trends
The previous section addressed general trends in ambient concentrations. Of particular
interest in this assessment is how concentrations have changed at a specific monitor over time.
This is an important consideration in determining how best to address alternative air quality
conditions. These alternative air quality conditions are useful in evaluating how varying
distributions of air quality might affect different exposure scenarios. In other recent NAAQS
reviews for NO2 (US EPA, 2008) and SO2 (US EPA, 2009) it was determined that the
relationship between high concentration and low concentration years of ambient monitoring data
was mainly proportional (Rizzo, 2008, 2009), that is all concentrations across the entire
distribution at a single monitor changed in equivalent amounts over time. We needed the
relationship to adjust current air quality because, at the time of the NAAQS reviews, the current
ambient NO2 and SO2 concentrations were far below that expected to just meet the current
standards.
Knowledge of this relationship for ambient CO concentrations is also needed to develop
alternative air quality conditions for use in some of the exposure scenarios investigated in this
REA. Ambient CO concentration data were obtained from AQS for several monitors in Los
3-11
-------
Angeles for two years: 1997 - representing a high concentration year and 2006 - representing a
low concentration year. In Denver, the year 1995 was selected to represent a high concentration
year, while 2006 was selected to represent a low concentration year. As was done for prior
NAAQS reviews (Rizzo, 2008, 2009), 75% completeness criteria were applied in selecting valid
monitoring data,11 the 1-hour daily maximum concentration for each day was identified, and the
0 through 100th percentiles of the distribution were calculated (by 1% increments). Then the
percentiles for the low concentration year were paired and plotted against those calculated for the
high concentration year at each individual monitor. Figure 3-4 illustrates the results for four
monitors in Denver, while Figure 3-5 illustrates a similar comparison for four monitors in Los
Angeles. A simple linear regression was also calculated and plotted, along with the regression
slope, intercept, and fit statistic (R2). As shown by the relationships and fit statistics in each
location, there is a very strong linear relationship when comparing each year of data within each
monitor. In general, the regression slopes and intercepts are similar for monitors within each
location, indicating a similarity in the rate of change in concentration occurring at the monitors
within each location. There are however, at most of the sites, instances where upper percentile
values deviate from linearity (i.e., > 99th percentile 1-hour daily maximum concentration).
Concentrations deviating above the best fit line indicate that these upper percentile
concentrations have not declined at the same rate as the middle of the distribution at that
particular monitor (e.g., Figure 3-5, monitor ID 060370113). In addition, many of the estimated
regression intercepts are positive, though most are < 0.1 ppm. A positive intercept also indicates
a larger percentage decline between high and low year concentrations in the upper end of the
distribution relative to that of the middle and lower percentiles. However, given that there are a
limited number of points deviating from linearity and that regression slopes and intercepts are
similar for most of the monitors within each location and having mainly small intercepts, this
analysis provides adequate support for adjusting air quality by a proportional method.
11 Monitoring sites first had to have 75% of hours reported in each day to be considered as a valid day.
Then each quarter had to have at least 75% valid days to be complete and all four quarters had to be complete across
the year for the site to be retained.
3-12
-------
080310002
080310013
E
Q.
Q.
7 -
6
8^5
II
= 5 3
Q ~
to 2 -
o
CM
1 -
0
E
Q.
Q.
3
85
I!
2
Q
S
1 -
y = 0.2846X + 0.2799
R2 = 0.9761
5 10 15 20
1995 Daily Maximum CO (ppm, 0-100 percentile)
080310014
y = 0.3487x + 0.1085
R2 = 0.9932
2 4 6 8 10
1995 Daily Maximum CO (ppm, 0-100 percentile)
E c 3
3
-------
060370113
060371103
E
Q.
Q.
3
is
1 "
•I £ 2
* Q.
1 -
10
E
Q.
a.
is
1 a
._
6
:,° 4
s~
= 0.4105x + 0.143
R2 = 0.992
1997 Daily Maximum CO (ppm, 0-100 percentile)
060371301
y=0.4106x +0.1019
R2 = 0.9945
I 3-|
81
ig
I I 2
«£
Sg
_>» *;•
-s o
1 -
E
Q.
o
O -y
2
1
5 10 15
1997 Daily Maximum CO (ppm, 0-100 percentile)
20
y = 0.4112x- 0.0053
R2 = 0.9901
10
1997 Daily Maximum CO (ppm, 0-100 percentile)
060591003
y=0.4422x-0.0148
R2 = 0.9927
234567
1997 Daily Maximum CO (ppm, 0-100 percentile)
Figure 3-5. Comparison of a high concentration year (1997) versus a low concentration year (2006) at four ambient monitors
in Los Angeles. The 0 through 100
year.
th
percentiles of the 1-hour daily maximum CO are plotted for each monitor-
3-14
-------
We were also interested in estimating the within-monitor temporal variability for three air
quality metrics. The first air quality metric was the current design value, that is, the 2nd highest
8-hour average CO concentration in a year. The next two air quality metrics we compared were
the 99th percentile 1-hour and 8-hour daily maximum CO concentrations. We evaluated the
within-monitor temporal variability using two comparisons: one using historical versus current
air quality data and the other comparing year-to-year variability of these upper percentile
concentrations within the air quality distribution. Two three-year periods (1995-1997 and 2005-
2007) were chosen to represent historical and recent air quality, respectively. We limited the
analysis to four monitors within the Denver CSA and ten monitors within the Los Angeles CSA,
with all monitor data meeting standard requirements for data completeness. In addition to the
temporal evaluation of the air quality metrics, a limited analysis of the spatial variability across
the two periods is also provided for the selected monitors in each area.
Tables 3-1 and 3-2 provide results for the current design value in Denver and Los
Angeles, respectively. As shown by the Tables, there is a wide range in the temporal variability
of the 2nd highest 8-hour average CO concentration in both locations, however, the relative
variability, as indicated by the coefficient of variation (COV),12 is slightly less for the recent air
quality when compared with the historical air quality. For example, in Denver the COV ranges
from 4-27 percent (mean = 13%) for the historical data, while the recent data temporal COV
ranges from 3-23 percent (mean = 10%) (Table 3-1). Note also that the design value decreases
with increasing monitoring year over the selected three-year periods in both Denver and Los
Angeles (and of course is consistent with Figures 3-1 and 3-2), though this trend is more
prevalent when considering the historical air quality data. In addition, the magnitude of the
spatial variability tends to vary from year-to-year as indicated by the COV, though there are
differences in the historical versus recent air quality pattern by location. In Denver, there was
generally less spatial variability in the 2nd highest 8-hour concentration when comparing the
recent to historical air quality data. There was no apparent trend in year-to-year spatial
variability for Los Angeles as both air quality periods had a mean COV of about 31% (Table 3-
2).
Similar temporal trends are observed with the 99th percentile 1-hour daily maximum
concentrations when comparing historical versus recent air quality (Tables 3-3 and 3-4 for
Denver and Los Angeles, respectively). The temporal variability in the recent air quality data
was also less than that of the prior air quality metric (i.e., the 2nd highest 8-hour average),
averaging about 4% COV in Denver and 7% COV in Los Angeles across that 3-year period. The
12 The COV is calculated here by dividing the standard deviation (std) by the mean, then multiplying by
100.
3-15
-------
year-to-year spatial variability for this metric is also consistent with that stated above. In
Denver, the COV on average was less for the recent air quality when compared to that of the
historical data. There was little difference in the year-to-year spatial variability in Los Angeles
when considering the two air quality periods. Results for the 99th percentile 8-hour daily
maximum concentrations were more similar to the results for the 2nd highest 8-hour average
concentration than the 99th percentile 1-hour daily maximum (Tables 3-5 and 3-6, for Denver and
Los Angeles, respectively).
3-16
-------
Table 3-1. Within monitor temporal variability in Denver using historical (1995-97) and recent (2005-07) air quality data -
»nd
2 highest 8-hour average.
Monitor
31-0002
31-0013
31-0014
59-0002
mean
std
cov
Historical Air Quality - 2nd highest 8-hour average
1995
9.5
6.2
5.9
4.6
6.6
2.1
32
1996
7.3
5.2
5.7
4.3
5.6
1.3
22
1997
5.5
4.7
6.2
4.9
5.3
0.6
12
mean
7.4
5.4
5.9
4.6
5.8
1.2
20
std
2.0
0.8
0.2
0.3
COV
27
14
4
7
Recent Air Quality - 2nd highest 8-hour average
2005
2.6
2.4
2.1
1.8
2.2
0.3
14
2006
3.1
2.5
3.0
2.0
2.6
0.5
19
2007
2.8
2.8
mean
2.8
2.4
2.5
1.9
2.4
0.4
16
std
0.3
0.1
0.6
0.1
COV
9
3
23
6
Table 3-2. Within monitor temporal variability in Los Angeles using historical (1995-97) and recent (2005-07) air quality
»nd
data - 2 highest 8-hour average.
Monitor
37-0113
37-1002
37-1103
37-1201
37-1301
37-2005
37-4002
59-0001/7
59-1003
59-5001
mean
std
COV
Historical Air Quality - 2nd highest 8-hour average
1995
9.4
10.9
7.9
9.4
11.7
8.6
6.3
7.3
5.3
6.4
8.3
2.1
25
1996
8.5
8.5
7.5
6.7
14.3
6.9
6.2
6.1
6.5
6.3
7.7
2.5
32
1997
4.1
7.2
5.9
7.7
15.0
5.4
6.4
5.4
5.0
5.7
6.8
3.1
45
mean
7.3
8.9
7.1
7.9
13.6
7.0
6.3
6.3
5.6
6.1
7.6
2.3
31
std
2.8
1.9
1.1
1.3
1.7
1.6
0.1
1.0
0.8
0.4
COV
39
21
15
17
13
23
2
16
14
6
Recent Air Quality - 2nd highest 8-hour average
2005
1.9
3.2
2.6
3.4
5.6
2.8
2.9
3.1
3.1
2.9
3.1
0.9
30
2006
1.9
3.4
2.5
3.4
5.6
2.7
3.3
2.9
2.5
2.9
3.1
1.0
32
2007
1.6
2.7
2.1
2.7
4.9
2.2
2.5
2.3
2.5
2.5
2.6
0.9
33
mean
1.8
3.1
2.4
3.2
5.3
2.6
2.9
2.8
2.7
2.8
3.0
0.9
31
std
0.2
0.4
0.3
0.4
0.4
0.3
0.4
0.4
0.3
0.2
COV
10
12
12
12
8
13
15
15
12
8
3-17
-------
Table 3-3. Within monitor temporal variability in Denver using historical (1995-97) and recent (2005-07) air quality data
99th percentile 1-hour daily maximum.
Monitor
31-0002
31-0013
31-0014
59-0002
mean
std
cov
Historical Air Quality - 99th percentile 1-hour daily maximum
1995
13.5
11.1
8.2
8.6
10.4
2.5
24
1996
13.4
9.0
7.3
6.8
9.1
3.0
32
1997
9.1
8.6
7.8
7.2
8.2
0.9
10
mean
12.0
9.6
7.8
7.5
9.2
2.1
22
std
2.5
1.3
0.5
0.9
COV
21
14
6
12
Recent Air Quality - 99th percentile 1-hour daily maximum
2005
3.8
3.5
3.3
3.4
3.5
0.2
6
2006
4.5
3.7
3.2
3.4
3.7
0.6
16
2007
4.4
4.4
mean
4.2
3.6
3.2
3.4
3.6
0.4
12
std
0.4
0.1
0.1
0.0
COV
9
3
3
0
Table 3-4. Within monitor temporal variability in Los Angeles using historical (1995-97) and recent (2005-07) air quality
data - 99th percentile 1-hour daily maximum.
Monitor
37-0113
37-1002
37-1103
37-1201
37-1301
37-2005
37-4002
59-0001/7
59-1003
59-5001
mean
std
COV
Historical Air Quality - 99th percentile 1-hour daily maximum
1995
13.9
11.6
9.0
10.6
16.2
10.3
7.6
9.1
7.3
10.7
10.6
2.8
26
1996
7.5
9.7
9.4
8.4
20.2
8.8
8.4
8.2
8.4
11.6
10.1
3.7
37
1997
6.1
8.2
7.4
8.4
18.5
6.2
7.6
7.7
6.9
10.3
8.7
3.6
42
mean
9.2
9.9
8.6
9.1
18.3
8.4
7.9
8.4
7.5
10.9
9.8
3.1
32
std
4.2
1.7
1.1
1.3
2.0
2.0
0.5
0.7
0.8
0.7
COV
45
17
13
14
11
24
6
9
11
6
Recent Air Quality - 99th percentile 1-hour daily maximum
2005
2.6
3.9
3.1
4.0
7.1
3.4
3.8
3.6
3.6
5.2
4.0
1.3
32
2006
2.7
4.1
2.9
3.9
7.4
3.3
3.8
3.6
3.2
5.4
4.0
1.4
35
2007
2.1
3.6
2.6
3.4
6.8
3.0
3.1
3.2
3.2
5.2
3.6
1.4
38
mean
2.5
3.9
2.9
3.8
7.1
3.2
3.6
3.5
3.3
5.3
3.9
1.3
34
std
0.3
0.3
0.3
0.3
0.3
0.2
0.4
0.2
0.2
0.1
COV
13
7
9
8
4
6
12
6
6
2
3-18
-------
Table 3-5. Within monitor temporal variability in Denver using historical (1995-97) and recent (2005-07) air quality data
99th percentile 8-hour daily maximum.
Monitor
31-0002
31-0013
31-0014
59-0002
mean
std
cov
Historical Air Quality - 99th percentile 8-hour daily maximum
1995
7.3
5.4
5.7
4.1
5.6
1.3
24
1996
7.2
5.2
5.5
3.8
5.4
1.4
26
1997
5.2
4.7
5.8
4.8
5.1
0.5
10
mean
6.6
5.1
5.7
4.2
5.4
1.0
18
std
1.2
0.4
0.1
0.5
COV
18
7
2
12
Recent Air Quality - 99th percentile 8-hour daily maximum
2005
2.4
2.2
2.1
1.8
2.1
0.3
12
2006
2.8
2.1
2.8
1.8
2.4
0.5
21
2007
2.7
2.7
mean
2.6
2.2
2.4
1.8
2.3
0.4
16
std
0.2
0.0
0.5
0.0
COV
9
2
22
2
Table 3-6. Within monitor temporal variability in Los Angeles using historical (1995-97) and recent (2005-07) air quality
data - 99th percentile 8-hour daily maximum.
Monitor
37-0113
37-1002
37-1103
37-1201
37-1301
37-2005
37-4002
59-0001
59-1003
59-5001
mean
std
COV
Historical Air Quality - 99th percentile 8-hour daily maximum
1995
8.6
9.7
7.5
9.0
11.2
8.5
5.9
6.5
4.7
6.3
7.8
2.0
25
1996
5.2
8.3
7.0
6.7
13.9
6.8
6.2
5.7
6.4
5.9
7.2
2.5
35
1997
3.7
6.8
5.6
7.3
13.1
5.0
5.9
5.1
4.9
5.3
6.3
2.6
42
mean
5.8
8.3
6.7
7.6
12.7
6.7
6.0
5.8
5.3
5.8
7.1
2.2
31
std
2.5
1.4
1.0
1.2
1.4
1.8
0.2
0.7
0.9
0.5
COV
43
17
15
15
11
26
3
12
17
9
Recent Air Quality - 99th percentile 8-hour daily maximum
2005
1.9
3.0
2.6
3.2
4.9
2.8
2.9
2.7
3.0
2.6
3.0
0.8
25
2006
1.8
3.2
2.4
3.1
5.1
2.6
2.7
2.7
2.2
2.7
2.9
0.9
31
2007
1.5
2.6
2.0
2.6
4.5
2.1
2.4
2.1
2.4
2.5
2.5
0.8
31
mean
1.8
3.0
2.4
2.9
4.8
2.5
2.7
2.5
2.6
2.6
2.8
0.8
29
std
0.2
0.3
0.3
0.3
0.3
0.3
0.2
0.3
0.4
0.1
COV
12
11
13
12
7
14
9
14
16
5
3-19
-------
3.2 STUDY AREAS SELECTED FOR CURRENT CO REA
We identified several criteria to select the exposure assessment study areas drawing from
information discussed in the earlier sections of this Chapter and additional scientific evidence in
the ISA. We selected Denver and Los Angeles as areas to focus the current assessment because
(1) both cities have been included in prior CO NAAQS exposure assessments and thus serve as
an important connection with past assessments, (2) they have historically had among the highest
CO ambient concentrations among urban areas in the U.S., and (3) Denver is at high altitude and
represents a scenario of interest due to the potentially increased susceptibility of visitors to high
altitude locations from exposure to CO. In addition, often urban areas across the US having
monitors meeting a 75% completeness criteria, the two locations were ranked 1st (Los Angeles)
and 2nd (Denver) regarding percent of elderly population within 5, 10, and 15 km of monitor
locations, and ranked 1st (Los Angeles) and 5th (Denver) regarding number of 1-hour and 8-hour
daily maximum CO concentration measurements (ISA, section 3.5.1.1).
3.3 KEY OBSERVATIONS
Presented below are key observations resulting from the air quality considerations.
• Mobile sources (i.e., gasoline powered vehicles) are the primary contributor to CO
emissions, particularly in urban areas due to greater vehicle and roadway densities.
• Recent (2005-2007) ambient CO concentrations across the US are lower than those
reported in the previous CO NAAQS review and are also well below the current CO
NAAQS levels. Further, a large proportion of the reported concentrations are below
the conventional instrument lower detectable limit of 1 ppm.
• The currently available information for CO monitors indicates that siting of microscale
and middle scale monitors in the current network is primarily associated with roads
having moderate traffic density (< 100,000 AADT), however, factors other than
reported AADT (e.g., orientation with regard to dense urban roadway networks) can
contribute to sites reporting higher CO concentrations.
• Ambient CO concentrations are highest at monitors sited closest to roadways (i.e.,
microscale and middle scale monitors) and exhibit a diurnal variation linked to the
typical commute times of day, with peak concentrations generally observed during
early morning and late afternoon during weekdays.
• Policy relevant background (PRB) concentrations across the US are generally less than
0.2 ppm, far below that of interest in this REA with regard to ambient CO exposures.
• Historical trends in ambient monitoring data indicate that at individual sites, ambient
concentrations have generally decreased in a proportional manner. This comparison
included air quality distributions with concentrations at or above the current 8-hour
standard and those reflecting current (as is) conditions.
3-20
-------
• The temporal variability in selected upper percentile ambient concentrations (e.g., 99th
percentile 1-hour daily maximum) at individual monitors in Denver and Los Angeles is
relatively small across a three-year monitoring period, particularly when considering
recent air quality. Much of the within-monitor temporal variability is due to a trend in
decreasing concentration from year-to-year.
• There is greater spatial variability in selected upper percentile ambient concentrations
(e.g., 99th percentile 1-hour daily maximum) at ten selected monitoring sites in Los
Angeles when compared with four selected monitoring sites Denver, particularly when
considering the recent air quality.
3-21
-------
REFERENCES
Baldauf R, Thoma E, Hays M, Shores R, Kinsey J, Gullett B, Kimbrough S, Isakov V, Long T, Snow R, Khlystov
A, Weinstein J, Chen FL, Seila R, Olson D, Gilmour I, Cho SH, Watkins N, Rowley P, Bang J. (2008).
Traffic and meteorological impacts on near-road air quality: Summary of methods and trends from the
Raleigh near-road study. J Air Waste ManagAssoc. 58:865-878.
Michie Jr RM, McElroy FF, Sokash JA, Thompson VL, Dayton DP, Sutcliffe CR. (1983). Performance Test
Results and Comparative for Designated Equivalence Methods for Carbon Monoxide. EPA-600/S4-83-
013.
Rizzo M. (2008). Investigation of how distributions of hourly nitrogen dioxide concentrations have changed over
time in six cities. Nitrogen Dioxide NAAQS Review Docket. Docket ID no. EPA-HQ-OAR-2006-
0922). Available at: http://www.epa.gOv/ttn/naaqs/standards/nox/s noxcrrea.html.
Rizzo M. (2009). Investigation of How Distributions of Hourly Sulfur Dioxide Concentrations Have Changed Over
Time in Six Cities. Sulfur Dioxide Review Docket. Docket ID no. EPA-HQ-OAR-2007-0352. Available
at: www.regulations.gov.
US EPA. (2000). Air Quality Criteria for Carbon Monoxide. National Center for Environmental Assessment,
Office of Research and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
27711; report no. EPA/600/P-99/001F. June 2000. Available at:
http ://cfpub .epa. gov/ncea/cfm/recordisplav .cfm?deid= 18163.
US EPA. (2006). 2002 National Emissions Inventory Data and Documentation. U.S. Environmental Protection
Agency, Office of Air and Radiation, Office of Air Quality Planning and Standards, Research Triangle
Park, NC. Available at: http://www.epa.gov/ttn/chief/net/2002inventorv.html.
US EPA. (2008). Risk and Exposure Assessment to Support the Review of the NO2 Primary National Ambient Air
Quality Standard. EPA-452/R-08-008a. November 2008. Available at:
http://www.epa.gov/ttn/naaqs/standards/nox/data/2008112 l_NO2_REA_final.pdf.
US EPA. (2009). Risk and Exposure Assessment to Support the Review of the SO2 Primary National Ambient Air
Quality Standard. EPA-452/R-09-007. August 2009. Available at:
http://www.epa.gov/ttn/naaqs/standards/so2/data/200908SO2REAFinalReport.pdf.
US EPA. (2010). Integrated Science Assessment for Carbon Monoxide. U.S. Environmental Protection Agency,
Washington, DC, report no. EPA/600/R-09/019F. Available at:
http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_isa.html.
Zhu Y, Hinds WC, Kim S, Shen S, Sioutas C. (2002). Study of ultrafine particles near a major highway with
heavy-duty diesel traffic. Atmos Environ. 36:4323-4335.
3-22
-------
4 OVERVIEW OF APEX MODELING SYSTEM FOR ESTIMATING
CO EXPOSURES AND COHB DOSE LEVELS
4.1 PURPOSE
This chapter presents an overview and description of the overall approach to estimating
human exposure and dose for past and recent NAAQS reviews. Section 4.2 provides a brief
overview of EPA's Air Pollutants Exposure model (APEX), the model used in this assessment to
estimate population exposure and dose. This overview is followed by a short history that
explains the evolution of exposure and dose models used by OAQPS to conduct exposure and
dose assessments for CO and other NAAQS reviews (section 4.3). Section 4.4 provides a
generalized description of the APEX simulation process, though having detailed focus on a few
of the important approaches used for modeling CO exposure and COHb dose. This includes
expanded discussion on the approach used to estimate microenvironmental concentrations
(section 4.4.4) and COHb dose levels (section 4.4.7).
4.2 MODEL OVERVIEW
The Air Pollutants Exposure model (APEX) is a personal computer (PC)-based program
designed to estimate human exposure to criteria and air toxic pollutants at the local, urban, and
consolidated metropolitan levels. APEX, also known as TREVI.Expo, is the human inhalation
exposure module of EPA's Total Risk Integrated Methodology (TRIM) model framework (US
EPA, 1999), a modeling system with multimedia capabilities for assessing human health and
ecological risks from hazardous and criteria air pollutants.1
APEX estimates human exposure using a
. . A microenvironment is a three-
stochastic, microenvironmental approach (see caption).
dimensional space in which human
The model randomly selects data for a sample of , , ... . , . „ , ,
contact with an environmental pollutant
hypothetical individuals from an actual population
takes place and which can be treated as
database and simulates each individual's movements a we|| characterized relatively
through time and space (e.g., indoors at home, inside
vehicles) to estimate his or her exposure to a pollutant.
APEX can account for travel to and from work locations
homogeneous location with respect to
pollutant concentrations for a specified
time period.
(i.e., commuting) and provide estimates of exposures at
both home and work locations for individuals who work away from home.
1 Additional information on the TRIM modeling system, as well as downloads of the APEX Model, user
guides (US EPA 2008a, 2008b), and other supporting documentation, can be found at http://www.epa.gov/ttn/fera.
4-1
-------
4.3 MODEL HISTORY AND EVOLUTION
APEX was derived from the National Ambient Air Quality Standards (NAAQS)
Exposure Model (NEM) series of models. The NEM series was developed to estimate
population exposures to the criteria pollutants (e.g., CO, ozone). In 1988, OAQPS first
incorporated probabilistic elements into the NEM methodology and used activity pattern data
based on available human activity diary studies to create an early version of probabilistic NEM
for ozone (i.e., pNEM/Os). In 1991, a probabilistic version of NEM was developed for CO
(pNEM/CO) that included a one-compartment mass-balance model to estimate CO
concentrations in indoor microenvironments. The first application of this model to Denver,
Colorado is summarized in Johnson et al. (1992). Between 1999 and 2001, updated versions of
pNEM/CO (versions 2.0 and 2.1) were developed that relied on detailed activity diary data
compiled in EPA's Consolidated Human Activities Database (CHAD) (McCurdy et al., 2000;
US EPA, 2002) and enhanced algorithms for simulating gas stove usage, estimating alveolar
ventilation rate (a measure of human respiration), and modeling home-to-work commuting
patterns. A draft report by Johnson et al. (2000) describes the application of Version 2.1 of
pNEM/CO to Denver and Los Angeles.
The first version of APEX was essentially identical to pNEM/CO (version 2.0) except
that it ran on a personal computer (PC) instead of a mainframe. The next version, APEX2, was
substantially different, particularly in the use of a personal profile approach rather than the
previously used cohort simulation approach. APEX3 introduced a number of new features
including automatic site selection from national databases, a series of new output tables
providing summary exposure and dose statistics, and a thoroughly reorganized method of
describing microenvironments and their variable parameters. Johnson and Capel (2003) describe
a case study in which the PC-based Version 3.1 of APEX was used to estimate population
exposure to CO in Los Angeles.
The current version of APEX (Version 4.3) (US EPA, 2008a; 2008b) was used to
estimate CO exposure and dose as described in chapter 5 of this document. This version was
also recently used to estimate Os exposures in 12 urban areas for the Os NAAQS review (US
EPA, 2007), to estimate population exposures to nitrogen dioxide (NO2) in Atlanta as part of the
NO2 NAAQS review (US EPA, 2008c), and to estimate sulfur dioxide (802) exposures for
asthmatics and asthmatic children in two study areas in Missouri as part of the SO2 NAAQS
review (US EPA, 2009a). There have been several recent enhancements to APEX since the prior
1994 CO NAAQS review, including:
• Algorithms for the assembly of multi-day (longitudinal) activity diaries that model intra-
individual variance, inter-individual variance, and day-to-day autocorrelation in diary
properties;
4-2
-------
• Methods for adjusting diary-based energy expenditures for fatigue and excess post-
exercise oxygen (EPOC) consumption;
• New equations for estimation of ventilation (i.e., breathing rate);
• The ability to use air quality data and model exposures over flexible time scales;
• New output files containing diary event-level, time-step level, and hourly-level exposure,
dose, and ventilation data, and hourly-level microenvironmental data;
• The ability to model the prevalence of disease states such as asthma or heart disease;
• New output exposure tables that report exposure statistics for population groups and life-
stages such as children and active people at varying ventilation rates;
• The inclusion of tract-level commuting data from the 2000 census; and
• Expanded options for modeling microenvironments.
4.4 MODEL SIMULATION PROCESS
APEX4.3 is designed to simulate population exposure to criteria and air toxic pollutants
at local, urban, and regional scales. The user specifies the geographic area to be modeled and the
number of individuals to be simulated to represent a population of interest. APEX4.3 then
generates a personal profile for each simulated person that specifies various parameter values
required by the model to estimate their exposure and dose. The model next uses diary-derived
time/activity data matched to each personal profile to generate an exposure event sequence (also
referred to as a time-location-activity pattern or composite diary) for the modeled individual that
spans a specified time period, such as a calendar year. Each event in the sequence specifies a
start time, exposure duration, a geographic location, a microenvironment inhabited, and an
activity performed. Probabilistic algorithms are used to estimate the pollutant concentration and
ventilation (respiration) rate associated with each exposure event. The estimated pollutant
concentrations account for the effects of ambient (outdoor) pollutant concentration, penetration
factor, air exchange rate, decay/deposition rate, and proximity to emission sources, each
depending on the microenvironment, available data, and the estimation method selected by the
user. The ventilation rate is derived from an energy expenditure rate estimated for each
individual when performing the specified activity. Because the simulated individuals represent a
random sample of the population of interest and are proportionally derived from actual
population distributions, the distribution of modeled individual exposures can then be
extrapolated to the larger population of interest.
The model simulation generally includes up to seven steps as follows:
• Characterize study area: APEX4.3 selects sectors (e.g., census tracts) within a study
area—and thus identifies the potentially exposed population — usually based on the
4-3
-------
user-defined center and radius of the study area and availability of air quality and
meteorological input data for the area (section 4.4.1).
• Generate simulated individuals: APEX4.3 stochastically generates a sample of
simulated individuals based on the census data for the study area and human profile
distribution data (such as age-specific employment probabilities or disease prevalence)
(section 4.4.2)
• Construct activity sequences: APEX4.3 constructs an exposure event sequence (time-
location-activity pattern) spanning the simulation period for each of the simulated
persons based on the CHAD diaries (section 4.4.3).
• Calculate microenvironmental concentrations: APEX4.3 enables the user to define
microenvironments that people in a study area would visit (e.g., by grouping location
codes included in the activity pattern database). The model then calculates time-
averaged concentrations (e.g., hourly) of each pollutant in each of the
microenvironments for each simulated person for the period of simulation based on the
user-provided ambient air quality data (section 4.4.4).
• Estimate energy expenditure and ventilation rates: APEX4.3 constructs a time-
series of energy expenditures for each individual's exposure profile based on the
sequence of activities performed. The sequence of energy expenditures are adjusted to
ensure that they are physiologically realistic and then used to estimate activity-specific
alveolar ventilation rates (section 4.4.5).
• Calculate exposure: APEX4.3 assigns a concentration to each exposure event based
on the microenvironment occupied during the event and the person's activity. These
values are time-averaged (e.g., hourly) to produce a sequence of exposures spanning
the specified exposure period (typically one year). The hourly values may be further
aggregated to produce 8-hour, daily, monthly, and annual average exposure values
(section 4.4.6).
• Calculate dose: APEX4.3 optionally calculates hourly, daily, monthly, and annual
average dose values for each of the simulated individuals. For the application of
APEX to CO, a module within the model estimates the percent COHb level in the
blood at the end of each hour based on the time-series of CO concentrations and
alveolar ventilation rates experienced by the simulated person (section 4.4.7).
The model simulation continues until exposures (and associated COHb dose levels) are
calculated for the user-specified number of simulated individuals. Figure 4-1 presents a
conceptual model and simplified data flow diagram illustrating the implementation of APEX4.3
to estimate CO exposure and dose. The following sections provide additional details on the
general procedures and algorithms used in each of the seven simulation steps listed above,
though more complete discussion can be found in US EPA (2008a, 2008b). The specific input
data and microenvironmental factors used in applying APEX4.3 to CO for the current assessment
are further described in section 5.1.
4-4
-------
Air Exchange Rates
(AER) & Prevalence
US Census
Tract-Level Population
& Commuting Data
Meteorological Data
Temperature
Outdoor CO Concentration
Algorithm
(spatial & temporal adjustment)
Ambient Air Quality
1 -Hour Ambient Monitor
CO Concentrations
USDHHS/CDC
Disease Prevalence
CHAD
Time-Location-Activity
Patterns
Exposure Algorithms
(mass balance or factors)
Microenvironmental
Factors/Distributions
J
Persons & Person-Days
At or Above
Exposure Levels
Energy Expenditure &
Ventilation Algorithm
(alveolar respiration rate)
Dose Algorithm
(CFK model)
J
Persons & Person-Days
At or Above
Benchmark Dose Levels
Figure 4-1. Conceptual model and simplified data flow for estimating population exposure and dose using APEX4.3.
4-5
-------
4.4.1 Characterize Study Area
An initial study area in an APEX4.3 assessment consists of a set of basic geographic units
called sectors, typically defined by US census data reported at the census tract level. The user
may provide the geographic center (latitude/longitude) and radius of the study area. Then
APEX4.3 calculates the distances to the center of the study area of all the sectors included in the
sector location database and selects the sectors within the radius of the study area. APEX4.3
then maps the user-provided air quality and meteorological data for specified monitoring districts
to the selected sectors. The sectors identified as having acceptable air quality and meteorological
data within the radius of the study area are selected to comprise a final study area for the
APEX4.3 simulation analysis. This final study area determines the population make-up of the
simulated persons (profiles) to be modeled.
4.4.2 Generate Simulated Individuals
APEX4.3 stochastically generates a user-specified number of simulated persons to
represent the population in the study area. Each simulated person is represented by a personal
profile. APEX4.3 generates the simulated person by probabilistically selecting values for a set of
profile variables. The profile variables include:
• Demographic variables that are generated based on US census data (e.g., age, gender,
home sector, work sector);
• Residential variables that are generated based on sets of distribution data (e.g., air
conditioning prevalence);
• Physiological variables that are generated based on age- and gender-specific distribution
data (e.g., blood volume, body mass, resting metabolic rate); and
• Daily varying variables that are generated based on distribution data that change daily
during the simulation period (e.g., daily work status).
APEX4.3 first selects and calculates demographic, residential, and physiological
variables (except for daily values) for each of the user-specified number of simulated individuals.
APEX4.3 then follows each simulated individual over time and calculates exposures (and
optionally doses) for the individual over the duration of the assessment period. The complete
listing of profile variables used by APEX4.3 and detailed description can be found in section 5 of
US EPA (2008b). An overview of the data sources used and their implementation in APEX4.3 is
provided below.
4.4.2.1 Population Demographics
APEX4.3 takes population characteristics into account to develop accurate
representations of study area demographics. Specifically, population counts by area and
4-6
-------
employment probability estimates are used to develop representative profiles of hypothetical
individuals for the simulation.
APEX4.3 is flexible in the resolution of population data provided. As long as the data are
available, any resolution can be used (e.g., county, census tract, census block). For this
application of the model, census tract level data were used. Census tract level population counts
are obtained from the 2000 Census of Population and Housing Summary File 1 (SF-1). This file
contains data compiled from the questions asked of all respondents and about every housing unit.
As part of the population demographics inputs, it is important to integrate working
patterns into the assessment. In the 2000 US Census, estimates of employment were developed
by census information (US Census Bureau, 2007). The employment statistics are broken down
by gender and age group, so that each gender/age group combination is given an employment
probability fraction (ranging from 0 to 1) within each census tract. The age groupings used are:
16-19, 20-21, 22-24, 25-29, 30-34, 35-44, 45-54, 55-59, 60-61, 62-64, 65-69, 70-74, and >75.
Children under 16 years of age were assumed by the model to not be employed.
4.4.2.2 Commuting Database
In addition to using estimates of employment by census tract, APEX4.3 also incorporates
home-to-work commuting data. Commuting data were derived from the 2000 Census and were
collected as part of the Census Transportation Planning Package (CTPP) (US DOT, 2007). The
data used contain counts of individuals commuting from home to work locations at a number of
geographic scales. These data were processed to calculate fractions for each tract-to-tract flow to
create the national commuting data distributed with APEX4.3. This database contains
commuting data for each of the 50 states and Washington, D.C.
Several assumptions were made in the development of the database and with the
modeling of a person's commute in this assessment as follows.
• Commutes within tracts and home workers: There is no differentiation between
people that work at home from those that commute within their home tract.
• Commute distance cutoff: All persons in home-work flows up to 120 km are assumed
to be daily commuters and no persons in more widely separated flows would commute
daily. This means that the list of destinations for each home tract was restricted to only
those work tracts that are within 120 km of the home tract. This distance is based on
the presence of a near-constant relationship between commute flows and distance
traveled up to 120 km.
• Eliminated Records: Tract-to-tract pairs representing workers who either worked
outside of the US (9,631 tract pairs with 107,595 workers) or worked in an unknown
location (120,830 tract pairs with 8,940,163 workers) were eliminated from the
database. An additional 515 workers in the commuting database whose data were
4-7
-------
missing from the original files, possibly due to privacy concerns or errors, were also
deleted.
• Commuting outside the study area: APEX4.3 allows for some flexibility in the
treatment of persons in the modeled population who commute to destinations outside
the study area. Users can either retain these persons and include them as part of the
population exposed or have them eliminated from the model simulation. In the first
instance (i.e., "KeepLeavers = Yes"), APEX4.3 can assign input concentrations based
on the available ambient concentration data within the model domain. For the second
option (i.e., "KeepLeavers = No"), people who work inside the study area but live
outside of it are not modeled, nor are people who live in the study area but work
outside of it.
4.4.2.3 Profile Functions File
A Profile Functions file contains settings used to generate results for variables related to
simulated individuals. While certain settings for individuals are generated automatically by
APEX4.3 based on other input files, including demographic characteristics, others can be
specified using this file. For example, the file may contain settings for determining whether the
profiled individual's residence has an air conditioner, a gas stove, etc.
4.4.2.4 Physiology File
The APEX4.3 physiology.txt file contains age- and gender-based information for several
physiological parameters used in human exposure modeling. This information includes various
equations, distributional shapes, and parameters for all age and gender cohorts from age 0 to 100
years for variables such as normalized maximal oxygen uptake, body mass, resting metabolic
rate (RMR), and blood hemoglobin content. Appendix A provides an evaluation of a few
important variables used by APEX4.3 in this exposure and dose assessment as well as their
updated values or distributions (e.g., new age-gender body mass distributions derived from 1999-
2004 National Health and Nutrition Examination Survey data). Details regarding any other
physiology variable distributions and their parameters not discussed in this CO REA and
associated appendices can be found in US EPA (2008a, 2008b).
4.4.3 Construct Activity Sequences
Different human activities, such as spending time outdoors, indoors, or driving, will be
associated with varying pollutant concentrations. Therefore, to accurately model individuals and
their exposure to pollutants, it is critical to understand people's daily activities and use such data
in the exposure model. EPA's Consolidated Human Activity Database (CHAD) provides diary-
derived data indicating where people spend time and the activities they perform at each location
(US EPA, 2002). CHAD was designed to provide a basis for conducting multi-route, multi-
media exposure assessments (McCurdy et al., 2000). The data contained within CHAD originate
4-8
-------
from multiple activity pattern surveys with varied structures (Table 4-1), however the surveys
have commonality in that they contain daily diaries of human activities performed, locations
visited, and the personal attributes of survey participants (e.g., age and gender).
There are four CHAD-related input files used in APEX4.3. The first three can be
considered standard input files for most model simulations; the user typically does not modify
their contents. These include the human activity diaries file, the personal data file, and a
metabolic information file, each of which are discussed briefly below. The fourth CHAD-related
input file maps the five-digit location codes used in the diary file to APEX4.3
microenvironments; this file is commonly modified by the user and is discussed in section 5.8
(i.e., it is most relevant for the specific microenvironments modeled in this CO REA). And
finally, section 4.4.3.4 discusses how these diaries are linked together to form a continuous time-
location-activity pattern for each individual across the entire simulation period.
4-9
-------
Table 4-1. Summary of activity pattern studies comprising the recent version of CHAD.
Study Name
Baltimore
GARB: Adults
GARB: Adolescents
GARB: Children
Cincinnati (EPRI)
Denver (EPA)
Los Angeles:
Elementary
Los Angeles: High
School
NHAPS A
NHAPS B
PSID1
(U Michigan I)
PSID2
(U Michigan II)
Valdez
Washington, DC
RTI Ozone Averting
Behavior
RTP Panel Study
Seattle Study
Internal EPA Study
2006-2007
EPA Longitudinal 1
EPA Longitudinal 2
EPA Longitudinal 3
CHAD
Prefix
BAL
CAA
CAY
CAC
CIN
DEN
LAE
LAH
NHA
NHW
UMC
ISR
VAL
WAS
OAB
RTP
SEA
EPA
EPA
EPA
EPA
Study
Years
1997-1998
1987-1988
1987-1988
1989-1990
1985
1982-1983
1989
1990
1992-1994
1992-1994
1997
2002-2003
1990-1991
1982-1983
2002-2003
2000-2001
1999-2002
2006-2007
1999,2002
2000
2008
Number of
Diary Days
391
1579
183
1200
2614
805
51
43
4723
4663
5616
4782
397
699
2907
1003
1693
434
736
197
62
Reference
Williams et al. (2000)
Wiley etal. (1991 a)
Wiley etal. (1991 a)
Wiley etal. (1991b)
Johnson (1989)
Johnson (1984); Akland et al. (1985)
Spier etal. (1992)
Spier etal. (1992)
Klepeis et al. (1996); Tsang and Klepeis (1996)
Klepeis et al. (1996); Tsang and Klepeis (1996)
University of Michigan (2010)
University of Michigan (2010)
Goldstein et al. (1992)
Hartwell et al. (1984); Akland et al. (1985)
Mansfield and Corey (2003); Mansfield et al.
(2004; 2006)
Williams et al. (2003a, 2003b)
Liu et al. (2003)
Isaacs et al. (2009)
Isaacs et al. (2009)
Isaacs et al. (2009)
Isaacs et al. (2009)
4.4.3.1 Personal Information file
Personal attribute data are contained in the CHAD questionnaire file that is distributed
with APEX4.3. This file also has information for each day individuals have diaries. The
different variables in this file are:
• The study, person, and diary day identifiers
• Day of week
4-10
-------
• Gender
• Employment status
• Age in years
• Maximum temperature in degrees Celsius for the diary day
• Mean temperature in degrees Celsius for the diary day
• Occupation code (if requested in survey)
• Time, in minutes, during this diary day for which no data are included in the database
4.4.3.2 Diary Events File
The human activity diary data are contained in the events file that is distributed with
APEX4.3. This file contains the locations visited and the activities performed for the nearly
35,000 person-days of data with event intervals ranging from a minimum of one minute upwards
to a one hour maximum duration. Typically, a study individuals' diary can vary in length from
one to 15 days (i.e., referring to the number of person-days) though a few recent surveys have
upwards of hundreds of diary days for a few individuals. Nevertheless, the diary events file
contains the following variables:
• The study, person, and diary day identifiers
• Start time of the event
• Number of minutes for the event
• Activity code (a record of what the individual was doing)
• Location code (a record of where the individual was)
4.4.3.3 Activity-Specific Metabolic File
The metabolic file contains the distributional forms and parameters for the activity-
specific metabolic equivalents (METs) used to quantitatively assign exertion levels to each
activity performed by simulated individuals (McCurdy, 2000). Some activities are specified as a
single point value (for instance, watching TV), while others, such as athletic endeavors or
manual labor, are represented by normal, lognormal, or other statistical distributions. APEX4.3
samples from these distributions and calculates values to simulate the variable nature of activity
levels among different people. The CHAD User's guide provides details on the distributions
used, parameters, and sources for each activity (US EPA, 2002).
4.4.3.4 Longitudinal Diary Processing
APEX4.3 probabilistically creates a composite longitudinal diary for each of the
simulated persons by selecting a 24-hour diary record - or diary day - from an activity database
for each day of the simulation period. The EPA's CHAD data (US EPA, 2002) are supplied with
4-11
-------
APEX4.3 for this purpose. A composite diary is a sequence of events that simulates the
movement of a modeled person through varying geographical locations and microenvironments
for the duration of the simulation period. Each diary event is defined by geographic location,
start time, duration, microenvironment visited, and an activity performed.
The activity database input to APEX4.3 contains the following information for each diary
day: age, gender, employment status, occupation, day-of-week (or day-type), and maximum
hourly average temperature. This information enables APEX4.3 to select data from the activity
database that tend to match the characteristics of the simulated person, the study area, and the
specified time period. APEX4.3 develops a composite diary for each of the simulated
individuals according to the following steps.
• Divide diary days in the CHAD database into user-defined activity pools, based on
day-type and temperature categories.
• Assign an activity pool number to each day of the simulation period, based on the user-
provided daily maximum/average temperature data.
• Calculate a selection probability for each of the diary days in each of the activity pools,
based on age/gender/employment similarity of a simulated person to a diary day.
• Probabilistically select a diary day from available diary days in the activity pool
assigned to each day of the simulation period.
• Estimate a MET value for each activity performed while in a location, based on a
random sampling of the particular distribution of each specific activity. The METs
values are used to calculate an activity-specific ventilation rate (see section 4.4.5) for
the simulated person.
• Map the CHAD locations in the selected diary to the user-defined modeled
microenvironments.
• Concatenate the selected diary days into a sequential longitudinal diary for a simulated
individual covering all days in the simulated period.
APEX4.3 provides an optional longitudinal diary-assembly algorithm that enables the
user to create composite diaries that reflect the tendency of individuals to repeat activities on a
day-to-day basis. The user specifies values for two statistical variables (i.e., D and A) that relate
to a key daily variable, typically the time spent per day in a particular microenvironment (e.g., in
a motor vehicle). The D statistic reflects the relative importance of within-person variance and
between-person variance in the key variable. The^4 statistic quantifies the lag-one (day-to-day)
variable autocorrelation. APEX4.3 then constructs composite diaries that exhibit the statistical
properties defined by the specified values ofD and A. The longitudinal diary assembly
algorithm is described in greater detail by Glen et al. (2008) and in section 6.3 of US EPA
(2008b).
4-12
-------
4.4.4 Calculate Microenvironmental Concentrations
Probabilistic algorithms are used by APEX4.3 to estimate the pollutant concentration
associated with each exposure event. The estimated pollutant concentrations account for the
effects of ambient (outdoor) pollutant concentration, penetration factor, air exchange rate,
decay/deposition rate, and proximity to emission sources, depending on the microenvironment,
available data, and the estimation method selected by the user.
APEX4.3 calculates air concentrations in the various microenvironments visited by the
simulated person by using the ambient air data for the relevant census tracts and the user-
specified method and parameters that are specific to each microenvironment. In typical
applications, APEX4.3 calculates hourly concentrations in all the microenvironments at each
hour of the simulation for each of the simulated individuals, based on the hourly ambient air
quality data specific to the geographic locations visited by the individual. APEX4.3 provides
two methods for calculating microenvironmental concentrations: the mass balance method and
the transfer factors method (each are described briefly below). The user is required to specify a
calculation method for each of the microenvironments; there are no restrictions on the method
specified for each microenvironment (e.g., some microenvironments can use the mass balance
method while the others can use the transfer factors method). Each of these approaches is
described in sections 4.4.4.1 and 4.4.4.2, respectively.
When using an exposure model to estimate population exposures to CO based on
exposures to concentrations in microenvironments, it is best to use estimates of the outdoor
(ambient) CO concentration in the immediate vicinity of each microenvironment to address the
ambient contribution to that microenvironment. These concentrations may need to be derived
because concentrations measured at a fixed-site monitor may not adequately represent the spatial
and temporal heterogeneity in concentrations expected with distance from the ambient monitor
location. There can be different ways to derive the ambient concentration in the immediate
vicinity of the microenvironment. For example, one can use an emission-based dispersion model
to estimate ambient concentrations at a fine temporal (e.g., hourly) and spatial scale (e.g., census
block-level or 500 meter grid cells). Another method is to use a statistically-based approach that
addresses the variability in concentrations in a similar manner as a dispersion model, only that
important physical factors that influence concentration levels are represented by and/or possibly
combined with a series of regression equation coefficients and are related to an ambient monitor
CO concentration. Ultimately, it is this estimated outdoor CO concentration that is then used as
input to the microenvironmental algorithm (either the mass balance model or factors method)
employed to estimate CO microenvironmental concentrations.
For this APEX application, staff selected a statistically-based approach to estimate
ambient concentrations in the immediate vicinity of each microenvironment based on the
4-13
-------
ambient monitor concentrations. The approach was designed to reflect both the spatial and
temporal variability expected to occur outside microenvironments, while also appropriately
linking the estimated microenvironmental concentrations to observed concentrations at a fixed-
site ambient monitor. The approach was developed using personal exposure, fixed-site monitor,
and outdoor concentration measurement data and first implemented in the pNEM/CO model for
use in the most recent CO exposure assessment (Johnson et al., 2000). This approach was
proposed as a method to address spatial and temporal variability in outdoor and
microenvironmental concentrations in the draft scope and methods plan (US EPA, 2009b),
though not fully described there as is done here.
To provide both historical perspective and context regarding the current application, this
microenvironmental algorithm and the data that were used in the past with pNEM/CO to estimate
values for the algorithm variables is described in section 4.4.4.3. The pNEM/CO approach was
then adapted and implemented in APEX3.1, a model more similar in structure to the current
version of APEX (version 4.3) than pNEM/CO. This approach as applied to APEX3.1 is then
described in section 4.4.4.4. The details regarding selection of specific microenvironments and
parameters used by APEX4.3 in this assessment is provided in section 5.9.
4.4.4.1 Overview of the Mass Balance Model
The mass balance method models an enclosed microenvironment as a well-mixed volume
in which the air concentration is spatially uniform at any specific time. The concentration of an
air pollutant in such a microenvironment is estimated using the following four processes:
• Inflow of air into the microenvironment;
• Outflow of air from the microenvironment;
• Removal of a pollutant from the microenvironment due to deposition, filtration, and/or
chemical degradation; and
• Emissions from sources of a pollutant inside the microenvironment.
Table 4-2 lists the parameters required by the mass balance method to calculate
concentrations in a microenvironment. The proximity factor (fpr0ximity) is used to account for
differences in ambient concentrations between the geographic location represented by the
ambient air quality data (e.g., a fixed-site monitor) and the geographic location of the
microenvironment (e.g., near a roadway). This factor could take a value either greater than or
less than 1. Emission source (ES) represents the emission rate for the emission source, and
concentration source (CS) is the mean air concentration resulting from the source (these are not
used in the current assessment. The factor Rremovai is defined as the removal rate of a pollutant
4-14
-------
from a microenvironment due to deposition, filtration, and chemical reaction. The air exchange
rate (Rair exchange) is expressed in air changes per hour.
Table 4-2. Variables used by APEX4.3 in the mass balance model.
Variable
' proximity
cs
ES
" removal
" air exchange
V
Definition
Proximity factor
Concentration source
Emission source
Removal rate due to
deposition, filtration, and
chemical reaction
Air exchange rate
Volume of
microenvironment
Units
unitless
ppm
ug/hr
1/hr
1/hr
m3
Value Range
' proximity ^ U
CS>0
ES>0
^removal — *-*
p > n
nair exchange — u
V>0
The mass balance equation for a pollutant in a microenvironment is described by the
differential equation
where:
dCME(t)
ACm
AC
out
removal
dt
•= AC,. -AC ,-AC
remoml
AC
(4-1)
Change in concentration in a microenvironment at time t (ppm),
Rate of change in microenvironmental concentration due to influx
of air (ppm/hour),
Rate of change in microenvironmental concentration due to outflux
of air (ppm/hour),
Rate of change in microenvironmental concentration due to
removal processes (ppm/hour), and
Rate of change in microenvironmental concentration due to an
emission source inside the microenvironment (ppm/hour).
4-15
-------
Within the time period of an hour each of the rates of change, ACin, ACout, ACremova/, and
rce, is assumed to be constant. The change in microenvironmental concentration due to
influx of air is represented by the following equation:
=
j
=
ambient J proximity J penetration air exchan
ge
(4-2)
where:
{-•ambient
Jproximity
Jpenetration
J^air exchange
Ambient hourly outdoor concentration (ppm)
Proximity factor
Penetration factor
Air exchange rate (I/hour)
The change in microenvironmental concentration due to outflux of air is described by:
•CME(t) (4-3)
ACout = dCo,4t)_=R
dt
air exchange '
The change in concentration due to deposition, filtration, and chemical degradation in a
microenvironment is simulated by a first-order equation:
_ dCremoval(t) M=R
removal 7j (deposition filtration chemical/ ME\/ remova
dt
where:
-^deposition
Infiltration
^chemical
^removal
Removal rate of a pollutant from a microenvironment due to
deposition (I/hour)
Removal rate of a pollutant from a microenvironment due to
filtration (I/hour)
Removal rate of a pollutant from a microenvironment due to
chemical degradation (I/hour)
Removal rate of a pollutant from a microenvironment due to
overall removal (I/hour)
As discussed in Section 2.2, EPA has not modeled indoor emissions of CO in the current
exposure assessment; consequently, the optional term ACTOMrce was uniformly set equal to 0.0 for
this study.
4-16
-------
Combining equation 4-1 with equations 4-2, 4-3, and 4-4 yields
ME
,
- \r -R *C (f}-R
~^^in ^-air exchange ^^ME\l) ^removal
The solution to this differential equation is
)^p(-Rcombimdt) (4-6)
combined combined
where:
CME(O) = Concentration of a pollutant in a microenvironment at the
beginning of a hour (ppm)
Cusft) = Concentration of a pollutant in a microenvironment at time t within
the time period of a hour (ppm)
-^combined t\air exchange ' -^removal
Based on equation 4-6, the following three hourly concentrations in a microenvironment
are calculated:
-, L
„
combined
RconUned) (4-8)
\C(t)dt
s~i hourlymean _ 0 /"> equil , ( f~i ( f\\ f~i equil \ ^ CAp ^ ^combined)
E V / MB / 7-,
f ij combined
where:
Equilibrium concentration in a microenvironment (ppm)
Concentration in a microenvironment at the beginning of an hour
(ppm)
4-17
-------
end
Concentration in a microenvironment at the end of an hour (ppm)
Hourly mean concentration in a microenvironment (ppm)
At each hour time step of the simulation period, APEX4.3 uses equations 4-7, 4-8, and 4-
9 to calculate the hourly equilibrium, hourly ending, and hourly mean concentrations. APEX4.3
reports hourly mean concentration as the hourly concentration for a specific hour. The
calculation continues to the next hour by using C^rlyend for the previous hour as CME(O).
4.4.4.2 Overview of the Factors Model
The factors model approach is conceptually simpler than the mass balance method and
has fewer user-specified parameters. It estimates the concentration in a microenvironment as a
linear function of ambient concentration of that hour, regardless of the concentration in the
microenvironment during the preceding hour. Table 4-3 lists the parameters required by the
factors model approach to calculate concentrations in a microenvironment in the absence of
indoor emissions sources.
Table 4-3. Variables used by APEX4.3 in the factors model.
Variable
' proximity
I penetration
Definition
Proximity factor
Penetration factor
Units
unitless
unitless
Value Range
' proximity ^ U
0 ^ f penetration - 1
The factors model approach uses the following equation to calculate hourly mean
concentration in a microenvironment from the user-provided hourly air quality data:
fi hourlymean /~i r r
ME ambient J proximity J penetration
where
' = Hourly concentration in a microenvironment (ppm)
= Hourly concentration in ambient environment (ppm)
/proximity = Proximity factor (unitless)
/penetration = Penetration factor (unitless)
The proximity factor (/proximity) is used to account for differences in ambient
concentrations between the geographic location represented by the ambient air quality data (e.g.,
a fixed-site monitor) and the geographic location of the particular microenvironment. For
example, persons travelling inside motor vehicles may be located on a heavily-trafficked
4-18
-------
roadway, whereby the ambient air outside the vehicle would likely have elevated levels of
mobile source pollutants such as carbon monoxide relative to the ambient monitor. In this case,
a value greater than one for the proximity factor would be appropriate to represent the increase in
concentrations outside the vehicle relative to the ambient monitor. Additionally, for some
pollutants the process of infiltration may remove a fraction of the pollutant from the air. The
fraction that is retained in the indoor/enclosed microenvironment is given by the penetration
factor (/penetration) and is dependent on the particular pollutant's physical and chemical removal
rates.
4.4.4.3 Description of the Original pNEM/CO Microenvironmental Algorithm
Version 2.1 of pNEM/CO determined the hourly outdoor CO concentration applicable to
each microenvironment through a Monte Carlo process based on the following equation
COout(c,m,d,h) =M(m) x L(c, m, d) x T(c,m,d,h) x [COmon(d,h)}A (4-11)
where,
CO0ut(c,m,d,h) = outdoor CO concentration (ppm) for cohort c with respect to
microenvironment m in district d during hour /z,
M(m) = multiplier (> 0) specific to microenvironment m,
L(c,m,d) = location factor (> 0) specific to cohort c, microenvironment m, and
district d (held constant for all hours),
T(c,m,d,h) = time-of-day factor (> 0) specific to cohort c, microenvironment m,
district J, and hour /z,
COmon(d,h) = ambient monitor-derived CO concentration (ppm) for hour h in
district d, and
A = exponent (A > 0).
This equation was used to generate a year-long sequence of outdoor one-hour CO
concentrations for each combination of cohort (c), microenvironment (m\ and district (d) by
Johnson et al. (2000). The exponent^ was set equal to 0.621 and held constant for all
sequences. The value ofM(m) varied only with microenvironment as indicated in Table 4-4 [and
is identical to Table 2-6 in Johnson et al. (2000)].
A value of the location factor L(c, m, d) was specified for each individual sequence and
held constant for all hours in the sequence. The value was randomly selected from a lognormal
distribution with geometric mean (GML) equal to 1.0 and geometric standard deviation
4-19
-------
equal to 1.5232. The natural logarithms of this distribution can be characterized by a normal
distribution with an arithmetic mean (UL) equal to 0 and an arithmetic standard deviation (GL)
equal to 0.4208.
A value of the time-of-day factor T(c, m, d, h) was randomly selected for each hour
within a sequence from a lognormal distribution with geometric mean (GMx) equal to 1.0 and
geometric standard deviation (GSDT) equal to 1.6289. The natural logarithms of this distribution
follow a normal distribution with an arithmetic mean (UT) equal to 0 and an arithmetic standard
deviation (GT) equal to 0.4879.
The COout(c, m, d, h) term is interpreted as the outdoor CO concentration in the
immediate vicinity of microenvironment m in district d during hour h. COmon(d, h) is the CO
concentration reported for hour h by a nearby fixed-site monitor selected to represent district d.
The mass balance model in pNEM/CO included a penetration factor that was set equal to
1.0 for CO. Consequently, this predicts no change in CO concentration associated ambient
(outdoor) air as it moves into a microenvironment, though the CO concentration within the
microenvironment will be affected by inputs for air exchange rate and indoor sources.
4-20
-------
Table 4-4. Estimated values of distribution parameters and variables in equation 4-11 as
implemented in the application of pNEM/CO to Denver and Los Angeles
(Johnson et al., 2000).
Microenvironment3
Code
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
General
location
Indoors
Indoors
Indoors
Indoors
Indoors
Indoors
Indoors
Indoors
Indoors
Outdoors
Outdoors
Vehicle
Vehicle
Vehicle
Outdoor
Specific location
Residence
Nonresidence A
Nonresidence B
Nonresidence C
Nonresidence D
Nonresidence E
Nonresidence F
Nonresidence G
Residential garage
Near road
Other locations
Automobile
Truck
Mass transit vehicles
Public parking or
fueling facility
Activity diary
locations included in
microenvironment
Indoors - residence
Service station
Auto repair
Other repair shop
Shopping mall
Restaurant
Bar
Other indoor location
Auditorium
Store
Office
Other public building
Health care facility
School
Church
Manufacturing facility
Residential garage
Near road
Bicycle
Motorcycle
Outdoor res. garage
Construction site
Residential grounds
School grounds
Sports arena
Park or golf course
Other outdoor
Automobile
Truck
Bus
Train/subway
Other vehicle
Indoor parking garage
Outdoor parking garage
Outdoor parking lot
Outdoor service station
Parameter Estimates for Equation 4-1 1
A
0.621
0.621
0.621
0.621
0.621
0.621
0.621
0.621
0.621
0.621
0.621
0.621
0.621
0.621
0.621
O-L
0.4208
0.4208
0.4208
0.4208
0.4208
0.4208
0.4208
0.4208
0.4208
0.4208
0.4208
0.4208
0.4208
0.4208
0.4208
CTT
0.4879
0.4879
0.4879
0.4879
0.4879
0.4879
0.4879
0.4879
0.4879
0.4879
0.4879
0.4879
0.4879
0.4879
0.4879
M(m)
1.034
2.970
1.213
1.213
1.213
1.213
1.213
0.989
1.034
1.607
1.436
3.020
3.020
3.020
2.970
Notes:
a Aggregate microenvironments defined for statistical analysis of Denver PEM data: residence (1 and 9), service/parking (2 and 15), commercial (3
through 7), and vehicle (12 through 14).
4-21
-------
4.4.4.3.1 Data Used To Estimate pNEM/CO Microenvironmental Algorithm
Parameters
The parameter values for the location factor (aL\ time-of-day (err), and ambient
concentration exponent (A) were based on data collected during a residential monitoring study
described by Wilson et al. (1995). Ten-minute CO concentrations were measured outside 293
residences throughout California in 1992 including customers of Pacific Gas and Electricity
(PG&E) (129 residences in Northern California), San Diego Gas and Electric Company (89
residences in the San Diego area), and Southern California Gas Corporation (75 residences in the
Los Angeles area). After excluding the PG&E data (i.e., not part of the Los Angeles study area)
and homes for which valid CO data were not available, analysts used a remaining subset of 156
residences, 70 from Los Angeles and 86 from San Diego, as the basis for estimating values of OL,
OT, and A applicable to the Los Angeles study area.2 This data subset contained 44,726 valid 10-
minute averages measured outside of residences, of which less than 1% were negative (smallest
value = -1.0 ppm), 14,817 (33%) were equal to 0 ppm, and the remainder were positive
(maximum = 68.7 ppm). These valid 10-minute CO concentrations were then averaged by clock
hour to permit comparison with hourly CO concentrations reported by nearby fixed-site
monitors.
An assumption was made by the original data analysts to maximize the number of hourly
averaged outdoor residential samples available to use in determining the algorithm parameters.
It was proposed that the negative concentrations in this data set were most likely caused by the
subtraction of an offset from all measured values to account for monitor drift. To adjust for this
offset and to prevent the occurrence of negative and zero values in the hourly-averaged data
(which could not be used in fitting equation 4-11), analysts added a constant offset of 0.5 ppm to
all hourly-averaged values measured outside a residence. In addition, seventeen (0.2%) of the
original hourly averages < -0.5 ppm were removed from the data set (i.e., the offset adjustment
would still yield a concentration of <0 ppm). Each of the resulting one-hour outdoor residential
CO concentrations was paired with the one-hour CO concentration measured simultaneously at
the nearest fixed-site monitor [based on data obtained from EPA's Aerometric Information
Retrieval System (AIRS)]. The fixed-site ambient monitoring data were used as reported. This
approach yielded a final database containing 6,330 pairs of hourly average CO concentrations, in
2 Note these same coefficient values derived from the California measurement data were also applied to
estimate exposures in the pNEM/CO Denver study area, as researchers were unable to identify a usable data set
specific to Denver.
4-22
-------
which each pair was indexed by date, time, residence identifier, fixed-site monitor identifier, and
fixed-site monitor scale (e.g., neighborhood scale).
The parameters for the microenvironmental factors (orM(m)) in equation 4-11 were
derived from data generated through the Denver Personal Monitoring Study (Akland et al, 1985;
Johnson, 1984). During this study, each of approximately 450 subjects carried a personal
exposure monitor (PEM) for two 24-hour periods. Each PEM measured CO concentration
continuously. The PEM readings were averaged by exposure event such that each event was
associated with a single microenvironment and a single clock hour (e.g., 1 pm to 2 pm). Event
durations ranged from one minute to one hour. The microenvironment assigned to each PEM
reading was determined from entries made in a real-time diary carried by the subject.
Researchers created a database in which each PEM CO concentration was matched to the
corresponding hourly-average CO concentration reported by the nearest fixed-site monitor. The
data were first processed by excluding data with missing measurements, where measurements
failed a quality control check, and instances in which applicable diary data indicated the presence
of smokers or gas stoves. Each PEM CO concentration was then assigned to a
microenvironment, m, based on entries in the activity dairy. In some cases, data for two or more
similar microenvironments were aggregated to provide more stable estimates than those based on
the very limited amount of data available for specific microenvironments (see Table 4-4
footnote). For consistency with the above described Wilson et al. (1995) database, all cases with
a zero PEM measurement were excluded, as were all cases in which the fixed-site monitor
concentration was zero after rounding to the nearest integer ppm. Note that the Denver fixed-site
data were recorded to the nearest 0.1 ppm, whereas the Los Angeles fixed-site data were only
recorded to the nearest integer.
4.4.4.3.2 Development of the pNEM/CO Microenvironmental Algorithm Form
Equation 4-11 was based on the results of data analyses that suggested that the
relationship between COout(c, m, d, h) and COmon(d, h) should account for the specific
microenvironment, the geographic location of the microenvironment, and the time-of-day.
Analysts recognized that numerous statistical algorithms could have been developed. In
specifying the algorithm that was ultimately used (i.e., equation 4-11), the analysts attempted to
balance the need for simplicity and parsimony with the need to represent the patterns in
concentration variability observed in the available data. The bulk of the algorithm development
was based on the Wilson et al. (1995) database, that is, hourly average 10-minute CO
concentrations measured outside residences in southern California paired with hourly average
CO concentrations measured at the nearest fixed-site monitor. For this case and consistent with
equation 4-11 nomenclature, m represented the residence microenvironment in the district d.
4-23
-------
The district d was initially taken to be the entire study region where measurements were
collected (i.e., San Diego and Los Angeles areas).
Analysts began by considering a simple linear regression model of the form
C0out(c,m,d,h) = a(m,d) + A x [COmon(d,h)] + e(c,m,d,h) (4-12)
where the residual term e(c,m,d,h) was assumed to be independent and normally distributed with
a mean of zero. For simplicity and parsimony, the slope coefficient^ was assumed to be the
same for all microenvironments (m) and districts (d).
Although the coefficient of determination (R2) for this linear regression model was
moderate (0.53),3 the model was found to be unacceptable because it does not properly reflect
the strong correlations that were observed between concentrations measured outside the same
location. Instead, this form of regression model assumes that the residuals associated with a
particular residential location are independent. In other words, this model does not properly
separate out the variation between locations from the variation within locations. Analysts
identified two other deficiencies in this model: (1) large negative values of the randomly-selected
e(c,m,d,h) term could produce negative outdoor concentrations, an unrealistic exposure scenario,
and (2) the model did not generate outdoor concentrations characterized by lognormal
distributions. Various researchers (e.g., Ott, 1995) have demonstrated that ambient CO
concentrations tend to be characterized by lognormal distributions rather than normal
distributions.
To better address these latter concerns, analysts evaluated an alternative model where the
natural logarithm of outdoor concentration was expressed as a linear function of the natural
logarithm of monitor concentration:
l^[C0out(c,m,d,h)] = a(m,d) + A x U^[COmon(d,h)] + e(c,m,d,h) (4-13)
In this equation and those that follow, LN[ ] indicates the natural logarithm of the
quantity in brackets. To properly separate the variability between and within locations, the
intercept term a(m,d) was also permitted to vary with the cohort location, c, leading to the final
selected algorithm:
t(c,m, d,h)] = a(c,m,d) + A* U^[COmon(d,h)] + e(c,m,d,h) (4-14)
3 Note that the R2 goodness-of-fit statistic is not an appropriate measure of model adequacy when the true,
underlying errors are highly correlated.
4-24
-------
Exponentiating both sides of equation 4-14 yields the equivalent formulation to that
presented above in equation 4-11:
C0out(c,m,d,h) = M(m) x L(c,m,d) * T(c,m,d,h) x [COmon(d,h)]A (4-15)
where
M(m) = expjmean [a(c,m,d)]}, averaged over cohorts,
L(c,m,d) = exp{a(c,m,d) - mean [a(c,m,d)]}, and
T(c,m,d,h) = exp[e(c,m,d,h)].
Several alternative statistical models were considered by analysts during the development
of the selected algorithm formulation. Early in the process, analysts evaluated a series of
autoregressive time series models, in which model predictions were influenced by the past
history of CO concentrations at the monitor and outdoors of the microenvironment. These
models were rejected for several reasons: (1) they were inherently complex, (2) they yielded a
wide variation in model coefficients which did not always produce reasonable estimates when
applied to specific California residences, and (3) they required microenvironment-specific time
series data for coefficient estimation which were not readily available for non-residential
microenvironments.
Analysts also evaluated algorithms similar to equation 4-11 in which the exponent^
varied with microenvironment. These algorithms were rejected due to the need for parsimony
and perhaps more importantly, the lack of sufficient, suitable data for estimating
microenvironment-specific values of A A simpler model in which the exponent^ is fixed at 1
was rejected because fits of equation 4-11 to the California data indicated that^4 differed
significantly from 1 (p<0.01). In addition, the assumption that^4 = 1 produced unrealistically
high predictions for outdoor CO concentrations when the model was applied to monitoring data
obtained from the Denver Broadway site (ID 08310002). These high values were found to be a
direct result of setting^ = 1, which forced the geometric standard deviation of the estimated
outdoor concentrations to significantly exceed the geometric standard deviation of the monitor
values.
Analysts ultimately arrived at equation 4-11 (equivalent to equation 4-15), which permits
the A exponent to differ from 1.0. The model was fitted using statistical software for a mixed
(random and fixed effects) model which employed restricted maximum likelihood estimation.
4-25
-------
The fit yielded estimates of OL = 0.4208, OT = 0.4879, and A = 0.621, the values subsequently
used in the pNEM/CO runs described by Johnson et al. (2000). The fitted value ofM(m),
representing residences in Los Angeles during 1992, was actually 0.9706. An alternative value
(1.034), based on the additional analyses described below, was applied to the indoor-residence
microenvironment in the pNEM/CO runs (see Table 4-4).
This algorithm, considered a reasonable compromise between model simplicity and
performance, is completely specified by four parameters \M(m), OL, OT, and^4]. Note that OL, OT,
and^4 are defined to be independent of the microenvironment, whereas M(m) is
microenvironment-specific. At the time of the initial algorithm development, researchers were
unable to find a single data source capable of providing estimates of all four parameters.
Consequently, values for OL, OT, and A were estimated by analyzing data obtained from the
Wilson et al. (1995) database, whereas the specificM(m) values were based on data provided by
the Denver PEM database (Akland et al, 1985; Johnson, 1984).
Researchers conducted a series of sensitivity analyses to evaluate the potential effects on
parameter estimates of variations in the regional location and scale of the fixed-site monitor.
Equation 4-11 was fitted to a series of data subsets defined by region (Los Angeles or San
Diego) or by the scale of the fixed-site monitor (based on the estimated maximum distance from
the monitor represented by the measured concentrations: micro, middle, neighborhood, or urban
scale). The fitted values of OL, OT, A, andM(m) were very similar across the different subsets,
supporting the assumption that these parameters can be assumed to be representative of
concentration patterns outside residences in other regions and for other time periods, and can be
chosen to be the same value for all monitoring scales. Due to a lack of additional suitable data,
the values of OL, q^, and^4 are also assumed to be applicable to concentrations outside all other
microenvironments, although M(m) varies with the particular microenvironment (see below).
In equation 4-11, the COout(c, m, d, h) term represents the outdoor CO concentration
associated with a particular microenvironment m, even when the microenvironment is an indoor
location. Few of the Denver outdoor PEM concentrations could be reliably associated with
particular indoor microenvironments. Consequently, researchers employed a simplified
procedure for estimating M(m) values which assumed that the mean of the indoor PEM values
associated with each indoor microenvironment was approximately equal to the mean of the
outdoor concentration for the microenvironment.4 This assumption is consistent with the results
of applying mass-balance modeling to non-reactive pollutants in enclosed spaces where the only
4 Because the simplified approach was also less sensitive to the wide variation in averaging times exhibited
by the PEM values (i.e., one minute to 60 minutes), analysts were able to use the majority of PEM values in the
statistical analysis. Limiting the analysis to one-hour PEM values would have significantly reduced the pool of
usable data.
4-26
-------
source of the pollutant is the outside air. In such cases, the mean indoor concentration
approximates the mean outdoor concentration, with the instantaneous indoor concentration
exhibiting a lower degree of variability than the corresponding outdoor concentration.
When equation 4-11 is expressed in a logarithmic form (i.e., as in equation 4-14) and
averaged over cohorts, one obtains the equation
Mean{LN[CCWc, m, d, h)}
= Mean[a(c, m,d)] + A* Mean{LN[COmon(d, h)]} + Mean[efc, m, d, h)]
= LN[M(m)] +A x Mean{LN[COmo/J, h)]}.
Therefore, the value ofM(m) equals
M(m) = expjMean U*[COout(c, m, d, h)}~A* Mean LN[COmon(d, h)]} (4-16)
where A = 0.621 (as above). This equation was then used to obtain estimates ofM(m) for each
particular microenvironment, or aggregate of microenvironments, as indicated in Table 4-4 using
the available Denver PEM study data (Akland et al, 1985; Johnson, 1984). The same value of
M(m) was applied to each specific microenvironment within an aggregate.
4.4.4.4 The Micronenvironmental Algorithm as Implemented by APEX3.1
As discussed in section 4-3, the pNEM/CO model effectively evolved into what is known
today as the APEX model. In APEX3.1, the portion of the outdoor concentration affecting the
indoor concentration is determined by the formula
CO out = Ambient x Proximity x Penetration (4-17)
Note that we can represent Proximity and Penetration as distributions in APEX3.1.
These distributions can be sampled hourly, daily, or yearly. Let us make the following
substitutions of the variables used to estimate the outdoor concentrations:
Ambient = [COmon(d,h)}A (4-18)
Proximity = M(m) x L(c, m, d) (4-19)
Penetration = T(c,m,d,h) (4-20)
which yields
4-27
-------
COout = M(m) x L(c, m, d) x T(c,m,d,h) x [COmon(d,h)}A (4-21)
and is identical to equation 4-11 above.
To obtain results from APEX3.1 that are comparable to that generated by pNEM/CO,
Johnson and Capel (2003) preprocessed the hourly ambient monitor data assigned to the district
containing the microenvironment using the formula
Ambient = [COmon(d,h)]°'621 (4-22)
where COmon(d,h) is expressed in ppm. For each profile, a value for the Proximity term was
selected for each microenvironment from a lognormal distribution with geometric mean equal to
M(m) and geometric standard deviation equal to 1.5232. The natural logarithms of this
distribution were characterized by a normal distribution with an arithmetic mean (JUL) equal to
LN[M(m)] and an arithmetic standard deviation (OL) equal to 0.4208. Consistent with the
pNEM/CO algorithm, Proximity values were not permitted to fall below the 5th percentile of the
specified distribution or above the 95th percentile of the distribution. Table 4-5 lists the
parameter values applicable to the 15 microenvironments defined by Johnson and Capel (2003).
Penetration values were randomly selected for each hour from a lognormal distribution
with geometric mean (GMr) equal to 1.0 and geometric standard deviation (GSDT) equal to
1.6289. As indicated above, the natural logarithms of this distribution followed a normal
distribution with an arithmetic mean (JUT) equal to zero and an arithmetic standard deviation (
-------
Table 4-5. Parameters of bounded lognormal distributions defined for proximity factors
used in applications of APEX3.1 to Los Angeles (Johnson and Capel, 2003).
Microenvironment
Code
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
General
Location
Indoors
Indoors
Indoors
Indoors
Indoors
Indoors
Indoors
Indoors
Indoors
Outdoors
Outdoors
Vehicle
Vehicle
Vehicle
Outdoor
Specific location
Residence
Nonresidence A
Nonresidence B
Nonresidence C
Nonresidence D
Nonresidence E
Nonresidence F
Nonresidence G
Residential
garage
Near road
Other locations
Automobile
Truck
Mass transit
vehicles
Public parking or
fueling facility
Activity diary
locations included in
microenvironment
Indoors - residence
Service station
Auto repair
Other repair shop
Shopping mall
Restaurant
Bar
Other indoor location
Auditorium
Store
Office
Other public building
Health care facility
School
Church
Manufacturing facility
Residential garage
Near road
Bicycle
Motorcycle
Outdoor res. garage
Construction site
Residential grounds
School grounds
Sports arena
Park or golf course
Other outdoor
Automobile
Truck
Bus
Train/subway
Other vehicle
Indoor parking garage
Outdoor parking
garage
Outdoor parking lot
Outdoor service station
Parameters of bounded lognormal
distribution
GM
1.034
2.970
1.213
1.213
1.213
1.213
1.213
0.989
1.034
1.607
1.436
3.020
3.020
3.020
2.970
GSD
1.5232
1.5232
1.5232
1 .5232
1.5232
1.5232
1.5232
1.5232
1.5232
1.5232
1.5232
1.5232
1.5232
1.5232
1.5232
Lower
Bound
(5th pet)
0.5175
1 .4864
0.6071
0.6071
0.6071
0.6071
0.6071
0.4950
0.5175
0.8042
0.7187
1.5114
1.5114
1.5114
1 .4864
Upper
Bound
(95th pet)
2.0661
5.9345
2.4237
2.4237
2.4237
2.4237
2.4237
1 .9762
2.0661
3.2110
2.8693
6.0344
6.0344
6.0344
5.9345
4-29
-------
4.4.5 Estimate Energy Expenditure and Ventilation Rates
APEX4.3 includes a module that estimates COHb levels in the blood as a function of
alveolar ventilation rate, the CO concentration of the respired air, endogenous CO production
rate, and various physiological variables such as blood volume and pulmonary CO diffusion rate.
Alveolar ventilation rate is estimated as a function of oxygen uptake rate, which in turn is
estimated as a function of energy expenditure rate. This section provides a brief summary of the
algorithm used to estimate alveolar ventilation rate. A detailed description of the algorithm,
based on the nonlinear solution to the Coburn-Forster-Kane (CFK) equation (Coburn et al.,
1965), together with the distributions and estimating equations used in determining the value of
each parameter in the algorithm, can be found in Appendix B of this document.
4.4.5.1 Energy Expenditure
McCurdy (2000) has recommended that measures of human ventilation (respiration) rate
be estimated as functions of energy expenditure rate. The energy expended by an individual
during a particular activity can be expressed as
EE = METSxRMR (4-23)
where EE is the average energy expenditure rate (kcal min"1) during the activity and RMR is the
resting metabolic rate of the individual expressed in terms of number of energy units expended
per unit of time (kcal min"1). METS (i.e., metabolic equivalent of work) is a ratio specific to the
activity and is dimensionless.
The METS concept provides a means for estimating the alveolar ventilation rate
associated with each activity. For convenience, \QiEE(iJ,k) indicate the energy expenditure rate
associated with the ith activity on dayy for person k. Equation 4-23 can now be expressed as
EE(iJ,k) = METS(iJ,k) x RMR(k) (4-24)
where RMR(k) is the average value for resting metabolic rate specific to person k. Note that
METS(iJ,k) is specific to a particular activity performed by person k.
4.4.5.2 Oxygen Requirements for Energy Expenditure
Energy expenditure requires oxygen which is supplied by ventilation (respiration).
ECF(k) represents an energy conversion factor defined as the volume of oxygen required to
produce one kilocalorie of energy in person k. The oxygen uptake rate (VOl) associated with a
particular activity can be expressed as
V02(i,j,k) = ECF(k) x EE(ij,k) (4-25)
4-30
-------
where VO2(i,j,k) has units of liters oxygen min"1, ECF(k) has units of liters oxygen kcal"1, and
EE(iJ,k) has units of kcal min"1. The value of VO2(i,j,k) can now be determined from MET(iJ,k)
by substituting equation 4-24 into equation 4-25 to produce the relationship
VO2(i,j,k) = ECF(k) x METS(iJ,k) x RMR(k) (4-26)
4.4.5.3 Excess Post-Exercise Oxygen Consumption
At the beginning of exercise, there is a lag between work expended and oxygen
consumption. During this work/ventilation mismatch, an individual's energy needs are met by
anaerobic processes. The magnitude of the mismatch between expenditure and consumption is
termed the oxygen deficit. During heavy exercise, further oxygen deficit (in addition to that
associated with the start of exercise) may be accumulated. At some point, oxygen deficit reaches
a maximum value, and performance and energy expenditure deteriorate. After exercise ceases,
ventilation and oxygen consumption will remain elevated above baseline levels. This increased
oxygen consumption was historically labeled the oxygen debt or recovery oxygen consumption.
However, the term excess post-exercise oxygen consumption (EPOC) has been adopted here to
represent this phenomenon. APEX4.3 has an algorithm for adjusting the MET values to account
for EPOC. This algorithm is described in detail in section 7.2 of US EPA (2008b).
4.4.5.4 Alveolar Ventilation Rate
Alveolar ventilation (VA) represents the portion of the minute ventilation that is involved
in gaseous exchange with the blood. VO2 is the oxygen uptake that occurs during this exchange.
The absolute value of VA is known to be affected by total lung volume, lung dead space, and
respiration frequency - parameters that vary according to the person and/or exercise rate.
However, it is reasonable to assume that the ratio of VA to VO2 is relatively constant regardless of
a person's physiological characteristics or energy expenditure rate. Consistent with this
assumption, APEX4.3 converts each estimate of VO2(i,j,k) to an estimate of VA(i,j,k) by the
proportional relationship
VA(i,j,k) = 19.63 x V02(ij,k) (4-27)
where both VA and VO2 are expressed in units of liters min"1. This relationship was obtained
from Joumard et al. (1981), who based it on research by Galetti (1959). Equation 4-15 can also
be expressed by the equivalent equation
VA(iJ,k) = 19.63 x METS(iJ,k) x ECF(k) x RMR(k) (4-28)
4-31
-------
IfECF and RMR are specified for an individual, then equation 4-28 requires only an
activity-specific estimate ofMETSto produce an estimate of the energy expenditure rate for a
given activity. APEX4.3 processes time-location-activity data obtained from the CHAD to
create a sequence of activity-specific METS values for each simulated individual. APEX4.3
estimates RMR as a function of body mass based on probabilistic equations specific to age and
gender using equations reported by Schofield (1985). A value ofECF is selected for each
individual from a uniform distribution (minimum = 0.20, maximum = 0.21) based on data
provided by Esmail et al. (1995). Using equation 4-28 and these inputs, APEX4.3 calculates a
sequence of VA values for each simulated individual. These values are provided to the algorithm
that estimates the percent COHb in the blood resulting from the simulated exposure (see section
4.4.7 and Appendix B).
4.4.6 Calculate Exposure
APEX4.3 calculates exposure as a time series of exposure concentrations that a simulated
individual experiences during the simulation period. APEX4.3 determines the exposure using
hourly ambient air concentrations, calculated concentrations in each microenvironment based on
these ambient air concentrations, and the minutes spent in a sequence of microenvironments
visited according to the composite diary. The hourly exposure concentration at any clock hour
during the simulation period is determined using the following equation:
N
^ 1 ^~i hourly-mean ,
C> = — (4-29)
where
Ci = Hourly exposure concentration at clock hour /' of the simulation period
(ppm)
N = Number of events (i.e., varied microenvironments visited/activities
performed) in clock hour /' of the simulation period.
C^r^"ean = Hourly mean concentration in microenvironment y' (ppm)
t(j) = Time spent in microenvironment y' (minutes)
T = 60 minutes
From the hourly exposures, APEX4.3 calculates time series of 8-hour and daily average
exposure concentrations that a simulated individual would experience during the simulation
period. APEX4.3 then statistically summarizes and tabulates the number of persons and person-
days at or above selected hourly, 8-hour, and daily average exposure concentrations in a series of
output tables.
4-32
-------
4.4.7 Calculate Dose
Using time-location-activity pattern data obtained from several diary studies, APEX4.3
constructs a composite diary for each simulated person in the specified population. The
composite diary consists of a sequence of events spanning the specified period of the exposure
assessment (typically one calendar year). Each event is defined by a start time, duration, a
geographic location, a microenvironment, and an activity. Using the algorithms described above
in sections 4.4.4, 4.4.5, and 4.4.6, APEX4.3 provides estimates of CO microenvironmental
concentrations and the persons' alveolar ventilation rate for each event in the composite diary,
for each simulated individual. APEX4.3 then uses these data, together with estimates of various
physiological parameters specific to the simulated individual, to estimate the percent COHb in
the blood at the end of each event. The percent COHb calculation is based on the solution to the
nonlinear Coburn-Forster-Kane (CFK) equation (Coburn et al., 1965), as detailed in Appendix B.
Briefly, the CFK module in APEX4.3 describes the rate of change in COHb blood levels as a
function of the following quantities:
• Inspired CO pressure;
• COHb level;
• Oxyhemoglobin (O2Hb) level;
• Hemoglobin (Hb) content of blood;
• Blood volume;
• Alveolar ventilation rate;
• Endogenous CO production rate;
• Mean pulmonary capillary oxygen pressure;
• Pulmonary diffusion rate of CO;
• Haldane coefficient (M);
• Barometric pressure; and
• Vapor pressure of water at body temperature (47 torr).
If all of the listed quantities except COHb level are constant over some time interval, the
CFK equation has a linear form over the interval and is readily integrated. The solution to the
linear form gives reasonably accurate results for lower levels of COHb (ISA section 4.2.1).
However, CO and oxygen can compete for binding with the available hemoglobin and, therefore,
are not independent of each other. If this dependency is taken into account, the resulting
differential equation is no longer linear. Peterson and Stewart (1975) proposed a heuristic
approach to account for this dependency which assumed the linear form and then adjusted the
4-33
-------
O2Hb level iteratively based on the assumption of a linear relationship between COHb and
O2Hb. This approach was used in the COHb module of the original CO-NEM exposure model
(Biller and Richmond, 1982; Johnson and Paul, 1983).
Alternatively, it is possible to determine COHb at any time by numerical integration of
the nonlinear CFK equation if one assumes a particular relationship between COHb and O2Hb.
Muller and Barton (1987) demonstrated that assuming a linear relationship between COHb and
O2Hb leads to a form of the CFK equation equivalent to the Michaelis-Menten kinetic model that
can be analytically integrated. However, the analytical solution in this case cannot be solved
explicitly for COHb. Muller and Barton (1987) demonstrated a binary search method for
determining the COHb value.
The COHb module used in pNEM/CO employed a linear relationship between COHb and
O2Hb which was consistent with the basic assumptions of the CFK model. The approach
differed from the linear forms used by other modelers in that the Muller and Barton (1987)
solution was employed. However, instead of the simple binary search described in the Muller
and Barton paper, a combination of the binary search and Newton-Raphson root finding methods
was used to solve for COHb (Press et al., 1986).
As mentioned above, the current COHb module included in APEX4.3 is based on the
solution to the nonlinear CFK equation using the assumption adopted by Muller and Barton
(1987) which employs a linear relationship between O2Hb and COHb. The CFK equation does
not have an explicit solution, so an iterative solution or approximation is needed to calculate each
percent COHb value. APEX4.3 solves the CFK equation using a 4th-order Taylor's series with
subintervals. This method, first incorporated in APEX3 (Glen, 2002), is summarized in
Appendix B. The selected method (4th-order Taylor series with subintervals) was chosen
because of its simplicity, fast execution speed, and ability to produce relatively accurate
estimates of percent COHb at both low and high levels of CO exposure.
While there may be other approaches proposed as improvements to the standard CFK
equation (e.g., Bruce and Bruce (2003) multi-compartment model), at this time both the
nonlinear and linear CFK models remain the most widely accepted and validated approaches
used to estimate COHb levels (ISA, section 4.2.3). Before any such future module modifications
could be planned and implemented, a more thorough and balanced evaluation of the uncertainties
needs to be performed to include those uncertainties that may be reduced, as well as those
uncertainties introduced by the model modification,. Briefly as an example, the Bruce and Bruce
(2003) model accounts for distribution of CO to five modeled compartments: the lungs, arterial
blood, mixed venous blood, muscle tissue, and other soft tissues. In accounting for these
additional compartments in a new APEX/COHb module, a number of variables would need to be
introduced. Some of these variables may have data or equations available in the extant literature
4-34
-------
to parameterize the variable (e.g., Q or cardiac output), while others may not be measureable or
are unknown (i.e., the distribution of Q between two tissue compartments). When data do
become available to support such model modifications, one would need to evaluate the
appropriateness of these data sets for estimating the parameter values used for the selected at-risk
population. Further, a comparison of estimated COHb levels using the standard CFK equation
with that of the 5-compartment model indicated that, consistently, the estimated COHb levels
would be lower when considering uptake and storage within muscle tissue, however these
differences were very small, particularly at the lowest (and most relevant) exposure
concentration level evaluated (see Figure 7 of Bruce and Bruce, 2003). This preliminary
comparison indicates that while adding multiple compartments to the COHb model may be more
physiologically representative, the extent of any overall benefit in adding such modeling
complexity to the current approach used is unclear at this time. Given the extremely tight
timeframe for this assessment and the relative strength of the dose modeling approach used, we
elected to use the nonlinear CFK model to best approximate population-based end-of-hour
COHb levels for this current CO NAAQS review.
And finally, the current structure of APEX allows the user to control the random
sampling of model input parameters, such that, the same persons, their personal attributes, and
microenvironmental factors will be identical from one simulation to another. Modelers can then
vary a particular input to evaluate the impact to exposure and dose results. This is being used in
this REA to develop estimates of the contribution of ambient exposure to an individual's COHb
levels, an additional metric of interest in this current APEX application. Results for simulations
that are identical in all respects except their CO exposure can be used to separate the contribution
of endogenous CO production to an individual's maximum end-of-hour COHb level from that of
the ambient exposure contribution. For such an analysis, two simulations are performed: the first
is a typical simulation that generates exposure and dose in the presence of ambient CO and the
second simulation uses ambient concentrations equal to zero at all monitors and all hours of the
day. In this first simulation, the exposures persons experience will be a result of their contact
with ambient and microenvironmental CO concentrations, while end-of-hour dose levels will
reflect both the contribution from CO exposure and endogenous CO production. In the second
simulation, exposure concentrations will be zero for all hours and for all persons, while end-of-
hour COHb dose levels will be that resultant from endogenous CO production alone. The
difference in the event-level time series for each individual (and entire population) can thus be
used to approximate the contribution from ambient CO for all exposure events throughout the
simulation period. See section 5.10 and Appendix B for details.
4-35
-------
4.4.8 Model Output
All of the output files written by APEX4.3 are ASCII text files; the complete list and
their descriptions can be found in Table 5-1 of the APEX4.3 User's Guide (US EPA, 2008a). In
general, the simulation output files most relevant to results generated for the assessment include
summary tabulations of population exposure concentrations and maximum end-of-hour COHb
levels. Detailed event-level (minute to 1-hour in duration) or hourly-average information can
also be output for each of the exposure and dose metrics of interests as well as activity specific
ventilation rates and energy expenditures. For example, both the hourly and events APEX files
were needed to estimate a distribution of microenvironmental-to-ambient concentration ratios
(see section 5.10). However, given the potential size of the files that can be generated for a large
population and assessment duration, it is not common to generate event-level files outside of
research purposes. Specific outputs generated for the purposes of the current CO exposure and
dose assessment are discussed in section 6.1.
4.5 KEY OBSERVATIONS
Presented below are key observations related to the modeling system used for the
population assessment of CO exposure and dose.
• APEX, an EPA human exposure and dose model, has a long history of use in estimating
exposure and dose for many of the criteria pollutants including CO, Os, SO2, and NO2.
Over time, EPA has improved and developed new model algorithms, incorporated newer
available input data and parameter distributions, as well as performed several model
evaluations, sensitivity analyses, and uncertainty characterizations for the above
pollutants. Based on this analysis, APEX was judged to be an appropriate model to use
for assessing CO exposure and dose.
4-36
-------
4.6 REFERENCES
Akland GG, Hartwell TD, Johnson TR, Whitmore RW. (1985). Measuring human exposure to carbon monoxide in
Washington, D. C. and Denver, Colorado during the winter of 1982-83. Environ Sci Technol. 19:911-918.
Biller WF and Richmond HM. (1982). Sensitivity Analysis on Coburn Model Predictions of COHb Levels
Associated with Alternative CO Standards. Report to Strategies and Air Standards Division of the Office
of Air quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park,
NC. November, 1982.
Bruce EN and Bruce MC. (2003). A multicompartment model of carboxyhemoglobin and carboxymyoglobin
responses to inhalation of carbon monoxide. JAppl Physiol. 95:1235-1247.
Coburn RF, Forster RE, Kane RB. (1965). Considerations of the physiology and variables that determine blood
carboxyhemoglobin concentration in man. JClin Invest. 44:1899-1910.
Esmail S, Bhambhani Y, Brintnell S. (1995). Gender differences in work performance on the Baltimore therapeutic
equipment work simulator. Amer J Occup Therapy. 49:405-411.
GalettiPM. (1959). Respiratory exchanges during muscular effort. Helv Physiol Acta. 17:34-61.
Glen G. (2002). Programmer's Guide for the APEX3 Model. Prepared by ManTech Environmental Technology,
Inc., for the U.S. Environmental Protection Agency, Research Triangle Park, NC. April, 2002.
Glen G, Smith L, Isaacs K, McCurdy T, Langstaff J. (2008). A new method of longitudinal diary assembly for
human exposure modeling. J Expos Sci Environ Epidem. 18:299-311.
Goldstein BR, Tardiff G, Hoffnagle G, Kester R. (1992). Valdez Health Study: Summary Report. Alyeska Pipeline
Service Company, Anchorage AK.
Hartwell TD, Clayton CA, Ritchie RM, Whitmore RW, Zelon HS, Jones SM, Whitehurst DA. (1984). Study of
Carbon Monoxide Exposure of Residents of Washington, DC and Denver, Colorado. Research Triangle
Park, NC: U.S. Environmental Protection Agency, Office of Research and Development, Environmental
Monitoring Systems Laboratory. EPA-600/4-84-031.
Isaacs K, McCurdy T, Errickson A, Forbes S, Glen G, Graham S, McCurdy L, Nysewander M, Smith L, Tulve N,
Vallero D. (2009). Statistical properties of longitudinal time-activity data for use in EPA exposure models.
Poster presented at the American Time Use Research Conference; College Park MD, June 26, 2009.
Johnson T and Capel J. (2003). Total Risk Integrated Methodology. TRIM Expolnhalation User's Document.
Volume III: Air Pollutant Exposure Model Criteria Air Pollutants Case Study (Carbon Monoxide
Exposures in Los Angeles). Report prepared by TRJ Environmental, Inc., for the Office of Air Quality
Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, North Carolina.
Johnson T and Paul R. (1983). The NAAQS Exposure Model (NEM) Applied to Carbon Monoxide. EPA Report
No. 450/5-83-003. Prepared by PEDCo Environmental, Inc. for the Office of Air Quality Planning and
Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC.
Johnson T. (1984). A Study of Personal Exposure to Carbon Monoxide in Denver, Colorado. Research Triangle
Park, NC: U.S. Environmental Protection Agency, Environmental Monitoring Systems Laboratory. EPA-
600/4-84-014.
Johnson T. (1989). Human Activity Patterns in Cincinnati, Ohio. Palo Alto, CA: Electric Power Research Institute.
EPRI EN-6204.
4-37
-------
Johnson T, Capel J, Olaguer E, Wijnberg L. (1992). Estimation of Carbon Monoxide Exposures and Associated
Carboxyhemoglobin Levels in Denver Residents Using a Probabilistic Version of NEM. Report prepared
for the Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research
Triangle Park, North Carolina.
Johnson T, Mihlan G, LaPointe J, Fletcher K, Capel J. (2000). Estimation of Carbon Monoxide Exposures and
Associated Carboxyhemoglobin Levels for Residents of Denver and Los Angeles Using pNEM/CO
(Version 2.1). Report prepared by ICF Consulting and TRJ Environmental, Inc., under EPA Contract No.
68-D6-0064. U.S. Environmental Protection Agency, Research Triangle Park, North Carolina. Available
at: http://www.epa.gov/ttn/fera/human_related.html. June 2000.
Joumard R, Chiron M, Vidon R, Maurin M, Rouzioux JM. (1981). Mathematical models of the uptake of carbon
monoxide on hemoglobin at low carbon monoxide levels. Environ Health Persp. 41:277-289.
Klepeis ME, Tsang AM, Behar JV. (1996). Analysis of the National Human Activity Pattern Survey (NHAPS)
Respondents from a Standpoint of Exposure Assessment. Washington, DC: U.S. Environmental Protection
Agency, Office of Research and Development. EPA/600/R-96/074.
Liu L-JS, Box M, Kalman D, Kaufman J, Koenig J, Larson T, Lumley T, Sheppard L, Wallace L. (2003). Exposure
assessment of paniculate matter for susceptible populations in Seattle. Environ Health Persp. Ill :909-
918.
Mansfield CA and Corey C. (2003). Analysis of Survey Data on Ozone Alert Days. Research Triangle Park:
Research Triangle Institute. (Task 4 Report).
Mansfield C, Johnson FR, Van Houtven G. (2006). The missing piece: averting behavior for children's ozone
exposures. Resource Energy Econ. 28:215-228.
Mansfield C, Van Houtven G, et al. (2004). Parental Averting Behavior With Respect To Ozone Alerts. Research
Triangle Park NC: Research Triangle Institute.
McCurdy T. (2000). Conceptual basis for multi-route intake dose modeling using an energy expenditure approach.
J Expos Anal Environ Epidemiol. 10:1-12.
McCurdy T, Glen G, Smith L, and Lakkadi Y. (2000). The National Exposure Research Laboratory's Consolidated
Human Activity Database. J Expos Anal and Environ Epidemiol. 10:566-578.
Muller KE and Barton CN. (1987). A nonlinear version of the Coburn, Forster and Kane model of blood
Carboxyhemoglobin. Atmos Environ. 21:1963-1967.
OttW. (1995). Environmental Statistics and Data Analysis. CRC Press, Boca Raton.
Peterson JE and Stewart RD. (1975). Predicting the Carboxyhemoglobin levels resulting from carbon monoxide
exposures. JApplPhysiol. 39(4):633-638.
Press WH, Flannery BP, Teukolsky SA, Vettering WT. (1986). Numerical Recipes. Cambridge University Press.
SchofieldWN. (1985). Predicting basal metabolic rate, new standards, and review of previous work. HumNutr
ClinNutr. 39C(S1):5-41.
Spier CE, Little DE, Trim SC, Johnson TR, Linn WS, Hackney JD. (1992). Activity patterns in elementary and
high school students exposed to oxidant pollution. J Expo Anal Environ Epidemiol. 2:277-293.
Tsang AM and Klepeis NE. (1996). Descriptive Statistics Tables from a Detailed Analysis of the National Human
Activity Pattern Survey (NHAPS) Data. U.S. Environmental Protection Agency. EPA/600/R-96/148.
4-38
-------
US Census Bureau. (2007). Employment Status: 2000- Supplemental Tables. Available at:
http://www.census.gov/population/www/cen2000/phc-t28.html.
US DOT. (2007). Part 3-The Journey To Work files. Bureau of Transportation Statistics (BTS). Available at:
http://transtats.bts.gov/.
US EPA. (1999). Total Risk Integrated Methodology. Available at:
http://www.epa.gov/ttnatw01/urban/trim/trimpg.html.
US EPA. (2002). EPA's Consolidated Human Activities Database. Data and associated documentation available
at: http://www.epa.gov/chad/.
US EPA. (2007). Ozone Population Exposure Analysis for Selected Urban Areas (July 2007). Office of Air
Quality Planning and Standards, Research Triangle Park, NC. EPA-452/R-07-010. Available at:
http://epa.gOv/ttn/naaqs/standards/ozone/s o3crtd.html.
US EPA. (2008a). Total Risk Integrated Methodology (TRIM) Air Pollutants Exposure Model Documentation
(TRIM.Expo/APEX, Version 4.3). Volume 1: Users Guide. Report no. EPA-452/B-08-00la. Office of
Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC.
Available at: http://www.epa.gov/ttn/fera/human apex.html
US EPA. (2008b). Total Risk Integrated Methodology (TRIM) Air Pollutants Exposure Model Documentation
(TRIM.Expo/APEX, Version 4.3). Volume 2: Technical Support Document. Report no. EPA-452/B-08-
00Ib. Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research
Triangle Park, NC. Available at: http://www.epa.gov/ttn/fera/human apex.html
US EPA. (2008c). Risk and Exposure Assessment to Support the Review of the NO2 Primary National Ambient
Air Quality Standard. Report no. EPA-452/R-08-008a. November 2008. Available at:
http://www.epa.gov/ttn/naaqs/standards/nox/data/2008112 l_NO2_REA_final.pdf.
US EPA. (2009a). Risk and Exposure Assessment to Support the Review of the SO2 Primary National Ambient Air
Quality Standard. Report no. EPA-452/R-09-007. August 2009. Available at:
http://www.epa.gov/ttn/naaqs/standards/so2/data/200908SO2REAFinalReport.pdf.
US EPA. (2009b). Carbon Monoxide National Ambient Air Quality Standards: Scope and Methods Plan for Health
Risk and Exposure Assessment. Report no. EPA-452/R-09-004. Available at:
http://www.epa.gOv/ttn/naaqs/standards/co/s co cr_pd.html.
University of Michigan. (2010). The Panel Study of Income Dynamics (PSID). Data and documentation Available
at: http://www.psidonline.isr.umich.edu/Data/.
Wiley JA, Robinson JP, Piazza T, Garrett K, Cirksena K, Cheng Y-T, Martin G. (1991a). Activity Patterns of
California Residents: Final Report. California Air Resources Board, Sacramento, CA. ARB/R93/487.
Available from: NTIS, Springfield, VA., PB94-108719.
Wiley JA, Robinson JP, Cheng Y-T, Piazza T, Stork L, Pladsen K. (1991b). Study of Children's Activity Patterns:
Final Report. California Air Resources Board, Sacramento, CA. ARB-R-93/489.
Williams R, Suggs J, Creason J, Rodes C, Lawless P, Kwok R, Zweidinger R, Sheldon L. (2000). The 1998
Baltimore paniculate matter epidemiology-exposure study: Part 2. Personal exposure associated with an
elderly population. J Expo Anal Environ Epidemiol. 10(6):533-543.
Williams R, Suggs J, Rea A, Leovic K, et al. (2003a). The Research Triangle paniculate panel study: PM mass
concentrations relationships. AtmosEnviron. 37:5349-5363.
4-39
-------
Williams R, Suggs J, Rea A, Sheldon L, et al. (2003b). The Research Triangle paniculate panel study: modeling
ambient source contributions to personal and residential PM mass concentrations. Atmos Environ.
37:5365-5378.
Wilson AL, Colome SD, Tian Y. (1995). California Residential Indoor Air Quality Study. Volume III: Ancillary
and Exploratory Analysis. Integrated Environmental Services, Irvine, California.
4-40
-------
5 APPLICATION OF APEX4.3 IN THIS ASSESSMENT
5.1 PURPOSE
This chapter presents detailed information regarding the varied input data sources, the
APEX model settings, and input variable parameterizations used in estimating population
exposure and dose in the Denver and Los Angeles study areas. In particular, this chapter (and its
associated appendices) describes the:
• geographic study areas and time periods defined for the exposure and dose analyses,
• method and parameters used to construct a composite diary for each simulated individual,
• study area population, the modeled at-risk population and associated CHD prevalence
rates,
• exposure scenarios under evaluation,
• air quality and meteorological data used for each study area and exposure scenario,
• method used to estimate local outdoor and microenvironmental CO concentrations, and
• additional output data files generated for this particular assessment.
Note that the APEX model version used in this assessment was APEX4.3, but for
simplicity will be referred to as APEX in much of the discussion that follows.
5.2 OVERVIEW
As summarized above in section 1.3, the previous analysis of population CO exposure
employed the pNEM/CO model in Denver and Los Angeles study areas, comprising the majority
of census tracts within those metropolitan areas (Johnson et al., 2000). In this earlier exposure
assessment, air quality data were obtained from multiple fixed-site monitors within the study
areas, and the exposure assessment accounted for the effects of geographic location, a diverse set
of microenvironments, commuting within the study area, and selected indoor sources (e.g.,
passive smoking, gas stoves). In the specific application of APEX described in this CO REA, a
similar exposure and dose modeling approach has been developed by staff, though without
inclusion of indoor source emissions. The detailed approach presented here was designed in
consideration of comments and recommendations made by the CASAC and public regarding the
earlier draft CO REAs (US EPA, 2009a; 2010).
The general description of APEX, the standard databases used, modeling capabilities, as
well as the history of the pNEM/APEX series of exposure models, can be found in chapter 4.
This includes use of the national data files obtained from the US Census Bureau (i.e., the 2000
Census data) for the following types of information:
5-1
-------
• Population data and employment probabilities by gender, age, and census tract;
• Locations of census tracts (latitude and longitude); and
• Commuting flows for combinations of home and work census tracts.
Other default input files provided within APEX include tables of age- and gender-specific
physiological parameters (e.g., body weight) and activity-specific metabolic equivalents (METs).
The contents of each of these default files and their use were summarized in chapter 4. They are
described in greater detail in the APEX Users Guide (US EPA, 2008a) and the APEX Technical
Support Document (US EPA, 2008b). The typical output files (e.g., number of persons at or
above a selected exposure or dose level) were also summarized in chapter 4, though additional
exposure and dose outputs were generated for this assessment using the APEX hourly and events
files (US EPA, 2008a, 2008b) and are described in section 5.10.
5.3 STUDY AREAS
As discussed in section 3.2, areas within Denver, Colorado, and Los Angeles, California,
were selected for the current exposure and dose assessment. Briefly, considerations in selection
of these areas included: the prior analysis of these locations in CO NAAQS reviews, the areas
having historically elevated CO concentrations, and the areas currently having some of the most
complete ambient monitoring data available. The monitors selected for use in defining the air
quality in each urban area are listed in Tables 5-1 (Denver) and 5-2 (Los Angeles).
The actual study areas were defined as including all census tracts within 10 km of the
selected fixed-site monitors. These areas are illustrated in Figures 5-1 and 5-2, which indicate
the locations of the fixed-site monitors and the circular 10-km region surrounding each ambient
monitor. Each 10 km region defines the aforementioned air district that includes the geographic
area (i.e., the census tracts) represented by data from the associated CO monitor. Note that all air
districts have the same radius (10 km), a value specified by the "AirRadius" input parameter of
APEX. Any tracts residing within overlapping monitor radii were assigned to the closest
monitor.
In addition to defining the air districts, the model user must specify a location for the
center of the study area and a value for "CityRadius." The circular area defined by the city
center location and the value of "CityRadius" must be large enough to include all census tracts
included in the air districts. For Denver, staff used the location of monitor ID 08310014 (Denver
-Carriage) for the city center and set the "CityRadius" equal to 20 km (Figure 5-1). Staff used
the location of monitor ID 06371103 (Los Angeles) for the center city of Los Angeles and set the
"CityRadius" equal to 65 km (Figure 5-2).
5-2
-------
5.4 EXPOSURE PERIODS
EPA selected the following calendar years as the study periods for each area:
Denver: 1995 and 2006
Los Angeles: 1997 and 2006
The year 2006 was selected for both cities because it was the most recent year of
monitoring data that met the 75% completeness requirement for the ambient monitors listed
above. Note, the CO levels reported for 2006 were well below the 8-hour NAAQS (see Tables
5-1 and 5-2) and are considered representative of the as is air quality in each study area for
purposes of this assessment. The year 1995 for Denver and the year 1997 for Los Angeles were
selected as periods for which the ambient monitor concentrations were near or exceeding the 8-
hour average CO NAAQS of 9 ppm. Staff judged that these historical monitoring data would be
most useful in representing air quality that just meets the current or alternative CO standards and,
following an appropriate concentration level adjustment, would represent a particular air quality
scenario (see sections 5.6 and 5.7.3).
Table 5-1. Attributes of fixed-site monitors selected for the Denver study area.
Monitor ID
City
Local Name
Latitude
Longitude
Elevation (m)
Scale
Objective
19952™
Highest 8-hour
avg. CO (ppm)
2006 2na
Highest 8-hour
avg. CO (ppm)
031-0002a
Denver
CAMP
39.751184
-104.987625
1593
Microscale
Highest
Concentration
9.5
3.1
031-00133
Denver
NJH-E
39.738578
-104.939925
1620
Neighborhood
Population
Exposure
6.2
2.5
031-00143
Denver
Carriage
39.800333
-105.099973
1640
-
Unknown
5.9
3
059-00023
Arvada
-
39.751761
-105.030681
1621
Neighborhood
Population
Exposure
4.6
2
Notes:
3 Identified monitor was used in the 2000 pNEM/CO analysis (Johnson et al., 2000).
5-3
-------
1 Table 5-2. Attributes of fixed-site monitors selected for the Los Angeles study area.
Monitor ID
City
Local Name
Latitude
Longitude
Elevation (m)
Scale
Objective
19972"°
Highest 8-hour
avg. CO (ppm)
2006 2na
Highest 8-hour
avg. CO (ppm)
037-01 13a
West LA
-
34.05111
-118.45636
91
-
Unknown
4.1
1.9
037-1 002a
Burbank
-
34.17605
-118.31712
168
-
Unknown
7.2
3.4
037-1 103a
Los
Angeles
-
34.06659
-118.22688
87
-
Unknown
5.9
2.5
037-1201
Reseda
-
34.19925
-118.53276
226
-
Unknown
7.7
3.4
037-13013
Lynwood
-
33.92899
-118.21071
27
Middle
Highest
Cone.
15
5.6
037-20053
Pasadena
-
34.1326
-118.1272
250
-
Unknown
5.4
2.7
037-40023
Long
Beach
-
33.82376
-118.18921
6
-
Unknown
6.4
3.3
059-0001 /7a'b
Anaheim
-
33.83062
-117.93845
45
Neighborhood
Population
Exposure
5.4
2.9
059-1003
Costa
Mesa
-
33.67464
-117.92568
0
Middle
Unknown
5
2.5
059-5001 a
La Habra
-
33.92513
-117.95264
82
-
Population
Exposure
5.7
2.9
Notes:
3 Identified monitor was used in the 2000 pNEM/CO analysis (Johnson et al., 2000).
b When considering the two monitoring periods (1997 and 2006), two separate ambient monitor IDs were noted (059-0001 and 059-0007) though
effectively the locations of both monitors were the same.
5-4
-------
.: 08-031-0014
x. vr ^ [.
Legend
• Study Area Center
O Selected MET Station
*** Selected CO Monitor
J Circle Radius (Study Area Center) = 20 km
Population (2007) Density
(people per square km)
0-900
900-1,900
'1,900-3,100
Circle Radius (Selected MET Station) = 15.5 km
j "j Circle Radius (Selected CO Monitor) = 10 km
3,100-5,300
5,300 - 9,500
Figure 5-1. Ambient monitor locations, air districts (black circles), meteorological zones
(blue circles), and study area (red circle) for the Denver exposure modeling
domain.
5-5
-------
No: 06-037-1301
No: 06-037-4002
No: 06-059-5001
06-059-0001/0007
No: 06-059-1003
Legend
• Study Area Center
Selected MET Station for 2006
Selected MET Station for 1997
Population (2007) Density
(people per square km)
0 - 2,500
A, 2,500 - 5,500
*"* Selected CO Monitor 5,500-10,300
Circle Radius (Selected MET Station for 1997) = 70.5 km ^| 10,300 - 20,400
Circle Radius (Selected MET Station for 2006) = 70.5 km • 20,400 - 38,400
Circle Radius (Study Area Center) = 65 km
L... J circ'e Radius (Selected CO Monitor) = 10 km
100 Kilometers
Figure 5-2. Ambient monitor locations, air districts (black circles), meteorological zones
(blue and pink circles), and study area (red circle) for the Los Angeles
exposure modeling domain.
5-6
-------
5.5 STUDY POPULATION
Population estimates were obtained from the 2000 US Census for Denver and Los
Angeles study areas.1 In light of the health outcome of interest and characteristics of the
susceptible population, the population in each area was first restricted to persons aged 18 years
or older. Next, the populations were adjusted to remove residents commuting outside of the
study area. The resulting population for the Denver study area was 617,020 persons (or 81.1%
of the total population >18 years of age residing in modeled census tracts). The corresponding
figure for Los Angeles was 5,017,551 persons (or 88.5% of the total population >18 years old
within the modeled census tracts). These populations are referred to below as the total simulated
population in each area. To obtain adequate representation of the simulated population while
also keeping the model runs tractable, fifty-thousand exposure profiles (or simulated individuals)
were run by APEX for each study area and exposure scenario.2
5.5.1 Simulated at-Risk Subpopulations
As mentioned above, the simulated at-risk populations within each study area focused on
adults (ages 18 and older), consistent with the previous CO exposure assessment (Johnson et. al,
2000) and the completed 1994 CO NAAQS review (US EPA, 1992), as the incidence of heart
disease in younger individuals is extremely small (CDC, 2009). For this assessment, we
identified two at-risk subpopulations using disease prevalence rates characterized in the National
Health Interview Survey (NHIS): (1) the potential population comprised of persons with
"coronary heart disease" and (2) the potential population comprised of persons with "all types of
heart disease".
This first category (i.e., coronary heart disease) is limited to those persons with diagnosed
coronary heart disease, angina pectoris, and heart attack (CDC, 2009) in addition to an estimate
of those persons having undiagnosed CHD (US EPA, 2010). The second category (i.e., all types
of heart disease) is inclusive of those with diagnosed coronary heart disease, angina pectoris,
heart attack, and any other heart condition or disease (CDC, 2009), in addition to an estimate of
those persons with undiagnosed coronary heart disease. The specific data and method used to
estimate the prevalence rates for each of these subpopulations is provided in the following
subsections.
The disease prevalence rates (stratified by age and gender) are used to generate a
population-weighted representative sample of simulated individuals that are then used to
calculate exposure and associated COHb for the simulated at-risk subpopulations. While this
1 No adjustments were made to census estimates to reflect alternate years.
2 There were a few APEX simulations performed for purposes of obtaining the exposure and dose time-
series for each individual. These runs were 5,000 persons only. See section 5.10 for details.
5-7
-------
provides estimates of exposure and COHb levels for subpopulations in the two study areas
having the demographic characteristics (i.e., age and gender) of the two at-risk populations of
interest, we note that the simulation does not include any other characteristics specific to the at-
risk populations of interest. For example, the CHAD diaries used to represent the simulated
individuals are not exclusively drawn from a pool of diaries of persons identified by the original
activity pattern survey as having the disease state(s) of interest.3 This limitation and its effect on
exposure and dose estimates for representing the simulated at-risk subpopulations are discussed
in chapter 7 below.
5.5.1.1 Coronary Heart Disease
For estimates of adults with diagnosed CHD, staff obtained CHD prevalence data from
the NHIS for 2007 (CDC, 2009). The CHD prevalence for the population at or above 18 years of
age is about 6% (ISA, section 5.7.2.1).4 We assumed the national prevalence rates for CHD
were appropriate to use in each of the two study areas because there was a general similarity in
the reported regional rates. Although we desired the prevalence rates to be stratified by age and
gender, the available data were stratified by age or gender. Table 5-3 provides national
prevalence data for CHD by age and Table 5-4 provides CHD stratified by gender. The gender-
only data were used to estimate gender-specific adjustment factors to apply to the age-only data
set. For males, the adjustment factor = 0.080/0.061 = 1.31; for females, the adjustment factor =
0.045/0.061 = 0.74. Table 5-5 provides the estimated national prevalence rates for CHD by age
range adjusted for gender using these adjustment factors.
Table 5-3. National prevalence rates for diagnosed coronary heart disease by age range.
Age range
18 to 44
45 to 64
65 to 74
75+
Prevalence rate (fraction) for diagnosed coronary
disease3
heart
0.009
0.067
0.186
0.236
Notes:
3 Source: Coronary heart disease statistics in Table 2 of NHIS (CDC, 2009),
which include coronary heart disease, angina pectoris and heart attack.
Note though, that the activity database did include a few activity pattern studies (e.g., NHAPS) where a
disease state was identified (including having a lung or heart condition). Therefore, in sampling for diaries using
attributes such as age and gender, some of the APEX simulated individuals would have their activity sequence
constructed of diaries obtained from persons having a heart condition.
4 Note that in the last CO NAAQS review completed in 1994, the estimated number of individuals with
CHD represented about 3% of the entire (all ages) US population (US EPA, 1992).
-------
Table 5-4. National prevalence rates for diagnosed coronary heart disease by gender.
Age range
18+
Prevalence rate (fraction) for diagnosed coronary heart
disease3
Total
0.061
Males
0.080
Females
0.045
Notes:
3 Source: Coronary heart disease statistics in Table 2 of NHIS (CDC, 2009),
which include coronary heart disease, angina pectoris and heart attack.
Table 5-5. Estimated national prevalence rates for diagnosed coronary heart disease,
stratified by age and gender.
Age range
18 to 44
45 to 64
65 to 74
75+
Prevalence rate (fraction) for diagnosed
coronary heart disease
Males3
0.012
0.088
0.244
0.310
Females'3
0.007
0.050
0.138
0.175
Notes:
3 Values listed in Table 5-3 were multiplied by 1 .31 .
b Values listed in Table 5-4 were multiplied by 0.74.
The selected at-risk population was then expanded to also include undiagnosed cases of
coronary heart disease using a method similar to that developed by OAQPS for use in the 2000
exposure assessment (see Appendix F of Johnson et al., 2000). Briefly, in the prior assessment
the prevalence estimates of diagnosed IHD5 were stratified by age and sex (Adams and Marano,
1995) and constituted approximately 8.0 million individuals in the civilian, non-institutionalized
population.6 In addition, as many as three to four million persons were estimated by the
American Heart Association as having silent ischemia or undiagnosed IHD (AHA, 1990).7 We
used this information to provide estimates of the undiagnosed IHD population for use in the
pNEM/CO model. We assumed 3.5 million persons had undiagnosed IHD and assumed the
prevalence to be distributed by age and gender in the same manner as diagnosed IHD. These
data yield an adjustment factor of 0.438 (i.e., 3.5 million/8.0 million) to apply to the diagnosed
The NHIS prevalence rates used in the 2000 assessment used the term IHD, rather than CHD (Adams and
Marano, 1995).
6 These estimates did not include individuals in the military or individuals in nursing homes or other
institutions.
7 Note that the size of this undiagnosed IHD population (i.e., 3-4 million persons) is the same as that
reported by AHA (2003).
5-9
-------
prevalence for use in estimating the undiagnosed prevalence. Consequently, this factor can be
interpreted as the undiagnosed cases may be 43.8% of the diagnosed prevalence.
Table 5-6 lists the results of applying the 0.438 factor to the age and gender stratified
prevalence rates listed in Table 5-5. This assumes that CHD and IHD are identical with respect
to the ratio of undiagnosed cases to diagnosed cases and this ratio has not changed since reported
in 1990 (and 2003) and assumes that undiagnosed prevalence rates would not vary by gender.8
This total prevalence for coronary heart disease (diagnosed and undiagnosed combined) stratified
by gender was used by APEX in estimating the first simulated at-risk subpopulation.
When using these CHD prevalence rates in the APEX model runs, there were 383,040
simulated persons (or 7.6% of the total simulated population) with either diagnosed or
undiagnosed CHD in the Los Angeles study area, while in Denver there were 53,656 simulated
persons (or 8.7% of the total simulated population) within the CHD simulated at-risk population.
Table 5-6. Estimated national prevalence rates for coronary heart disease, including
diagnosed and undiagnosed cases, stratified by age and gender.
Age range
18 to 44
45 to 64
65 to 74
75+
Prevalence rate (fraction) for coronary heart disease
Males
Diagnosed
0.012
0.088
0.244
0.310
Undiagnosed3
0.005
0.038
0.107
0.135
Total
0.017
0.127
0.351
0.446
Females
Diagnosed
0.007
0.050
0.138
0.175
Undiagnosed3
0.003
0.022
0.060
0.077
Total
0.010
0.072
0.198
0.252
Notes:
a Values listed in Table 5-5 (diagnosed CHD) were multiplied by 0.438 to estimate the undiagnosed
prevalence. This calculation assumed CHD and IHD are identical with respect to the ratio of undiagnosed
cases (3.5 million) to diagnosed cases (8.0 million), that this ratio has been constant since reported in
1990 and 2003, and that there is no gender difference in undiagnosed prevalence rates.
5.5.1.2 All Heart Disease
For estimates of adults with heart disease (HD), we also obtained prevalence data from
the NHIS for 2007 (CDC, 2009). The HD prevalence for the population above 18 years of age is
about 11% (Table 2, CDC, 2009). The national prevalence rates for HD were used in each of the
two study areas because there was a general similarity in the reported regional rates. As
described in section 5.5.2.1, although staff desired the prevalence rates to be stratified by age and
gender, the available data were stratified by age or gender. Table 5-7 provides national
8 Specific data on which to base development of differing prevalence estimates by gender for undiagnosed
CHD as compared to diagnosed CHD were not identified in the limited time available for this assessment.
5-10
-------
prevalence data for HD by age and Table 5-8 provides HD stratified by gender. These gender-
only data were used to estimate gender-specific adjustment factors to apply to the age-only data
set. For males, the adjustment factor = 0.125/0.112= 1.12; for females, the adjustment factor =
0.102/0.112 = 0.91. Table 5-9 provides the estimated national prevalence rates for HD by age
range adjusted for gender using these adjustment factors.
Table 5-7. National prevalence rates for all types of diagnosed heart disease by age range.
Age range
18 to 44
45 to 64
65 to 74
75+
Prevalence rate (fraction) for all
disease
types of diagnosed heart
a
0.041
0.122
0.271
0.358
Notes:
3 Source: Statistics for all types of heart disease listed in Table 2 of NHIS
(CDC, 2009), which include coronary heart disease, angina pectoris, heart
attack, or any other heart condition or disease.
Table 5-8. National prevalence rates for all types of diagnosed heart disease by gender.
Age range
18+
Prevalence rate (fraction) for all types of diagnosed
heart disease3
Total
0.112
Males
0.125
Females
0.102
Notes:
a Source: Statistics for all types of heart disease listed in Table 2 of NHIS
(CDC, 2009), which include coronary heart disease, angina pectoris,
heart attack, or any other heart condition or disease.
Table 5-9. Estimated national prevalence rates for all types of diagnosed heart disease,
stratified by age and gender.
Age range
18 to 44
45 to 64
65 to 74
75+
Prevalence rate (fraction) for all types of
diagnosed heart disease3
Males
0.046
0.137
0.304
0.401
Females
0.037
0.111
0.247
0.326
Notes:
a Values listed in Table 5-7 were multiplied by 1 .12 for males
and 0.91 for females using data from Table 5-8.
The same approach described in section 5.5.2.1 was used to include an estimate of the
5-11
-------
percent of persons with undiagnosed coronary heart disease in addition to the estimate of
population estimated having HD. The undiagnosed CHD prevalence from Table 5-6 was simply
added to the HD to generate the prevalence rates summarized in Table 5-10. The total
prevalence listed for each gender was used by APEX to estimate the second simulated at-risk
population.
When using these prevalence rates in the APEX model runs, there were 630,807
simulated persons (or 12.6% of the total simulated population) with HD in the Los Angeles study
area, while in Denver there were 85,926 simulated persons (or 13.9% of the total simulated
population) comprising the same simulated at-risk population.
Table 5-10. Estimated national prevalence rates for all types of diagnosed heart disease
plus undiagnosed coronary heart disease, stratified by age and gender.
Age range
18 to 44
45 to 64
65 to 74
75+
Prevalence rate (fraction) for all types of heart disease
Males
Diagnosed
heart
disease
0.046
0.137
0.304
0.401
Undiagnosed
coronary
heart
disease3
0.005
0.038
0.107
0.135
Total
0.051
0.175
0.410
0.536
Females
Diagnosed
heart
disease
0.037
0.111
0.247
0.326
Undiagnosed
coronary
heart
disease3
0.003
0.022
0.060
0.077
Total
0.040
0.133
0.307
0.402
Notes:
a Values obtained from Table 5-6 (i.e., undiagnosed CHD).
5.5.2 Time-Location-Activity Patterns
APEX constructs a 365-day longitudinal diary for each simulated individual by selecting
24-hour diaries from those available in CHAD. In performing the exposure assessments
described in this report, all available diaries for persons above age 17 in the CHAD database
were used.
5.5.3 Construction of Longitudinal Diaries
As discussed in section 4.4.3.4, APEX provides a longitudinal diary assembly algorithm
that enables the user to create composite diaries that reflect the tendency of individuals to repeat
day-to-day activities (Glen et al., 2008). The user specifies values for two statistical variables (D
and A) that relate to a key daily variable, typically the time spent per day in a particular
microenvironment (e.g., in a motor vehicle). The D statistic reflects the relative importance of
intra- and inter-personal variance within the selected key daily variable. The A variable
5-12
-------
quantifies the day-to-day autocorrelation in the selected key daily variable. APEX then
constructs composite diaries that exhibit the statistical properties defined by the specified values
ofD and A
In this exposure assessment, we used the longitudinal diary algorithm to construct year-
long activity patterns for each simulated individual to reflect the day-to-day correlation of time
spent inside motor vehicles. Each diary day in the CHAD database was tagged with the number
of minutes spent in the vehicle microenvironment. Parameter settings ofD = 0.31 and A = 0.19
were specified to control the day-to-day repetition of time spent in motor vehicles in the
constructed composite diaries. These particular D and A values were obtained from Isaacs et al.
(2009) (see Appendix C).
In selecting particular diaries to represent the simulated population, the CHAD data are
categorized or separated by APEX into data pools. The pools were defined by three ranges for
o o o
the maximum temperature of the diary day (< 55.0 F, between 55.0 and 83.9 F, and>84.0 F)
and two day-types (i.e., weekend and weekday); thus, there were 3x2 = 6 diary pools. The
window for age was set at 15%. For example, diaries can be selected for a simulated individual
of age 60 from CHAD individuals ranging from ages 51 though 69 (i.e., 60 +/- 15 percent).
5.6 EXPOSURE SCENARIOS
In this CO REA, the exposure scenario refers to the air quality conditions considered for
each APEX simulation. Staff evaluated five exposure scenarios for each study area. The first
exposure scenario used unadjusted 2006 ambient air quality as input to APEX; this is designated
as the as is air quality exposure scenario. The purpose of this scenario is to determine the
number of persons that may experience COHb levels at or above selected benchmarks when
considering current air quality conditions. The next four exposure scenarios used ambient data
from a high concentration year in each location (i.e., the 1995 monitoring data in Denver and the
1997 monitoring data in Los Angeles) adjusted to represent different air quality conditions. The
purpose of these scenarios is to determine the number of persons that may experience COHb
levels at or above selected benchmark levels when considering air quality conditions that just
meet a selected level, form, and averaging time of interest. This is not the same as considering
exposures associated with the as is air quality conditions.
The first of these four adjusted air quality exposure scenarios considered ambient
concentrations adjusted to just meet the current 8-hour CO NAAQS of 9 ppm. The 8-hour
standard was selected when considering the two current standards (8-hour and 1-hour) because it
5-13
-------
is the controlling standard.9 The second of these exposure scenarios using the historical
monitoring data also considered the form of the current 8-hour CO standard, but with the
ambient concentrations in each study adjusted to meet an alternative standard level of 5 ppm.
The next two scenarios considered percentile forms of potential alternative standards, consistent
with the alternative standards investigated for other criteria pollutants (e.g., NC>2 (US EPA,
2008c); and SO2 (US EPA, 2009b)). The first of these potential percentile forms considered a
99th percentile daily maximum 8-hour average CO concentration of 5.0 ppm, while the second
considered the same form though with a 1-hour averaging time and a 1-hour level of 8.0 ppm.
Details regarding the concentration adjustments associated with each of the current and potential
alternative standards are provided in section 5.7.3.
Tables 5-11 and 5-12 provide perspective on the selected levels and the air quality used
to represent each scenario in Denver and Los Angeles, respectively. An array was constructed
using the varying air quality scenarios to indicate how a single APEX run using a particular air
quality input data set might reflect different levels and forms of potential alternative standards.
For example, in Denver, the exposure and dose results for the as is scenario would be the same
as a standard level of 3.1 ppm in terms of second highest non-overlapping 8-hour average (Table
5-11). The generated results would also represent exposures and doses experienced considering
a 99th percentile 1-hour daily maximum concentration of 4.5 ppm.
9 The controlling standard by definition would be the standard that allows air quality to have either a 2nd
highest 8-hour average concentration of < 9.4 ppm (i.e., the 8-hour standard is the controlling standard) or to have a
2nd highest 1-hour concentration of < 35.4 ppm (i.e., the 1-hour standard is the controlling standard).
5-14
-------
Table 5-11. Array of alternative standard forms and levels informed by modeled exposure
scenarios in Denver.
Denver Design Values (ppm)
Averaging Time & Form
Air Quality Scenario
8-hour
0)
O)
X
re
E
re
•c
0)
?
o>
1-hour
0)
O)
X
re
E
re
0)
O
0)
Q.
o>
o>
As Is
3.1
2.8
4.6
4.5
Current 8-hour standard (9 ppm)
9.4
7.2
16.2
13.3
2 highest 8-hour average (5 ppm)
5.4
4.1
9.3
7.7
99 percentile daily max 8-hour (5.0 ppm)
6.5
5.0
11.2
9.2
99 percentile daily max 1 -hour (8.0 ppm)
5.6
4.3
9.7
8.0
Notes:
3 This is the form of the current standards.
b Note that the rounding convention for the current standard allows for
concentrations of up to the given standard level plus 0.4 ppm.
5-15
-------
Table 5-12. Array of alternative standard forms and levels informed by modeled exposure
scenarios in Los Angeles.
Los Angeles Design Values (ppm)
Averaging Time & Form
Air Quality Scenario
As Is
Current 8-hour standard (9 ppm)
2 highest 8-hour average (5 ppm)
99 percentile daily max 8-hour (5.0 ppm)
99 percentile daily max 1 -hour (8.0 ppm)
Notes:
3 This is the form of the current standards.
8-hour
0)
O)
5.6
9.4
5.4
5.7
6.5
x
re
E
re
•c
0)
?
95th percentiles) in Denver when compared with the Los Angeles
ambient concentration distribution, for either year.
5-16
-------
5.7.2 Method for Estimating Missing 1-Hour Ambient Concentrations
APEX requires that each site-year of monitoring data be complete (i.e., it is free of hourly
gaps in concentration levels). The missing values in each data set were estimated by the
sequential application of the following four methods.
1) If the data gap was less than six continuous missing values, the missing values were
estimated by linear interpolation using the valid values at the ends of the gap.
2) Where possible, data gaps of at least six hours were estimated as linear functions of
hourly values reported by other ambient CO monitors in the area. Linear regression
was used to develop a set of models that were specific to time-of-day and monitor. The
model selected to estimate missing values for a particular time of day was the model
that maximized the variance explained (R2) for that hour, subject to the constraints that
regression model R2 was greater than 0.5 and the number of available measurements
used in constructing the model was at least 50.
3) In cases where method 2 (above) could not be used (i.e., no regression models were
available for a particular time-of-day) and the gap was less than nine hours, the missing
values were estimated by linear interpolation between the valid values at the ends of
the gap.
4) All remaining missing values were substituted with the 1-hour concentration from the
same day and hour as the nearest monitor. The hourly concentration used was
normalized to the respective monitors' monthly mean concentrations.
Tables 5-13 to 5-16 provide the descriptive statistics for 1-hour CO concentrations in
each data set, before and after estimating missing values, and considering the two years of
ambient monitoring data in each study area. The excellent agreement between concentrations at
the various percentiles of the distribution (before and after substitution) indicates that the
addition of the estimated missing-value concentrations did not significantly affect the overall
distributions of the hourly CO concentrations.
5-17
-------
Table 5-13. Descriptive statistics for hourly carbon monoxide concentrations before and after estimation of missing values -
Denver 1995.
Monitor
ID
31-0002
31-0013
31-0014
59-0002
Missing
values
filled?
No
Yes
No
Yes
No
Yes
No
Yes
1 -hour values
(n)
Present
8697
8760
8647
8760
8701
8760
8680
8760
Missing
63
0
113
0
59
0
80
0
CO concentration (ppm)
Mean
1.50
1.50
1.25
1.25
1.09
1.09
0.96
0.96
SD
1.20
1.20
1.08
1.08
1.05
1.05
0.93
0.93
Min
0.0
0.0
0.1
0.1
0.0
0.0
0.1
0.1
Percentile
25th
0.8
0.8
0.6
0.6
0.5
0.5
0.4
0.4
50th
1.2
1.2
0.9
0.9
0.7
0.7
0.6
0.6
75th
1.8
1.8
1.5
1.5
1.3
1.3
1.1
1.1
90th
2.7
2.7
2.5
2.5
2.3
2.3
2.0
2.0
95th
3.4
3.4
3.4
3.4
3.2
3.2
2.7
2.7
99th
6.1
6.1
5.5
5.5
5.3
5.3
4.8
4.8
99.9th
13.1
13.1
8.9
8.8
7.7
7.8
7.5
7.5
Max
24.5
24.5
14.6
14.6
10.4
10.4
11.9
11.9
Table 5-14. Descriptive statistics for hourly carbon monoxide concentrations before and after estimation of missing values -
Denver 2006.
Monitor
ID
31-0002
31-0013
31-0014
59-0002
Missing
values
filled?
No
Yes
No
Yes
No
Yes
No
Yes
1 -hour values
(n)
Present
8672
8760
8635
8760
8557
8760
8603
8760
Missing
88
0
125
0
203
0
57
0
CO concentration (ppm)
Mean
0.62
0.62
0.49
0.49
0.47
0.47
0.40
0.40
SD
0.39
0.39
0.36
0.36
0.38
0.38
0.37
0.37
Min
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Percentile
25th
0.4
0.4
0.3
0.3
0.3
0.3
0.2
0.2
50th
0.5
0.5
0.4
0.4
0.4
0.4
0.3
0.3
75th
0.7
0.7
0.6
0.6
0.5
0.5
0.5
0.5
90th
1.0
1.0
0.9
0.9
0.9
0.9
0.8
0.8
95th
1.3
1.3
1.2
1.2
1.2
1.2
1.1
1.1
99th
2.2
2.1
1.8
1.8
2.0
2.0
1.9
1.9
99.9th
4.1
4.1
3.4
3.4
3.1
3.1
2.8
2.8
Max
6.4
6.4
4.4
4.4
3.9
3.9
3.6
3.6
5-18
-------
Table 5-15. Descriptive statistics for hourly carbon monoxide concentrations before and after estimation of missing values -
Los Angeles 1997.
Monitor
ID
37-0113
37-1002
37-1103
37-1201
37-1301
37-2005
37-4002
59-0001/7
59-1003
59-5001
Missing
values
filled?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
1 -hour values
(n)
Present
8360
8760
8025
8760
8292
8760
8245
8760
8302
8760
8250
8760
8347
8760
8354
8760
8325
8760
8230
8760
Missing
400
0
735
0
468
0
515
0
458
0
510
0
413
0
406
0
435
0
530
0
CO concentration (ppm)
Mean
0.84
0.84
1.75
1.73
1.36
1.36
1.15
1.17
2.35
2.34
1.11
1.10
1.11
1.11
1.11
1.11
0.74
0.74
1.36
1.36
SD
0.86
0.85
1.27
1.24
1.19
1.17
1.25
1.24
2.19
2.17
0.84
0.83
1.10
1.11
0.91
0.90
1.01
1.00
1.21
1.19
Min
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Percentile
25th
0.2
0.2
0.9
0.9
0.5
0.5
0.4
0.4
1.1
1.1
0.6
0.6
0.4
0.4
0.6
0.6
0.2
0.2
0.6
0.6
50th
0.6
0.6
1.4
1.4
0.9
1.0
0.7
0.7
1.7
1.7
0.9
0.9
0.7
0.7
0.8
0.8
0.3
0.3
1.0
1.0
75th
1.2
1.2
2.2
2.1
1.9
1.9
1.5
1.5
2.8
2.8
1.4
1.4
1.3
1.4
1.4
1.4
0.9
0.9
1.7
1.7
90th
2.0
2.0
3.5
3.5
3.1
3.0
2.8
2.8
4.9
4.9
2.1
2.1
2.7
2.7
2.3
2.3
2.1
2.1
2.8
2.8
95th
2.6
2.6
4.5
4.4
3.9
3.8
3.8
3.8
6.8
6.7
2.8
2.8
3.6
3.6
2.9
2.9
3.0
3.0
3.7
3.7
99th
3.7
3.6
6.1
6.0
5.4
5.4
6.0
5.9
11.3
11.2
4.2
4.2
5.2
5.2
4.6
4.6
4.7
4.6
6.2
6.2
99.9th
5.1
5.1
7.8
7.7
7.2
7.1
8.4
8.3
17.2
17.2
6.1
6.0
7.3
7.2
6.9
6.9
6.3
6.2
9.9
9.9
Max
7.3
7.3
8.8
8.8
8.9
8.9
11.7
11.7
19.2
19.2
8.1
8.1
9.0
9.0
8.4
8.4
7.3
7.3
11.9
11.9
5-19
-------
Table 5-16. Descriptive statistics for hourly carbon monoxide concentrations before and after estimation of missing values -
Los Angeles 2006.
Monitor
37-0113
37-1002
37-1103
37-1201
37-1301
37-2005
37-4002
59-0001/7
59-1003
59-5001
Missing
values
filled?
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
1 -hour values
(n)
Present
8365
8760
8345
8760
8265
8760
8375
8760
8275
8760
8258
8760
8216
8760
8342
8760
8358
8760
8227
8760
Missing
395
0
415
0
495
0
385
0
485
0
502
0
544
0
418
0
402
0
533
0
CO concentration (ppm)
Mean
0.42
0.43
0.67
0.67
0.55
0.56
0.55
0.56
1.00
1.01
0.73
0.73
0.74
0.75
0.43
0.43
0.33
0.33
0.64
0.64
SD
0.37
0.37
0.61
0.61
0.50
0.50
0.54
0.53
0.89
0.90
0.49
0.49
0.55
0.54
0.47
0.47
0.45
0.45
0.57
0.56
Min
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Percentile
25th
0.2
0.2
0.3
0.3
0.2
0.2
0.2
0.2
0.5
0.5
0.4
0.4
0.4
0.4
0.1
0.1
0.1
0.1
0.3
0.3
50th
0.3
0.3
0.5
0.5
0.4
0.4
0.4
0.4
0.7
0.7
0.6
0.6
0.6
0.6
0.3
0.3
0.1
0.1
0.4
0.4
75th
0.6
0.6
0.8
0.8
0.7
0.8
0.6
0.7
1.1
1.1
1.0
1.0
0.9
0.9
0.5
0.5
0.4
0.4
0.7
0.7
90th
0.9
0.9
1.5
1.5
1.3
1.3
1.2
1.2
2.0
2.0
1.4
1.3
1.5
1.5
1.0
1.0
0.9
0.9
1.3
1.3
95th
1.2
1.2
2.0
2.0
1.6
1.6
1.7
1.7
2.9
2.9
1.7
1.7
1.9
1.9
1.4
1.4
1.4
1.4
1.8
1.8
99th
1.7
1.7
2.9
2.9
2.3
2.2
2.7
2.7
4.7
4.6
2.4
2.4
2.7
2.7
2.3
2.3
2.1
2.1
3.0
2.9
99.9th
2.5
2.5
4.0
3.9
2.9
2.9
3.8
3.7
6.9
6.8
3.2
3.1
3.7
3.7
3.4
3.4
3.1
3.0
4.7
4.6
Max
2.9
2.9
4.3
4.3
3.5
3.5
4.8
4.8
8.4
8.4
4.1
4.1
4.2
4.2
4.5
4.5
3.5
3.5
6.0
6.0
5-20
-------
5.7.3 Adjusted 1-Hour Ambient Concentrations
In addition to modeling exposures based on recent as is air quality (i.e., ambient
monitoring data for year 2006), exposures and resulting dose were estimated for air quality
conditions that just meet the current 8-hour CO NAAQS and various alternative standards under
evaluation. Because CO concentrations in recent years were significantly lower than the current
NAAQS, staff first selected an earlier year for each city (1995 for Denver and 1997 for Los
Angeles) to represent air quality conditions that were near the current 8-hour CO standard.
Consistent with the data adjustment approach employed in the previous draft CO exposure
assessment (Johnson et al., 2000), and approaches used in prior REAs supporting other pollutant
NAAQS reviews (e.g., US EPA, 2008c; US EPA, 2009b), staff concluded (1) that the policy-
relevant background levels of CO were negligible in each area (section 3.1.4), and (2) that the
fixed-site monitoring data could be adjusted to simulate just meeting the current CO standards by
use of a simple proportional adjustment of all hourly values (section 3.1.5). Consequently, the
following adjustment equation was employed:
COadj(w,/0 = (NAAQS/DV) x CO(m,h). (5-1)
CO(m,h) is the 1-hour CO concentration at hour h for monitor m. It follows that COadj(m,h) is
the adjusted CO concentration for hour h at monitor m through the use of the specific design
value (DV) for monitor m. Although the current 8-hour NAAQS for CO specifies a maximum
concentration of 9 ppm, which is not to be exceeded more than one time in a year, the NAAQS
term in Equation 5-1 is equivalent to 9.4 ppm due to the application of a standard data rounding
convention used in calculating design values (DVs) for CO.10
The DVs for Denver (1995) and for Los Angeles (1997) were 9.5 ppm and 15 ppm,
respectively. The Denver DV is calculated as the second-highest 8-hour average CO
concentration reported by monitor ID 080310002 for 1995. The adjustment factor (or
NAAQS/DV) that was applied equally to all 8,760 hourly ambient CO concentrations at that
monitor is thus 9.4/9.5, or 0.989. In a similar manner, the DV used in Los Angeles is the second-
highest 8-hour average CO concentration reported at monitor ID 060371301 for 1997, giving an
ambient concentration adjustment factor of 9.4/15, or 0.627 which was applied equally to all
8,760 hourly ambient CO concentrations from the Los Angeles monitor.
10 A design value is a statistic that describes the air quality status of a given area or monitor relative to the
level of the NAAQS. For the CO 8-hour NAAQS, the design value is the highest annual second maximum non-
overlapping 8-hour concentration during the most recent two years. The design value for the 1-hour CO NAAQS is
the highest annual second maximum 1-hour concentration during the most recent two years. The latest update
(2007-2008) on the CO design values can be found at: http://www.epa.gov/airtrends/pdfs/dv co 2006 2008.pdf
5-21
-------
Staff evaluated three additional air quality scenarios considering potential alternative
standard levels, averaging times, and forms. Assuming a similar form and averaging time of the
current 8-hour standard (2nd highest non-overlapping 8-hour average CO concentration), staff
selected a level of 5 ppm for the first potential alternative standard.11 As was done for other
recent NAAQS reviews (US EPA, 2008c; US EPA, 2009b), staff selected percentiles of the air
quality distribution and averaging times to identifying potential levels associated with alternative
standards. The second potential alternative standard considered by staff also uses an 8-hour
average concentration, though having a 99th percentile daily maximum CO concentration of 5.0
ppm.12 The final potential alternative standard that staff evaluated was a 99th percentile daily
maximum 1-hour CO concentration of 8.0 ppm. Table 5-17 summarizes the adjustment factors
that were developed from equation 5-1 and used to adjust the high concentration year air quality
data in each study area.
Table 5-17. Design values and adjustment factors used to represent air quality just meeting
the current and potential alternative standards.
Study Area
Denver
Los Angeles
Standard
Averaging
Time
8-hour
1-hour
8-hour
1-hour
Form
2nd highest
99th pet daily max
99th pet daily max
2nd highest
99th pet daily max
99th pet daily max
Level
(ppm)
9
5
5.0
8.0
9
5
5.0
8.0
Design Value3
(ppm)
9.5
7.3
13.5
15
13.1
18.5
Adjustment
Factor
0.989b
0.568
0.685
0.593
0.627b
0.360
0.382
0.432
Notes:
3 All design values were obtained from monitor ID monitor ID 080310002 in Denver (1995 data) and
monitor ID 060371301 in Los Angeles (1997 data).
b Adjustment factor for just meeting the current 8-hour average CO standard.
11 Note that this would allow a 2nd highest non-overlapping 8-hour concentration up to 5.4 ppm (hence the
design value).
12 It was assumed that there are an infinite number of zeros, that is, the level is exactly 5.0 ppm. This
rounding convention also applies to the other potential alternative standard selected; the level of the 99th percentile
1-hour daily maximum is exactly 8.0 ppm.
5-22
-------
Tables 5-18 and 5-19 provide the descriptive statistics for the Denver and Los Angeles
ambient monitor 1-hour CO concentrations, respectively, after applying the appropriate
adjustment factor to simulate just meeting the current standard. As expected, the adjusted
monitoring concentrations for Denver 1995 are very similar to the unadjusted data set given that
the adjustment factor used was close to unity. For example, the maximum concentration at the
design monitor was reduced from 24.5 ppm to 24.2 ppm. The change in CO concentrations was
much greater in Los Angeles compared with that of Denver as a result of differences in the
adjustment factor used in each study area. For example, the maximum CO concentration at the
design monitor in Los Angeles was reduced from 19.2 ppm to 12.0 ppm. Considering the
patterns described above in section 5.7.1 for the unadjusted air quality and given that the
concentration adjustment was proportional, additional remarks can be made regarding
differences in the air quality adjusted to just meet the current 8-hour CO NAAQS. When
comparing the adjusted concentrations in Denver and Los Angeles, there is still a sharper rate of
increase in CO concentrations at and above the 95th percentiles of the distribution, only now all
of the Denver monitors have greater CO concentrations at these upper percentiles when
compared with concentrations observed at all of the Los Angeles monitors (excluding
concentrations at the Los Angeles design monitor).
Given the proportional approach used to adjust ambient concentrations for each of the
other exposure scenarios (e.g., 99th percentile daily maximum 1-hour concentration of 8.0);
similar patterns in concentrations were expected and are therefore not summarized here.
5-23
-------
Table 5-18. Descriptive statistics for hourly carbon monoxide concentrations after adjusting to just meet the current 8-hour
standard - Denver (adjusted 1995 data).
Monitor
ID
31-0002
31-0013
31-0014
59-0002
Hourly-average CO concentration (ppm)
Mean
1.5
1.2
1.1
1.0
SD
1.2
1.1
1.0
0.9
25.0
0.8
0.6
0.5
0.4
50.0
1.2
0.9
0.7
0.6
75.0
1.8
1.5
1.3
1.1
90.0
2.7
2.5
2.3
2.0
95.0
3.4
3.4
3.2
2.7
99.0
6.0
5.4
5.3
4.8
99.5
7.6
6.4
6.4
5.7
99.9
13.0
8.7
7.7
7.4
Max
24.2
14.4
10.3
11.8
DV
(ppm)
9.4
6.1
5.8
4.5
Table 5-19. Descriptive statistics for hourly carbon monoxide concentrations after adjusting to just meet the current 8-hour
standard - Los Angeles (adjusted 1997 data).
Monitor
ID
37-0113
37-1002
37-1103
37-1201
37-1301
37-2005
37-4002
59-0001
59-1003
59-5001
Hourly-average CO concentration (ppm)
Mean
0.5
1.1
0.9
0.7
1.5
0.7
0.7
0.7
0.5
0.9
SD
0.5
0.8
0.7
0.8
1.4
0.5
0.7
0.6
0.6
0.7
25.0
0.1
0.6
0.3
0.3
0.7
0.4
0.3
0.4
0.1
0.4
50.0
0.4
0.9
0.6
0.4
1.1
0.6
0.4
0.5
0.2
0.6
75.0
0.8
1.3
1.2
0.9
1.7
0.9
0.8
0.9
0.6
1.1
90.0
1.3
2.2
1.9
1.8
3.1
1.3
1.7
1.4
1.3
1.8
95.0
1.6
2.8
2.4
2.4
4.2
1.8
2.3
1.8
1.9
2.3
99.0
2.3
3.8
3.4
3.7
7.0
2.6
3.3
2.9
2.9
3.8
99.5
2.6
4.1
3.6
4.3
8.5
2.9
3.7
3.4
3.2
4.5
99.9
3.2
4.8
4.5
5.2
10.8
3.8
4.5
4.3
3.9
6.1
Max
4.6
5.5
5.6
7.3
12.0
5.1
5.6
5.3
4.6
7.5
DV
(ppm)
2.6
4.5
3.7
4.8
9.4
3.4
4.0
3.4
3.1
3.6
5-24
-------
5.8 METEOROLOGICAL DATA
A few algorithms within APEX require meteorological data (primarily temperature) from
stations located within the study area. For example, in selecting a CHAD diary to simulate an
individual's daily activities, a range of daily maximum temperatures is used to categorize diaries
for sampling purposes so as to best match the temperature observed on the simulation day within
the study area (section 5.5.4). In addition, mean temperatures are used by APEX to select from
an appropriate air exchange rate distribution to estimate indoor microenvironmental
concentrations (section 5.8). For the analyses described in this report, hourly temperature data
were obtained from meteorological stations located at or near the fixed-site CO monitor specified
for each study area.
Tables 5-20 and 5-21 list the meteorological stations we used in modeling the Denver and
Los Angeles study areas, respectively. Ideally, staff would have used the same station (Long
Beach: 37-4002) matched for both monitoring years (1997 and 2006) in Los Angeles. Because
this station did not report a complete year of data for 1997, we substituted data reported by the
Long Beach Daugherty Field station located approximately 3.6 km from the 37-4002 station.
The same two stations (31-0002 and 59-0002) were used for the Denver study area for 1995 and
Denver 2006, because there were adequate data for both years for both sites.
To run APEX, a "ZoneRadius" is specified by the user as the maximum radius for the
region surrounding each meteorological station that will be represented by the temperature data
provided by the station. In this assessment, we set this to a value that includes all census tracts
within the air districts. A radius of 15.5 km met this requirement for Denver (Figure 5-1), while
Los Angeles required a larger radius of 70.5 km (Figure 5-2).
Table 5-20. Locations of meteorological stations selected for Denver.
Meteorological station
Monitor ID
31-0002
59-0002
County
Denver
Jefferson
Location coordinates
Latitude
39.751184
39.800333
Longitude
-104.987625
-105.099973
Monitoring Year
1995
1-hour
values
(n)
8742
8702
Mean
temp
(°F)
53.3
49.7
2006
1-hour
values
(n)
8749
8758
Mean
temp
(°F)
55.2
51.5
5-25
-------
Table 5-21. Locations of meteorological stations selected for Los Angeles.
Meteorological station
Monitor ID
Daugherty
Field
37-4002
County
Long Beach
Long Beach
Coordinates
Latitude
33.81667
33.82376
Longitude
-118.15
-118.18921
Monitoring Year
1997
1-hour
values
(n)
8751
~
Mean
temp
(°F)
65.8
~
2006
1-hour
values
(n)
~
8759
Mean
temp
(°F)
~
63.8
5.8.1 Method for Estimating Missing 1-Hour Temperature Data
APEX also requires a complete (full) meteorological data set to run properly. In
checking the meteorological data for completeness, staff noted all stations and years had at least
one missing hourly value for temperature (Tables 5-20 and 5-21). To generate the complete one-
year temperature data sets, we estimated the missing values for the selected meteorological
(MET) stations in Denver and Los Angeles as follows.
For the Denver study area, we selected two MET stations for use in 1995 and 2006. All
missing values in year 2006 were filled using linear interpolation. For the missing values in
1995, staff used linear interpolation to fill in short gaps. Where there were long gaps in the data
(e.g., more than 16 continuous hours of missing values), linear interpolation was judged as
inappropriate because this method would likely not produce reasonable estimates of the potential
variability in temperature (particularly the daily maximum) that might occur during the gap. In
these instances, staff applied an alternative approach in which the average temperature of the
previous day and the latter day were averaged and then substituted for the corresponding hours.
For example, if the temperature data were missing from 1 am to 11 pm on 2/8/1995, staff
averaged the hourly temperature of 2/7/1995 and 2/9/1995 for 1 am, 2 am ..., 11 pm to fill the
missing hours (all eleven hours have an individual value).13
For Los Angeles, we evaluated the two sites identified here as site 1 (ID 037-4002) and
site 2 (located at Daugherty Field). Both locations reported temperature in both years of interest;
however, the degree of completeness for each varied. Given their close proximity to one another
(3.6 km), we decided that a complete data set would be best generated by using a composite of
the two monitors, using the monitor with the greatest number of measurements as the primary
13 Calculating the average temperature using this method does not apply if 1) the long gap occurs on
January 1 or December 31, or if 2) the temperature data in the previous day or the latter day are not available. In
such cases, we used the non-missing values in the previous day or the latter day, whichever was available.
5-26
-------
data set. Because site 1 had fewer missing values than site 2 for 2006, site 1 was selected as the
primary meteorological site to represent the Los Angeles area for that year. For the one missing
value on Site 1 in 2006, the corresponding temperature from Site 2 was used to fill the missing
value for 2006. For 1997, there were 2,263 missing values on Site 1 while only 9 missing values
on Site 2. As a result, Site 2 was selected as the primary meteorological station for 1997. Two
of the nine missing values from Site 2 were available from Site 1. Therefore, these temperatures
were directly substituted with values from the corresponding hours of the Site 1 data set. To fill
the remaining seven missing values, we used linear interpolation by connecting successive
straight line segments and fitting a continuous curve to the data.14
The temperature distributions before and after filling missing values were compared at
for each station in each year to assess the impact (if any) of the substitution method. Given the
limited number of missing values in the original data sets, there were negligible differences when
comparing mean, median, variance and percentile statistics.
5.9 MICROENVIRONMENTS MODELED
This section briefly discusses the approach and specific factors used to estimate CO
microenvironmental concentrations in the current assessment. As described in section 4.4.4.3,
the approach was originally developed for pNEM/CO and used in the previous assessment
(Johnson et al., 2000).
5.9.1 The Microenvironmental Model as Implemented by APEX4.3
Section 8.2.2 of US EPA (2008b) indicates that the mass balance model in APEX4.3
models the portion of outdoor air that enters the microenvironment as
'-''-'out tproximity X Ipenetration X ^'Jambient \-^~^)
Since this is effectively equivalent to the method used by APEX3.1 described in section
4.4.4 A, we used the same method here with respect to application of the proximity and
penetration factors in APEX4.3 to implement equation 4-11. First, to obtain the appropriate CO
concentrations outside each microenvironment, ambient CO concentration were adjusted by an
exponential factor of 0.621 (see equation 4-22). Then for each profile, a value for fproximity term
would be sampled for each microenvironment from a lognormal distribution with geometric
mean (GM) equal ioM(m) and geometric standard deviation (GSD) equal to 1.5232. A value for
/penetration for each hour would also be sampled from a lognormal distribution with geometric
14 This was done in SAS using a procedure, PROC EXPAND, along with the JOIN option.
5-27
-------
mean (GMT) equal to 1.0 and geometric standard deviation (GSDT) equal to 1.6289. As
described in section 4.4.4.4, the /penetration values were bounded at the lower and upper tails of the
distribution by 0.4482 and 2.2313, respectively.
Table 5-16 presents the algorithm parameters proposed for the eight microenvironments
currently defined for the application of APEX to Los Angeles and Denver. These eight
microenvironments were selected rather than the fifteen selected in earlier assessments (see
Tables 4-4 and 4-5) based on the locations having the same proximity factors and air exchange
rates distributions, or when using a similar microenvironmental approach (see section 5.9.5).
Note that when this algorithm is implemented within the APEX framework, the
application of Equation 4-11 will produce a "compression" effect in which the ratio of COout to
COmon tends to become smaller (on average) as COmon increases. This effect is consistent with
data reported by field studies such as Wilson et al. (1995) which have compared outdoor
concentrations with simultaneously measured fixed-site concentrations. Note also that the
effective microenvironment-to-ambient concentration ratios will differ from the proximity factor
distributions provided in Table 5-22 given the influence of other variables used in equation 4-11.
An analysis of these ratios (e.g., in-vehicle to ambient monitor concentration) is provided in
section 6.1.
5-28
-------
Table 5-22. Parameters of bounded lognormal distributions defined for proximity factors
used in the application of APEX4.3 to Los Angeles and Denver.
Microenvironment
Code
1
2
3
4
5
6
7
8
General
location
Indoors
Indoors
Indoors
Indoors
Outdoors
Outdoor
Outdoors
Vehicle
Specific
location
Residence
Service
station and
auto repair
Other
indoor
locations A
Other
indoor
locations B
Near road
locations
Public
parking or
fueling
facility
Other
outdoor
locations
Automobile
and mass
transit
Activity diary
locations
included in
microenvironment
Indoors - residence
Service station
Auto repair
Other re pair shop
Shopping mall
Other indoor location
Auditorium
Store
Office
Other public building
Bars
Restaurants
Health care facility
School
Church
Manufacturing facility
Bus stop
Bicycle
Motorcycle
Other near road
Indoor parking
garage
Outdoor parking
garage
Outdoor parking lot
Outdoor service
station
Outdoor res. garage
Construction site
Residential grounds
School grounds
Sports arena
Park or golf course
Other outdoor
Automobile
Truck
Bus
Train/subway
Other vehicle
Parameters of bounded lognormal
distribution for proximity factor
GM
1.034
2.970
1.213
0.989
1.607
2.970
1.436
3.020
GSD
1 .5232
1 .5232
1 .5232
1 .5232
1 .5232
1 .5232
1 .5232
1 .5232
Minimum
(5th pet)
0.5175
1 .4864
0.6071
0.4950
0.8042
1 .4864
0.7187
1.5114
Maximum
(95th pet)
2.0661
5.9345
2.4237
1 .9762
3.2110
5.9345
2.8693
6.0344
5-29
-------
5.9.2 Microenvironmental Mapping
In APEX, microenvironments represent the exposure locations for simulated individuals.
For exposures to be estimated accurately, it is important to have realistic microenvironments that
match closely to the locations where actual people spend time on a daily basis. It is necessary to
map the CHAD location codes to one of the eight specific microenvironments selected for this
exposure assessment or to a supplemental category (either -1 or 0}. As a reminder, these eight
microenvironments were selected based on having suitable data to use for proximity factors and
air exchange rates (when using a mass balance approach). The -1 code is assigned to events
where the location code is missing (X) or the location is classified as uncertain (U); the -1 code
instructs APEX to use the last known microenvironment for that person's diary in determining
the exposure concentration. The 0 code is assigned to an airplane microenvironment (CHAD
location code: 31160) and instructs APEX to set the exposure concentration equal to 0 ppm.
Figure D-l in Appendix D describes the specific mapping of CHAD codes to
microenvironments.
The microenvironment mapping file also permits the user to assign a home/work/other
(H/W/O) location to each CHAD location code. The home/work/other location determines the
source of the hourly-average monitoring data that will represent the ambient CO concentration
for the microenvironment: the home district monitor, the work district monitor, or other.
The initial APEX assignments of H/W/O to the CHAD location codes were used as a
starting point (see Figure D-l in Appendix D) and modified using a few of the options available
in APEX. First, staff overrode the H/W/O designations listed in the microenvironment mapping
file for selected activities by compiling a list of CHAD activity codes that will always be
associated with the work district (regardless of the CHAD location code). This list is inserted in
the "CustomWork" parameter found in the simulation control file. The default list of work
activity codes, which were used in this application, includes codes 10000 through 10300 (see
Appendix D Table D-l). As a result of using this option, APEX will assign the simulated person
to the work district whenever the activity code falls between 10000 and 10300. This assignment
will override the home/work/assignment associated with the applicable CHAD location code.
There will still be exposure events in which the simulated person is assigned to the
"other" location. In the default mode, APEX uses an average of all monitor values to determine
the ambient concentration for these events. Note that this averaging approach will tend to
smooth the data; that is, it will produce ambient CO concentrations that have slightly less
variance than a comparable set of ambient concentrations obtained from a single monitor. To
avoid this effect, staff chose to specify the option OtherDistricts = 1, so that only one monitor is
used to represent "other." The monitor used in the model application is randomly selected from
the set of all monitors.
5-30
-------
5.9.3 Selection of Microenvironmental Method Used
As discussed in chapter 4, the two approaches available in APEX for calculating pollutant
levels within microenvironments are (1) the mass balance method and (2) the factors method.
Table 5-23 lists the microenvironments used in this study and the calculation method used.
Table 5-23. List of microenvironments modeled and calculation methods used.
Microenvironment
Code
1
2
3
4
5
6
7
8
Location
Indoors
Indoors
Indoors
Indoors
Outdoors
Outdoors
Outdoors
Vehicle
Name
Residence
Service station and auto repair
Other indoor locations A
Other indoor locations B
Near road locations
Public parking or fueling facility
Other outdoor locations
Automobile and mass transit
Calculation
Method
Mass balance
Mass balance
Mass balance
Mass balance
Factors
Factors
Factors
Factors
5.9.4 Air Exchange Rates and Air Conditioning Prevalence
For the microenvironments using the mass balance method (i.e., all indoor
microenvironments), air exchange rate (AER) and air conditioning prevalence data are needed to
estimate microenvironmental concentrations. Air exchange rate data used for the indoor
residential microenvironment were the same used in APEX for the most recent Oj NAAQS
review (US EPA, 2007). As part of that earlier review, AER data were reviewed, compiled, and
evaluated from the extant literature to generate location-specific AER distributions15 categorized
by influential factors, namely temperature and presence of air conditioning. In general,
lognormal distributions provided the best fit, and are defined by a geometric mean (GM) and
standard deviation (GSD). To avoid unusually extreme simulated AER values, bounds of 0.1
and 10 were selected for minimum and maximum AER, respectively. Tables 5-24 and 5-25
15 There were AER measurement data specific to the Los Angeles study area; these were used by US EPA
(2007) to develop AER distributions. Denver was not a location of interest in US EPA (2007); therefore there were
no Denver-specific AER developed for this study area. Consistent with what was done in US EPA (2007) for cities
not having location-specific AER data available, the composite AER distributions developed using data from cities
outside California were applied in this study to Denver.
5-31
-------
summarize the AER distributions used in modeling indoor exposures in Denver and Los
Angeles, respectively, each classified by A/C prevalence and temperature categories. For all
other indoor microenvironments, the AER distributions used here (Tables 5-24 and 5-25) were
based on data provided by an indoor air quality study (Persily et al., 2005). These are the same
AER distributions used for the APEX assessments in the most recent Os NAAQS review (US
EPA, 2007), NO2 REA (US EPA, 2008c) and SO2 REA (US EPA, 2009b).
Because the selection of an air exchange rate distribution is conditioned on the presence
or absence of an air-conditioner (A/C), the air conditioning status of the residential
microenvironments in each modeled area is simulated randomly using the probability that a
residence has an air conditioner. A value of 55% was used to represent the A/C prevalence rate
in Los Angeles, based on data obtained from US EPA (2007). For Denver, residential A/C
prevalence was estimated to be 69% of homes, a value obtained from AHS (2005). Air
conditioning prevalence is noted as being distinct from usage rate, the latter being represented by
the air exchange rate distribution and dependent on temperature.
Table 5-24. Lognormal distributions of indoor air exchange rates used in Denver.
Micro-
environment
Indoors -
residence
Indoors - other
Classification category
A/C present?
Yesb
No
-
Mean Temp
(degrees F)
<50
50-68
68-77
77-86
86+
<50
50-68
68+
-
Parameters of bounded lognormal distribution3
GM
0.919
0.564
0.468
0.424
0.567
0.926
0.733
1.378
1.109
GSD
1.859
1.940
2.201
2.037
1.945
2.084
2.330
2.276
3.015
Minimum
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
Maximum
10.0
10.0
10.0
10.0
10.0
10.0
10.0
10.0
10.0
Notes:
3 Obtained from Table D-4 of US EPA (2007) and derived from locations outside California.
b Estimated air conditioning prevalence rate for Denver = 69% (see Table 1-4 in AHS, 2005).
c Assumed here to be consistent with other approximated lower and upper bounds.
5-32
-------
Table 5-25. Lognormal distributions of indoor air exchange rates used in Los Angeles.
Micro-
environment
Indoors -
residence
Indoors - other
Classification category
A/C present?
Yesb
No
-
Mean Temp
(degrees F)
<50
50-67
68-76
77-85
86+
<50
50-67
68-76
77-85
86+
-
Parameters of bounded lognormal distribution3
GM
0.589
0.589
1.100
0.813
0.266
0.543
0.747
1.372
0.988
0.988
1.109
GSD
1.894
1.894
2.365
2.415
2.790
3.087
2.085
2.283
1.967
1.967
3.015
Minimum
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1C
Maximum
10.0
10.0
10.0
10.0
10.0
10.0
10.0
10.0
10.0
10.0
10.0C
Notes:
3 Obtained from Table D-4 of US EPA (2007).
b Estimated air conditioning prevalence rate for Los Angeles = 55 percent (see page 47 and Table A-3
of US EPA, 2007).
c Assumed here to be consistent with other approximated lower and upper bounds.
5.10 ADDITIONAL EXPOSURE AND DOSE OUTPUT GENERATED USING
REDUCED APEX SIMULATIONS
In a typical model run, APEX generates a complete time-series of exposure and dose for
each simulated individual based on the microenvironmental concentrations they contact and the
activities they perform. Because there are usually thousands of simulated persons and thus
thousands of exposure and dose profiles, it is common practice that only summary output data
are generated. As there are 8,760 hours in a year and potentially multiple events within an hour,
the large size of these time-series profile files presents computational challenges for the data
analyst, and increases the model run time. In this assessment however, we were interested in
additional exposure and dose output data not summarized by the typical APEX output files. The
APEX events and hourly files provide event-level and hourly-level exposure and dose profiles
for all simulated individuals (US EPA, 2008a, 2008b). The time-series of exposure and dose for
each individual are useful in performing three important analyses in this assessment, each
described in the following sections.
5-33
-------
5.10.1 Estimate of Microenvironment Contribution to At-Risk Population Exposure
Levels
The first analysis using the individual-level exposure and dose output was designed to
identify the microenvironments that are most influential to different levels of population
exposure.16 To simplify the presentation of the data in this analysis, the total time spent in each
of the modeled eight microenvironments was aggregated into five microenvironments to evaluate
time spent by simulated persons at selected exposure levels. The aggregation was based on
microenvironments having similar proximity factors (see Table 5-22), resulting in aggregated
microenvironments defined as follows: indoors-low (indoor-residence and indoor-other A&B),
indoors-high (indoors-auto service/repair), outdoors-low (outdoors-other and near-road),
outdoors-high (outdoor-parking/gas station), inside-vehicles. In addition, the total time the CHD
population spent at or above each exposure level was calculated.
5.10.2 Estimate of Microenvironment-to-Ambient Concentration Ratios
The second analysis using the individual-level exposure and dose output was designed to
estimate the effective microenvironmental factors, or the ratio of modeled microenvironmental
concentrations to associated ambient monitor concentrations. The distributions of these effective
microenvironmental factors are then compared to commonly reported microenvironment-to-
ambient concentration ratios that are based on measurement data. The use of these estimated
microenvironmental factors for comparison is more informative than simply using the
distribution of proximity factors given in Table 5-22. As noted earlier, the series of factors in
equation 4-11 are designed to spatially and temporally adjust the ambient concentrations to
reflect concentrations immediately proximal to the microenvironment. The proximity factors
listed in Table 5-22 would effectively reflect microenvironment-to-spatially and temporally
adjusted ambient concentration ratios, not microenvironment-to-fixed site ambient monitor
concentration ratios commonly reported in the extant human exposure literature.
The microenvironment-to-ambient concentration ratios were calculated by dividing each
estimated event-level microenvironmental CO concentration by its corresponding hourly CO
ambient monitor concentration.17 In summarizing these microenvironment-to-ambient
concentration ratios, we excluded those ratios that were associated with ambient concentrations
16 The standard APEX summary output table only generates microenvironment contributions for the
general population (i.e., it is not necessarily representative of the demographics of the CHD orHD population).
17 Note that event-level ambient concentration is not a variable that can be output from APEX at the time of
analysis. The 1-hour ambient concentration in the APEX hourly file does account for when a person might
experience ambient concentrations outside their home tract (e.g., when commuting). When this does occur, the 1-
hour concentration would be time-averaged based on the time spent in each air quality district for each event during
the hour, effectively approximating an average event-level ambient concentration.
5-34
-------
less than 1 ppm. These calculated ratios using the event-level modeled concentrations tended to
be extremely large, particularly when considering the as is air quality, as a result of dividing by
extremely low ambient concentrations rather than the calculation method employed in estimating
the microenvironmental concentrations. Inclusion of these ratios (while valid) would be of little
practical use in interpreting the microenvironment-to-ambient concentration ratios and high
microenvironmental concentrations because the design of the microenvironmental algorithm
results in the highest concentrations being driven largely by the high ambient concentrations, not
by these extreme ratios. For example, Figure 5-3b illustrates an inverse relationship between the
microenvironment-to-ambient concentration ratios and estimated microenvironmental
concentration. Any large proximity factors that might have occurred when randomly sampling
from the distributions used in Table 5-22 were effectively modified by the ambient concentration
exponential adjustment (equation 4-11), thus controlling for extreme ratio and high concentration
combinations in estimating the microenvironmental concentrations. Although the calculated
ratios can reach extremely high values (> 100), they are not responsible for estimating the
highest microenvironmental concentrations. The full distribution representing all event-level
microenvironmental concentrations estimated to be experienced by the CFID population is also
included in this analysis.
ME= ME1: In-Resid
ME= ME8: In-vsti
MEJWT1O
1CXH
5 10 15 20 25 30 35 40 0 5 10 15 20 25 X 35 40
Mcnoenvironmerrtal CO (ppm) Mictoenvironmental CO (ppm)
Figure 5-3. Relationship between microenvironment-to-ambient concentration ratios
using estimated indoor-residential concentrations (left panel) and inside-
vehicle concentrations (right panel) in Denver - as is air quality.
5-35
-------
5.10.3 Estimate of Ambient Exposure Contribution to Total COHb Level
In the third analysis, we were interested in the contribution of ambient CO exposure to
each individual's COHb dose level above their endogenous CO production. As summarized in
section 4.4.7 and described fully in Appendix B, we estimated COHb levels in each simulated
individual using the CFK dose module within APEX. Theoretically, in the absence of ambient
concentrations or other sources of CO, one can perform an APEX simulation to estimate
endogenous CO production and its effect on COHb levels. Two APEX model simulations were
performed for each location and air quality scenario evaluation: one using ambient
concentrations and the second using ambient concentrations equal to zero. Each of the new
runs simulated the complete dose time-series for each individual. By design, the simulated
persons in each of these two model runs line up perfectly in terms of physiology and activities
performed, enabling staff to compare the COHb levels experienced across the two runs event-by-
event. We calculated all event-by-event ambient contributions (COHb ambient contribution) to
corresponding COHb levels (i.e., COHb ambient contribution = % COHb with ambient exposure
minus % COHb for zero exposure), effectively giving the ambient contribution to estimated
COHb levels. Consistent with the dose metric of interest in this assessment, the daily maximum
end-of-hour COHb level is calculated using each individual's entire dose profile, only in this
instance it is the daily maximum end-of-hour contribution to COHb associated with ambient CO
exposure.
5.10.4 Comparison of the Exposure and Dose Results Generated Using 50,000
Persons Versus 5,000 Persons Simulation
For each of these particular model runs, APEX generated the complete time-series of
exposure and dose for 5,000 persons, of which there were approximately 400 CHD simulated
individuals.18 The CHD prevalence rates were used to simulate the at-risk population for two air
quality scenarios in each study area: as is air quality and just meeting the current standard. The
complete output and analysis of these data are provided in chapter 6; however, a comparison of
the summary results is provided here to demonstrate the representativeness of the smaller sample
run size. The summary exposure and dose results from the smaller sample size runs (i.e., the
percent of persons and person-days at or above selected levels) were very similar to the results
generated using a 50,000 person simulation for either exposure scenario and study area. Table 5-
26 summarizes the exposure output using as is air quality, while Table 5-27 summarizes the
summary dose output for air quality just meeting the current standard; as noted above, little
difference can be observed when comparing the 50,000 person simulation to the 5,000 person
18 Of the 5,000 person model simulations, APEX simulated 438 CHD persons in Denver and 394 CHD
persons in Los Angeles.
5-36
-------
simulation for either exposure scenario. Therefore, the time-series results generated for the
smaller sample size simulations are considered representative of the larger modeled CHD
population.19
Table 5-26. Comparison of exposure summary output generated when simulating 50,000
persons versus that of simulating 5,000 persons - as is air quality.
Daily
Maximum
Exposure
(ppm)
1-
hour
8-
hour
>0
>3
>6
>9
>12
>15
>20
>30
>40
>0
>3
>6
>9
>12
>15
>20
>30
>40
Denver CHD Population3'"'0
Percent of Persons
50K Rund
100
99.5
60.6
19.9
5.7
1.6
0.1
0
0
100
57.8
3.2
0.1
<0.1
0
0
0
0
SKRun
100
99.3
59.6
16.4
5.0
1.1
0
0
0
100
55.3
3.0
0.2
0
0
0
0
0
Percent of
Person-days
50K Rund
100
11.2
0.9
0.1
<0.1
<0.1
<0.1
0
0
100
1.0
<0.1
<0.1
<0.1
0
0
0
0
SKRun
100
10.5
0.8
0.1
<0.1
<0.1
0
0
0
100
0.9
<0.1
<0.1
0
0
0
0
0
Los Angeles CHD Population3'"'0
Percent of Persons
50K Run6
100
99.9
75.1
32.0
11.2
3.6
0.6
<0.1
0
100
76.9
9.5
0.8
<0.1
0
0
0
0
SKRun
100
100
78.2
28.4
10.9
4.3
1.0
0
0
100
78.4
9.1
0.8
0.3
0
0
0
0
Percent of
Person-days
50K Run6
100
16.9
1.7
0.3
<0.1
<0.1
<0.1
<0.1
0
100
2.7
<0.1
<0.1
<0.1
0
0
0
0
SKRun
100
17.1
1.7
0.3
<0.1
<0.1
<0.1
0
0
100
2.7
<0.1
<0.1
<0.1
0
0
0
0
Notes:
3 Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1).
c Unadjusted ambient concentrations from four monitors (Denver) and ten monitors (Los Angeles) in 2006
were used to represent the As Is air quality scenario.
d Exposure results obtained from Table 6-1 .
e Exposure results obtained from Table 6-4.
19 Given that the results generated using the CHD and HD populations were also generally similar when
comparing the percent of persons and person-days (e.g., see Table 6-1), the microenvironmental contributions (and
microenvironmental factors) estimated from the smaller sample size CHD population simulations are considered
representative estimated exposures for the simulated HD population.
5-37
-------
Table 5-27. Comparison of dose summary output generated when simulating 50,000
persons versus that of simulating 5,000 persons -just meeting the current
standard.
COHb
Level
(%)
0.0
1.0
1.5
1.75
2.0
2.5
3.0
4.0
Denver CHD Population3'"'0
Percent of
Persons
50K Rund
100
82.3
23.4
10.8
4.2
0.8
0.2
<0.1
SKRun
100
82.6
21.2
10.0
4.1
0.7
0
0
Percent of
Person-days
50K Rund
100
6.8
0.4
<0.1
<0.1
<0.1
<0.1
<0.1
SKRun
100
6.1
0.4
<0.1
<0.1
<0.1
0
0
Los Angeles CHD Population3'"'0
Percent of
Persons
50K Run6
100
41.2
4.7
1.6
0.6
<0.1
<0.1
0
SKRun
100
44.4
5.1
1.3
0.8
0.3
0.3
0
Percent of
Person-days
50K Run6
100
1.8
<0.1
<0.1
<0.1
<0.1
<0.1
0
SKRun
100
2.1
<0.1
<0.1
<0.1
<0.1
<0.1
0
Notes:
3 Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1).
c Ambient concentrations from 1 995 (Denver) and 1 997 (Los Angeles) were adjusted to just meet a
2nd highest 8-hour average concentration of 9.4 ppm.
d Dose results obtained from Table 6-18.
e Dose results obtained from Table 6-19.
5.11 KEY OBSERVATIONS
The following presents the key observations for this chapter:
• Two exposure model domains (Denver and Los Angeles study areas) were defined by
overlaying ambient monitor locations having 10 km radii with US census tract
population data. Monitors selected comprised the bulk of the urban core in each
location, where ambient monitoring data exist.
• Two simulated at-risk subpopulations were identified by combining the census tract-
specific age and gender population distributions with HD and CHD prevalence rates,
each also stratified by age and gender. In using this approach, staff can represent the
variability that exists in the simulated at-risk HD and CHD subpopulations that reside
in each census tract and within each study area.
• Both simulated at-risk subpopulations include an estimate of persons with
undiagnosed CHD.
• To represent spatial variability in ambient concentrations in Denver, a total of four
monitors were used; in Los Angeles, the total number of monitors was ten. Temporal
variability was represented by use of hourly ambient concentrations in each study area.
5-38
-------
• The exposure and dose model simulations included 8 microenvironments in each
location to represent the expected variability in microenvironmental CO
concentrations.
• All indoor microenvironments were modeled using a mass balance model to represent
temporal variability in indoor CO concentrations with respect to the outdoor CO
concentration variability. In addition, distributions of microenvironmental factors were
used for all microenvironments rather than point estimates. Using distributions of
microenvironmental factors will better represent both spatial and temporal variability in
estimated microenvironmental CO concentrations.
• Additional analyses using output from individual-level simulations were performed to
provide information on the microenvironments most influential to population exposure
at different exposure levels. This included an analysis of the effective ratios of
microenvironment to ambient concentrations and the contribution of ambient CO
exposure to total COHb level estimates. The smaller sample sizes generated for these
analyses were found to be representative of the larger simulations employed for
estimating exposure and dose in the different air quality exposure scenarios.
5-39
-------
5.12 REFERENCES
AHS. (2005). American Housing Survey for the Denver Metropolitan Area: 2004. Available at:
http://www.census.gov/prod/2005pubs/hl70-04-46.pdf
Adams PF and Marano MA. (1995). Current Estimates from the National Health Interview Survey, 1994. National
Center for Health Statistics. Vital Health Stat. 10(193).
American Heart Association. (1990). Heart and Stroke Facts. American Heart Association, Dallas, TX. p. 13. As
cited by Johnson et al. (2000).
American Heart Association. (2003). Heart and Stroke Facts. American Heart Association, Dallas, TX. p. 14.
Available at: http://www.americanheart.org/downloadable/heart/1056719919740HSFacts2003text.pdf
CDC. (2009). Summary Health Statistics for U.S. Adults: National Health Interview Survey, 2007. Series 10,
Number 240. U.S. Department of Health and Human Services, Hyattsville, MD, May 2009.
Glen G, Smith L, Isaacs K, McCurdy T, Langstaff J. (2008). A new method of longitudinal diary assembly for
human exposure modeling. J Expos Sci Environ Epidem. 18:299-311.
Isaacs K, McCurdy T, Erickson A, Forbes S, Glen G, Graham S, McCurdy L, Nysewander, M, Smith L, Tulve N,
Vallero D. (2009). Statistical properties of longitudinal time-activity data for use in EPA exposure models.
Poster presented at the American Time Use Research Conference; College Park MD, June 26, 2009.
Johnson T, Mihlan G, LaPointe J, Fletcher K, Capel J. (2000). Estimation of Carbon Monoxide Exposures and
Associated Carboxyhemoglobin Levels for Residents of Denver and Los Angeles Using pNEM/CO
(Version 2.1). Report prepared by ICF Consulting and TRJ Environmental, Inc., under EPA Contract No.
68-D6-0064. U.S. Environmental Protection Agency, Research Triangle Park, North Carolina. Available
at: http://www.epa.gov/ttn/fera/human_related.html. June 2000.
Persily A, Gorfain J, Brunner G. (2005). Ventilation design and performance in U.S. Office buildings. ASHRAE
Journal. April 2005, 30-35.
US EPA. (1992). Review of the National Ambient Air Quality Standards for Carbon Monoxide: Assessment of
Scientific and Technical Information. Office of Air Quality Planning and Standards Staff Paper. Report no.
EPA/452/R-92-004.
US EPA. (2003). Total Risk Integrated Methodology TRIM.ExpoInhalation User's Document. Volume I: Air
Pollutants Exposure Model (APEX, version 3). Available at:
http://www.epa.gov/ttn/fera/data/apex322/apexusersguidevoli4-24-03.pdf
US EPA. (2007). Ozone Population Exposure Analysis for Selected Urban Areas. Office of Air Quality Planning
and Standards, Research Triangle Park, NC. July 2007. Report no. EPA-452/R-07-010. Available at:
http://www.epa.gov/ttn/naaqs/standards/ozone/data/2007 07 o3 exposure tsd.pdf.
US EPA. (2008a). Total Risk Integrated Methodology (TRIM) Air Pollutants Exposure Model Documentation
(TRIM.Expo/APEX, Version 4.3). Volume 1: Users Guide. Report no. EPA-452/B-08-00la. Office of
Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC.
US EPA. (2008b). Total Risk Integrated Methodology (TRIM) Air Pollutants Exposure Model Documentation
(TRIM.Expo/APEX, Version 4.3). Volume 2: Technical Support Document. Report no. EPA-452/B-08-
00Ib. Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research
Triangle Park, NC.
5-40
-------
US EPA. (2008c). Risk and Exposure Assessment to Support the Review of the NC>2 Primary National Ambient
Air Quality Standard. Report no. EPA-452/R-08-008a. November 2008. Available at:
http://www.epa.gov/ttn/naaqs/standards/nox/data/2008112 l_NO2_REA_final.pdf.
US EPA. (2009a). Risk and Exposure Assessment to Support the Review of the Carbon Monoxide Primary
National Ambient Air Quality Standard. First External Review Draft. Report no. EPA-452/P-09-008.
Available at: http://www.epa.gov/ttn/naaqs/standards/co/data/COREAlstDraftOct2009.pdf
US EPA. (2009b). Risk and Exposure Assessment to Support the Review of the SO2 Primary National Ambient Air
Quality Standard. Report no. EPA-452/R-09-007. August 2009. Available at
http://www.epa.gov/ttn/naaqs/standards/so2/data/200908SO2REAFinalReport.pdf.
US EPA. (2010). Risk and Exposure Assessment to Support the Review of the Carbon Monoxide Primary National
Ambient Air Quality Standard. Second External Review Draft. Report no. EPA-452/P-10-004. Available
at: http://www.epa.gov/ttn/naaqs/standards/co/data/COREA2ndDraftFeb2010.pdf
Wilson AL, Colome SD, Tian Y. (1995). California Residential Indoor Air Quality Study. Volume III: Ancillary
and Exploratory Analysis. Integrated Environmental Services, Irvine, California.
5-41
-------
6 SIMULATED EXPOSURE AND COHB RESULTS
This chapter summarizes the CO exposure and dose results for the Denver and Los
Angeles study areas that were generated using EPA's APEX model described in chapters 4 and
5. Staff considered exposures associated with five air quality scenarios; air quality (1)
unadjusted or as is, (2) adjusted to just meet the current 8-hour standard of 9 ppm, (3) adjusted to
just meet a 2nd highest 8-hour average concentration of 5 ppm, (4) adjusted to just meet a 99th
percentile daily maximum 8-hour average of 5.0 ppm, and (5) adjusted to just meet a 99th
percentile daily maximum 8-hour average of 8.0 ppm (see section 5.7 for details regarding the air
quality adjustment procedure). This chapter is divided into four main sections, with each
described briefly below.
The first section (6.1) summarizes the estimated exposures associated with each of the
five air quality scenarios. As described in section 5.5.1, two at-risk subpopulations were
simulated in this assessment. The first simulated at-risk subpopulation includes individuals with
diagnosed CHD as well as those persons with potentially undiagnosed CHD.1 For simplicity,
they will be combined and referred to as the CHD population in this chapter. The second
simulated at-risk population includes individuals with diagnosed heart disease (HD) as well as
those persons with potentially undiagnosed CHD. For simplicity, this subpopulation will be
combined and referred to as the HD population in this chapter. The primary exposure metrics of
interest in this REA and generated by APEX include the number and percent of persons at or
above staff-selected exposure levels and the corresponding number of person-days.2 Two
exposure averaging times were also selected: 1-hour and 8-hour daily maximum exposures.
Section 6.2 summarizes the estimated COHb levels for persons in the simulated at-risk
population residing in each study area. The primary dose metric of interest in this REA and
generated by APEX includes the number and percent of persons at or above staff selected COFtb
levels and the corresponding number of person-days. Consistent with prior CO exposure
assessments, the daily maximum end-of-hour COFtb level was reported. This section also
1 As described in section 5.5.2, in characterizing the population of interest with regard to demographics
(age and gender), the assessment drew from estimates of the prevalence of coronary heart disease (CHD, which
includes CHD, angina pectoris and heart attack) and all types of heart disease (HD, which includes coronary heart
disease, angina pectoris, heart attack, or any other heart condition or disease) provided by the National Health
Interview Survey, each combined with estimates of undiagnosed ischemia developed by EPA.
2 Because the duration of the exposure assessment is one year, there are opportunities for individuals to
experience more than one day in the year above a selected exposure concentration, hence use of the term person-
days.
6-1
-------
presents an evaluation of the ambient contribution to COHb levels for APEX simulated
individuals for three of the air quality scenarios.
In section 6.3, staff compares the dose estimates in this CO REA with those estimated in
the 2000 exposure assessment (Johnson et al., 2000). Finally, key observations are presented in
the final section (6.4). As mentioned in Chapter 1, due to the extremely tight timeline for this
NAAQS review, the exposure and risk results are provided here without substantial
interpretation. Rather, interpretative discussion of these results is provided in the CO Policy
Assessment.
6.1 ESTIMATED EXPOSURES
This section summarizes the estimated exposures for the simulated individuals in a series
of tables, classified by the five air quality scenarios and two study areas considered. Given the
complexity of the simulations and output data requirements, a limited number of additional
modeling runs were added to this final assessment. First, in expanding the at-risk population to
include persons with all types of heart disease, the standard APEX output data (i.e., number and
percent of persons and person-days at selected exposure levels) were generated for all five air
quality scenarios and summarized. Additional exposure output data sets (i.e.,
microenvironmental contributions to selected exposure levels and the evaluation of
microenvironmental factors distributions) were generated using a smaller simulation size (i.e.,
5,000 persons) and the CFID population for two air quality scenarios (i.e., as is and just meeting
the current standard). The smaller simulation size was used to reduce the size of the output file
under analysis. See section 5.10 for details on the approach used and evaluation of the
representativeness of the simulation size and at-risk population used.
6.1.1 Air Quality "As Is"
As described in section 5.6, ambient monitoring data from each study area for the year
2006 were used to represent the as is air quality. Table 6-1 summarizes the distribution of the 1-
hour and 8-hour daily maximum CO exposures experienced by the CFID and HD populations in
the Denver Study area. About 80% of the simulated CHD population did not experience a 1-
hour daily maximum exposure at or above 9 ppm; 99.9% did not experience a 1-hour daily
maximum exposure concentration at or above 20 ppm. Of the nearly 20 million CUD person-
days, over 99% were associated with a 1-hour daily maximum exposure below 6 ppm. Very few
individuals were estimated to experience an 8-hour daily maximum exposure at or above 9 ppm
(0.1% of the CHD population). Approximately 99% of simulated CHD person-days were
associated with 8-hour daily maximum exposure concentrations of less than 3 ppm.
6-2
-------
While there were a greater number of persons and person-days at each of the exposure
levels for the HD as compared to the CHD population, consistent with the larger size of that
population, there was little difference when considering the percentages of each population at or
above selected exposure levels and for either averaging time. For example, about 20% of both
the CHD and HD populations experienced a 1-hour daily maximum exposure at or above 9 ppm.
About 89% of person-days for either population were associated with 1-hour daily maximum
exposures of less than 3 ppm.
Table 6-1. Estimated daily maximum 1-hour or 8-hour exposure for simulated at-risk
populations in the Denver study area - as is air quality.
Daily
Maximum
Exposure
(ppm)
1-
hour
8-
hour
>0
>3
>6
>9
>12
>15
>20
>30
>40
>0
>3
>6
>9
>12
>15
>20
>30
>40
Coronary Heart Disease3'0
Persons
Number
53,656
53,397
32,517
10,662
3,048
876
62
0
0
53,656
31,036
1,715
62
12
0
0
0
0
Percent
100
99.5
60.6
19.9
5.7
1.6
0.1
0
0
100
57.8
3.2
0.1
<0.1
0
0
0
0
Person-days
Number
19,580,000
2,188,000
170,400
24,560
4,677
1,061
62
0
0
19,580,000
189,500
2,851
86
12
0
0
0
0
Percent
100
11.2
0.9
0.1
<0.1
<0.1
<0.1
0
0
100
1.0
<0.1
<0.1
<0.1
0
0
0
0
All Heart Disease"'0
Persons
Number
85,926
85,494
52,274
17,412
5,010
1,431
99
0
0
85,926
50,361
2,604
99
12
0
0
0
0
Percent
100
99.5
60.8
20.3
5.8
1.7
0.1
0
0
100
58.6
3.0
0.1
<0.1
0
0
0
0
Person-days
Number
31,360,000
3,582,000
281,700
39,550
7,441
1,666
99
0
0
31,360,000
309,400
4,171
136
12
0
0
0
0
Percent
100
11.4
0.9
0.1
<0.1
<0.1
<0.1
0
0
100
1.0
<0.1
<0.1
<0.1
0
0
0
0
Notes:
Unadjusted ambient concentrations from four monitors in 2006 were used to represent the As Is air
quality scenario.
3 Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Inclusive of those persons with diagnosed coronary heart disease, angina pectoris, heart attack, and any
other heart condition or disease (CDC, 2009).
c Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1).
These exposure results are consistent with the ambient concentration distributions used to
represent this scenario, where upper percentile concentrations range from about 2 to 6.4 ppm (see
Table 5-14). Note also that the highest estimated 1-hour daily maximum exposures are likely a
function of microenvironmental concentrations (e.g., exposures occurring while inside vehicles
or immediately near roads) that, in general, may be a factor of two to five times higher than
6-3
-------
ambient CO concentrations at monitors that are not immediately near roads (2000 AQCD; ISA
section 3.6.1).
As mentioned above, a smaller subset of the at-risk CHD population was simulated to
generate additional exposure results. First, we were interested in determining the important
microenvironments that contribute to CHD population exposures at each of the selected levels.3
Figure 6-1 illustrates such an analysis, beginning with the total minutes per year spent by the
simulated CHD population in Denver at or above each exposure level. Note that total time spent
at or above a particular exposure concentration decreases with increasing exposure level. This
pattern is consistent with the exposure results above (Table 6-1); that is, for most of the time, the
population is exposed to concentrations less than 6 ppm (about 99.9% of the total time), with
very little time spent exposed to concentrations at or above 20 ppm. For this scenario, in fact,
there were 50 total minutes out of the 230 million minutes in the year simulated where the
population was exposed at or above a level of 20 ppm. Note that when considering the zero
exposure concentration level, the distribution of microenvironmental contributions effectively
approximates the time spent by the population in each of the microenvironments across the
simulation period. Not surprisingly, the simulated population spends over 85% of their time
within indoor microenvironments and is consistent with the reported activity pattern survey data
(Graham and McCurdy, 2004). At the lowest exposure levels (e.g., < 2 ppm), much of the
estimated population exposure occurs within indoor microenvironments. For exposure
concentrations at or above 2 ppm, most of the population exposure occurs while inside vehicles
or during time spent within outdoor high-concentration microenvironments. These two
aggregate microenvironments predominantly contribute to exposure concentrations at or above 4
ppm (> 90% of all personal time at these levels).
These smaller CHD population simulations also served to approximate the effective
microenvironmental factors, the distributions of which are more useful to compare with literature
reported microenvironment-to-ambient concentration ratios rather than simply using the
distribution of proximity factors given in Table 5-22. Table 6-2 summarizes these ratios that
were calculated by dividing the estimated event-level microenvironmental concentration by its
corresponding ambient concentration. Values for the estimated microenvironmental ratios
correspond reasonably well with reported personal exposure to ambient concentration ratios
(ISA, Figure 3-46). The distribution of estimated microenvironmental concentrations (Table 6-
3) also generally reflect the range of concentrations reported in personal exposure studies,
3 The default APEX summary output table only generates microenvironmental contributions for the general
population.
6-4
-------
particularly those where measurements were made inside-vehicles and near-roadways (ISA
section 3.6.6.2).
c
+j *r
C O
0) 'f
Q. TO
W i=
l§
i §
O in
i- o
• -
0) >
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
>0
Denver: AQ As Is
op rj- CD
LJJ LJJ LJJ
ro in T-
cxi in CT>
¥
LJJ
r^
cxi
¥
LU
in
LU
in
LJJ
ro
cxi
T
LJJ
in
ro
ro
LU
ro
LJJ
CD
CO Exposure Concentration (ppm)
LJJ LJJ
p p
in o
LJJ LJJ
p p
o o
• inside-vehicles
• outdoors (low)
n outdoors (high)
n indoors (low)
a indoors (high)
>20 >30 >40 >60
Figure 6-1. Estimated microenvironmental contributions to time spent at or above
selected exposure concentrations using the Denver CHD population - as is air
quality. The total minutes spent at or above each exposure concentration are
presented above each bar.
6-5
-------
Table 6-2. Estimated distribution of microenvironment-to-ambient concentration ratios
using the Denver CHD population - as is air quality.
Microenvironment3
1: In-Residence
2: In-Service Station
3: In-Other A
4: In-Other B
5: Out-Near Road
6: Out-Parking Lot/
Refueling
7: Out-Other
8: In-Vehicle
Events"'0
(n)
401,098
1,425
39,348
15,956
8,822
3,079
24,515
44,998
Mean
Ratioc'd
(unitless)
0.9
2.9
1.1
0.9
1.6
2.9
1.5
2.9
std
0.4
1.3
0.5
0.4
0.8
1.4
0.8
1.4
Distribution Percentiles°'di
min
0
0.5
0.1
0.1
0.1
0.4
0
0
p1
0.3
0.9
0.4
0.3
0.5
0.8
0.4
0.8
P5
0.4
1.3
0.5
0.4
0.6
1.2
0.6
1.2
p50
0.8
2.7
1.0
0.9
1.5
2.6
1.3
2.6
p95
1.8
5.4
2.1
1.7
3.1
5.5
2.9
5.7
unitless)
p99
2.4
6.6
2.6
2.2
4.1
7.4
3.9
7.5
max
5.1
10.0
4.2
4.2
6.1
10.2
6.1
14.1
Notes:
aSee section 5.9 and Table 5-22 for details.
b This is the number of times the population experienced an exposure event in that microenvironment.
c Data set used to calculate the ratios represent the CHD population extracted from a 5,000 person
simulation and was screened to eliminate ambient concentrations less than 1 ppm. See section 5.10
for details.
dThe mean, standard deviation, and percentiles (p) were calculated using all events regardless of
event duration. Note that based on the activity pattern diaries used, the length of an event can range
from 1 minute to 1 hour.
Table 6-3. Estimated distribution of microenvironmental concentrations using the Denver
CHD population - as is air quality.
Microenvironment3
1: In-Residence
2: In-Service Station
3: In-Other A
4: In-Other B
5: Out-Near Road
6: Out-Parking Lot/
Refueling
7: Out-Other
8: In-Vehicle
Events"'0
(n)
4,046,011
17,715
486,641
183,535
100,303
39,801
354,654
546,573
Mean
C0c'd
(ppm)
0.7
2.2
0.8
0.7
1.1
2.0
1.0
2.0
Stdc'd
0.4
1.2
0.5
0.4
0.8
1.3
0.7
1.4
Distribution Percentiles°'d (ppm)
min
0
0
0
0
0
0
0
0
p1
0.1
0.3
0.1
0.1
0
0
0
0
p5
0.2
0.8
0.3
0.2
0.3
0.6
0.3
0.6
p50
0.6
1.9
0.7
0.6
1.0
1.7
0.8
1.7
p95
1.5
4.5
1.7
1.5
2.6
4.4
2.2
4.6
p99
2.2
6.4
2.5
2.1
3.8
6.7
3.4
6.9
Max
7.8
14.3
7.3
6.2
15.4
16.6
12.3
23.0
Notes:
aSee section 5.9 and Table 5-22 for details.
b This is the number of times the population experienced an exposure event in that microenvironment.
c These include all exposure events for the CHD population extracted from a 5,000 person simulation.
See section 5.1 0 for details.
dThe mean, standard deviation, and percentiles (p) were calculated using all events regardless of
event duration. Note that based on the activity pattern diaries used, the length of an event can range
from 1 minute to 1 hour.
6-6
-------
In Los Angeles, there was a greater number of individuals experiencing exposures at or
above each of the selected exposure levels (Table 6-4) when compared with exposures in Denver
(Table 6-1) and considering either simulated at-risk population. This is expected given that the
exposure modeling domain in Los Angeles encompasses a larger area than Denver and therefore
comprises a larger total simulated population. The estimated percentage of persons exposed in
Los Angeles is also greater when compared with the corresponding exposure levels evaluated for
the Denver study area. For example, approximately 32% of the CHD population was estimated
to experience a 1-hour daily maximum exposure of at least 9 ppm in Los Angeles (Table 6-6)
while in Denver this same level was experienced by approximately 20% of the CHD population
(Table 6-1). This result is likely driven by the differences noted in the as is air quality data,
where in Los Angeles, the 2006 ambient concentrations were generally higher than those
observed for Denver (see section 5.7.1; Tables 5-14 and 5-16).
In addition, the highest 1-hour daily maximum exposure was estimated to be at or above
30 ppm (but less than 40 ppm) in the Los Angeles study area, though limited to a small fraction
of either simulated at-risk population (<0.1%). The corresponding highest 1-hour daily
maximum exposure in the Denver study area was at or above 20 ppm (but less than 30 ppm) and
was experienced by approximately 0.1% of either simulated at-risk population. Therefore, the
overall range of the exposure distribution was wider in Los Angeles when compared with that of
Denver when considering the as is air quality scenario.
Similar to that estimated for either at-risk population in Denver, over 98% of the person-
days in Los Angeles were associated with 1-hour daily maximum exposures below 6 ppm and
very few persons (<1%) experienced 8-hour daily maximum exposures at or above 9 ppm.
These exposure results are also consistent with the distributions of ambient air quality used to
represent this scenario, where the ambient monitor upper percentile concentrations extend from
about 2 to 8.4 ppm (Table 5-16).
6-7
-------
Table 6-4. Estimated daily maximum 1-hour or 8-hour exposure for simulated at-risk
populations in the Los Angeles study area - as is air quality.
Daily
Maximum
Exposure
(ppm)
1-
hour
8-
hour
>0
>3
>6
>9
>12
>15
>20
>30
>40
>0
>3
>6
>9
>12
>15
>20
>30
>40
Coronary Heart Disease3'0
Persons
Number
383,040
382,739
287,606
122,428
42,850
13,949
2,208
100
0
383,040
294,430
36,528
3,011
301
0
0
0
0
Percent
100
99.9
75.1
32.0
11.2
3.6
0.6
<0.1
0
100
76.9
9.5
0.8
<0.1
0
0
0
0
Person-days
Number
139,800,000
23,620,000
2,423,000
408,300
83,990
20,170
2,509
100
0
139,800,000
3,793,000
72,150
3,412
301
0
0
0
0
Percent
100
16.9
1.7
0.3
<0.1
<0.1
<0.1
<0.1
0
100
2.7
<0.1
<0.1
<0.1
0
0
0
0
All Heart Disease"'0
Persons
Number
630,807
630,305
471,951
194,681
65,730
21,074
3,211
100
0
630,807
481,986
57,100
4,616
301
0
0
0
0
Percent
100
99.9
74.8
30.9
10.4
3.3
0.5
<0.1
0
100
76.4
9.1
0.7
<0.1
0
0
0
0
Person-days
Number
230,200,000
38,680,000
3,880,000
651,500
133,500
31,510
3,613
100
0
230,200,000
6,178,000
114,300
5,319
301
0
0
0
0
Percent
100
16.8
1.7
0.3
<0.1
<0.1
<0.1
<0.1
0
100
2.7
<0.1
<0.1
<0.1
0
0
0
0
Notes:
Unadjusted ambient concentrations from ten monitors in 2006 were used to represent the As Is air quality
scenario.
a Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Inclusive of those persons with diagnosed coronary heart disease, angina pectoris, heart attack, and any
other heart condition or disease (CDC, 2009).
c Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1 .1).
The microenvironmental contributions to time spent at or above selected exposure levels
in Los Angeles (Figure 6-2) are also similar to that estimated for Denver, though indoor
microenvironments contribute a somewhat greater percentage in Los Angeles at each of the
exposure levels. This is likely the result of generally higher as is ambient concentrations across
the entire distribution in Los Angeles when compared with the Denver ambient concentrations
distribution (Table 5-14). As expected, the distributions of event-level microenvironment-to-
ambient concentration ratios (Table 6-5) are also similar to those calculated for Denver (Table 6-
3), though estimated microenvironmental concentrations are slightly higher in Los Angeles
(Table 6-6). This is also a function of the generally higher ambient concentrations measured at
the Los Angeles monitors when compared with ambient monitor concentrations in Denver.
6-8
-------
" §
0
>2 >3 >4 >5 >6 >9
>20 >30 >40 >60
CO Exposure Concentration (ppm)
Figure 6-2. Estimated microenvironmental contributions time spent at or above selected
exposure concentrations using the Los Angeles CHD population - as is air
quality. The total minutes spent at or above each exposure concentration are
presented above each bar.
6-9
-------
Table 6-5. Estimated distribution of microenvironment-to-ambient concentration ratios
using the Los Angeles CHD population - as is air quality.
Microenvironment3
1: In-Residence
2: In-Service Station
3: In-Other A
4: In-Other B
5: Out-Near Road
6: Out-Parking Lot/
Refueling
7: Out-Other
8: In-Vehicle
Events"'0
(n)
631,366
2,190
45,134
18,735
12,954
4,307
40,000
59,191
Mean
Ratioc'd
(unitless)
0.9
2.8
1.2
1.0
1.5
2.9
1.4
2.9
Std
0.4
1.2
0.5
0.4
0.8
1.5
0.7
1.5
Distribution Percentiles°'d (unitless)
min
0
0.4
0
0.1
0
0.3
0.1
0
p1
0.3
1.0
0.4
0.3
0.3
0.7
0.4
0.7
P5
0.4
1.3
0.5
0.4
0.6
1.1
0.5
1.1
p50
0.9
2.6
1.0
0.9
1.4
2.6
1.3
2.5
p95
1.8
5.1
2.2
1.8
3.0
5.9
2.9
5.7
p99
2.3
6.6
2.8
2.3
3.9
7.6
3.8
7.5
max
5.7
9.4
5.4
3.8
9.0
10.8
7.0
21.1
Notes:
aSee section 5.9 and Table 5-22 for details.
b This is the number of times the population experienced an exposure event in that
microenvironment.
c Data set used to calculate the ratios represents the CHD population extracted from a 5,000
person simulation and was screened to eliminate ambient concentrations less than 1 ppm. See
section 5.10 for details.
dThe mean, standard deviation, and percentiles (p) were calculated using all events regardless of
event duration. Note that based on the activity pattern diaries used, the length of an event can
range from 1 minute to 1 hour.
Table 6-6. Estimated distribution of microenvironmental concentrations using the Los
Angeles CHD population - as is air quality.
Microenvironment3
1: In-Residence
2: In-Service Station
3: In-Other A
4: In-Other B
5: Out-Near Road
6: Out-Parking Lot/
Refueling
7: Out-Other
8: In-Vehicle
Events"'0
(n)
3,620,332
19,825
433,366
164,167
102,844
41,611
359,727
523,967
Mean
C0c'd
(ppm)
0.8
2.2
0.9
0.8
1.2
2.1
1.0
2.1
Std
0.6
1.5
0.6
0.5
0.9
1.6
0.8
1.6
Distribution Percentiles°'d (ppm)
min
0
0
0
0
0
0
0
0
p1
0
0
0
0
0
0
0
0
p5
0.2
0.6
0.2
0.2
0.3
0.5
0.2
0.5
p50
0.7
1.8
0.7
0.6
0.9
1.6
0.8
1.7
p95
1.9
5.1
2.0
1.7
2.9
5.2
2.6
5.3
p99
2.9
7.5
3.0
2.6
4.5
8.2
4.1
8.1
max
9.7
14.6
10.7
7.8
14.7
26.8
14.1
34.3
Notes:
aSee section 5.9 and Table 5-22 for details.
b This is the number of times the population experienced an exposure event in that
microenvironment.
c These include all exposure events for the CHD population extracted from a 5,000 person
simulation. See section 5.10 for details.
dThe mean, standard deviation, and percentiles (p) were calculated using all events regardless of
event duration. Note that based on the activity pattern diaries used, the length of an event can
range from 1 minute to 1 hour.
6-10
-------
6.1.2 Air quality adjusted to just meet the current 8-hour standard
As described in section 5.6, historical ambient monitoring data from each study area were
adjusted to represent air quality that just meets the current 8-hour standard. For both Denver
(year 1995) and Los Angeles (year 1997), air quality data were adjusted downwards to meet a 2nd
highest 8-hour average concentration of 9.4 ppm. Note that even with a downward proportional
adjustment, these adjusted ambient concentrations remain much higher than as is ambient air
quality. Table 6-7 summarizes the exposure results for the simulated at-risk populations in the
Denver study area when using these adjusted ambient CO concentrations as an input to APEX
and using the same modeling assumptions and parameter distributions described in chapters 4
and 5.
Over half of the Denver at-risk population was estimated to experience a 1-hour daily
maximum exposure at or above 12 ppm. This is nearly a factor of 10 greater than that estimated
when using the as is air quality (Table 6-1). The highest 1-hour daily maximum exposure was
estimated to be at or above 40 ppm (but below 60 ppm) when considering air quality adjusted to
just meet the current standard, though only experienced by less than 0.2% of the simulated at-
risk populations. Thus, there is a wider range in the exposure levels experienced by the
simulated at-risk populations when considering this exposure scenario.
The number and percent of persons experiencing 8-hour daily maximum exposures is
also greater for this scenario when compared with corresponding levels using the as is air
quality. Nearly 10% of the simulated at-risk population was estimated to experience an 8-hour
daily maximum exposure at or above 9 ppm (Table 6-7) when considering air quality just
meeting the current 8-hour standard. Most of the CHD or HD population (>99%) would not
experience an 8-hour daily maximum concentration at that same level when considering the as is
air quality scenario (Table 6-1).
Indoor microenvironments simulated in Denver were estimated to contribute to a greater
percentage of time spent at each selected exposure level using the adjusted air quality when
compared with the results using as is air quality, though the difference is most notable at
exposures less than 9 ppm (Figure 6-3). For example, about 50% of exposures that occurred at
or above 3 ppm were experienced within indoor microenvironments when air quality just meets
the current standard (Figure 6-3), while indoor microenvironments account for less than 20% of
exposures when considering as is air quality (Figure 6-1). Indoor microenvironments would be
expected to play a larger role in low level exposures when using air quality adjusted to just meet
the current standard given the higher ambient concentrations across the entire air quality
distribution when compared with as is air quality.
6-11
-------
As was observed for both locations using as is air quality, the distributions of
microenvironment to ambient concentration ratios are largely the same when comparing
microenvironments (Table 6-8), likely a function of the same algorithm and parameter inputs
used for each scenario and location. Also as expected, the estimated microenvironmental
concentrations (Table 6-9) are higher across the entire distribution when compared with those
estimated when considering the as is exposure scenario (Table 6-3).
Table 6-7. Estimated daily maximum 1-hour and 8-hour exposures for simulated at-risk
populations in the Denver study area - air quality just meeting the current 8-
hour standard.
Daily
Maximum
Exposure
(ppm)
1-
hour
8-
hour
>0
>3
>6
>9
>12
>15
>20
>30
>40
>60
>0
>3
>6
>9
>12
>15
>20
>30
>40
>60
Coronary Heart Disease3'0
Persons
Number
53,656
53,656
53,039
44,598
28,469
16,610
6,022
691
86
0
53,656
52,706
23,879
5,060
1,037
309
37
0
0
0
Percent
100
100
98.9
83.1
53.1
31.0
11.2
1.3
0.2
0
100
98.2
44.5
9.4
1.9
0.6
<0.1
0
0
0
Person-days
Number
19,580,000
8,638,000
1,625,000
404,800
127,300
46,710
10,290
802
86
0
19,580,000
2,690,000
97,760
9,724
1,382
346
37
0
0
0
Percent
100
44.1
8.3
2.1
0.7
0.2
<0.1
<0.1
<0.1
0
100
13.7
0.5
<0.1
<0.1
<0.1
<0.1
0
0
0
All Heart Disease"'0
Persons
Number
85,926
85,926
84,964
71,426
45,919
26,840
9,885
1,283
136
0
85,926
84,445
38,132
7,861
1,666
457
62
0
0
0
Percent
100
100
98.9
83.1
53.4
31.2
11.5
1.5
0.2
0
100
98.3
44.4
9.1
1.9
0.5
<0.1
0
0
0
Person-days
Number
31,360,000
13,960,000
2,660,000
666,400
206,700
75,350
16,500
1,493
136
0
31,360,000
4,318,000
157,000
14,850
2,197
531
62
0
0
0
Percent
100
44.5
8.5
2.1
0.7
0.2
<0.1
<0.1
<0.1
0
100
13.8
0.5
<0.1
<0.1
<0.1
<0.1
0
0
0
Notes:
Ambient concentrations from 1995 were adjusted to just meet a 2nd highest 8-hour average
concentration of 9.4 ppm using a relationship derived from the design monitor (ID 080310002).
a Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Inclusive of those persons with diagnosed coronary heart disease, angina pectoris, heart attack, and
any other heart condition or disease (CDC, 2009).
c Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1 .1).
6-12
-------
§
3 UJ «
.E a> c
c
+j *r
C O
a) ^
o. TO
C/) *=
18
i §
Denver: AQ Just Meeting Current Standard
oo
LJJ
ro
cxi
LJJ LJJ LJJ
ro o> "3-
CD
LJJ
^
CD
CD
LJJ
¥
LU
O>
in
LU
in
LJJ
CD
T
LJJ
CD
100% :
90% :
80% : - ^
70% \
60% \
50% :
40% : -
30% \ -
20% :
10% :
0% '-
>0 >1
T ?
LJJ LJJ
LJJ
CN
CD
LJJ
p
cxi
• inside-vehicles
• outdoors (low)
n outdoors (high)
n indoors (low)
0 indoors (high)
>3 >4 >5 >6 >9
CO Exposure Concentration (ppm)
>20 >30 >40 >60
Figure 6-3. Estimated microenvironmental contributions to time spent at or above
selected exposure concentrations using the Denver CHD population - air
quality just meeting the current 8-hour standard. The total minutes spent at
or above each exposure concentration are presented above each bar.
6-13
-------
Table 6-8. Estimated distribution of microenvironment-to-ambient concentration ratios
using the Denver CHD population - air quality just meeting the current 8-hour
standard.
Microenvironment3
1: In-Residence
2: In-Service Station
3: In-Other A
4: In-Other B
5: Out-Near Road
6: Out-Parking Lot/
Refueling
7: Out-Other
8: In-Vehicle
Events"'0
(n)
1,633,520
7,342
202,514
76,797
43,108
16,563
130,733
228,813
Mean
Ratioc'd
(unitless)
0.9
2.8
1.1
0.9
1.5
2.6
1.4
2.6
Std
0.5
1.3
0.5
0.4
0.8
1.3
0.7
1.4
Distribution Percentiles°'di
min
0.1
0.4
0.1
0.1
0.1
0.2
0.1
0
P1
0.2
0.8
0.3
0.3
0.4
0.7
0.4
0.7
P5
0.3
1.2
0.5
0.4
0.6
1.0
0.5
1.0
p50
0.8
2.6
1.0
0.8
1.3
2.4
1.2
2.3
p95
1.8
5.3
2.1
1.7
2.9
5.2
2.7
5.3
unitless)
p99
2.3
7.0
2.7
2.3
4.0
6.9
3.7
7.1
max
5.5
10.5
6.2
6.0
9.6
13.6
8.5
17.7
Notes:
aSee section 5.9 and Table 5-22 for details.
b This is the number of times the population experienced an exposure event in that microenvironment.
c Data set used to calculate the ratios represents the CHD population extracted from a 5,000 person
simulation and was screened to eliminate ambient concentrations less than 1 ppm. See section 5.10
for details.
dThe mean, standard deviation, and percentiles (p) were calculated using all events regardless of
event duration. Note that based on the activity pattern diaries used, the length of an event can range
from 1 minute to 1 hour.
Table 6-9. Estimated distribution of microenvironmental concentrations using the Denver
CHD population - air quality just meeting the current 8-hour standard.
Microenvironment3
1: In-Residence
2: In-Service Station
3: In-Other A
4: In-Other B
5: Out-Near Road
6: Out-Parking Lot/
Refueling
7: Out-Other
8: In-Vehicle
Events"'0
(n)
4,047,786
17,781
489,984
186,897
100,520
39,895
351,163
548,978
Mean
C0c'd
(ppm)
1.2
3.7
1.4
1.2
2.0
3.5
1.7
3.5
Std
0.8
2.2
0.9
0.8
1.4
2.5
1.3
2.5
Distribution Percentiles°'d(ppm)
min
0
0.4
0.1
0.1
0.1
0
0
0
p1
0.3
1.0
0.4
0.3
0.4
0.7
0.3
0.8
p5
0.4
1.3
0.5
0.4
0.6
1.0
0.5
1.1
p50
1.0
3.2
1.2
1.0
1.6
2.8
1.3
2.8
p95
2.7
8.0
3.2
2.6
4.7
8.0
4.1
8.2
p99
4.0
11.4
4.7
3.9
7.1
12.4
6.4
12.8
max
16.6
25.0
16.6
11.4
22.5
63.4
25.5
56.9
Notes:
aSee section 5.9 and Table 5-22 for details.
b This is the number of times the population experienced an exposure event in that microenvironment.
c These include all exposure events for the CHD population extracted from a 5,000 person simulation.
See section 5.1 0 for details.
dThe mean, standard deviation, and percentiles (p) were calculated using all events regardless of
event duration. Note that based on the activity pattern diaries used, the length of an event can range
from 1 minute to 1 hour.
6-14
-------
Similarly in Los Angeles, the number and percent of persons exposed above selected
exposure concentrations is greater when considering the air quality adjusted to just meet the
current standard than when using as is air quality. For example, nearly 50% of the CHD
population was estimated to experience a 1-hour daily maximum exposure of 9 ppm when
considering air quality just meeting the current standard (Table 6-8), while only 32% were
estimated to experience a similar concentration using as is air quality (Table 6-2). The range of
the 1-hour daily maximum exposure distribution extends upward to 30 ppm, but less than 40
ppm for this scenario in Los Angeles. This estimate of an upper level is below the maximum in-
vehicle concentration of 46 ppm measured by Shikiya (1989) during 112 southern California
commutes in wintertime in 1987-88, of average duration shorter than an hour.4 For a time period
closer to the present scenario, Rodes et al. (1998) reported maximum in-vehicle and on-road CO
concentrations of only 7.6 and 9.0 ppm, respectively during Los Angeles commutes in 1997.
Note however the scripted commutes in this latter study are not necessarily directly comparable
to this modeled data, as the measurements were time-averaged for two hours, the sample size
was limited to about 30 total samples, and data were collected over nine days in the fall.
When comparing the overall population exposure distribution for Los Angeles to Denver
for this exposure scenario, there are greater percentages of persons and person-days estimated for
the Denver simulated at-risk populations at each corresponding exposure level. For example,
only 2.7% of the CFtD population was estimated to experience an 8-hour daily maximum
exposure at or above 9 ppm in Los Angeles (Table 6-10), while in Denver, the estimated percent
of the CFID population exposed at this level was over a factor of three greater (9.4%) (Table 6-
7). This result is likely driven by differences observed at the upper tails of the air quality
distribution noted in section 5.7.3, even though both study areas have ambient concentrations
adjusted to just meet the same 8-hour average CO concentration of 9.4 ppm.
4 On average, the commute time associated with the collection of these samples was 33 minutes. The
reported mean 4-hour integrated ambient monitor concentrations was 3.6 ppm (std = 2.1; max = 8.6 ppm).
6-15
-------
Table 6-10. Estimated daily maximum 1-hour and 8-hour exposures for simulated at-risk
populations in the Los Angeles study area - air quality just meeting the current
8-hour standard.
Daily
Maximum
Exposure
(ppm)
1-
hour
8-
hour
>0
>3
>6
>9
>12
>15
>20
>30
>40
>60
>0
>3
>6
>9
>12
>15
>20
>30
>40
>60
Coronary Heart Disease3'0
Persons
Number
383,040
383,040
335,975
189,563
83,693
36,126
8,731
803
0
0
383,040
342,598
75,966
10,336
1,505
301
100
0
0
0
Percent
100
100
87.7
49.5
21.9
9.4
2.3
0.2
0
0
100
89.4
19.8
2.7
0.4
0.1
<0.1
0
0
0
Person-days
Number
139,800,000
36,430,000
4,826,000
982,300
257,200
80,180
14,450
803
0
0
139,800,000
8,655,000
262,600
18,670
2,308
401
100
0
0
0
Percent
100
26.1
3.5
0.7
0.2
<0.1
<0.1
<0.1
0
0
100
6.2
0.2
<0.1
<0.1
<0.1
<0.1
0
0
0
All Heart Disease"'0
Persons
Number
630,807
630,807
553,536
305,268
130,356
55,293
13,547
1,004
0
0
630,807
562,166
122,328
16,157
2,007
502
100
0
0
0
Percent
100
100
87.8
48.4
20.7
8.8
2.1
0.2
0
0
100
89.1
19.4
2.6
0.3
<0.1
<0.1
0
0
0
Person-days
Number
230,200,000
59,960,000
7,828,000
1,550,000
396,600
119,900
20,970
1,004
0
0
230,200,000
14,190,000
405,100
28,600
3,011
602
100
0
0
0
Percent
100
26.0
3.4
0.7
0.2
<0.1
<0.1
<0.1
0
0
100
6.2
0.2
<0.1
<0.1
<0.1
<0.1
0
0
0
Notes:
Ambient concentrations from 1997 were adjusted to just meet a 2nd highest 8-hour average
concentration of 9.4 ppm using a relationship derived from the design monitor (ID 060371301).
3 Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Inclusive of those persons with diagnosed coronary heart disease, angina pectoris, heart attack, and
any other heart condition or disease (CDC, 2009).
c Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1 .1).
6-16
-------
r C
0 2
Los Angeles: AQ Just Meeting Current Standard
oo
CD CD
LJJ LJJ LJJ LJJ LJJ
T- in ^ CN ^
CN OO CN OO CO
CD
LU
in
LU
CT>
oo
LU
CT>
LU
CN
in
LU
LU
ro
ro
LU
OO
LU LU
O O
O O
o ;=
"I
+j *r
C O
a) ?
a. TO
C/) •!=
C O
i °
_ 0)
c 0
a >
y o
5 •
• inside-vehicles
• outdoors (low)
n outdoors (high)
D indoors (low)
indoors (high)
>0
>2
>4 >5 >6
>9
>20 >30 >40 >60
CO Exposure Concentration (ppm)
Figure 6-4. Estimated microenvironmental contributions to time spent at or above
selected exposure concentrations using the Los Angeles CHD population - air
quality just meeting the current 8-hour standard. The total minutes spent at
or above each exposure concentration are presented above each bar.
The percent contribution of the aggregated microenvironments to the selected exposure
levels in Los Angeles (Figure 6-4) is nearly identical to that estimated for Denver, given air
quality adjusted to just meeting the current standard. The Los Angeles data differ in that upper-
level exposure concentrations only extend upwards to about 30 ppm (occurring for about 180
event-level minutes), while in Denver there were a greater number of exposure events with
concentrations at or above 40 ppm (occurring for about 1,100 event-level minutes). Not
surprisingly, estimated microenvironment-to-ambient concentration ratios for this scenario
(Table 6-11) were also similar to those for the Denver air quality scenarios (Tables 6-2 and 6-8)
and those derived for the Los Angeles as is air quality (Table 6-5). Note also that estimated
upper level concentrations here for the in-vehicle microenvironment are within the maximum
measured peak level (one minute average) concentration reported by Rodes et al. (1998) of 67
ppm during rush hour commutes in Los Angeles (ISA, section 3.6.6.2).
6-17
-------
Table 6-11. Estimated distribution of microenvironment-to-ambient concentration ratios
using the Los Angeles CHD population - air quality just meeting the current 8-
hour standard.
Microenvironment3
1: In-Residence
2: In-Service Station
3: In-Other A
4: In-Other B
5: Out-Near Road
6: Out-Parking Lot/
Refueling
7: Out-Other
8: In-Vehicle
Events"'0
(n)
1,087,377
4,516
92,255
37,297
23,676
8,614
78,546
112,592
Mean
Ratioc'd
(unitless)
1.0
2.8
1.1
1.0
1.5
2.9
1.4
2.8
Std
0.5
1.4
0.5
0.4
0.8
1.5
0.7
1.5
Distribution Percentiles°'d (unitless)
min
0
0.4
0
0
0
0
0
0
p1
0.3
0.8
0.4
0.3
0.3
0.7
0.3
0.7
P5
0.4
1.2
0.5
0.4
0.5
1.1
0.5
1.0
p50
0.9
2.6
1.0
0.9
1.3
2.5
1.2
2.5
p95
1.8
5.4
2.2
1.8
2.9
5.8
2.9
5.7
p99
2.4
7.0
2.9
2.4
4.0
7.7
3.8
7.7
max
6.6
12.6
6.4
4.2
7.4
16.9
10.8
18.9
Notes:
aSee section 5.9 and Table 5-22 for details.
b This is the number of times the population experienced an exposure event in that microenvironment.
c Data set used to calculate the ratios represents the CHD population extracted from a 5,000 person
simulation and was screened to eliminate ambient concentrations less than 1 ppm. See section 5.10
for details.
dThe mean, standard deviation, and percentiles (p) were calculated using all events regardless of
event duration. Note that based on the activity pattern diaries used, the length of an event can range
from 1 minute to 1 hour.
Table 6-12. Estimated distribution of microenvironmental concentrations using the Los
Angeles CHD population - air quality just meeting the current 8-hour
standard.
Microenvironment3
1: In-Residence
2: In-Service Station
3: In-Other A
4: In-Other B
5: Out-Near Road
6: Out-Parking Lot/
Refueling
7: Out-Other
8: In-Vehicle
Events"'0
(n)
3,621,050
19,803
432,894
162,031
102,771
41,803
358,006
522,406
Mean
C0c'd
(ppm)
1.0
2.7
1.0
0.9
1.4
2.5
1.3
2.6
Std
0.7
1.8
0.7
0.6
1.1
2.0
1.0
2.0
Distribution Percentiles0 d (ppm)
min
0
0
0
0
0
0
0
0
p1
0
0
0
0
0
0
0
0
p5
0.2
0.6
0.2
0.2
0.2
0.4
0.2
0.4
p50
0.8
2.2
0.9
0.8
1.2
2.0
1.0
2.1
p95
2.4
6.1
2.4
2.1
3.5
6.3
3.1
6.4
p99
3.5
9.0
3.6
3.1
5.4
10.1
4.9
9.9
max
15.7
25.2
12.0
9.6
15.2
25.4
22.5
39.8
Notes:
aSee section 5.9 and Table 5-22 for details.
b This is the number of times the population experienced an exposure event in that microenvironment.
c These include all exposure events for the CHD population extracted from a 5,000 person simulation.
See section 5.1 0 for details.
dThe mean, standard deviation, and percentiles (p) were calculated using all events regardless of
event duration. Note that based on the activity pattern diaries used, the length of an event can range
from 1 minute to 1 hour.
6-18
-------
6.1.3 Air quality adjusted to just meet alternative air quality scenarios
Three potential alternative air quality scenarios were investigated to observe how the
averaging times, forms, and levels for the simulated alternative standards would affect the
estimated exposure concentrations (section 5.6). The data for the 1-hour and 8-hour daily
maximum exposure concentrations are presented here, with a focus on the number and percent of
persons exposed at selected concentrations. As observed in the above two sections summarizing
exposures associated with as is air quality and air quality adjusted to just meet the current 8-hour
standard, the distributions of the microenvironment-to-ambient concentration ratios are expected
to be the same and are thus not provided here for the alternative air quality scenarios. The air
quality associated with these alternative standards is generally similar to as is conditions.
Therefore, the microenvironmental contributions to exposure and microenvironmental
concentrations associated with these alternative standard scenarios are expected to be similar to
those estimated for the as is scenario and are not included here.
Table 6-13 summarizes the 1-hour and 8-hour daily maximum exposures for each of the
three alternative standards scenarios in the Denver study area, while Table 6-14 presents the
same information for the Los Angeles study area. In comparing the exposure results for each
potential alternative scenario within each study area and exposure averaging time, generally
similar numbers of persons and their respective percentages of the simulated at-risk populations
are observed at the same level. This was by general design, that is, to investigate differing forms
of the potential alternative standards that would generate potentially similar exposure (and dose)
results. Again, there is a wider range in the 1-hour exposure levels experienced by the simulated
at-risk populations in Denver (Table 6-13) when compared with those of Los Angeles (Table 6-
14) when considering the same potential alternative standard, consistent with the differing
distribution of CO concentrations. There are also consistent patterns in the estimated distribution
of 8-hour daily maximum exposures experienced by the simulated at-risk populations, though the
upper range of that 8-hour maximum exposure is of course less than that of the 1-hour daily
maximum in each respective location.
There is some variability in the percent of persons exposed when considering a particular
level, form, and study area. For example, the 2nd highest 8-hour CO concentration of 5 ppm
most limited the number and percent of exposed persons in each location when compared to
results for the other potential alternative standard, though in Denver there were still a few
persons estimated to experience a 1-hour daily maximum at or above 30 ppm (Table 6-13). In
Los Angeles, the upper level of the 1-hour daily maximum exposure concentration experienced
by the simulated CHD population was just at or above 20 ppm, though below 30 ppm.
6-19
-------
Table 6-13. Estimated daily maximum 1-hour and 8-hour exposures for simulated at-risk populations in the Denver study
area - alternative air quality scenarios.
Daily
Maximum
Exposure
(ppm)
1-
hour
8-
hour
>0
>3
>6
>9
>12
> 15
>20
>30
>40
>60
>0
>3
>6
>9
>12
> 15
>20
>30
>40
>60
2nd highest 8-hour average of 5 ppm
CHD Persons3'0
Number
53,656
53,656
47,264
25,174
11,082
4,850
975
62
0
0
53,656
44,574
6,590
839
111
25
0
0
0
0
Percent
100
100
88.1
46.9
20.7
9.0
1.8
0.1
0
0
100
83.1
12.3
1.6
0.2
<0.1
0
0
0
0
HD Persons"'0
Number
85,926
85,926
75,671
40,489
18,140
7,886
1,802
74
0
0
85,926
71,488
10,637
1,296
173
37
0
0
0
0
Percent
100
100
88.1
47.1
21.1
9.2
2.1
<0.1
0
0
100
83.2
12.4
1.5
0.2
<0.1
0
0
0
0
99th pet 8-hour Daily Max of 5.0 ppm
CHD Persons3'0
Number
53,656
53,656
50,534
32,147
16,326
7,676
1,888
136
0
0
53,656
48,819
10,650
1,555
296
49
0
0
0
0
Percent
100
100
94.2
59.9
30.4
14.3
3.5
0.3
0
0
100
91.0
19.8
2.9
0.6
<0.1
0
0
0
0
HD Persons"'0
Number
85,926
85,926
80,928
51,645
26,384
12,501
3,209
247
0
0
85,926
78,189
17,412
2,369
432
74
0
0
0
0
Percent
100
100
94.2
60.1
30.7
14.5
3.7
0.3
0
0
100
91.0
20.3
2.8
0.5
<0.1
0
0
0
0
99th pet 1-hour Daily Max of 8.0 ppm
CHD Persons3'0
Number
53,656
53,656
48,239
26,877
12,192
5,282
1,222
74
0
0
53,656
45,808
7,380
926
123
25
0
0
0
0
Percent
100
100
89.9
50.1
22.7
9.8
2.3
0.1
0
0
100
85.4
13.8
1.7
0.2
<0.1
0
0
0
0
HD Persons"'0
Number
85,926
85,926
77,263
43,241
19,917
8,688
2,123
111
0
0
85,926
73,376
11,995
1,469
197
37
0
0
0
0
Percent
100
100
89.9
50.3
23.2
10.1
2.5
0.1
0
0
100
85.4
14.0
1.7
0.2
<0.1
0
0
0
0
Notes:
Ambient concentrations from 1995 were adjusted to just meet the level of the potential alternative standard indicated using a relationship
derived from the design monitor (ID 080310002).
a Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Inclusive of those persons with diagnosed coronary heart disease, angina pectoris, heart attack, and any other heart condition or
disease (CDC, 2009).
c Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1 .1).
6-20
-------
Table 6-14. Estimated daily maximum 1-hour and 8-hour exposures for simulated at-risk populations in the Los Angeles study
area - alternative air quality scenarios.
Daily
Maximum
Exposure
(ppm)
1-
hour
8-
hour
>0
>3
>6
>9
>12
> 15
>20
>30
>40
>60
>0
>3
>6
>9
>12
> 15
>20
>30
>40
>60
2nd highest 8-hour average of 5 ppm
CHD Persons3'0
Number
383,040
378,624
215,454
68,540
19,769
6,322
903
0
0
0
383,040
214,149
17,060
903
301
0
0
0
0
0
Percent
100
98.8
56.2
17.9
5.2
1.7
0.2
0
0
0
100
55.9
4.5
0.2
<0.1
0
0
0
0
0
HD Persons"'0
Number
630,807
624,183
348,720
105,469
30,507
9,533
1,606
0
0
0
630,807
350,626
26,192
1,204
502
0
0
0
0
0
Percent
100
99.0
55.3
16.7
4.8
1.5
0.3
0
0
0
100
55.6
4.2
0.2
<0.1
0
0
0
0
0
99th pet 8-hour Daily Max of 5.0 ppm
CHD Persons3'0
Number
383,040
379,728
229,302
77,571
24,285
7,827
1,305
0
0
0
383,040
230,807
20,672
1,204
301
100
0
0
0
0
Percent
100
99.1
59.9
20.3
6.3
2.0
0.3
0
0
0
100
60.3
5.4
0.3
<0.1
<0.1
0
0
0
0
HD Persons"'0
Number
630,807
626,090
373,105
120,321
36,728
11,841
2,208
0
0
0
630,807
376,918
31,811
1,706
502
100
0
0
0
0
Percent
100
99.3
59.1
19.1
5.8
1.9
0.3
0
0
0
100
59.8
5.0
0.3
<0.1
<0.1
0
0
0
0
99th pet 1-hour Daily Max of 8.0 ppm
CHD Persons3'0
Number
383,040
381,535
260,913
98,244
34,721
10,738
2,709
0
0
0
383,040
264,425
28,801
2,007
301
100
0
0
0
0
Percent
100
99.6
68.1
25.6
9.1
2.8
0.7
0
0
0
100
69.0
7.5
0.5
<0.1
<0.1
0
0
0
0
HD Persons"'0
Number
630,807
628,498
426,090
154,641
53,186
16,759
4,315
0
0
0
630,807
430,707
44,355
3,011
502
201
0
0
0
0
Percent
100
99.6
67.5
24.5
8.4
2.7
0.7
0
0
0
100
68.3
7.0
0.5
<0.1
<0.1
0
0
0
0
Notes:
Ambient concentrations from 1997 were adjusted to just meet the level of the potential alternative standard indicated using a relationship
derived from the design monitor (ID 060371 301).
a Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Inclusive of those persons with diagnosed coronary heart disease, angina pectoris, heart attack, and any other heart condition or
disease (CDC, 2009).
c Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1 .1).
6-21
-------
6.2 ESTIMATED COHB LEVELS
Consistent with section 6.1, this section summarizes the estimated COHb levels for the
simulated at-risk populations in a series of tables, classified by the air quality scenarios and study
areas considered. As was done in presenting the exposure results, we summarized the dose
results corresponding to both the CHD and HD populations. For all five air quality scenarios, we
report the number and percentage of persons and person-days estimated to have experienced
selected levels of the dose metric of interest (daily maximum end-of-hour COHb level). In
addition, we include an evaluation of the number of days in the year that individuals are
estimated to experience a dose at or above the selected COHb level. For three of the scenarios
(i.e., as is conditions, air quality adjusted to just meet the current 8-hour standard, and air quality
adjusted to just meet a 99th percentile daily maximum 8-hour average concentration of 5.0 ppm),
additional dose output data sets were generated using a smaller simulation size (i.e., 5,000
persons) and for the CHD population, to estimate the contribution of ambient CO exposure alone
to selected COHb levels. See section 5.10 for details on the approach used and evaluation of the
representativeness of the simulation size and at-risk population used.
6.2.1 Air Quality "As Is"
Table 6-15 provides the COHb levels (%) for the simulated at-risk populations in Denver
when considering the as is air quality. No persons were estimated to have experienced a daily
maximum end-of-hour COHb level at or above 2.5%, while only a few (<0.1%) were estimated
to have experienced a COHb level >2.0%. Nearly 99% of the simulated at-risk populations had
their highest daily maximum end-of-hour COHb level below 1.5%. Most of the simulated
person-days (about 97.5%) were associated with daily maximum end-of-hour COHb levels
below 1.0%.
Similarly in Los Angeles, very few persons (1.6%) were estimated to experience a daily
maximum end-of-hour COHb level at or above 1.5% when considering the as is air quality
(Table 6-16). Similar to that observed for the at-risk population in Denver, very few persons
(<0.1%) were estimated to experience at least one daily maximum end-of-hour COHb levels at
or above 2.0% in Los Angeles. Of these few hundred simulated individuals, all were estimated
to have only one person-day per person at that level (i.e., to have experienced this COHb level on
only one day in the year). As was observed with the Denver dose results, the majority of the
simulated person-days (>98%) were limited to COHb levels at or below 1.0%.
6-22
-------
Table 6-15. Portion of the simulated at-risk populations in the Denver study area estimated
to experience a daily maximum end-of-hour COHb at or above specified levels-
as is air quality.
COHb
Level
(%)
>0.0
> 1.0
>1.5
>1.75
>2.0
>2.5
>3.0
>4.0
Coronary Heart Disease3'0
Persons
Number
53,656
10,773
654
111
12
0
0
0
Percent
100
20.1
1.2
0.2
<0.1
0
0
0
Person-days
Number
19,580,000
494,600
17,170
2,616
86
0
0
0
Percent
100
2.5
<0.1
<0.1
<0.1
0
0
0
All Heart Disease"'0
Persons
Number
85,926
17,807
1,074
234
12
0
0
0
Percent
100
20.7
1.2
0.3
<0.1
0
0
0
Person-days
Number
31,360,000
781,100
29,050
3,579
86
0
0
0
Percent
100
2.5
<0.1
<0.1
<0.1
0
0
0
Notes:
Unadjusted ambient concentrations from four monitors in 2006 were used to represent the As Is air
quality scenario.
a Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Inclusive of those persons with diagnosed coronary heart disease, angina pectoris, heart attack, and
any other heart condition or disease (CDC, 2009).
c Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1 .1).
Table 6-16. Portion of the simulated at-risk populations in the Los Angeles study area
estimated to experience a daily maximum end-of-hour COHb at or above
specified levels - as is air quality.
COHb
Level
(%)
>0.0
>1.0
> 1.5
>1.75
>2.0
>2.5
>3.0
>4.0
Coronary Heart Disease3'0
Persons
Number
383,040
99,348
6,021
1,907
301
0
0
0
Percent
100
25.9
1.6
0.5
<0.1
0
0
0
Person-days
Number
139,800,000
1,725,000
91,120
18,870
301
0
0
0
Percent
100
1.2
<0.1
<0.1
<0.1
0
0
0
All Heart Disease"'0
Persons
Number
630,807
165,880
9,834
3,011
502
0
0
0
Percent
100
26.3
1.6
0.5
<0.1
0
0
0
Person-days
Number
230,200,000
3,112,000
165,200
28,100
502
0
0
0
Percent
100
1.4
<0.1
<0.1
<0.1
0
0
0
Notes:
Unadjusted ambient concentrations from four monitors in 2006 were used to represent the As Is air
quality scenario.
3 Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Inclusive of those persons with diagnosed coronary heart disease, angina pectoris, heart attack, and
any other heart condition or disease (CDC, 2009).
c Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1 .1).
6-23
-------
The population-based person-day dose metric was evaluated further by analyzing the
number of days in the year each individual experiences a daily maximum end-of hour COHb at
or above a selected level. As observed in the above analyses, there were little differences
between the two simulated at-risk populations in the percentage estimated to experience a
%COHb at or above selected levels. Therefore, we have chosen to use the CHD population as a
base case for this analysis. Figure 6-5 presents the percent of the CHD populations in Denver
(top) and Los Angeles (bottom) experiencing repeated COHb levels using as is air quality. As a
point of reference in the figure, the percent of persons with the number of occurrences > 1
corresponds to the data summarized in Tables 6-15 and 6-16.5 Note that we have included only
those COHb levels between 1.5 and 2.0%, though in 0.1% COHb increments.
Consistent with the summary tables, a small percentage of the CHD population (about
1.2% in Denver; 1.6% in Los Angeles) was estimated to experience a single COHb level at or
above 1.5% (Figure 6-15). Even fewer persons experienced two or more occurrences of COHb
above 1.5% (about 0.7% of the CHD population in Denver; 0.5% in Los Angeles), generally
about a factor of two or three lower than when considering persons experiencing at least one
COHb dose at or above that level. There were very few persons experiencing 3 or more COHb
levels at or above 1.5% in either study area. Even fewer persons experienced multiple
occurrences of higher COHb levels. For example less than 0.1% of the population experienced 3
or more COHb levels at or above 1.8% in either study area.
As discussed in chapter 2, we also evaluated the contribution of ambient CO exposure
alone to each simulated person's COHb level. The complete time-series of exposure for each
individual was used to generate each person's maximum end-of-hour COHb level attributable to
ambient CO exposure. This analysis also focused on the CHD population as a base-case in each
study area, given the limited differences between the percentage of persons at or above specific
COHb levels for either simulated at-risk population. Table 6-17 summarizes the estimated
COHb levels experienced by the CHD population for both study areas, using the as is air quality.
None of the persons experienced a COHb level at or above 1.8% due to ambient CO exposure
alone for this scenario in either study area. Estimated levels of maximum end-of-hour COHb
attributable to ambient CO exposure were at or below 1.3% COHb for approximately 99% of the
simulated CHD population. This is consistent with the above results for total COHb (Tables 6-
15 and 6-16)6 and the finding that endogenous CO production, on average, can contribute to an
5 The number of occurrences in Figure 6-5 corresponds to the number of days per year each person(s)
experienced a maximum end-of-hour COHb dose at or above the given level.
6 We used the term total COHb here as the combined dose from ambient CO exposure and endogenous CO
production.
6-24
-------
end-of-hour COHb level of approximately 0.3% for the simulated at-risk population in either
study area (Appendix B, Table B-4).
6-25
-------
Denver As Is Air Quality
> 6
Number of
Occurrences
Per Year
1.5 1.6 1.7 1.8 1.9 2.0
COHb Dose Level (%)
Los Angeles As Is Air Quality
1.6 1.7 1.8 1.9
COHb Dose Level ""-
2.0
Figure 6-5. Estimated percent of the CHD population in Denver (top) and Los Angeles
(bottom) experiencing repeated COHb levels - as is air quality.
6-26
-------
Table 6-17. Percentages of the simulated CHD populations in the Denver and Los Angeles
study areas estimated to experience a daily maximum end-of-hour COHb
contribution from ambient exposure alone at or above specified levels - as is
air quality.
Ambient-Exposure
Contribution to
COHb Level (%)
>1.0
>1.1
> 1.2
>1.3
> 1.4
>1.5
>1.6
> 1.7
>1.8
> 1.9
>2.0
Percent of CHD Population3
Denver
2.1
1.4
0.9
0.7
0.2
0.2
0.2
0
0
0
0
Los Angeles
7.1
5.3
3.0
1.5
0.5
0.5
0.5
0.3
0
0
0
Notes:
Unadjusted ambient concentrations from 2006 were used to represent
the As Is air quality scenario.
a Persons with diagnosed coronary heart disease, angina pectoris, and
heart attack (CDC, 2009). Includes estimate of persons with
undiagnosed ischemia developed by EPA (see section 5.5.1 .1).
6.2.2 Air Quality Adjusted to Just Meet the Current 8-hour Standard
Consistent with the estimated exposure concentrations, COHb levels estimated to be
experienced by the simulated at-risk populations in each study area were greater when
considering exposures associated with air quality adjusted to just meet the current standard than
when using as is air quality. For example, in Denver, just over 4% of the simulated at-risk
populations were estimated to have experienced a daily maximum end-of-hour COHb level at or
above 2.0% (Table 6-18). Note there were fewer than 0.1% of persons in Denver estimated to
have experienced COHb at a level above 2.0% based on estimated ambient exposures associated
with as is air quality (Table 6-15). A similar pattern is observed for the simulated at-risk
population in Los Angeles (Table 6-19), though a lower percentage of persons (0.6%) were
estimated to have experienced a daily maximum end-of-hour COHb level at or above 2.0% when
compared to the results for Denver. In both study areas, a few persons were estimated to have
experienced a daily maximum end-of-hour COHb levels as high as 3.0%. However, most of the
persons that did experience these higher COHb levels (>2.0%) experienced them for fewer than
2 days in a year (Table 6-19). A general pattern in experiencing multiple days at or above
selected COHb levels is evident in Figure 6-6. The percent of the population that experiences
two or more occurrences of %COHb at or above selected levels is reduced by about a factor of
two when compared with those experiencing at least one. Even fewer persons experience three
6-27
-------
or more days per year, also reduced by about a factor of about 2 when compared with those
experiencing at least two. When considering three or more occurrences in a year, the reduction
rate in the percent of persons at or above the selected COHb levels lessens with increasing
number of occurrences. In general, this pattern indicates a few simulated individuals, given
specific conditions of their estimated exposure and dose, experienced multiple days (e.g.,
possibly 6 or more) at or above selected COHb levels.
6-28
-------
Table 6-18. Portion of the simulated at-risk populations in the Denver study area estimated
to experience a daily maximum end-of-hour COHb at or above specified levels
- air quality just meeting the current 8-hour standard.
COHb
Level
(%)
>0.0
> 1.0
>1.5
>1.75
>2.0
>2.5
>3.0
>4.0
Coronary Heart Disease3'0
Persons
Number
53,656
44,166
12,563
5,800
2,258
444
111
12
Percent
100
82.3
23.4
10.8
4.2
0.8
0.2
<0.1
Person-days
Number
19,580,000
1,330,000
79,050
15,240
3,677
494
111
12
Percent
100
6.8
0.4
<0.1
<0.1
<0.1
<0.1
<0.1
All Heart Disease"'0
Persons
Number
85,926
71,710
21,028
9,502
3,826
802
222
25
Percent
100
83.5
24.5
11.1
4.5
0.9
0.3
<0.1
Person-days
Number
31,360,000
2,125,000
124,800
24,820
6,047
901
247
25
Percent
100
6.8
0.4
<0.1
<0.1
<0.1
<0.1
<0.1
Notes:
Ambient concentrations from 1995 were adjusted to just meet a 2nd highest 8-hour average concentration
of 9.4 ppm using a relationship derived from the design monitor (ID 080310002).
a Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Inclusive of those persons with diagnosed coronary heart disease, angina pectoris, heart attack, and
any other heart condition or disease (CDC, 2009).
c Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1 .1).
Table 6-19. Portion of the simulated at-risk populations in the Los Angeles study area
estimated to experience a daily maximum end-of-hour COHb at or above
specified levels - air quality just meeting the current 8-hour standard.
COHb
Level
(%)
>0.0
> 1.0
>1.5
> 1.75
>2.0
>2.5
>3.0
>4.0
Coronary Heart Disease3'0
Persons
Number
383,040
157,852
17,963
6,222
2,107
301
100
0
Percent
100
41.2
4.7
1.6
0.6
<0.1
<0.1
0
Person-days
Number
139,800,000
2,584,000
125,400
30,510
3,211
401
100
0
Percent
100
1.8
<0.1
<0.1
<0.1
<0.1
<0.1
0
All Heart Disease"'0
Persons
Number
630,807
260,612
31,410
10,537
3,613
401
100
0
Percent
100
41.3
5.0
1.7
0.6
<0.1
<0.1
0
Person-days
Number
230,200,000
4,598,000
227,400
50,280
4,716
502
100
0
Percent
100
2.0
<0.1
<0.1
<0.1
<0.1
<0.1
0
Notes:
Ambient concentrations from 1997 were adjusted to just meet a 2nd highest 8-hour average concentration
of 9.4 ppm using a relationship derived from the design monitor (ID 060371301).
a Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Inclusive of those persons with diagnosed coronary heart disease, angina pectoris, heart attack, and
any other heart condition or disease (CDC, 2009).
c Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1 .1).
6-29
-------
Denver (8-hr Std) Air Quality
>6
Number of
Occurrences
Per Year
2.0
1.6 1.7 1.8 1.9
COHb Dose Level (%)
Los Angeles (8-hr Std) Air Quality
1.6
1.7
1.8 1.9
2.0
COHb Dose Level (%)
Figure 6-6. Estimated percent of the CHD population in Denver (top) and Los Angeles
(bottom) experiencing repeated COHb levels - air quality just meeting the
current 8-hour standard.
6-30
-------
Table 6-20. Percentage of simulated CHD populations in the Denver and Los Angeles study
areas estimated to experience daily maximum end-of-hour COHb contribution
from ambient exposure alone at or above specified levels - air quality just
meeting the current 8-hour standard.
Ambient-Exposure
Contribution to
COHb Level (%)
>1.0
>1.1
>1.2
>1.3
>1.4
>1.5
>1.6
>1.7
>1.8
>1.9
>2.0
Percent of CHD Population3
Denver
43.6
30.4
21.0
16.0
12.8
9.4
6.8
4.1
3.4
3.0
2.7
Los Angeles
17.3
11.2
7.6
4.1
2.0
1.8
1.3
0.8
0.8
0.5
0.5
Notes:
Ambient concentrations from were adjusted to just meet a 2nd highest 8-
hour average concentration of 9.4 ppm.
a Persons with diagnosed coronary heart disease, angina pectoris, and
heart attack (CDC, 2009). Includes estimate of persons with
undiagnosed ischemia developed by EPA (see section 5.5.1.1).
Table 6-20 summarizes the percentages of the CHD population for both study areas
estimated to have experienced COHb levels at or above selected levels, when considering
ambient CO exposure alone (i.e., COHb level in the absence of endogenous CO production) and
using air quality adjusted to just meet the current 8-hour standard. As observed above (Tables 6-
18 and 6-19), there are differences between the two study areas when comparing the percent of
the CHD population estimated to experience a given COHb level. Approximately 9% of the
Denver population is estimated to have experienced a daily maximum end-of-hour COHb level
at or above 1.5% due to ambient CO exposure alone, while it was estimated that about 2% of the
simulated at-risk population in Los Angeles had at least one occurrence at this level.
6.2.3 Air Quality Adjusted to Just Meet Alternative Air Quality Scenarios
Consistent with the exposure results described above, the percentage of persons estimated
to experience a daily maximum end-of-hour COHb at or above the selected levels is generally
similar across the three potential alternative standard scenarios. For example, in Denver most of
the population (>99%) was estimated to not experience a daily maximum end-of-hour COHb
level at or above 2.0% (Table 6-21). Further, the potential alternative standard of a 2nd highest 8-
6-31
-------
hour average of 5 ppm resulted in the lowest number and percent of the simulated population
experiencing at least one occurrence of %COHb at or above the selected levels.
There are a few study area differences worthy of note. As expected, the corresponding
estimated percent of the CHD population experiencing a particular dose level in Denver is
greater than that estimated for Los Angeles, even when considering the same potential alternative
standard form and air quality level. For example, when considering a 99th percentile daily
maximum 8-hour average CO concentration of 5.0 ppm, 1.0% of the CHD population in the
Denver study area was estimated to have experienced an estimated daily maximum end-of-hour
COHb level at or above 2.0% (Table 6-21); in Los Angeles this dose level was estimated to have
been experienced by fewer than 0.1% of the CHD population (Table 6-22). Again, this is largely
a function of the differences observed between the upper percentile ambient concentrations used
to simulate these air quality scenarios in each study area (i.e., greater spatial variability in
ambient concentrations in the larger Los Angeles study area) and the large differences in altitude
between the two study areas.
Table 6-23 summarizes the percentage of the CHD population in both study areas
estimated to have experienced selected levels of COHb, when considering ambient CO exposure
alone (i.e., COHb level in the absence of endogenous CO production) and using air quality
adjusted to just meet a 99th percentile daily maximum 8-hour average concentration of 5.0 ppm.
As observed above (Tables 6-21 and 6-22), there is a similar pattern when comparing the two
study areas and the percent of the CHD population estimated to have experienced a given COHb
level. Approximately 3.2% of the Denver population experienced a daily maximum end-of-hour
COHb level at or above 1.5% due to ambient CO exposure alone, while in Los Angeles, it was
estimated that about 0.3% of the simulated at-risk population had at least one occurrence at this
level.
6-32
-------
Table 6-21. Portion of the simulated at-risk populations in the Denver study area estimated to experience a daily maximum
end-of-hour COHb at or above specified levels- air quality just meeting potential alternative standards.
COHb
Level (%)
>0.0
>1.0
>1.5
>1.75
>2.0
>2.5
>3.0
>3.5
>4.0
2nd highest 8-hour average of 5 ppm
CHD Persons3'0
Number
53,656
24,212
2,764
802
284
62
0
0
0
Percent
100
45.1
5.2
1.5
0.5
0.1
0
0
0
HD Persons"'0
Number
85,926
39,773
4,640
1,382
494
86
12
0
0
Percent
100
46.3
5.4
1.6
0.6
0.1
<0.1
0
0
99th pet 8-hour Daily Max of 5.0 ppm
CHD Persons3'0
Number
53,656
31,690
4,961
1,703
555
111
12
0
0
Percent
100
59.1
9.2
3.2
1.0
0.2
<0.1
0
0
HD Persons"'0
Number
85,926
51,459
8,256
2,789
975
210
37
12
0
Percent
100
59.9
9.6
3.2
1.1
0.2
<0.1
<0.1
0
99th pet 1-hour Daily Max of 8.0 ppm
CHD Persons3'0
Number
53,656
25,656
3,171
1,074
284
74
12
0
0
Percent
100
47.8
5.9
2.0
0.5
0.1
<0.1
0
0
HD Persons"'0
Number
85,926
42,031
5,282
1,740
518
111
25
12
0
Percent
100
48.9
6.1
2.0
0.6
0.1
<0.1
<0.1
0
Notes:
Ambient concentrations from 1995 were adjusted to just meet the level of the potential alternative standard indicated using a relationship
derived from the design monitor (ID 080310002).
3 Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Inclusive of those persons with diagnosed coronary heart disease, angina pectoris, heart attack, and any other heart condition or
disease (CDC, 2009).
c Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1 .1).
6-33
-------
Table 6-22. Portion of the simulated at-risk populations in the Los Angeles study area estimated to experience a daily
maximum end-of-hour COHb at or above specified levels - air quality just meeting potential alternative
standards.
COHb
Level (%)
>0.0
>1.0
> 1.5
>1.75
>2.0
>2.5
>3.0
>3.5
>4.0
2nd highest 8-hour average of 5 ppm
CHD Persons3'0
Number
383,040
54,491
2,509
803
301
0
0
0
0
Percent
100
14.2
0.7
0.2
<0.1
0
0
0
0
HD Persons"'0
Number
630,807
91,319
4,917
1,305
401
0
0
0
0
Percent
100
14.5
0.8
0.2
<0.1
0
0
0
0
99th pet 8-hour Daily Max of 5.0 ppm
CHD Persons3'0
Number
383,040
62,017
3,613
1,004
301
0
0
0
0
Percent
100
16.2
0.9
0.3
<0.1
0
0
0
0
HD Persons"'0
Number
630,807
102,258
6,121
1,706
401
0
0
0
0
Percent
100
16.2
1.0
0.3
<0.1
0
0
0
0
99th pet 1-hour Daily Max of 8.0 ppm
CHD Persons3'0
Number
383,040
78,776
5,319
1,706
401
0
0
0
0
Percent
100
20.6
1.4
0.4
0.1
0
0
0
0
HD Persons"'0
Number
630,807
130,858
9,232
3,011
502
0
0
0
0
Percent
100
20.7
1.5
0.5
<0.1
0
0
0
0
Notes:
Ambient concentrations from 1997 were adjusted to just meet the level of the potential alternative standard indicated using a relationship
derived from the design monitor (ID 060371301).
a Persons with diagnosed coronary heart disease, angina pectoris, and heart attack (CDC, 2009).
b Inclusive of those persons with diagnosed coronary heart disease, angina pectoris, heart attack, and any other heart condition or
disease (CDC, 2009).
c Includes estimate of persons with undiagnosed ischemia developed by EPA (see section 5.5.1 .1).
6-34
-------
Table 6-23. Percentage of simulated CHD populations in the Denver and Los Angeles study
areas estimated to experience daily maximum end-of-hour COHb contribution
from ambient exposure alone at or above specified levels - air quality just
meeting a 99th percentile daily maximum 8-hour average concentration of 5.0
ppm.
Ambient-Exposure
Contribution to
COHb Level (%)
>1.0
>1.1
>1.2
>1.3
>1.4
>1.5
>1.6
>1.7
>1.8
>1.9
>2.0
Percent of CHD Population3
Denver
18.0
13.2
8.9
5.9
3.7
3.2
2.7
2.1
0.9
0.7
0.2
Los Angeles
2.8
1.8
1.3
0.8
0.5
0.3
0.3
0.3
0.3
0.3
0.3
Notes:
Ambient concentrations from were adjusted to just meet a 99th percentile
daily maximum 8-hour average concentration of 5.0 ppm.
a Persons with diagnosed coronary heart disease, angina pectoris, and
heart attack (CDC, 2009). Includes estimate of persons with
undiagnosed ischemia developed by EPA (see section 5.5.1.1).
6.3 COMPARISON OF COHB ESTIMATES OBTAINED FROM THE 2000
PNEM/CO AND 2010 APEX/CO ASSESSMENTS
As described above in chapters 2 and 4, population exposure and dose were estimated in
2000 using pNEM/CO, a predecessor to APEX, for adults with ischemic heart disease (IHD)
residing in a defined study area within the same two urban areas (Johnson et al., 2000). As
described in section 1.2 above, IHD is also termed CHD, and with regard to characterizing the
population of interest with regard to demographics (age and gender), the 2000 assessment, like
the current assessment, drew from estimates of the prevalence provided by the NHIS (which
includes CHD or IHD, angina pectoris, and heart attack) and corresponding estimates of
undiagnosed ischemia developed by EPA. As part of this current (2010) CO REA, staff has used
APEX to estimate CO exposures and resulting COHb levels using a largely similar approach,
6-35
-------
modeling domains, years of ambient concentration data,7 and defined at-risk population.8 There
are some differences that exist when comparing details of the methodologies and data sets used:
• number of ambient monitors used (e.g., previously six in Denver versus four used
here),9
• location of ambient monitors used (e.g., eight of the same monitors used previously
were used here for Los Angeles),
• number of microenvironments modeled (previously 15 versus the 8 modeled here),
• use of mass balance modeling (previously all 12 enclosed microenvironments used
mass balance, here only indoor microenvironments used a mass balance approach)
• use of a cohort approach (pNEM) versus individual approach (APEX), and
• inclusion of two indoor emission sources of CO in the 2000 pNEM/CO assessment for
residential microenvironments: gas stoves and passive smoking.
Despite these differences and a few others not listed, staff still did not expect to see
greatly different results when comparing the two assessments given the similarities in the most
likely influential variables (i.e., ambient CO concentrations, microenvironmental approach, CFK
module used, etc.). Table 6-24 presents estimates for the percentage of Denver adults with CHD
estimated to experience a daily maximum end-of-hour %COHb at or above the selected level
under the specified air quality conditions for 1995. Table 6-25 presents similar estimates for Los
Angeles using the 1997 ambient air quality data adjusted to just meet the current 8-hour standard.
Each table provides two sets of COHb level estimates for the 2000 pNEM/CO assessment (one
with and the other without indoor source emissions) and one set generated from the current
(2010) APEX/CO REA.
As expected, the estimated percent of persons at or above selected COHb levels from the
2000 pNEM/CO assessment is greatest when indoor source emissions are included in the
exposure modeling simulation. It is clear by comparing the two estimates from Johnson et al.
(2000) that the presence of indoor sources had a significant impact on COHb levels - much more
so than the ambient air contributions, as the percent of persons at selected COHb levels increased
by large margins (about 10-30 percentage points) where data are available and comparable from
both model simulations. The range of COHb levels also extends upwards to at or above 6.0%
7 When considering the exposure scenario that uses air quality just meeting the current standard.
8 When considering the CHD population.
9 The actual number of monitors used in the 2000 assessment was seven, though air quality data from two
monitors in Boulder CO (monitor IDs 080130010 and 080131001) were averaged to create a composite air quality
district (Johnson et al., 2000).
6-36
-------
COHb for 0.2% of the CHD population when considering indoor source emissions and air
quality adjusted to just meet the current 8-hour standard.
While these results regarding indoor sources from Johnson et al. (2000) are generally
informative, they cannot be directly applied to the current dose assessment results. This is
because the data used for simulating the indoor source emissions, while some of it may be
readily available for use in the currently used APEX model, are considered not necessarily
reflective of current conditions. For example, the indoor gas stove emissions data used were
generated at a time where pilot lights (a continuous low-level combustion scenario) and limited
external ventilation conditions existed. In addition, while tobacco smoking prevalence rates have
not necessarily changed much over the past two decades, the prevalence of smoking indoors has
been substantially reduced in public buildings and likely within many residential
microenvironments. It is these important changes in indoor source emissions, the limited
availability of current and relevant input data, and the limited time and resources allocated for
this assessment, that preclude a current quantitative assessment of the impact of indoor source
emissions on population COHb levels.
The range of dose estimates without simulated indoor sources are generally similar in
both study areas when comparing results from the Johnson et al. (2000) assessment with those
generated in the current CO REA. However at selected COHb levels in Denver, the current
approach estimated a higher percent of the CHD population than when compared with the
previous Johnson et al. (2000) assessment. For example, approximately 4.2% of the CHD
population was estimated to have a daily maximum end-of-hour COHb level at or above 2.0% in
this current assessment. The corresponding value estimated in the Johnson et al. (2000)
assessment was approximately 0.5% of the IHD population. One factor contributing to the
difference in the results for Denver at this benchmark level is the air quality data used for each
assessment. The two additional air quality districts used in the 2000 assessment had consistently
lower hourly CO concentrations when compared with concentrations measured at the other four
ambient monitors that were consistent for both assessments. For example, at selected upper
percentiles of the air quality distribution for these two monitoring sites, concentrations are
reduced by a factor of 1.5 to 6.8 (Table 3-8 of Johnson et al., 2000). As a result, the simulated
persons residing within these low CO concentration air quality districts10 would have
consistently lower estimated exposure concentrations and thus experience lower COHb levels.
While there were differences in two of the ten monitors used to define the Los Angeles study
10 These persons comprised 20.6% of the total simulated IHD population in the Denver study area (see
Table 2-8 of Johnson et al., 2000).
6-37
-------
areas in each assessment, a 1.5 to 6.8 factor difference in the upper percentiles of the air quality
distributions used is not present.
6-38
-------
Table 6-24. Percentage of Denver adults with coronary heart disease (CHD) estimated to
experience a daily maximum end-of-hour COHb level - air quality just meeting
the current 8-hour standard.
COHb
Level
(%)
>0.0
>1.0
>1.5
>2.0
>2.5
>3.0
>4.0
>5.0
>6.0
Percentage of CHD Adults at or Above COHb Level
Johnson et al. (2000) pNEM/COa
Includes Indoor
Source Emissions
100
83.2
37.6
19.9
10.4
5.5
1.6
0.6
0.2
No Indoor Source
emissions
100
65.0
6.7
0.5
0.2
<0.1
0
0
0
2010REAAPEX/COb
No Indoor Source
Emissions
100
82.3
23.4
4.2
0.8
0.2
<0.1
na
na
Notes:
3 Used Denver 1995 CO ambient concentrations with no adjustment (2nd highest 8-hour
CO concentration was 9.5 ppm, close in value to the design value of 9.4 ppm).
b Denver 1995 ambient CO concentrations adjusted to just meet the current 8-hour
standard (9.4 ppm).
na - benchmark level was not evaluated in the current exposure and dose simulations.
Table 6-25. Percentage of Los Angeles adults with coronary heart disease (CHD) estimated
to experience a daily maximum end-of-hour COHb level - air quality just
meeting the current 8-hour standard.
COHb
Level
(%)
>0.0
>1.0
>1.5
>2.0
>2.5
>3.0
>4.0
>5.0
>6.0
Percentage of CHD Adults at or Above COHb Level
Johnson et al. (2000) pNEM/COa
Includes Indoor
Source Emissions
100
79.0
32.3
16.8
9.0
5.1
2.2
0.9
0.3
No Indoor Source
emissions
100
58.1
5.2
0.5
<0.1
<0.1
0
0
0
2010REAAPEX/C03
No Indoor Source
Emissions
100
41.2
4.7
0.6
<0.1
<0.1
0
0
0
Notes:
3 Los Angeles 1997 ambient CO concentrations adjusted to just meet the current 8-
hour standard (9.4 ppm).
6-39
-------
6.4 KEY OBSERVATIONS
Presented below are key observations resulting from the exposure and dose assessment
for ambient CO.
• Ambient CO exposures and resulting COHb levels in the blood of two simulated at-risk
populations in the Los Angeles and Denver study areas were estimated considering five
air quality scenarios: as is air quality, air quality adjusted to simulate just meeting the
current 8-hour CO NAAQS, and air quality adjusted to just meet three potential
alternative standards.
• The two at-risk populations simulated were: (1) persons with diagnosed CHD,
including those estimated to have undiagnosed CHD, and (2) the larger group of
persons with any type of HD including those estimated to have undiagnosed CHD.
While the number of persons and person-days at or above selected COHb levels
differed between the two populations, reflecting their differing size, the percentage of
each population's persons and person-days were similar.
• The relative contribution of various microenvironments to exposure concentrations was
generally similar between the two study areas. When considering as is air quality,
indoor microenvironments contributed mostly to low level exposures (at or above 1
ppm and 2 ppm), comprising between 40 - 80% of the time spent at those exposure
levels, while time spent inside vehicles contributed to most exposures at or above 3
ppm (70 - 100%). In comparison, when considering air quality just meeting the
current standard, the percent contribution from indoor microenvironments was
generally higher for low level exposures (about 65 - 85% of exposure concentrations at
or above 1 ppm and 2 ppm), though again higher level exposures were dominated by
the contributions from inside-vehicle microenvironments.
• The relationship between the two study areas with regard to estimated distribution of
maximum end-of-hour COHb levels differed with the different air quality scenarios.
Under as is air quality conditions, the simulated at-risk populations in the Los Angeles
study area were estimated to experience a slightly higher distribution of maximum end-
of-hour COHb levels than the Denver populations. Under conditions of air quality
adjusted from historical air quality data to just meet the current or alternative standards,
however, appreciably larger percentages of the Denver populations were estimated to
experience COHb at or above specific levels than the Los Angeles populations.
• For as is air quality conditions, the highest daily maximum end-of-hour COHb
estimated to be experienced over the course of the simulated year was below 1.5% for
more than 98% of the at-risk populations simulated in each study area; it was below
2% COHb for more than 99.9% of these simulated populations. A lower percentage of
the simulated at-risk populations in Denver were estimated to experience daily
maximum end-of hour COHB below these benchmarks than were the populations in
Los Angeles.
• Under as is air quality conditions, the highest incremental contribution of
ambient CO exposure to maximum end-of-hour COHb levels estimated in the
simulated populations was 1.7% COHb, and more than 99% of both study area
populations were estimated to have ambient CO contributions to COHb below
6-40
-------
1.4%. As with estimates of total COHb (i.e., COHb from endogenous CO
production and ambient exposure together), a larger percentage of the Los
Angeles population was estimated to experience the higher ambient
contributions to maximum end-of-hour COHb compared to the Denver
population. For example, the percentage of the population estimated to
experience ambient contributions to COHb at or above COHb levels above
1.0% was approximately 2 to 3 times as high in Los Angeles than in Denver.
For simulations of air quality adjusted to just meet the current 8-hour standard of 9
ppm, the highest estimated daily maximum end-of-hour COHb over the course of the
simulated year was below 1.5% for 95% of both simulated at-risk populations in the
Los Angeles study area, and was below 2% COHb for 99.4% of these populations. In
contrast, the percentage of the simulated at-risk populations in the Denver study area
estimated to experience daily maximum end-of-hour COHb levels that did not exceed
1.5% was about 80%. The percentage of the Denver populations with their highest
estimated daily maximum end-of-hour COHb below 2% was approximately 95%.
• As with estimates of total COHb (i.e., COHb from endogenous CO production
and ambient exposure together), the percentage of the simulated population
estimated to experience the higher ambient contributions to maximum end-of
hour COHb was appreciably greater in Denver as compared to Los Angeles.
While estimated ambient CO contributions to daily maximum end-of-hour
COHb were below 1.4% for nearly 98% of the Los Angeles simulated
population, the corresponding percentage of the Denver population was about
87%.
In addition to the simulations for as is and just meeting the current 8-hour standard air
quality conditions, three simulations of air quality just meeting potential alternative
standards were performed. The alternatives comprise different combinations of form,
averaging time and level which were expected to achieve somewhat similar exposure
and dose results. The combinations that were selected were based on consideration of
exposure and dose results obtained for the as is air quality conditions and the higher
just meeting the current 8-hour standard conditions. The combinations of form,
averaging time and level that were simulated include: (1) a second-highest 8-hour
average of 5 ppm, (2) a 99th percentile daily maximum 8-hour average of 5.0 ppm, and
(3) a 99th percentile daily maximum 1-hour average of 8.0 ppm.
• These three simulations generated generally similar percentages of the at-risk
populations estimated to be exposed at selected concentrations and experience
maximum end-of hour COHb levels at or above selected levels. Each of these
three simulations generated fewer persons and a lower percent of the at-risk
populations at or above selected COHb levels than did simulations for air
quality adjusted to just meet the current 8-hour standard. For example, about
1% or less of the Denver populations had a highest daily maximum end-of-
hour COHb at or above the 2.0% COHb level in the alternative standards'
simulations as compared to approximately 4.2% in simulations of air quality
just meeting the current standard. When considering the potential alternative
standards in Los Angeles, generally fewer than 0.1% of the simulated at-risk
populations were estimated to experience a maximum end-of-hour COHb level
6-41
-------
at or above 2.0 % COHb, compared to 0.5% at that same COHb level
associated with air quality adjusted to just meet the current standard.
• Results for the five air quality scenarios are further analyzed in the Policy Assessment
to inform consideration of the level of public health protection that might be provided
by alternative standard levels associated with different combinations of averaging time
and form.
• Results generated in the current assessment for the air quality conditions just meeting
the current NAAQS were compared with estimates from the assessment conducted in
2000 (Johnson et al., 2000) for similar conditions in the Denver and Los Angeles study
areas (section 6.3). The two assessments employed similar approaches, similar,
although not identical air quality data for this scenario, and they used different
exposure models (APEX vs. pNEM). Results were similar for the 1.5% and 2% COHb
level for the simulated Los Angeles study area population and somewhat different for
the Denver study area population. For example, the two assessments' Los Angeles
estimates of population percentages with highest daily maximum end-of hour COHb at
or above these COHb levels were similar and within about 10% of one another. For
the Denver simulated study populations, however, the estimates for these two COHb
levels were about 3 and 8 times greater, respectively, in the current assessment when
compared with that estimated in the prior assessment.
6-42
-------
6.5 REFERENCES
Graham SE and McCurdy T. (2004). Developing meaningful cohorts for human exposure models. J Expos Anal
Environ Epidemiol. 14(l):23-43.
Johnson T, Mihlan G, LaPointe J, Fletcher K, Capel J. (2000). Estimation of Carbon Monoxide Exposures and
Associated Carboxyhemoglobin Levels for Residents of Denver and Los Angeles Using pNEM/CO
(Version 2.1). Report prepared by ICF Consulting and TRJ Environmental, Inc., under EPA Contract No.
68-D6-0064. U.S. Environmental Protection Agency, Research Triangle Park, North Carolina. Available
at: http://www.epa.gov/ttn/fera/human_related.html. June 2000.
Rodes C, Sheldon L, Whitaker D, Clayton A, Fitzgerald K, Flanagan J, DiGenova F, Hering S, Frazier C. (1998).
Measuring Concentrations of Selected Air Pollutants Inside California Vehicles. Final Report. Prepared
by Research Triangle Institute under Contract No. 95-339. California Air Resources Board. Sacramento,
California.
Shikiya D, Liu C, Kahn M, Juarros J, Barcikowski W. (1989). In-Vehicle Air Toxics Characterization Study in the
South Coast Air Basin. Office of Planning and Rules, South Coast Air Quality Management District.
October 1989.
6-43
-------
7 VARIABILITY ANALYSIS AND UNCERTAINTY
CHARACTERIZATION
An important issue associated with any population exposure or risk assessment is the
characterization of variability and uncertainty. Variability refers to the inherent heterogeneity in
a population or variable of interest (e.g., residential air exchange rates). The degree of variability
cannot be reduced through further research, only better characterized with additional
measurement. Uncertainty refers to the lack of knowledge regarding the values of model input
variables (i.e., parameter uncertainty), the physical systems or relationships used (i.e., use of
input variables to estimate exposure or risk or model uncertainty), and in specifying the scenario
that is consistent with purpose of the assessment (i.e., scenario uncertainty). Uncertainty is,
ideally, reduced to the maximum extent possible through improved measurement of key
parameters and iterative model refinement. The approaches used to assess variability and to
characterize uncertainty in this REA are discussed in the following two sections. Each section
also contains a concise summary of the identified components contributing to uncertainty and
how each source may affect the estimated exposures.
7.1 ANALYSIS OF VARIABILITY
The purpose for addressing variability in this REA is to ensure that the estimates of
exposure and risk reflect the variability of ambient CO concentrations, population characteristics,
associated CO exposure and dose, and potential health risk across the study area and for the
simulated at-risk populations. In this CO REA, there are several algorithms that account for
variability of input data when generating the number of estimated benchmark exceedances or
health risk outputs. For example, variability may arise from differences in the population
residing within census tracts (e.g., age distribution) and the activities that may affect population
exposure to CO and the resulting dose (e.g., time spent inside vehicles, performing moderate or
greater exertion level activities outdoors). A complete range of potential exposure levels and
associated risk estimates can be generated when appropriately addressing variability in exposure
and risk assessments; note however that the range of values obtained would be within the
constraints of the input parameters, algorithms, or modeling system used, not necessarily the
complete range of the true exposure or risk values.
Where possible, staff identified and incorporated the observed variability in input data
sets to estimate model parameters within the exposure and dose assessment rather than
employing standard default assumptions and/or using point estimates to describe model inputs.
The details regarding variability distributions used in data inputs are described in chapter 5. To
the extent possible given the data available for the assessment, staff accounted for variability
within the exposure and dose modeling. APEX has been designed to account for variability in
7-1
-------
some of the input data, including the physiological variables that are important inputs to
determining ventilation rates and COHb levels. As a result, APEX addresses much of the
variability in factors that affect human exposure and dose. Important sources of the variability
accounted for in this analysis are summarized in Table 7-1.
Table 7-1. Summary of how variability was incorporated into the assessment.
Component
Simulated
Individuals
Ambient Input
Microenvironmental
Approach
Variability Source
Population data
Commuting data
Activity patterns
Longitudinal profiles
Coronary heart
disease (CHD)
prevalence
Measured ambient
CO concentrations
Meteorological data
Microenvironments
Proximity factors
Mass balance model
Comment
Individuals are randomly sampled from US census tracts
used in model domains, by age (single years) and
gender (US Census Bureau, 2007).
Individuals are probabilistically assigned ambient
concentrations originating from either their home or work
tract based on US Census derived commuter data (US
Census Bureau, 2007).
Data diaries are randomly selected from CHAD master
(35,000 diaries) using six diary pools stratified by two
day-types (weekday, weekend) and three temperature
ranges (< 55.0 °F, between 55.0 and 83.9°F, and >84.0
F). The CHAD diaries capture real locations that
people visit and the activities they perform, ranging from
1 minute to 1 hour in duration (US EPA, 2002).
A sequence of diaries is linked together for each
individual that preserves both the inter- and intra-
personal variability in human activities (Glen et al.,
2008).
CHD prevalence is stratified by four age groups (18-44,
45-64, 65-74, and 75+) and both genders (CDC, 2009)
Temporal: 1 -hour CO for an entire year predicted using
ambient monitoring data.
Spatial: Four monitors were used to represent ambient
conditions in Denver; ten monitors used in Los Angeles;
each monitor was assigned a 10 km zone of influence.
Spatial: Local surface NWS stations used.
Temporal: 1-hour NWS temperature data for each year.
Eight total microenvironments were represented,
including those expected to be associated with high
exposure concentrations (i.e., in-vehicle and near-road).
This results in differential exposure estimates for each
individual (and event) when spending time within each
microenvironment.
In the current APEX approach, microenvironmental
concentrations were estimated using proximity factors to
adjust the outdoor CO concentrations. All proximity
factors were represented by lognormal distributions
whose values are randomly selected for every individual
exposure event.
For the indoor microenvironments, using a mass
balance model accounts for CO concentrations
occurring during a previous hour (and of ambient origin)
to calculate current indoor CO concentrations.
7-2
-------
Component
Physiological
Factors Relevant to
Ventilation Rate and
Estimation of COHb
Levels
Variability Source
Air exchange rates
Resting metabolic
rate
Metabolic
equivalents by
activity (METS)
Oxygen uptake per
unit of energy
expended
Weight (body mass)
Height
Blood volume
Hemoglobin content
of the blood
Pulmonary CO
diffusion rate
Endogenous CO
production rate
Comment
Several lognormal distributions are sampled based on
five daily mean temperature ranges, two regions, and
location specific A/C prevalence rates.
Regression equations for three age-group (18-29, 30-
59, and 60+) by two genders were used with body mass
as the independent variable (Johnson et al., 2000).
Values randomly sampled from distributions developed
for specific activities (some age-specific) (McCurdy,
2000; US EPA, 2002).
Values randomly sampled from a uniform distribution
(Johnson etal., 2000).
Randomly selected from population-weighted lognormal
distributions with age- and gender-specific geometric
mean (GM) and geometric standard deviation (GSD)
derived from data from the National Health and Nutrition
Examination Survey (NHANES), for the years 1999-
2004 (Isaacs and Smith (2005) in Appendix A).
Values randomly sampled from distribution based on
equations developed for each gender developed by
Johnson (1998) using height and weight data from
Brainard and Burmaster (1992) (see Appendix B for
details).
Values determined according to gender using equations
developed from Allen et al. (1956) (see Appendix B for
details).
Values randomly selected from distributions developed
by gender and age categories based on NHANES study
(see Isaacs and Smith (2005) in Appendix A).
Values selected according to gender, height, and age
based on equations adapted from Salorinne (1976) (see
Appendix B for details).
Values randomly selected from lognormal distributions
according to equations specific to age, gender, and
menstrual phase (data obtained from eight independent
studies; see Appendix B for details).
7-3
-------
7.2 CHARACTERIZATION OF UNCERTAINTY
While it may be possible to capture a range of exposure or risk values by accounting for
variability inherent to influential factors, the true exposure or risk for any given individual within
a study area is largely unknown. To characterize health risks, exposure and risk assessors
commonly use an iterative process of gathering data, developing models, and estimating
exposures and risks, given the goals of the assessment, scale of the assessment performed, and
limitations of the input data available. However, significant uncertainty often remains and
emphasis is then placed on characterizing the nature of that uncertainty and its impact on
exposure and risk estimates.
We have used such an iterative process in characterizing the uncertainty associated with
the approach and data used in developing this final CO REA. Following a review of the draft
REA's by CAS AC and the public, a few sources of uncertainty were identified as important in
improving the approach used to estimate exposure and dose. These included spatial
representation of the monitors used, the number of microenvironments and approach used to
estimate exposure, and representation of the at-risk population, among a few others (e.g., Brain
and Samet, 2010a). Major approach modifications and analyses conducted throughout the
evaluation of this final assessment included the following:
• Expanding the number of monitors used to better address spatial variability in ambient
CO concentrations;
• Increasing the number of microenvironments modeled from two to eight;
• Using distributions of proximity factors to estimate all microenvironmental
concentrations rather than simple point estimates;
• Expanding analysis of historical trends in ambient CO concentrations at individual
monitors;
• Including two simulated at-risk populations based on prevalence rates for CHD and all
types of heart disease (including estimates of undiagnosed CHD prevalence);
• Evaluating endogenous CO production and the ambient contribution to individual and
population COHb levels;
• Identifying the specific microenvironments that contribute to low- and high-level
exposures;
• Estimating the percent of simulated at-risk persons experiencing multiple occurrences
per year at or above selected COHb levels;
• Evaluating the distribution of microenvironmental factors used to estimate exposure
concentrations, and;
• Performing sensitivity analyses including
• Evaluating the impact of additional monitoring data on estimated exposures
and COHb levels;
7-4
-------
• Evaluating the impact of varying undiagnosed prevalence rates by gender to
estimated population COHb levels, and;
• Evaluating the impact of using alternative hemoglobin content distributions to
represent a hypothetical population with anemia.
These additional analyses and approaches used are not without their own uncertainties,
and following this iterative process, these uncertainties also need to be characterized. This
characterization of uncertainty can include either qualitative or quantitative evaluations, or a
combination of both. The approach can also be tiered; that is, the analysis can begin with a
simple qualitative uncertainty characterization and then progress to a complex probabilistic
uncertainty analysis. This second level of analysis may be appropriate when a lower tier analysis
indicates there is a high degree of uncertainty for certain identified sources, the sources of
uncertainty are highly influential variables in estimating the exposure and risk, and sufficient
information and other resources are available to conduct a quantitative uncertainty assessment.
This is not to suggest that quantitative uncertainty analyses should always be performed in all
exposure and risk assessments. The decision regarding the type of uncertainty characterization
performed is also informed by the intended scope and purpose of the assessment, whether the
selected analysis will provide additional information to the overall decision regarding health
protection, whether sufficient data are available to conduct a complex quantitative analysis, and
whether time and resources are available for higher tier characterizations (US EPA, 2004; WHO,
2008).
The primary purpose of the uncertainty characterization approach selected in this CO
REA is to identify and compare the relative impact that important sources of uncertainty may
have on the estimated potential health effect endpoints. The approach used to characterize
uncertainty was adapted from guidelines outlining how to conduct a qualitative uncertainty
characterization (WHO, 2008) and applied in the most recent NO2 (US EPA, 2008) and SO2
NAAQS reviews (US EPA, 2009). While it may be considered ideal to follow a tiered approach
in the REA to quantitatively characterize all identified uncertainties, staff selected the mainly
qualitative approach given the extremely limited data available to inform probabilistic analyses.
The qualitative approach used in this REA varies from that of WHO (2008) in that a
greater focus was placed on evaluating the direction and the magnitude1 of the uncertainty; that
is, qualitatively rating how the source of uncertainty, in the presence of alternative information,
may affect the estimated exposures and health risk results. In addition and consistent with the
WHO (2008) guidance, staff discuss the uncertainty in the knowledge base (e.g., the accuracy of
the data used, acknowledgement of data gaps) and decisions made where possible (e.g., selection
1 This is synonymous with the "level of uncertainty" discussed in WHO (2008), section 5.1.2.2.
7-5
-------
of particular model forms), although qualitative ratings were assigned only to uncertainty
regarding the knowledge base.
First, staff identified the key aspects of the assessment approach that may contribute to
uncertainty in the exposure and risk estimates and provided the rationale for their inclusion.
Then, staff characterized the magnitude and direction of the influence on the assessment results
for each of these identified sources of uncertainty. Consistent with the WHO (2008) guidance,
staff subjectively scaled the overall impact of the uncertainty by considering the degree of
uncertainty as implied by the relationship between the source of uncertainty and the exposure
concentrations and COHb levels.
Where the magnitude of uncertainty was rated low, it was judged that large changes
within the source of uncertainty would have only a small effect on the exposure results. For
example, a statistical procedure was used to substitute missing ambient concentrations in each
ambient data set. Staff compared the air quality distributions and found negligible differences
between the substituted data set and the one with missing values (e.g., Tables 5-13 through 5-
16). There is still uncertainty in the approach used, since there may be alternative methods
available. However, staff judged that the quantitative comparison of the data sets indicates that
there would likely be little influence on exposure estimates by the data substitution procedure.
A magnitude designation of medium implies that a change within the source of
uncertainty would likely have a moderate (or proportional) effect on the results. For example,
the magnitude of uncertainty associated with using the historical data to represent a hypothetical
future scenario was rated as low-medium. While we do not have information regarding how the
ambient CO concentration distribution might look in the future, we do know however what the
distribution might look like based on historical trends and the primary emission sources. If these
trends in observed concentrations and emissions remain consistent in the future, then the
magnitude of the impact to estimated exposures in this assessment would be judged as likely low
or having negligible impact on the exposure and dose estimates. However, if there are entirely
new emission sources, the magnitude of influence might be greater. When adjusting air quality
in each location to simulate the various exposure scenarios, staff observed mainly proportional
differences (e.g., a factor of two or three) in the estimated exposure and dose levels. Assuming
that these types of ambient concentration adjustments could reflect the addition of a new source
in each area carries its own uncertainties; however, based on this information, staff also judged
the magnitude of influence in using the historical air quality data to represent a hypothetical
future scenario as medium. A characterization of high implies that a small change in the source
would have a large affect on results, potentially an order of magnitude or more. This rating
would be used where the model was extremely sensitive to the identified source of uncertainty.
In addition to characterizing the magnitude of uncertainty, staff also included the
direction of influence, indicating how the source of uncertainty was judged to affect estimated
7-6
-------
exposures or risk estimates; either the estimated values were likely over- or under-estimated. In
the instance where the component of uncertainty can affect the assessment endpoint in either
direction, the influence was judged as both. Staff characterized the direction of influence as
unknown when there was no evidence available to judge the directional nature of uncertainty
associated with the particular source. Staff also subjectively scaled the knowledge-base
uncertainty associated with each identified source using a three level scale: low indicated
significant confidence in the data used and its applicability to the assessment endpoints, medium
implied that there were some limitations regarding consistency and completeness of the data
used or scientific evidence presented, and high indicated the extent of the knowledge base was
extremely limited.
The output of the uncertainty characterization was a summary describing, for each
identified source of uncertainty, the magnitude of the impact and the direction of influence the
uncertainty may have on the exposure and risk characterization results. We identified sixteen
sources of uncertainty associated with this approach for modeling CO population exposure and
dose and associated potential health risk, each summarized in Table 7-2. Section 7.2.1 describes
the sources more fully and provides support for the ratings ultimately selected by staff, while
section 7.2.2 includes model sensitivity analyses referenced in section 7.2.1. As mentioned in
earlier chapters, given the significant time constraints of this review, results of the
characterization are provided in this document without substantial interpretation. Rather,
interpretative discussion of these results, including further consideration of public health
implications, is provided in the Policy Assessment.
7-7
-------
Table 7-2. Characterization of key uncertainties in the assessment.
Sources of Uncertainty
Category
Ambient Monitoring
Concentrations
Adjustment of Air
Quality to Simulate Just
Meeting the Current and
Potential Alternative
Standards
APEX Input Data and
Algorithms
Potential Health Effect
Benchmark Levels
Element
Database Quality
Missing Data Substitution Method
Zero Concentration Frequency
Temporal Representation
Spatial Representation
Historical Data Used
Proportional Approach Used
Population Database
Activity Pattern Database
Longitudinal Profile Algorithm
Meteorological Data
Microenvironmental Algorithm and
Input Data
Commuting Algorithm
At-Risk Prevalence Rates
Physiological Variables
Simulated At-Risk Populations a
Influence of Uncertainty
on Exposure/Dose
Estimates
Direction
Over
Under
Under
Both
Both
Unknown
Both
Both
Unknown
Both
Both
Unknown
Both
Both
Unknown
Unknown
Magnitude
Low
Low
Low
Low
Low- Medium
Low- Medium
Low
Low
Low- Medium
Low- Medium
Low
Medium
Low
Low
Low- Medium
Low
Knowledge-base
Uncertainty
Low
Low
Low
Medium
Medium
Medium
Low
Low
Medium
Medium
Low
Medium
Low
Medium
Medium
Medium
Notes:
a This entry focuses on the uncertainty associated with the benchmark levels in their application to estimated
COHb levels for the simulated at-risk populations (i.e., individuals with diagnosed HD or CHD combined with
an estimate of undiagnosed CHD). With regard to other potentially susceptible populations (as described in
section 2.4 above), we additionally note the lack of studies that describe COHb levels and health effects that
might be expected as a result of short-term elevations in CO exposure in those populations.
7-S
-------
7.2.1 Considerations in Characterizing Sources of Uncertainty
Staff considerations in reaching the judgments summarized in Table 7-2 above are noted
below in order of the presentation in Table 7-2.
7.2.1.1 Ambient Monitoring Concentrations
Five elements of uncertainty were identified regarding the ambient monitoring
concentrations used to estimate at-risk population exposures and COHb levels. These include
the following elements: database quality, the method used to substitute for missing ambient
concentrations, use of reported zero ambient concentrations, and the temporal and spatial
representation of the monitors used as input to the exposure model.
Database Quality
All ambient CO measurements available in AQS are quality-assured. There may be a
limited number of poor-quality, high concentration data within the ambient concentration data
sets, potentially influencing the number of benchmark dose-level exceedances. Note also that
any uncertainty regarding low level concentrations at or near the monitoring detection limit is
unlikely to influence high COHb levels, the levels of which are of greatest interest in this
assessment. Based on this, we judge there to be potential for overestimation in the number and
percent of persons at or above a given COHb level, though the magnitude of this potential
overestimation would be low. The source of ambient monitoring data used in the analyses,
EPA's Air Quality System, is of high quality. There is no other source of ambient monitoring
data as comprehensive. In addition, the data are being used in a manner consistent with one of
the defined objectives of ambient monitoring. Therefore, we judge uncertainty associated with
the knowledge base as low.
Missing Data Substitution Method
There were very few missing hourly concentration values when considering the years of
ambient monitoring data used in this assessment. This is because we first screened all available
monitoring data for a minimum of 75% completeness regarding hours/day and days/year for all
four quarters within each year. In meeting the completeness criteria, they are (by definition),
valid and appropriate for the purposes of this assessment. In Denver, the amount of missing
hourly data ranged from a low of 0.7% to a high of 2.3% of the potentially reportable
concentrations for a year (mean = 1.3%) (Tables 5-7 and 5-8). There was a greater percentage of
missing values in the Los Angeles ambient monitoring data set, with an average of about 5.4% of
the possible reportable concentrations for a year (minimum = 4.4%; maximum = 8.4%) (Tables
5-9 and 5-10). Therefore, both sets of data were well within the bounds set by the completeness
criteria.
7-9
-------
The method used to substitute for missing concentrations is informed by the
measurements available within and among each of the monitors. A variety of standard
techniques were used to fill missing values depending on the nature of the data gaps, including
the development of linear regression models and interpolation between two points, with each
method considering important factors that may influence concentration such as time-of-day, day-
of-week, or month-of-year. While a number of alternative methods might be available, it is
likely that the distribution of estimated concentrations would be similar to those generated here
given that most of these potential alternative data substitution methods are also informed by the
existing measurement data. Note also that there were negligible differences between the air
quality distributions when comparing the before-substitution and after-substitution data sets
(section 5.7.2).
Assuming there is an equal probability of missing either low or high concentration hourly
values, and that substituted data are limited by the bounds of the data substitution algorithm (i.e.,
as defined by limits in the measurement data), there still may be a few instances where missing
high concentration data would not be appropriately estimated. If this were the case, the selected
substitution method would lead to an underestimation in exposure concentrations and COHb
levels experienced by the simulated at-risk populations. This also assumes that the substitution
of missing low-level concentration data with potentially higher concentrations (though still
within the bounds of the algorithm) does not affect exposure and COHb results of interest. Very
few data values were substituted with respect to the number of measured values available in each
location.
Zero Concentration Frequency
The ambient monitoring data contain reported values equivalent to zero ppm, indicating
that the monitor was in operation, though concentrations were below a quantitative detection
limit. The minimum reported concentration in both study areas was 0.1 ppm, though reported
detection limits for many of the instruments are typically 0.5 ppm (section 3.1.2). This indicates
that reported concentrations between 0.1 and 0.5 ppm were quantified, though likely having
greater uncertainty in their assigned values compared with other reported concentrations > 0.5
ppm. There is a possibility that exposures and associated COHb levels are underestimated
because a reported zero value may represent a non-zero concentration less than 0.1 ppm.
Staff elected to use the reported monitoring data as a zero concentration rather than
substitute the zero concentrations with some other value. In Denver, there were very few
instances where a zero concentration was reported for either monitoring year of data (Table 7-3).
Given the limited occurrence of zero concentrations in Denver, it is highly unlikely that a data
substitution method employed to assign non-zero concentrations of less than 0.1 ppm would
change the number or percent of persons exposed or experiencing %COHb above levels of
7-10
-------
interest in this REA. There were, however, a greater number of instances where a zero ppm
monitoring concentration was reported in Los Angeles (Table 7-3). Staff judge that it would also
be of little consequence if all of these values were substituted with a non-zero concentration less
than or equal to 0.1 ppm, given that the COHb levels of interest are driven exclusively by upper
percentile ambient and microenvironmental exposure concentrations (For example, see Appendix
B.6).
Table 7-3. Frequency of CO concentrations reported as zero in Denver and Los Angeles
ambient monitoring data.
Monitor ID
0310002
0310013
0310014
0590002
Monitor ID
0370113
0371002
0371103
0371201
0371301
0372005
0374002
0590001
0591003
0595001
Frequency of Reported Zero CO Concentrations
Denver Ambient Monitoring Data
1995
n
3
0
1
0
%
0.0
0.0
0.0
0.0
2006
n
3
154
139
143
%
0.0
1.8
1.6
1.6
Los Angeles Monitoring Data
1997
n
980
81
126
393
74
430
206
7
908
61
%
11.2
0.9
1.4
4.5
0.8
4.9
2.4
0.1
10.4
0.7
2006
n
6
95
236
3
1
307
41
290
1288
2
%
0.1
1.1
2.7
0.0
0.0
3.5
0.5
3.3
14.7
0.0
Temporal Representation
As described above (section 7.2.1.2), staff used hourly ambient monitoring data that met
completeness criteria largely defined by temporal attributes (hours per day, days per year). The
hourly ambient CO concentrations are used to estimate event-level exposure concentrations that
can range from 1 minute to 1 hour in duration. The relationships used to estimate these exposure
concentrations are commonly derived from hourly measurements, thus are generally consistent
with how the ambient monitor data are being used in this REA. In addition, the
microenvironmental algorithm employed a factor that adjusts for temporal differences in outdoor
concentrations with respect to that measured at an ambient monitor (section 4.4.4.3). Therefore,
7-11
-------
staff judges the hourly ambient monitoring data of appropriate time resolution when modeling an
individual's maximum end-of-hour COHb level. Further, given the observed individual doses in
response to highly variable ambient/exposure concentrations (see Figure B-2 in Appendix B.6),
staff judges that there would likely be low impact to the estimated percent of persons
experiencing elevated COFtb levels with improved temporal representation (e.g., minute-by-
minute concentrations). Staff judges that there is a medium level of uncertainty in the
knowledge base regarding the temporal representation. This is because much of the data used to
derive the temporal adjustment factors are based on ambient, microenvironmental, and personal
exposure measurements conducted in the 1980's, although it is likely that while CO
concentrations have changed dramatically, the relationships among the measurements remain
constant.
Spatial Representation
In evaluating the uncertainty associated with representation of the spatial variability in
ambient CO concentrations in this REA, we have considered the impact of the improvements
made in representing spatial variability in ambient CO levels throughout each study area for this
REA as compared with the simplified approach used in the first draft CO REA.
Before considering the impact improving spatial representation of monitoring
concentrations had on dose estimates, we first considered the spatial variability in the air quality
data used for the study areas. Analysis of the full set of monitoring data indicates spatial
variability in monitoring concentrations across each area is relatively limited, particularly when
considering more recent years (Tables 3-1 through 3-6). This could indicate that ambient
monitored concentrations do not vary greatly across the study area or that the study area
monitors are sited in locations that measure similar (though still temporally variable) CO
concentration levels. Note also that the microenvironmental algorithm we used in this
assessment to estimate outdoor microenvironmental concentrations has an adjustment factor to
address spatial variability in ambient concentrations expected to occur when modeling human
exposures (section 4.4.4.3). Based on this limited analysis of air quality data from the existing
ambient monitoring network and observed relationships between monitor and outdoor
concentrations, it is possible that there would be a low magnitude of influence on estimated
COHb levels in the presence of an alternative or supplemental monitoring network.
Staff also considered the impact on dose estimates of using the full monitoring data set to
represent the spatial variability in each study area, as compared to a more limited approach such
as what was done for the first draft REA, in which a single high concentration ambient monitor
was used to represent the ambient concentrations across a larger study area (see section 7.2.2.1).
Based on these sensitivity results, staff judged improving spatial representativeness of the
monitoring network to potentially have a medium magnitude of impact on the estimated
7-12
-------
exposures and COHb levels of interest if the existing monitoring network does not adequately
represent high ambient concentrations to which people might be exposed. We note that the
limited, single monitor, approach described here assigned the highest monitor to all monitor
locations, which illustrated a potential impact associated with representation of higher ambient
concentrations. Conversely, a similar hypothetical scenario could be constructed to investigate
the potential impact associated with representation of lower ambient concentrations. Such a
comparison might find a similar, medium, magnitude of influence to the estimated number and
percent of persons at or above COHb levels (albeit lower COHb levels experienced by the at-risk
population compared with simulations that employed all monitors).
In considering uncertainty in the knowledge base, we note that each ambient monitor
comprising the existing monitoring network has specific objectives and monitoring scale that
may not appropriately capture the true spatial variability in CO concentrations. In the absence of
1) a monitoring network designed to better measure spatial variability in CO concentrations, 2)
performing air quality modeling to estimate fine scale spatial and temporal variability in CO
concentrations, and 3) analysis of additional monitoring data that can potentially indicate spatial
concentration gradients, staff judge the uncertainty in the knowledge base as medium.
7.2.1.2 Adjustment of Air Quality to Simulate Just Meeting the Current and
Potential Alternative Standards
Two elements of uncertainty were identified regarding the ambient monitoring
concentrations used to estimate at-risk population exposures and COHb levels. These include
the use of historical data to represent the hypothetical air quality scenarios (with adjustment) and
the proportional approach that was used for adjustment.
Historical Data Used
Even though the historical data represent air quality conditions that have existed in the
past, the conditions simulated using these data are hypothetical scenarios. There is uncertainty in
how the temporal and spatial distribution of CO concentrations represented by these historical
data might reflect the scenarios being simulated with these data given the air quality conditions
that affect ambient concentration levels. More specifically, there is uncertainty regarding how
influential factors such as emission levels per vehicle, vehicular traffic, and meteorology
associated with the historical monitoring data set might influence air quality conditions in future
situations which the simulated scenarios have been designed to represent.
We have noted differences between the two study areas in the percent of simulated
populations estimated to experience a daily maximum end-of-hour COHb level at or above 2.0%
COHb level when considering conditions just meeting the current standard (e.g., Tables 6-18 and
6-19). In these simulations involving adjusted historical hourly concentrations, there are a
7-13
-------
greater number of high 1-hour concentrations (e.g., above 8 ppm) in Denver (Tables 5-18 and 5-
19) as compared to the Los Angeles study area. It is evident that ambient concentrations at the
upper percentiles of the distribution likely have a strong influence on the number and percent of
persons experiencing the higher COHb levels. Whether these conditions that existed at the time
the Denver data were collected are appropriate to the modeled hypothetical scenario is largely
unknown. However, based on observed trends in air quality over time (Figures 3-4 and 3-5) and
the results generated using the adjusted ambient concentrations, staff judges that, at most, the
magnitude of potential influence to the estimated COHb levels in using these historical data
could be a medium level. It is possible that these historical patterns can serve as a reasonable
basis for predicting future air quality scenarios, though these patterns would not account for the
influence of a new CO emission source(s). Therefore, staff judges the magnitude of the
knowledge-base uncertainty as medium.
Proportional Approach Used
The magnitude of the adjustment applied to historical ambient concentration data was
wide ranging across the air quality scenarios. For example, in Denver, to simulate conditions
just meeting the current standard, 0.989 was the adjustment applied to the 1995 ambient
monitoring data. In comparison, in adjusting the 1997 ambient monitoring data to just meet a 2nd
highest 8-hour average CO concentration of 9.4 in Los Angeles, a greater adjustment to the data
was needed (i.e., a factor of 0.627). However, in comparing recent and historical ambient CO
concentrations for several ambient monitors in Los Angeles (Figure 3-4) and Denver (Figure 3-
5), a strong proportional relationship is present when comparing the recent and historic CO
concentrations. In general, the regression slopes and intercepts were similar for each of the
monitors used to represent the air quality within each study area, indicating a similarity in the
rate of change in concentration occurring at the monitors. This finding suggests there are likely
similar sources affecting each of the monitors and that their associated source emissions have
also changed at a similar rate over time.
The use of a proportional approach to simulate alternative air quality scenarios is not
uncommon. A similar proportional adjustment approach was judged adequate in simulating air
quality conditions just meeting the 8-hour CO NAAQS in prior assessments (US EPA, 1992;
Johnson et al., 2000). A proportional approach was also used in evaluating exposure scenarios
associated with just meeting the current and several alternative standards in the most recent NO2
and SO2 NAAQS reviews (US EPA, 2008; US EPA, 2009).
In addition, the simulations involving Los Angeles data indicate little difference in the
percentage of persons estimated to experience daily maximum end-of-hour COHb at or above
selected COHb levels between simulations using the historical data adjusted downwards to
conditions similar to as is air quality and simulations using the 2006 as is air quality. This is
7-14
-------
shown in Table 7-4 where COHb population estimates are presented for as is conditions (denoted
by the asterisk) and contrasted to results for simulations of potential alternative standards that
result in similar air quality conditions. As indicated by the presentation in Table 7-4, the as is
conditions in Los Angeles can be considered to just meet several potential alternative standard of
different form and level. For example, the 2nd highest non-overlapping 8-hour concentration at
the design monitor in Los Angeles was equal to 5.6 ppm (i.e., consistent with air quality just
meeting an 8-hour standard with a form of 2nd highest and level of 5.6; see Table 5-12). Table 7-
4 indicates, based on this Los Angeles analysis, that whether using as is air quality, or historical
air quality adjusted to a level similar to as is air quality, the estimated percent of persons at or
above selected COHb levels are similar, thus indicating the lack of a strong or unrealistic
influence of the air quality adjustment procedure on the results.
Table 7-4. Percentage of simulated at-risk CHD population in Los Angeles with highest
daily maximum end-of-hour COHb levels at or above indicated COHb level
considering potential alternative standards.
Form
Second Highest Non-
overlapping 8-hour
Concentration
99th Percentile of 8-
hour Daily Maximum
Concentration
Second Highest 1-hour
concentration
99th percentile of 1-
hour Daily Maximum
Concentration
Level
(ppm)
5.7
5.6*
5.4
5.7
5.1*
5.0
8.2*
8.1
8.0
7.4*
7.1
> 2.0 % COHb
<0.1
<0.1
<0.1
0.1
<0.1
<0.1
<0.1
0.1
0.1
<0.1
<0.1
> 1.75% COHb
0.3
0.5
0.2
0.4
0.5
0.3
0.5
0.4
0.4
0.5
0.3
> 1.5% COHb
0.9
1.6
0.7
1.4
1.6
0.9
1.6
1.4
1.4
1.6
0.9
Notes:
* Asterisk indicates simulation used as is (2006) air quality.
7.2.1.3 APEX Input Data and Algorithms
Eight elements of uncertainty were identified regarding the APEX input data and
algorithms used to simulate population activity. These include the population and activity
pattern databases, the longitudinal profile algorithm, meteorological data, microenvironmental
algorithm and input data, commuting algorithm, at-risk prevalence rates and physiological
variables.
7-15
-------
Population Database
Population data (tract population density, age/gender distributions) are from the US
Census Bureau, a reliable and quality-assured source. Data used are specifically for census tracts
modeled in Denver and Los Angeles. Staff assumed any remaining uncertainties in the database
would have negligible influence on exposure and %COHb results.
Activity Pattern Database
Data are actual records of the time spent in specific locations while performing specific
activities. While not specific to a particular study area, the activity patterns of a population are
generally well represented by the mainly population-based and nationally-representative survey
data. There is, however, uncertainty in how well the CHAD data represent the intended
simulated at-risk population, given that there may be local geographic attributes that influence a
person's exposure that are not accounted for by CHAD. For example, in each study area it was
observed that the in-vehicle microenvironment contributed greatly to concentrations at or above
3-6 ppm (Figures 6-1 through 6-4). This indicates that in-vehicle exposures are likely important
determinants in upper level exposures and, thus, the distribution of time spent commuting may
be an important influential factor.
To evaluate how well the CHAD data represented persons residing within each study
area regarding commute times, we obtained information on travel time to work for workers ages
16 years2 and over specific to Denver County, Colorado and Los Angeles, CA (US Census
Bureau, 2009, Table P31). We next isolated persons in CHAD that were >18 years of age and
spent at least one minute in a motor vehicle between 6 am and 9 am. The distributions of
commute times associated with these data are illustrated in Figure 7-1 (see Appendix E, Table E-
1 for the details). This comparison indicates that the available CHAD diaries3 reasonably
represent typical commute times in both the urban locations modeled in this assessment. Not
surprisingly, the percentage of CHAD diaries containing longer duration commute times (> 40
minutes) are more representative of Los Angeles than Denver, likely a function of the number of
CHAD diaries from that geographic area (Table 4-1).
This is how these data are reported by the US Census. We assumed that the distribution of commute
times for persons aged 18 years or older in Los Angeles and Denver counties would be similar to this.
3 Note, however, that this CHAD distribution is developed from an un-weighted sample of all possible
diaries that could be used in each study area. The actual diaries used, their frequency of use, and associated
distribution of commute times may be different from that presented in Figure 7-1. To obtain this information is not
a trivial computational undertaking and would involve generating the necessary data in the APEX daily output file
(which was not done here).
7-16
-------
• Denver County (US Census, 2009)
D LA County (US Census, 2009)
• CHAD (US EPA, 2002)
1to9 10 to 19 20 to 29 30 to 39 40 to 59 60 to 89
Commute or Travel Time (minutes)
90+
Figure 7-1. Comparison of commute and travel times for persons residing in Denver and
Los Angeles counties to those persons surveyed in CHAD.
We also recognize that the health status of an individual may affect their time-location-
activity patterns, and hence their estimated exposure and dose levels. This could include factors
such as time spent in certain locations as well as exertion levels attained by simulated
individuals. CHAD is comprised of data from individuals that have, or do not have, an identified
health condition4 and are assumed to represent the activities of persons with normal health status
as well as those with certain health conditions that may not significantly affect general activity
patterns. While some of the diary days are potentially reflective of the simulated at-risk
population, the majority of the diaries may be from persons identified as having a normal health
status. Thus, if there are differences in activities performed and locations visited that depend on
health status, there would be uncertainty in the representation of the simulated at-risk population
activity patterns modeled in this CO REA by the CHAD diaries.
A statistical analysis was performed on a subset of the CHAD data where persons were
specifically asked whether they had angina (see Appendix F of CO REA and Johnson et al.,
2000, for details). Activity patterns for persons with angina were compared to those for
4 Of the approximately 35,000 diary days within the CHAD master file, about 65% contain information on
whether the person was identified as asthmatic (or not) or whether the person was identified as having a heart or
lung condition (or not). The remaining 35% have unknown health status (i.e., the health status was not requested in
the original survey questionnaire).
7-17
-------
individuals not having angina using selected exertion level metrics and time spent outdoors or
inside vehicles. The percentages of time spent outdoors or in a vehicle were generally not
statistically significantly different between angina and non-angina subjects. While there were
statistically significant differences in the exertion level attained between angina and non-angina
subjects,5 actual differences were generally numerically small compared to the mean values.
The differences in activity and exertion level between angina and non-angina subjects, although
statistically significant, were judged not large enough to severely impact the validity of APEX
(or pNEM/CO) modeling results that do not adjust for an angina/non-angina difference in
activity patterns.
In characterizing the knowledge base uncertainty, we note that the CHAD time-location
activity diaries used are the most comprehensive source of such data and realistically represent
where individuals are located and what they are doing. Features of an individual's activity
pattern are well represented, adjustments are made to represent the population distribution in a
specific area (using age and gender), and temperature is used to link CHAD diaries with the
simulated individuals residing in a specific area. In addition, the CHAD data are from a reliable
and quality-assured source (US EPA, 2002) and have recently incorporated several thousand new
diary days (see Table 4-1). However, we judge the knowledge-base uncertainty as medium
given the overall limited number of diary days available to represent the large simulated at-risk
populations residing within each study area.
Longitudinal Profile Algorithm
Much of this assessment focused on persons having at least one exposure or dose above a
selected level; therefore, the potential magnitude of influence from the longitudinal profile
algorithm would be considered low. However, when considering multi-day exposures, staff
judges the magnitude of potential influence to COHb levels could range upwards to a medium
level. In developing the longitudinal method, the evaluation indicated that both the D and A
statistics are reasonably reproduced for the population (Glen et al., 2008). The approach was
also compared with two other independent methods used for constructing longitudinal activity
patterns (see Appendix B, Attachment 4 of US EPA, 2009). This particular comparison
indicated that, depending on the longitudinal profile method selected, the number of persons
experiencing multiple exposure events at or above a selected level could differ by about 15 to
50%.6 Note however, long-term diary profiles (i.e., monthly, annual) do not exist for a
5 Note there was a very large sample size for the non-angina subjects.
6 The comparison used simulated persons in Atlanta and evaluated the number of persons experiencing
three or more ozone 8-hour average exposures at or above 0.07 ppm concomitant with moderate or greater exertion.
7-18
-------
population, though a few persons surveyed have longitudinal diary profiles (Table 4-1).
Therefore, the knowledge-base uncertainty is judged by staff as medium.
Meteorological Data
Data are from the National Weather Service, a well-known and quality-assured source.
The daily maximum temperatures are used when selecting appropriate CHAD diaries to simulate
the at-risk population. The temperature bin ranges that were used (see section 5.5.3) are wide
such that any erroneous temperatures that do exist within the data set would likely have limited
impact to estimated exposure and dose results. Daily mean temperatures are used when selecting
air exchange rates (section 5.9.4). Given the overlap of the AER distributions and the wide
temperature ranges used to categorize them, it is also likely that there is limited impact by
erroneous temperature data that may exist. Therefore, staff judges uncertainty regarding
meteorological data and its potential magnitude of influence on COHb levels as low.
Microenvironmental Algorithm and Input Data
In this REA, the number of microenvironments selected captures the likely locations
persons spend time and where CO exposures would occur. Using distributions of proximity
factors derived from measurement data in Denver and applied to estimate microenvironmental
concentrations is judged by staff to be a reasonable approach. However, the extent to which
these Denver study data reflect similar relationships in Los Angeles likely has greater
uncertainty. Additionally, the Denver measurement data were collected in the 1980's; therefore,
there is also uncertainty as to how these data might reflect relationships observed for other years
modeled in this assessment (i.e., 2006). However, the distributions of microenvironmental
concentrations (Tables 6-8 and 6-11) and the effective microenvironmental-to-ambient
concentration ratios (e.g., Tables 6-9 and 6-12), in particular those used to estimate high-
exposure microenvironments, were found comparable to other measurement data and
relationships available (albeit limited in number) and generally support the algorithm and
distributions applied in this assessment.
Commuting Algorithm
In this REA, the commuting algorithm within APEX was implemented. Use of this
algorithm better represents individual exposures across each modeling domain. The data are
derived from the US Census, a well-known and quality-assured source. The data are used in
addressing home-to-work travel, certainly within the bounds of the objectives associated with the
original data collection; therefore, staff judged the knowledge-base uncertainty as low.
While there may be some uncertainties associated with the application of the database
(see US EPA, 2009), they are limited in the potential magnitude of influence they might have on
estimated exposures and COHb levels. For example, although several of the APEX
microenvironments account for time spent in travel, the travel is assumed to always occur in
7-19
-------
basically a composite of the home- and work-tracts. No other provision is made for the
possibility of passing through other census tracts during travel. This could contribute to either
over- or under-estimating exposure concentrations, dependent on the number and identity of
tracts the simulated individual would actually traverse and the spatial variability of the
concentration across different tracts. Given that most persons would likely experience ambient
concentrations from within their home air district (i.e., encompassing all tracts within a 10 km
radius of the monitor location), tracts existing between home and work tracts for these persons
would not have a different assigned ambient concentration. In addition, the commuting route
(i.e., which roads individuals are traveling on during the commute) is not accounted for. From a
practical perspective though, if staff was to consider multi-tract commuting, further complexity
would need to be added to the modeling while also requiring additional input data that are not
readily available (e.g., commuting route data for simulated individuals). These model
adjustments would come with a number of additional uncertainties and would require additional
time and resources not available for the assessment.
At-risk Prevalence Rates
Data are from the Centers for Disease Control, a well-known and quality-assured source.
Though prevalence data are not specific for each region, the national prevalence data were
stratified by selected age-groups and gender. Staff used gender-specific ratios and applied them
to all age groups uniformly even though there may be uncertainty in the accuracy of the
prevalence estimates for specific age and gender groups. In addition, potentially undiagnosed
individuals with CHD were included to expand the total CHD population considered. This was
based on several assumptions including using 1990 estimates of the population with undiagnosed
IHD. The percent of the population with undiagnosed CHD (i.e., 43.8%) was applied to the
diagnosed CHD prevalence, without any difference between genders. In comments on the
second draft REA, it was suggested that women with undiagnosed heart disease may be
underrepresented by this approach. In the limited time available, extant literature regarding this
topic were reviewed with regard to the suitability for developing an improved estimate of the
undiagnosed prevalence rate. Relevant data that could be used to generate new undiagnosed
prevalence rates and provide a greater degree of confidence than what was used in this
assessment were not identified.
We recognize there is uncertainty associated with the undiagnosed prevalence rate used.
To evaluate the potential impact the prevalence rate might have on the estimated COHb levels,
we performed simulations using alternative values for undiagnosed CHD prevalence rates.
These analyses are described in section 7.2.2.2 below. The COHb results generated from this
sensitivity analysis, as well as when using different prevalence rates to estimate COHb levels for
the HD population, had little impact to the percent of persons at or above the selected COHb
7-20
-------
levels. Only the total number of persons was affected by using the alternative prevalence rates.
This suggests that, if new data were uncovered with improved representation of either the
diagnosed or undiagnosed prevalence rates, there would be little change to the percent of persons
at or above selected COHb levels. Therefore, staff judges the magnitude of influence to COHb
as low for the estimated percent of persons at or above selected levels. The prevalence rates do
have a medium level of influence to the estimated number of persons at or above selected COHb
levels, though the observed impact was proportional across the range of selected COHb levels.
Physiological Variables
Many of the parameters used to estimate the physiological attributes of the simulated at-
risk population were developed from healthy individuals; there were no adjustments made to
account for a particular health condition. While the ISA notes some variability in some
parameters in individuals with specific health conditions that might affect CO uptake and
elimination, most of the identified health conditions that could affect the physiological variables
used in the CFK model may not necessarily be associated with the simulated at-risk populations,
i.e., HD or CHD individuals. In addition, there is uncertainty in some of the parameter values
used in the COHb algorithm due to the age of source publications cited (e.g., dating back to the
mid 20th century) or uncertainty as to their representativeness. As an example, alveolar
ventilation is represented in the simulations as a single point estimate of 19.63 and applied
directly to activity-specific oxygen consumption rates. This was based on an analysis by
Joumard et al. (1981), of data generated by Galetti (1959), which did not include measurements
made at elevated exertion levels, although theoretically one would expect there to be a nonlinear
relationship between VA and VO2 given the non-linear relationship of dead space volume (Vo) to
tidal volume (Vx) with increasing breathing rate.7 Thus, the point estimate of 19.63 used may
not adequately represent the VA to VO2 relationship at higher ventilation rates. We note however
that most of the upper level exposure concentrations in this assessment are associated with time
spent inside vehicles, where it is expected that the exertion level and breathing rate would be at a
relatively low level. Therefore, it may be that the point estimate is appropriately used for these
activities and the estimated maximum end-of-hour COHb associated with the in-vehicle
microenvironment may not be adversely affected. As described in this example, it is possible
that most of the data and or relationships still remain appropriate in modeling the current
population; however, in the absence of conducting a comprehensive review and comparing the
historical data to recent measurements, staff judges the knowledge base uncertainty as medium.
7 Dead space volume (VD) will remain relatively constant, increasing only slightly with increasing
ventilation rate. Tidal volume (VT) consistently increases with increasing ventilation rate. Alveolar ventilation (VA)
will approach that of total ventilation (VE) at higher ventilation rates, given that VA = VE (1-VD/VT). Note also, VE
is not linear with respect to VO2 (see Graham and McCurdy, 2005).
7-21
-------
The potential influence of another physiological parameter - hemoglobin content - was
evaluated in response to comments received on earlier drafts of this document. These analyses
are described in section 7.2.2.3 below.
7.2.1.4 Potential Health Effect Benchmark Levels for the Simulated At-risk
Populations
The potential health effect benchmark levels for considering the COHb estimates for the
simulated at-risk populations8 in this REA were identified (in section 2.6) based on data from a
well-conducted multi-center controlled human exposure study demonstrate cardiovascular effects
in subjects with moderate to severe coronary artery disease at study mean COHb levels as low as
2.0-2.4% of which were increased from a baseline mean of 0.6-0.7% as a result of short (~lhour)
experimentally controlled increases in CO exposures (study mean of 117 ppm CO). No
laboratory study has been specifically designed to evaluate the effect of experimentally increased
exposure to CO resulting in an increase in COHb levels to a study mean below 2.0%. However,
based on analysis of individual study subject responses at baseline and at the two increased
COHb levels, study authors concluded that each increase in COHb produced further changes in
the study response metric, without evidence of a measurable threshold effect. There is no
established "no adverse effect level" and, thus, there is greater uncertainty concerning the lowest
benchmark level identified (i.e., 1.5%). There is also uncertainty about whether individuals with
the most severe CHD are adequately represented. Additionally the COHb levels estimated in
this assessment result from CO exposure concentrations much lower than the experimental
exposure concentrations used to increase study subject COHb levels to the study targets (e.g.,
2.0%) and with which the responses were associated. Given that the evidence supporting the
choice of benchmark levels is based on controlled human exposure data, staff judged the
influence of this uncertainty on the risk characterization as being low.
7.2.2 Sensitivity Analyses
This section describes sensitivity analyses referenced in section 7.2.1 with regard to three
sources of uncertainty in the assessment, one in the category of ambient monitoring
concentrations and two in the category of APEX input data and algorithms.
8 Discussion here focuses on the uncertainty associated with benchmark levels in their application to
estimated COHb levels for the simulated at-risk populations (i.e., individuals with diagnosed HD or CHD combined
with an estimate of undiagnosed CHD). With regard to other potentially susceptible populations (as described in
section 2.4 above), we additionally note the lack of studies that describe COHb levels and health effects that might
be expected as a result of short-term elevations in CO exposure in those populations.
7-22
-------
7.2.2.1 Spatial Representation
We performed sensitivity analyses to illustrate the quantitative impact of the
improvements made by better representing the spatial variability in ambient CO concentrations.
Of particular interest was how expanding the number of monitors used in this REA for both
study areas compared with the simplified approach used in the first draft CO REA. In the
analysis shown here, however, the more spatially limited approach applied all hourly
concentrations from the single design monitor to all monitoring locations defined for each study
area to ensure comparable modeling domains.9 Four model simulations for the HD population
are shown: two air quality conditions (as is and just meeting the current 8-hour standard) in each
of the two study areas (Denver and Los Angeles). The results for the more spatially-limited
approach (single monitor) might be considered to provide an upper bound estimate of COHb
levels for this modeling domain, given that the design monitor represents the highest measured
concentrations in each study area and that there are restricted opportunities for persons to
experience ambient CO concentrations lower than that of the design monitor.
As shown in Tables 7-5 through 7-8, the use of multiple monitors (versus using the
design monitor alone) to represent the air quality input to the exposure model results in a lower
number and percent of persons at or above a given COHb level. In general, there were larger
differences in the number and percent of persons at or above the higher COHb levels (i.e., >
1.75% COHb). A similar pattern was observed when comparing air quality scenarios within
each study area, although expanding spatial variability in ambient monitoring concentrations had
a greater impact on the number of persons at or above selected COHb levels in the Los Angeles
study area (Tables 7-6 and 7-8). This is likely the result of the Los Angeles study area, which is
larger, having generally greater spatial variability in ambient monitoring concentrations (i.e.,
from ten monitors) when compared with that provided by the four monitors representing air
quality in the Denver study area. The spatial heterogeneity in ambient concentrations in the Los
Angeles study area allowed simulated persons to experience a wider range of ambient
concentrations when compared with the simulated persons Denver. As a result of the limited
spatial heterogeneity in Denver ambient concentrations (given that only four monitors were
selected, their close proximity to one another, and having similar concentration levels), a lesser
difference in the percent of persons at or above selected COHb levels was observed when using
the design monitor to represent all ambient concentrations in the model domain. This difference
9 In the first draft CO REA, the design monitor in each study area was used to represent air quality for all
census tracts within a 20 km radius of the monitor. In this current investigation, the 10 km radii for the 4 monitor
locations in Denver, and 10 monitor locations in Los Angeles were retained to define the respective exposure
modeling domains. All of these monitoring locations (and hence the air quality districts) used the CO
concentrations from the single design monitor for each respective study area (080310002 in Denver and 060371301
in Los Angeles).
7-23
-------
in spatial variability between the two study areas may also play a role in the study area
differences observed for the estimated percent of persons at or above the selected COHb levels
when using air quality just meeting the current standard (e.g., Table 6-18 and 6-19). This is
because a greater proportion of the simulated persons will experience generally similar exposure
levels in Denver (and at higher exposure concentrations, note Tables 6-7 and 6-9) than compared
with simulated persons in Los Angeles (Tables 6-10 and 6-12).
Table 7-5. Comparison of highest estimated daily maximum end-of-hour COHb levels for
Denver HD population for two model simulations - all monitor concentrations
versus the design monitor concentrations - as is air quality.
COHB
Level
(%)
>0.0
> 1.0
>1.5
>1.75
>2.0
>2.5
>3.0
>4.0
Number of HD Persons
All Monitors3
85,926
17,807
1,074
234
12
0
0
0
Design Monitor
only"
85,926
25,125
1,715
444
99
0
0
0
Percent of HD Persons
All Monitors3
100
20.7
1.2
0.3
<0.1
0
0
0
Design Monitor
only"
100
29.2
2.0
0.5
0.1
0
0
0
Ratio0
1.0
1.4
1.6
1.9
8.3
-
-
-
Notes:
3 Each monitor site used ambient concentrations from that site (dose results from Table 6-15).
b Each monitor site used ambient concentrations from design site (080310002).
c Ratio = (value for design site scenario) / (value for all sites scenario).
Table 7-6. Comparison of highest estimated daily maximum end-of-hour COHb levels for
Los Angeles HD population for two model simulations - all monitor
concentrations versus the design monitor concentrations - as is air quality.
COHB
Level
(%)
>0.0
> 1.0
>1.5
>1.75
>2.0
>2.5
>3.0
>4.0
Number of HD Persons
All Monitors3
630,807
165,880
9,834
3,011
502
0
0
0
Design Monitor
only"
630,807
406,020
61,816
22,378
7,526
401
0
0
Percent of HD Persons
All Monitors3
100
26.3
1.6
0.5
<0.1
0
0
0
Design Monitor
only"
100
64.4
9.8
3.5
1.2
<0.1
0
0
Ratio0
1.0
2.4
6.3
7.4
15.0
-
-
-
Notes:
3 Each monitor site used ambient concentrations from that site (dose results from Table 6-16).
b Each monitor site used ambient concentrations from design site (060371301).
c Ratio = (value for design site scenario) / (value for all sites scenario).
7-24
-------
Table 7-7. Comparison of highest estimated daily maximum end-of-hour COHb levels for
Denver HD population for two model simulations - all monitor concentrations
versus the design monitor concentrations - air quality just meeting the current
8-hour standard.
COHB
Level
(%)
>0.0
> 1.0
>1.5
>1.75
>2.0
>2.5
>3.0
>4.0
Number of HD Persons
All Monitors3
85,926
71,710
21,028
9,502
3,826
802
222
25
Design Monitor
only"
85,926
81,743
38,206
19,276
9,329
2,295
592
74
Percent of HD Persons
All Monitors3
100
83.5
24.5
11.1
4.5
0.9
0.3
<0.1
Design Monitor
only"
100
95.1
44.5
22.4
10.9
2.7
0.7
<0.1
Ratio0
1.0
1.1
1.8
2.0
2.4
2.9
2.7
3.0
Notes:
3 Each monitor site used ambient concentrations from that site (dose results from Table 6-18).
b Each monitor site used ambient concentrations from design site (080310002).
c Ratio = (value for design site scenario) / (value for all sites scenario).
Table 7-8. Comparison of highest estimated daily maximum end-of-hour COHb levels for
Los Angeles HD population for two model simulations - all monitor
concentrations versus the design monitor concentrations - air quality just
meeting the current 8-hour standard.
COHB
Level
(%)
>0.0
> 1.0
> 1.5
>1.75
>2.0
>2.5
>3.0
>4.0
Number of HD Persons
All Monitors3
630,807
260,612
31,410
10,537
3,613
401
100
0
Design Monitor13
630,807
558,955
168,289
68,941
29,302
4,817
602
0
Percent of HD Persons
All Monitors3
100
41.3
5.0
1.7
0.6
<0.1
<0.1
0
Design Monitor13
100
88.6
26.7
10.9
4.6
0.8
<0.1
0
Ratio0
1.0
2.1
5.4
6.5
8.1
12.0
6.0
-
Notes:
a Each monitor site used ambient concentrations from that site (dose results from Table 6-19).
b Each monitor site used ambient concentrations from design site (060371301).
c Ratio = (value for design site scenario) / (value for all sites scenario).
7-25
-------
7.2.2.2 At-risk Prevalence Rates
To evaluate the potential impact that the prevalence rate used for undiagnosed CHD
might have on the estimated COHb levels, we performed simulations using alternative values for
undiagnosed CHD prevalence rates. The alternative values were based on assuming that the total
CHD prevalence (diagnosed and undiagnosed combined) for females equaled that of the males,
in effect representing a greater proportion of undiagnosed CHD for females (Table 7-9). Two air
quality scenarios were considered in Denver; as is air quality and air quality adjusted to just meet
the current standard. All standard model settings previously described for these scenarios were
retained.
Table 7-9. Estimated alternative prevalence rates for CHD, stratified by age and gender.
Age range
18 to 44
45 to 64
65 to 74
75+
Prevalence rate (fraction) for coronary heart disease
Males3
Diagnosed
0.012
0.088
0.244
0.310
Undiagnosed
0.005
0.038
0.107
0.135
Total
0.017
0.127
0.351
0.446
Females
Diagnosed3
0.007
0.050
0.138
0.175
Undiagnosed13
0.010
0.077
0.213
0.271
Total
0.017
0.127
0.351
0.446
Notes:
3 Values obtained from Table 5-6.
b Generated assuming total CHD prevalence for females was equal to that of males.
The outputs generated from these new simulations were compared with output from the
previous simulations that applied an undiagnosed CHD uniformly to the two genders (termed
base-CHD). As expected, there are a greater number of persons at or above selected COHb
levels for both air quality scenarios due to the expansion of the simulated at-risk population
(Tables 7-10 and 7-11). However, there is little observed difference in the percentage of CHD
population at or above the selected COHb levels.
Interestingly, this was the same outcome when we expanded the simulated at-risk
population to include all persons with diagnosed heart disease (HD) along with undiagnosed
CHD. That is, the percentage of the HD population estimated to experience COHb at or above
selected levels was nearly identical to that estimated for the CHD population (e.g., Tables 6-18
and 6-19). Note that these three at-risk population simulations not only contain differing total
prevalence of a particular health condition, but contain variable prevalence when comparing
across age groups and the two genders. The variation in the prevalence rates used have only
resulted in differences in the number of persons exposed, indicating that if additional prevalence
7-26
-------
rates were available that varied across age groups and the two genders, there may be little impact
to the percentage of persons at or above selected COHb levels.
Table 7-10. Comparison of the portion of the simulated CHD population in the Denver
study area estimated to experience a daily maximum end-of-hour COHb at or
above specified levels using base and alternative undiagnosed CHD prevalence
rates - as is air quality.
COHb
Level
(%)
>0.0
>1.0
>1.5
> 1.75
>2.0
>2.5
CHD
Base Prevalence3'15'0
Number
53,656
10,773
654
111
12
0
Percent
100
20.1
1.2
0.2
<0.1
0
persons
Alternative Prevalencea'd
Number
71,093
15,512
963
185
12
0
Percent
100
21.8
1.4
0.3
<0.1
0
Notes:
Unadjusted ambient concentrations from four monitors in 2006 were used to
represent the As Is air quality scenario.
a Includes persons with diagnosed coronary heart disease, angina pectoris, and
heart attack (CDC, 2009).
b Includes estimate of persons with undiagnosed ischemia developed by EPA
(see section 5.5.2.1).
Dose results obtained from Table 6-15.
d Effectively, a hypothetical undiagnosed CHD prevalence was used to
generate an equivalent total CHD prevalence for both genders. See Table 7-9.
7-27
-------
Table 7-11. Comparison of the portion of the simulated CHD population in the Denver
study area estimated to experience a daily maximum end-of-hour COHb at or
above specified levels using base and alternative undiagnosed CHD prevalence
rates - air quality just meeting the current standard.
COHb
Level
(%)
>0.0
> 1.0
>1.5
£1.75
>2.0
>2.5
>3.0
>3.5
>4.0
CHD persons
Base Prevalence3'15'0
Number
53,656
44,166
12,563
5,800
2,258
444
111
62
12
Percent
100
82.3
23.4
10.8
4.2
0.8
0.2
0.1
<0.1
Alternative Prevalencea'd
Number
71,093
58,851
16,721
7,663
2,999
580
148
62
12
Percent
100
82.8
23.5
10.8
4.2
0.8
0.2
<0.1
<0.1
Adjusted ambient concentrations from four monitors in 1995 were used to
represent the air quality just meeting the current 8-hour standard.
a Includes persons with diagnosed coronary heart disease, angina pectoris,
and heart attack (CDC, 2009).
b Includes estimate of persons with undiagnosed ischemia developed by EPA
(see section 5.5.2.1).
Dose results obtained from Table 6-18.
d Effectively, a hypothetical undiagnosed CHD prevalence was used to
generate an equivalent total CHD prevalence for both genders. See Table 7-9.
7.2.2.3 Physiological Variables
Sensitivity analyses performed on one potentially influential physiological variable -
hemoglobin content - is described here. This physiological variable was identified and evaluated
in response to a comment suggesting that the some of the physiological variables used in the
simulations may not necessarily be adequately representing the at-risk populations or subgroups
having characteristics that may contribute to increased susceptibility. One variable identified
within the CFK module that could have an impact on estimated COHb levels, and for which data
were available to evaluate variation among population subgroups, is an individual's hemoglobin
content. For any simulated individual, the model currently samples from age- and gender-
stratified distributions developed from the 1999-2004 NHANES (see Appendices A and B).
These NHANES data are population-based; as such, represent a collection of individuals that
may be based on population age, gender, and race groupings. In addition, persons having been
diagnosed with anemia are represented in these data, although in a small fraction. Nevertheless,
7-28
-------
hemoglobin content estimates in this REA are drawn from a sample that includes persons with
anemia.
To evaluate the influence of hemoglobin on dose estimates, we obtained the most recent
data available from NHANES (2005-2008), combined these data with the previously used data
set (NHANES 1999-2004), and evaluated whether alternative distributions could be developed
based on influential factors such as race and health condition. Descriptive statistics were
generated using the hemoglobin content (Hb) of four data groups: African-Americans, pregnant
women, persons identified using WHO et al. (2001) guidelines as potentially having anemia
(Table 7-12),10 and all other persons.
Table 7-12. Hemoglobin levels below which anemia is present in a population (from WHO
et al., 2001).
Age or gender group
Children (0. 5-5. Oyrs)
Children (5-12 yrs)
Children (12-1 5 yrs)
Women, non-pregnant (>15yrs)
Women, pregnant
Men(>15yrs)
Hb threshold (g/dl)
11.0
11.5
12.0
12.0
11.0
13.0
The hemoglobin data for the four groups, as represented by normal distributions, are
provided in Table 7-13. Kolgomorov-Smirnov (KS) tests were performed using the SAS
procedure NPAR1WAY to test for differences between the subgroups and dataset comprised of
all other persons. All of these tests were significant (p<0.0001), indicating that the Hb
distributions were different for the compared groups. In knowing that the group of persons with
anemia contained the lowest values for hemoglobin content, we decided that this group would be
used to develop an alternative distribution for evaluating the effect it has on estimated COHb
levels.
Rather than identify anemic persons using the WHO (2001) thresholds and develop
distributions from these samples, we elected to apply the WHO (2001) thresholds to the existing
hemoglobin distributions when sampling for that variable within APEX. This way the original
population distribution shape is preserved, though truncated at the portion of the distribution of
interest to represent an anemic person's hemoglobin content. A model simulation was performed
10 While there were persons that responded to a question asking whether they were being treated for
anemia, we judged that these persons might not best represent the anemic population based on the medical
intervention. The WHO et al. (2001) thresholds used here are given in Table 7-12.
7-29
-------
as described previously, using the Denver study area, the CHD at-risk population, and air quality
just meeting the current standard.
The COHb levels experienced by this new simulated population (i.e., CHD persons with
anemic hemoglobin) were compared with the corresponding results generated when sampling for
hemoglobin content from the full distribution (base hemoglobin) (Table 7-14). Using the anemic
hemoglobin distribution resulted in a greater number of persons and hence a greater percent of
persons at or above selected COHb levels when compared with the base hemoglobin simulation.
The difference between results from the two simulations was not large when considering
differences in the percentage points, with the smallest differences observed at the upper COHb
levels. As far as the magnitude of the difference expressed as a percent increase in persons, there
was a 50 - 100% increase in the percent of persons at or above selected COHb levels when
considering this hypothetical anemic CHD population. This indicates that it is possible that the
number and percent of persons estimated to experience %COHb at or above selected benchmarks
may be underestimated when considering certain physiological attributes such as hemoglobin
content, but the overall magnitude of the effect is dependent on the benchmark level considered.
A quantitative characterization is not possible at this time due to the lack of readily available
information on the percent of HD or CHD population with anemia and regarding the extent to
which persons with anemia might have been included in the CAD/COHb clinical studies.
Table 7-13. Descriptive statistics of blood hemoglobin content measured in various groups
(from NHANES 1999-2008).
Group
African-American
Anemia
Pregnant
All Other Persons
Blood Hemoglobin Content (g/dL)
N
10,171
2,853
659
41,985
mean
13.4
11.3
12.5
14.2
stdev
1.5
1.0
1.1
1.5
min
5.8
5.8
8.6
5.8
max
18.5
12.9
16.4
19.7
7-30
-------
Table 7-14. Comparison of the portion of the simulated CHD population in the Denver
study area estimated to experience a daily maximum end-of-hour COHb at or
above specified levels when sampling from the base and anemic hemoglobin
content distributions - air quality just meeting the current standard.
COHb
Levels
(%)
0.0
1.0
1.5
1.75
2.0
2.5
3.0
3.5
4.0
CHD Persons3'"
Base Hemoglobin0
Number
53,656
44,166
12,563
5,800
2,258
444
111
62
12
Percent
100
82.3
23.4
10.8
4.2
0.8
0.2
0.1
<0.1
Anemic Hemoglobind
Number
53,656
46,918
15,845
7,639
3,344
827
197
74
25
Percent
100
87.4
29.5
14.2
6.2
1.5
0.4
0.1
<0.1
Notes:
Adjusted ambient concentrations from four monitors in 1995 were used to
represent the air quality just meeting the current 8-hour standard.
a Includes persons with diagnosed coronary heart disease, angina pectoris, and
heart attack (CDC, 2009).
b Includes estimate of persons with undiagnosed ischemia developed by EPA
(see section 5.5.2.1).
c Used hemoglobin distributions defined in Appendix A for general population.
Dose results were obtained from Table 6-1 8.
d Used hemoglobin distributions defined in Appendix A for general population,
only truncated the distributions by upper limits defined in Table 7-13.
7-31
-------
7.3 KEY OBSERVATIONS
Based on an overall qualitative judgment of the identified sources of uncertainty in the
assessment approach, selections made regarding input data, and algorithms used, and their
characterization as to direction and magnitude of influence on exposures and doses, staff
consider the exposure and dose estimates to be reasonable for the simulated population the
assessment is intended to represent (i.e., the CHD or HD population residing within the urban
core of each study area). This is because:
• Only three sources of uncertainty were associated with a potential directional influence -
data base quality (overestimation), missing data substitution (underestimation), and zero
concentration frequency (underestimation) - and all were judged to have a low magnitude
of influence on estimated exposures and doses.
• Thirteen of the identified sources of uncertainty were judged by staff to have either
bidirectional influence (eight sources) or unknown direction (five sources):
- One source of uncertainty (i.e., microenvironmental algorithm and data inputs)
was judged as having a potentially medium magnitude of influence on exposure
and dose estimates.
- Five of the remaining twelve sources (i.e., spatial representation, historical data
used, activity pattern database, longitudinal profile algorithm, physiological
factors) were judged as having low to medium magnitude of influence, the level
of which varied based on whether an identified condition existed.
- Ten of the sources were judged to have a low magnitude of influence on estimated
exposures and doses (i.e., database quality, missing data substitution method, zero
concentration frequency, proportional approach used, population database,
meteorological data, commuting database and algorithm, at-risk population
prevalence rates, and benchmark levels for the simulated at-risk population).
There was a wide-ranging level of uncertainty in the knowledge base for the identified
sources:
• Nine sources were judged by staff as having medium knowledge-base uncertainty
including: spatial and temporal representation, historical data used, activity pattern
database, longitudinal profile algorithm, microenvironmental algorithm and input data,
at-risk population prevalence rates, physiological factors, and the benchmark levels for
the simulated at-risk population.
• The knowledge-base uncertainty was judged as low for four of the identified sources
having either unknown or bidirectional influence. This included the proportional
approach used in adjusting air quality conditions, the population database, meteorological
data, and commuting data.
7-32
-------
• The knowledge-base uncertainty was also judged as low for the three sources identified
above as being associated with either under- or overestimating exposures, i.e., the data
base quality, missing data substitution, and zero concentration frequency.
The ratings of the knowledge-base uncertainty can indicate the need for additional data or
analyses to better characterize the uncertainty. When combined with the potential magnitude of
influence associated with each identified source, a prioritization can be given to the higher rated
influential sources. Based on the results of this uncertainty characterization, staff judges that
seven sources (i.e., the spatial and temporal representation of ambient monitoring data, historical
data used, activity pattern database, longitudinal profile algorithm, microenvironmental
algorithm and input data, and physiological factors) remain as the most important uncertainties in
this assessment.
7-33
-------
7.4 REFERENCES
Allen TH, Peng MT, Chen KP, Huang TF, Chang C, Fang HS. (1956). Prediction of blood volume and adiposity in
man from body weight and cube of height. Metabolism. 5:328-345.
Brain JD and Samet JM. (2010a). Letter from Drs. JD Brain and JM Samet to Administrator Lisa Jackson. Re:
Review of the Risk and Exposure Assessment to Support the Review of the Carbon Monoxide (CO)
Primary National Ambient Air Quality Standards: First External Review Draft. CASAC-10-006. February
12, 2010.
Brain JD and Samet JM. (2010b). Letter from Drs. JD Brain and JM Samet to Administrator Lisa Jackson. Re:
Review of the Risk and Exposure Assessment to Support the Review of the Carbon Monoxide (CO)
Primary National Ambient Air Quality Standards: Second External Review Draft. EPA-CASAC-10-012.
May 19,2010.
Brainard J and Burmaster D. (1992). Bivariate distributions for height and weight of men and women in the United
States. RiskAnalysis. 12(2):267-275.
CDC. (2009). Summary Health Statistics for U.S. Adults: National Health Interview Survey, 2007. Series 10,
Number 240. U.S. Department of Health and Human Services, Hyattsville, MD, May 2009.
GalettiPM. (1959). Respiratory exchanges during muscular effort. Helv PhysiolActa. 17:34-61.
Graham SE and McCurdy T. (2005). Revised ventilation rate (VE) equations for use in inhalation-oriented
exposure models. Report no. EPA/600/X-05/008. Report is found within Appendix A of US EPA (2009).
Metabolically Derived Human Ventilation Rates: A Revised Approach Based Upon Oxygen Consumption
Rates (Final Report). Report no. EPA/600/R-06/129F. Appendix D contains "Response to peer-review
comments on Appendix A", prepared by S. Graham (US EPA). Available at:
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=202543.
Glen G, Smith L, Isaacs K, McCurdy T, Langstaff J. (2008). A new method of longitudinal diary assembly for
human exposure modeling. J Expos Sci Environ Epidemiol. 18:299-311.
Issacs K and Smith L. (2005). New Values for Physiological Parameters for the Exposure Model Input File
Physiology.txt. Technical memorandum to Tom McCurdy, NERL WA10. December 20, 2005. Provided
in Appendix A of the CO REA.
Johnson T. (1998). Memo No. 5: Equations for Converting Weight to Height Proposed for the 1998 Version of
pNEM/CO. Memorandum Submitted to U.S. Environmental Protection Agency. TRJ Environmental, Inc.,
713 Shadylawn Road, Chapel Hill, North Carolina 27514.
Johnson T, Mihlan G, LaPointe J, Fletcher K, Capel J. (2000). Estimation of Carbon Monoxide Exposures and
Associated Carboxyhemoglobin Levels for Residents of Denver and Los Angeles Using pNEM/CO
(Version 2.1). Report prepared by ICF Consulting and TRJ Environmental, Inc., under EPA Contract No.
68-D6-0064. U.S. Environmental Protection Agency, Research Triangle Park, North Carolina. Available
at: http://www.epa.gov/ttn/fera/human_related.html. June 2000.
Joumard R, Chiron M, Vidon R, Maurin M, Rouzioux JM. (1981). Mathematical models of the uptake of carbon
monoxide on hemoglobin at low carbon monoxide levels. Environ Health Persp. 41:277-289.
McCurdy T. (2000). Conceptual basis for multi-route intake dose modeling using an energy expenditure approach.
J Expos Anal Environ Epidemiol. 10:1-12.
7-34
-------
SAIC. (2001). Technical Peer Review of "Estimation of Carbon Monoxide Exposures and Associated
Carboxyhemoglobin Levels for Residents of Denver and Los Angeles Using pNEM/CO (version 2.1)"
Prepared by Science Applications International Corporation under EPA Contract No. 68-D-98-113.
Available at: http://www.epa.gov/ttn/fera/human related.html.
Salorinne Y. (1976). Single-breath pulmonary diffusing capacity. ScandJResp Diseases. Supplementum 96.
US Census Bureau. (2007). Employment Status: 2000- Supplemental Tables. Available at:
http://www.census.gov/population/www/cen2000/phc-t28.html.
US Census Bureau. (2009). American Fact Finder. Census Summary File 3 (SF3) - custom tables. Available at:
www.factfinder.census.gov.
US EPA. (1992). Review of the National Ambient Air Quality Standards for Carbon Monoxide: Assessment of
Scientific and Technical Information. Office of Air Quality Planning and Standards Staff Paper. Report
no. EPA/452/R-92-004.
US EPA. (2002). EPA's Consolidated Human Activities Database. Available at: http://www.epa.gov/chad/.
US EPA. (2004). An Examination of EPA Risk Assessment Principles and Practices. Staff Paper prepared by the
US EPA Risk Assessment Task Force. Report no. EPA/100/B-04/001. Available at:
http://www.epa.gov/OSA/pdfs/ratf-final.pdf.
US EPA. (2008). Risk and Exposure Assessment to Support the Review of the NO2 Primary National Ambient Air
Quality Standard. Report no. EPA-452/R-08-008a. November 2008. Available at:
http://www.epa.gov/ttn/naaqs/standards/nox/data/2008112 l_NO2_REA_final.pdf.
US EPA. (2009). Risk and Exposure Assessment to Support the Review of the SO2 Primary National Ambient Air
Quality Standard. Report no. EPA-452/R-09-007. August 2009. Available
athttp://www.epa.gov/ttn/naaqs/standards/so2/data/200908SO2RE AFinalReport.pdf.
WHO, UNICEF, UNU. (2001). Iron Deficiency Anaemia: Prevention, Assessment and Control. Report of a joint
WHO/UNICEF/UNU consultation. World Health Organization, Geneva.
WHO. (2008). Harmonization Project Document No. 6. Part 1: Guidance document on characterizing and
communicating uncertainty in exposure assessment. Available at:
http://www.who.int/ipcs/methods/harmonization/areas/exposure/en/.
7-35
-------
8 SUMMARY OF KEY OBSERVATIONS
This document describes the quantitative human exposure assessment and risk
characterization being conducted to inform the U.S. Environmental Protection Agency's (EPA's)
current review of the National Ambient Air Quality Standards (NAAQS) for carbon monoxide
(CO). An assessment of ambient CO exposure/dose was developed in an earlier phase of this
review in the late 1990s. The design of this REA builds upon recommendations from CAS AC,
information presented in the final ISA, as well as comments made by the public. Presented
together below are the key observations made in each of the chapters.
Conceptual Overview: Assessing Ambient Carbon Monoxide Exposure and Risk
• Carbon monoxide in ambient air is formed primarily by the incomplete combustion of
carbon-containing fuels and photochemical reactions in the atmosphere, with on-road
mobile sources representing significant sources of CO to ambient air.
• Microenvironments influenced by on-road mobile sources are important contributors to
ambient CO exposures, particularly in urban areas. Where present, other (nonambient)
CO sources can also be important influences on total CO exposure and on the impact of
ambient CO exposure on COHb levels.
• The formation of COHb is a key step in the elicitation of various health effects by CO.
Further, COHb level is commonly used in exposure assessment and is considered the
best biomarker for evaluating CO exposure and potential for health effects of concern.
• An individual's COHb levels reflect their endogenous CO production, as well as CO
taken into the body during exposure to ambient and nonambient CO sources. CO
uptake into the bloodstream during exposure is influenced by a number of variables
including internal levels of CO and COHb, such that net uptake may be lower or
negligible in instances where a preceding exposure has been substantially higher than
the current one. Thus, the magnitude of the change in COHb level in response to
ambient CO exposure may decrease with the presence of concurrent or preceding
nonambient CO exposure.
• Individuals with CHD are the population with greatest susceptibility to short-term
exposure to CO, and the population for which the current evidence indicates health
effects occurring at the lowest exposures. The evidence further indicates a potential for
other underlying cardiovascular conditions, particularly other types of heart disease, to
contribute susceptibility to CO effects. Other populations potentially at risk include
those with diseases such as chronic obstructive pulmonary disease (COPD), anemia, or
diabetes, and those in prenatal or elderly life stages.
• Cardiovascular effects are the category of health effects for which the evidence is
strongest and indicative of a likely causal relationship with relevant short-term CO
exposures, particularly for people with CHD. Other endpoints for which the evidence
3-1
-------
is suggestive of causal relationships include effects on the central nervous system,
reproduction and prenatal development, and the respiratory system.
• The specific cardiovascular effects occurring at the lowest COHb levels studied in
CHD patients are reduced time to exercise-induced angina and other markers of
myocardial ischemia, in particular, specific changes to the ST-segment of an
el ectrocardi ogram.
• Risk is characterized in this REA through evaluation of COHb estimated in simulations
involving ambient CO exposures experienced by two target populations: (1)
individuals with CHD (including undiagnosed CHD persons) and (2) individuals with
HD, including CHD (diagnosed and undiagnosed).
• Two types of COHb estimates are considered for the two target populations: (1) daily
maximum end-of-hour COHb levels and (2) ambient contribution to daily maximum
end-of-hour COHb levels (i.e., the change in COHb associated with ambient CO
exposure alone).
• Results from simulations are reported in terms of percent of the simulated at-risk
population expected to experience daily maximum end-of-hour COHb levels (or
ambient CO contribution to daily maximum end-of-hour COHb levels) at or above a
series of levels that range as low as 1%. These results are interpreted in the Policy
Assessment document in light of potential health effects benchmarks.
• For daily maximum end-of-hour COHb levels (absolute), these benchmarks
range from 1.5%, which is below the lowest study mean COHb level resulting
from experimental CO exposure in controlled human exposures of subjects
with CAD, up to 3.0%, a level within the range associated with effects in those
studies. For ambient contribution to daily maximum end-of-hour COHb
levels, the comparison benchmarks include the range from 1.4% up to 2.4%,
which are the COHb increments associated with effects in those studies.
• Beyond the at-risk populations and myocardial ischemia-related effects that are the
focus of this quantitative REA, the current evidence regarding other potentially
susceptible populations and other health effects associated with CO exposures is
discussed and considered with regard to the review of the CO NAAQS in the Policy
Assessment.
Air Quality Considerations
• Mobile sources (i.e., gasoline powered vehicles) are the primary contributor to CO
emissions, particularly in urban areas due to greater vehicle and roadway densities.
• Recent (2005-2007) ambient CO concentrations across the US are lower than those
reported in the previous CO NAAQS review and are also well below the current CO
NAAQS levels. Further, a large proportion of the reported concentrations are below
the conventional instrument lower detectable limit of 1 ppm.
• The currently available information for CO monitors indicates that siting of microscale
and middle scale monitors in the current network is primarily associated with roads
having moderate traffic density (< 100,000 AADT), however, factors other than
8-2
-------
reported AADT (e.g., orientation with regard to dense urban roadway networks) can
contribute to sites reporting higher CO concentrations.
• Ambient CO concentrations are highest at monitors sited closest to roadways (i.e.,
microscale and middle scale monitors) and exhibit a diurnal variation linked to the
typical commute times of day, with peak concentrations generally observed during
early morning and late afternoon during weekdays.
• Policy relevant background (PRB) concentrations across the US are generally less than
0.2 ppm, far below that of interest in this REA with regard to ambient CO exposures.
• Historical trends in ambient monitoring data indicate that at individual sites, ambient
concentrations have generally decreased in a proportional manner. This comparison
included air quality distributions with concentrations at or above the current 8-hour
standard and those reflecting current (as is) conditions.
• The temporal variability in selected upper percentile ambient concentrations (e.g., 99th
percentile 1-hour daily maximum) at individual monitors in Denver and Los Angeles is
relatively small across a three-year monitoring period, particularly when considering
recent air quality. Much of the within-monitor temporal variability is due to a trend in
decreasing concentration from year-to-year.
• There is greater spatial variability in selected upper percentile ambient concentrations
(e.g., 99th percentile 1-hour daily maximum) at ten selected monitoring sites in Los
Angeles when compared with four selected monitoring sites Denver, particularly when
considering the recent air quality.
Overview of Approach Used for Estimating Co Exposure and COHb Dose Levels
• APEX, an EPA human exposure and dose model, has a long history of use in
estimating exposure and dose for many of the criteria pollutants including CO, Os,
SO2, and NO2. Over time, EPA has improved and developed new model algorithms,
incorporated newer available input data and parameter distributions, as well as
performed several model evaluations, sensitivity analyses, and uncertainty
characterizations for the above pollutants. Based on this analysis, APEX was judged to
be an appropriate model to use for assessing CO exposure and dose.
Application of APEX4.3 in this Assessment
• Two exposure model domains (Denver and Los Angeles study areas) were defined by
overlaying ambient monitor locations having 10 km radii with US census tract
population data. Monitors selected comprised the bulk of the urban core in each
location, where ambient monitoring data exist.
• Two simulated at-risk subpopulations were identified by combining the census tract-
specific age and gender population distributions with HD and CHD prevalence rates,
each also stratified by age and gender. In using this approach, staff can represent the
variability that exists in the simulated at-risk HD and CHD subpopulations that reside
in each census tract and within each study area.
• Both simulated at-risk subpopulations include an estimate of persons with
undiagnosed CHD.
8-3
-------
• To represent spatial variability in ambient concentrations in Denver, a total of four
monitors were used; in Los Angeles, the total number of monitors was ten. Temporal
variability was represented by use of hourly ambient concentrations in each study area.
• The exposure and dose model simulations included 8 microenvironments in each
location to represent the expected variability in microenvironmental CO
concentrations.
• All indoor microenvironments were modeled using a mass balance model to represent
temporal variability in indoor CO concentrations with respect to the outdoor CO
concentration variability. In addition, distributions of microenvironmental factors were
used for all microenvironments rather than point estimates. Using distributions of
microenvironmental factors will better represent both spatial and temporal variability in
estimated microenvironmental CO concentrations.
• Additional analyses using output from individual-level simulations were performed to
provide information on the microenvironments most influential to population exposure
at different exposure levels. This included an analysis of the effective ratios of
microenvironment to ambient concentrations and the contribution of ambient CO
exposure to total COHb level estimates. The smaller sample sizes generated for these
analyses were found to be representative of the larger simulations employed for
estimating exposure and dose in the different air quality exposure scenarios.
Simulated Exposure and COHb Dose Results
• Ambient CO exposures and resulting COHb levels in the blood of two simulated at-risk
populations in the Los Angeles and Denver study areas were estimated considering five
air quality scenarios: as is air quality, air quality adjusted to simulate just meeting the
current 8-hour CO NAAQS, and air quality adjusted to just meet three potential
alternative standards.
• The two at-risk populations simulated were: (1) persons with diagnosed CHD,
including those estimated to have undiagnosed CHD, and (2) the larger group of
persons with any type of HD including those estimated to have undiagnosed CHD.
While the number of persons and person-days at or above selected COHb levels
differed between the two populations, reflecting their differing size, the percentage of
each population's persons and person-days were similar.
• The relative contribution of various microenvironments to exposure concentrations was
generally similar between the two study areas. When considering as is air quality,
indoor microenvironments contributed mostly to low level exposures (at or above 1
ppm and 2 ppm), comprising between 40 - 80% of the time spent at those exposure
levels, while time spent inside vehicles contributed to most exposures at or above 3
ppm (70 - 100%). In comparison, when considering air quality just meeting the
current standard, the percent contribution from indoor microenvironments was
generally higher for low level exposures (about 65 - 85% of exposure concentrations at
or above 1 ppm and 2 ppm), though again higher level exposures were dominated by
the contributions from inside-vehicle microenvironments.
8-4
-------
• The relationship between the two study areas with regard to estimated distribution of
maximum end-of-hour COHb levels differed with the different air quality scenarios.
Under as is air quality conditions, the simulated at-risk populations in the Los Angeles
study area were estimated to experience a slightly higher distribution of maximum end-
of-hour COHb levels than the Denver populations. Under conditions of air quality
adjusted from historical air quality data to just meet the current or alternative standards,
however, appreciably larger percentages of the Denver populations were estimated to
experience COHb at or above specific levels than the Los Angeles populations.
• For as is air quality conditions, the highest daily maximum end-of-hour COHb
estimated to be experienced over the course of the simulated year was below 1.5% for
more than 98% of the at-risk populations simulated in each study area; it was below
2% COHb for more than 99.9% of these simulated populations. A lower percentage of
the simulated at-risk populations in Denver were estimated to experience daily
maximum end-of hour COHB below these benchmarks than were the populations in
Los Angeles.
• Under as is air quality conditions, the highest incremental contribution of
ambient CO exposure to maximum end-of-hour COHb levels estimated in the
simulated populations was 1.7% COHb, and more than 99% of both study area
populations were estimated to have ambient CO contributions to COHb below
1.4%. As with estimates of total COHb (i.e., COHb from endogenous CO
production and ambient exposure together), a larger percentage of the Los
Angeles population was estimated to experience the higher ambient
contributions to maximum end-of-hour COHb compared to the Denver
population. For example, the percentage of the population estimated to
experience ambient contributions to COHb at or above COHb levels above
1.0% was approximately 2 to 3 times as high in Los Angeles than in Denver.
• For simulations of air quality adjusted to just meet the current 8-hour standard of 9
ppm, the highest estimated daily maximum end-of-hour COHb over the course of the
simulated year was below 1.5% for 95% of both simulated at-risk populations in the
Los Angeles study area, and was below 2% COHb for 99.4% of these populations. In
contrast, the percentage of the simulated at-risk populations in the Denver study area
estimated to experience daily maximum end-of-hour COHb levels that did not exceed
1.5% was about 80%. The percentage of the Denver populations with their highest
estimated daily maximum end-of-hour COHb below 2% was approximately 95%.
• As with estimates of total COHb (i.e., COHb from endogenous CO production
and ambient exposure together), the percentage of the simulated population
estimated to experience the higher ambient contributions to maximum end-of
hour COHb was appreciably greater in Denver as compared to Los Angeles.
While estimated ambient CO contributions to daily maximum end-of-hour
COHb were below 1.4% for nearly 98% of the Los Angeles simulated
population, the corresponding percentage of the Denver population was about
87%.
• In addition to the simulations for as is and just meeting the current 8-hour standard air
quality conditions, three simulations of air quality just meeting potential alternative
$-5
-------
standards were performed. The alternatives comprise different combinations of form,
averaging time and level which were expected to achieve somewhat similar exposure
and dose results. The combinations that were selected were based on consideration of
exposure and dose results obtained for the as is air quality conditions and the higher
just meeting the current 8-hour standard conditions. The combinations of form,
averaging time and level that were simulated include: (1) a second-highest 8-hour
average of 5 ppm, (2) a 99th percentile daily maximum 8-hour average of 5.0 ppm, and
(3) a 99th percentile daily maximum 1-hour average of 8.0 ppm.
• These three simulations generated generally similar percentages of the at-risk
populations estimated to be exposed at selected concentrations and experience
maximum end-of hour COHb levels at or above selected levels. Each of these
three simulations generated fewer persons and a lower percent of the at-risk
populations at or above selected COHb levels than did simulations for air
quality adjusted to just meet the current 8-hour standard. For example, about
1% or less of the Denver populations had a highest daily maximum end-of-
hour COHb at or above the 2.0% COHb level in the alternative standards'
simulations as compared to approximately 4.2% in simulations of air quality
just meeting the current standard. When considering the potential alternative
standards in Los Angeles, generally fewer than 0.1% of the simulated at-risk
populations were estimated to experience a maximum end-of-hour COHb level
at or above 2.0 % COHb, compared to 0.5% at that same COHb level
associated with air quality adjusted to just meet the current standard.
Results for the five air quality scenarios are further analyzed in the Policy Assessment
to inform consideration of the level of public health protection that might be provided
by alternative standard levels associated with different combinations of averaging time
and form.
Results generated in the current assessment for the air quality conditions just meeting
the current NAAQS were compared with estimates from the assessment conducted in
2000 (Johnson et al., 2000) for similar conditions in the Denver and Los Angeles study
areas (section 6.3). The two assessments employed similar approaches, similar,
although not identical air quality data for this scenario, and they used different
exposure models (APEX vs. pNEM). Results were similar for the 1.5% and 2% COHb
level for the simulated Los Angeles study area population and somewhat different for
the Denver study area population. For example, the two assessments' Los Angeles
estimates of population percentages with highest daily maximum end-of hour COHb at
or above these COHb levels were similar and within about 10% of one another. For
the Denver simulated study populations, however, the estimates for these two COHb
levels were about 3 and 8 times greater, respectively, in the current assessment when
compared with that estimated in the prior assessment.
-------
Variability Analysis and Uncertainty Characterization
• Based on an overall qualitative judgment of the identified sources of uncertainty in the
assessment approach, selections made regarding input data, and algorithms used, and
their characterization as to direction and magnitude of influence on exposures and
doses, staff consider the exposure and dose estimates to be reasonable for the simulated
population the assessment is intended to represent (i.e., the CHD or HD population
residing within the urban core of each study area). This is because:
• Only three sources of uncertainty were associated with a potential directional
influence - data base quality (overestimation), missing data substitution
(underestimation), and zero concentration frequency (underestimation) - and
all were judged to have a low magnitude of influence on estimated exposures
and doses.
• Thirteen of the identified sources of uncertainty were judged by staff to have
either bidirectional influence (eight sources) or unknown direction(five
sources):
One source of uncertainty (i.e., microenvironmental algorithm and
data inputs) was judged as having a potentially medium magnitude
of influence on exposure and dose estimates.
Five of the remaining twelve sources (i.e., spatial representation,
historical data used, activity pattern database, longitudinal profile
algorithm, physiological factors) were judged as having low to
medium magnitude of influence, the level of which varied based
on whether an identified condition existed.
- Ten of the sources were judged to have a low magnitude of
influence on estimated exposures and doses (i.e., database quality,
missing data substitution method, zero concentration frequency,
proportional approach used, population database, meteorological
data, commuting database and algorithm, at-risk population
prevalence rates, and benchmark levels for the simulated at-risk
population).
• There was a wide-ranging level of uncertainty in the knowledge base for the identified
sources:
• Nine sources were judged by staff as having medium knowledge-base
uncertainty including: spatial and temporal representation, historical data used,
activity pattern database, longitudinal profile algorithm, microenvironmental
algorithm and input data, at-risk population prevalence rates, physiological
factors, and the benchmark levels for the simulated at-risk populations.
• The knowledge-base uncertainty was judged as low for four of the identified
sources having either unknown or bidirectional influence. This included the
proportional approach used in adjusting air quality conditions, the population
database, meteorological data, and commuting data.
3-7
-------
• The knowledge-base uncertainty was also judged as low for the three sources
identified above as being associated with either under- or overestimating
exposures, i.e., the data base quality, missing data substitution, and zero
concentration frequency.
The ratings of the knowledge-base uncertainty can indicate the need for additional data
or analyses to better characterize the uncertainty. When combined with the potential
magnitude of influence associated with each identified source, a prioritization can be
given to the higher rated influential sources. Based on the results of this uncertainty
characterization, staff judges that seven sources (i.e., the spatial and temporal
representation of ambient monitoring data, historical data used, activity pattern
database, longitudinal profile algorithm, microenvironmental algorithm and input data,
and physiological factors) remain as the most important uncertainties in this
assessment.
-------
APPENDICES
-------
Appendix A
Technical Memorandum on Updates To APEX Physiology.Txt File
(Isaacs And Smith, 2005)
The following contains a technical memo provided by Isaacs and Smith (2005) in its
original format. Staff included page numbers and performed some minor formatting to
text and table headers for the purposes of inclusion into the CO REA appendices.
A-l
-------
TECHNICAL MEMORANDUM
TO: Tom McCurdy, WA-COR, NERL WA 10
FROM: Kristin Isaacs and Luther Smith, Alion Science and Technology
DATE: December 20, 2005
SUBJECT: New Values for Physiological Parameters for the Exposure Model Input
File Physiology.txt.
Table of Contents
List of Figures 3
1. Introduction 4
2. Evaluation of the Current Physiology File Data 4
2.1 Normalized Maximal Oxygen Uptake (nvo2max) 4
2.2 Body Mass 5
2.3 Resting Metabolic Rate 5
2.4 Hemoglobin Content and Blood Volume Factor 5
2.5 Summary of Findings 5
3. Derivation of New Distributions for Body Mass 6
3.1 The NHANES Body Mass Dataset 6
3.2 Calculation of the New Sampling Weights for the Combined NHANES Dataset.
7
3.3 Fitting the Body Mass Data 7
4. Derivation of New Distributions for Normalized Vo2max 13
4.1 The Nvo2max Data 13
4.2 Determining the NVo2max Distributions 17
5. Derivation of New Distributions for Hemoglobin Content (Hemoglobin Density) 25
6. Blood Volume as a Function of Height and Weight 28
References 29
Appendix A. SAS Code for Estimating the Body Mass Distributions 40
Appendix B. SAS Code for Estimating the Normalized Vo2Max Distributions 41
Appendix C. SAS Code for Estimating the Hemoglobin Content Data 42
Appendix D. TheNewPhysiology.txt file 43
Appendix E. All Derived Physiological Parameters 54
A-2
-------
LIST OF FIGURES
Figure 1. Geometric Means for the Best-fit Lognormal Distributions for Body Mass as a
Function of Age, Derived from NHANES 1999-2004 Study Data 9
Figure 2. Geometric Standard Deviations for the Best-fit Lognormal Distributions for
Body Mass as a Function of Age, Derived from NHANES 1999-2004 Study Data 10
Figure 3. Minimums (1st Percentile) for Body Mass as a Function of Age, Derived from
NHANES 1999-2004 Study Data 11
Figure 4. Maximums (99th Percentile) for Body Mass as a Function of Age, Derived
from NHANES 1999-2004 Study Data 12
Figure 5. Individual Nvo2max Measurements for Males and Females, Derived from
Literature Studies and Experimental Measurements 14
Figure 6. Grouped Mean Nvo2max Measurements for Males and Females, Derived from
Literature Studies 15
Figure 7. Nvo2max Standard Deviations for Males and Females, Derived from Literature
Studies 16
Figure 8. Combined Nvo2max Group Means for Males and Females 19
Figure 9. Combined Nvo2max Group Standard Deviations 20
Figure 10. Nvo2max Normal Distribution Fits: Raw Fit Means and Smoothed Fits 21
Figure 11. Nvo2max Normal Distribution Fits: Raw Fit Standard Deviations and
Smoothed Fits 22
Figure 12. Nvo2max Minimums. 1st Percentile of the Best-fit Normal Distribution 23
Figure 13. Nvo2max Maximums. 99th Percentile of the Best-fit Normal Distribution... 24
Figure 14. Mean Values of Hemoglobin Content as Derived from the 1999-2002
NHANES Dataset, with Comparison to Current Physiology.txt Values 26
Figure 15. Values of Hemoglobin Content Standard Deviation as Derived from the 1999-
2002 NHANES Dataset, with Comparison to Current Physiology.txt Values 27
A-3
-------
1. INTRODUCTION
The purpose of this memo is to present an updated version of the physiological
parameters input file (Physiology.txt) for the APEX model. Portions of this file are also
used as input for SHEDS-PM and SHEDS-AirToxics.
The physiology file contains age- and gender-based information for several physiological
parameters used in human exposure modeling. This information includes distributional
shapes and parameters for all age and gender cohorts from age 0 to 100 years for
normalized maximal oxygen uptake (nvo2max), body mass, resting metabolic rate
(RMR), and blood hemoglobin content. In addition, a parameter called blood volume
factor (BVF), which is a cohort-dependent parameter in the equation for blood volume as
a function of body mass, is present in the file as well.
New age- and gender-dependent distributions were developed based the best available
physiological data from the literature. In this report, a summary of the current state of the
physiology file is presented, followed by the derivation of new physiological data for
body mass, normalized vo2max, and hemoglobin content. Portions of the SAS code used
for analysis are included (Appendices A-C), as is the new Physiology.txt file (Appendix
D). The final appendix (Appendix E) contains tables of all the derived physiological
parameters.
2. EVALUATION OF THE CURRENT PHYSIOLOGY FILE
DATA
The physiology.txt file was originally generated for the PNEM model by T. Johnson. It
was last updated 6/11/1998, as documented in the report User's Guide: Software for
Estimating Ventilation (Respiration) Rates for Use in Dosimetry Models, (T. Johnson and
J. Capel). In that report, the original references for the data in the file were provided. An
evaluation of the data in the file was included in a previous memo to the WA-COR under
this work assignment. A summary of those findings is repeated here.
2.1 2.1 NORMALIZED MAXIMAL OXYGEN UPTAKE (NVO2MAX).
The nvo2max data were derived from a number of sources. The data for males,
especially, were pieced together from a variety of studies (a total of 6), leading to
discontinuities in the distributional parameters. However, in each age and gender cohort,
the distributions parameters were derived from a single published study. Additionally,
much of the nvo2max data is quite old. The data for males at age 20 and at 28-69 came
A-4
-------
from a study from 1960 [1]. Data for males aged 0-8 and 16-19, and females 0-19 came
from a figure in a textbook from 1977 [2], which in turn was based on limited earlier
data. An additional issue with the 1977 data is (according to the report mentioned above)
that values for certain ages (very young or elderly) were acquired by simple tangential
extrapolation of the data in the figure.
In addition, in some cases it was not clear how the parameters were derived from the
referenced studies. For example, Heil et al. [3] was referenced as the source of the values
for females aged 66-100. However, an examination of that study provided no clues as to
how the values were actually determined. As far as can be determined, in no place did
the authors break down the means and SDs of their data into groups separated by both
gender and age simultaneously.
2.2 BODY MASS.
The current body mass data were derived from an in-depth analysis [4, 5] of the second
CDC National Health and Nutrition Examination Survey (NHANES II) body mass data
[6]. The data were relatively comprehensive, and the methods used to generate the
lognormal distributions were sound. However, the NHANES II data were compiled for
the years 1976-1980, so an analysis of more recent data is necessary to accurately
account for changes in human activity patterns in adults and especially children.
2.3 RESTING METABOLIC RATE.
Not included for evaluation, per discussion with WA-COR.
2.4 HEMOGLOBIN CONTENT AND BLOOD VOLUME FACTOR.
The original references for the hemoglobin content or blood volume factor values given
in the current physiology.txt file could not be identified. Therefore, their validity could
not be evaluated and it was desirable that new statistics be calculated.
2.5 SUMMARY OF FINDINGS
• In some cases, especially for nvo2max, the data are unnecessarily and confusingly
disjointed across ages.
• It is also unclear how some of the nvo2max values were derived from the referenced
studies.
• With the exception of the Schofield equations for the BM/RMR regression, parameter
distributions at each age and gender cohort were derived from data from a single
study.
A-5
-------
Many of the studies used are very old (ex. 1960, 1977).
Some the data is of questionable validity (for example, the extrapolation of a textbook
figure is used), although it may have been the best available at the time of the
compilation of the file.
The original source of the hemoglobin content and blood volume factor data could
not be identified.
Given these conclusions, we recommended a full review and update of the current
physiology.txt file data. Specifically, we recommended that where possible, new
distributions or equations should be developed based on thorough, compiled data
from appropriate studies.
3. DERIVATION OF NEW DISTRIBUTIONS FOR BODY MASS
3.1 THE NHANES BODY MASS DATASET.
New body mass distributions were generated from data from the National Health and
Nutrition Examination Survey (NHANES). This survey is an ongoing study carried out
by the National Center for Health Statistics of the Centers for Disease Control. EPA
recognizes the utility of this dataset in characterizing the American population for risk
assessment and policy support purposes [7].
Older NHANES data (for the years 1976-1980) have been used previously to develop
population estimates of body mass distributions [4,5]. The current Physiology.txt file
body mass distributions are based on this work. However, the analysis presented here is
based on the most recent NHANES data, for the years 1999-2004 [8].
Demographic (Demo) and Body Measurement (BMX) datasets for each of the NHANES
studies were downloaded from the NHANES website. The files were downloaded as
SAS xpt datasets. The downloaded files were as follows:
1999-2000 2001-2002 2003-2004
BMX.xpt BMX_b_r.xpt BMX_c.xpt
Demo.xpt Demo_b.xpt Demo_c.xpt
The Demographic datasets contained the age and gender values for each survey
participant, while the Body Measurement datasets contained the body weights for each
subject. The combined dataset comprised 31,126 individuals. This resulted in
approximately 400-500 persons in each age 0-18 year cohort, and approximately 80-150
persons in each age 19-85 year cohort (the NHANES studies more heavily sampled
children).
A-6
-------
3.2 CALCULATION OF THE NEW SAMPLING WEIGHTS FOR THE
COMBINED NHANES DATASET.
In the analysis of the NHANES data, sampling weights must be used to ensure that the
data are weighted to appropriately represent the national population. Sampling weights
for the combined NHANES body mass dataset were derived as recommended by the
documentation provided with the most recent NHANES release [9]. Specifically, the
sampling weight for each subject was calculated as:
W combined ~ T W 2003-2004 \
-
T
J
--
combined ~ T ^1999-2002
where wcombmed is the sampling weight for the combined dataset, W2003-2004 is the weight
for the subjects in the most recent study, and Wi999_2oo2 is the weight for subjects in
combined 4-year (1999-2000 and 2001-2002) NHANES dataset. (Both weights are
provided with the appropriate NHANES release. The combined 1999-2002 weight,
which is not a simply half of that for the corresponding 2-year periods, was explicitly
calculated for researcher use by CDC since the two 2-year periods use different census
data.)
By using the sampling weights, once can consider any 2-year NHANES dataset or any
combination of datasets as a nationally representative sample.
3.3 FITTING THE BODY MASS DATA.
In the current physiology file, body mass is modeled as a two-parameter lognormal
distribution. The NHANES body mass data were fit to several types of distributions
(including normal, beta, and three-parameter lognormal distributions). It was determined
that overall, the distribution that provided the best combination of good behavior over
ages and good fit to the data was a two-parameter lognormal distribution.
The data were fit to the lognormal distributions using the SAS PROC UNIVARIATE
procedure. The FREQ option of the procedure was used to apply the sampling weights.
The SAS code used to generate the body mass distributions is provided in Appendix A.
As the NHANES 1999-2003 studies only covered persons up to age 85, linear forecasts
were made for ages 86-100, as based on the data for ages 60 and greater.
A-7
-------
3.4 Body Mass Results.
Geometric means and standard deviations (SD) for the best-fit lognormal distributions for
body mass are given in Figures 1 and 2. The means behaved fairly smoothly across ages.
Note that for children age 0-18, the values of the new fits are similar, but slightly higher
than those in the current Physiology.txt file, which were derived from earlier NHANES
studies. The new means also capture the trend towards decreasing body weight in older
persons that was previously neglected in the Physiology.txt file.
The maximum and minimum values for the distributions are presented in Figures 3 and 4.
The minimums and maximums were calculated as the 1st and 99th percentile of the raw
body mass data for the cohort. (Note that these values differ from the 1st and 99th
percentiles of the fitted lognormals.) While the minimum value is consistent with the
current Physiology.txt (which was based on earlier NHANES studies), the new cohort
maximums are generally higher than before.
The behavior of several of the body mass parameters (especially the SD) is fairly noisy,
especially for adults. This is most likely due to the smaller number of samples for adults
as compared to children. Therefore, it may desirable to use age-grouped data or running
averages over years in these age ranges. While the attached prepared Physiology.txt file
uses the "raw" parameters, smoothed results using 5-year running averages are provided
in the attached data tables (Appendix E, plots not shown). These could be used at the
direction of EPA; changing the "official" release Physiology.txt file would be trivial.
A-8
-------
MALES: Body Mass Geometric Mean
an
100.000 -
90.000 -
80.000 -
70.000 -
60.000 -
50.000 -
40.000 -
30.000 -
20.000 -
10.000
0.000
0
Current Physiology
File
20
40 60
Age (years)
80
100
FEMALES: Body Mass Geometric Mean
New Values
Current Physiology
File
20
40 60
Age (years)
80
100
Figure 1.
Geometric Means for the Best-fit Lognormal Distributions for Body Mass as a
Function of Age, Derived from NHANES 1999-2004 Study Data.
A-9
-------
MALES: Body Mass GSD
New Values
Current Physiology
File
20
40 60
Age (years)
80
100
1.000
FEMALES: Body Mass GSD
New Values
Current Physiology
File
20
40 60
Age (years)
80
100
Figure 2. Geometric Standard Deviations for the Best-fit Lognormal Distributions for Body
Mass as a Function of Age, Derived from NHANES 1999-2004 Study Data.
A-10
-------
MALES: Body Mass Minimum
20
Current Physiology
File
40 60
Age (years)
80
100
ctf
FEMALES: Body Mass Minimum
20
New Values
Current Physiology File
40 60
Age (years)
80
100
Figure 3. Minimums (1 Percentile) for Body Mass as a Function of Age, Derived from
NHANES 1999-2004 Study Data.
A-ll
-------
MALES: Body Mass Maximum
250 n
200 -
150 -
I 100
50 -
0
0
•New Values
• Current Physiology
File
20
40 60
Age (years)
80
100
FEMALES: Body Mass Maximum
200 n
180 -
160 -
140 -
'55 120 -
¥ loo-
sj
S 80 -
60 -
40 -
20 -
0 -
0
•New Values
• Current Physiology
File
20
40 60
Age (years)
80
100
nth
Figure 4. Maximums (99 Percentile) for Body Mass as a Function of Age, Derived from
NHANES 1999-2004 Study Data.
A-12
-------
4. DERIVATION OF NEW DISTRIBUTIONS FOR
NORMALIZED VO2MAX
2.2 4.1 THE NVO2MAX DATA
The NHANES studies do report data for vo2max in individuals. However, the NHANES
vo2max values are estimated values, i.e. they are not measured directly. Such estimated
values are not appropriate for use in this context (as per discussion with the WA-COR).
Therefore, nvo2max distributional shapes were determined from a large database of
experimental and literature vo2max measurements for different age/gender cohorts.
A PubMed-based literature search located a number of studies in which vo2max was
directly measured. In addition, a large number of scientific papers (-350) reporting
vo2max were also provided to Alion by the WA-COR. All the studies were evaluated for
use by determining if: 1) any normalized vo2max data for individuals were reported or 2)
any group means for narrow age-gender cohorts were reported. Studies in which the
studied age group was very broad or contained both males and females were discarded.
Also discarded were any studies in which vo2max was not normalized by body mass, or
for which no age data were reported. Data for ill or highly-trained individuals were not
used; however, studies in which subjects underwent mild or moderate exercise training
were included. Two large databases, one of individual vo2max data and one of grouped
means and SDs, were constructed from the valid studies.
The database of individual data comprised age versus nvo2max data for 1949 men and
1558 women. The data were pulled from either tables or graphs in 20 published studies
[11-30]. Additional raw experimental data were provided by the WA-COR [31]. In the
case of the graphical data, the original source was digitized and the data points were
pulled from the digital figure using graphics software. (This was accomplished by
calibrating the pixels of the digitized image with the range of age and nvo2max values.)
The individual nvo2max data for males and females are shown in Figure 5.
The grouped mean and SD data were derived from 136 studies [32-167]. These data
comprised approximately 550 means and SDs for different age/gender cohorts. Single
age/gender cohort means and SD values for the Adams data [31] were also included in
this dataset. Only data for subject groups having an age SD of less than approximately 2-
3 years were considered. The grouped mean values for men and women are shown in
Figure 6, while the group SD values are shown in Figure 7.
A-13
-------
MALES :Nvo2max
oo
-------
MALES: Study Means, Nvo2max
80.0 n
70.0 -
oo 60.0 -
1, 5°-°"
ce 40.0 -
e
<3 30.0 -
« 20.0 -
10.0 -
o.o -
0
0
20.0
40.0
Age (Years)
60.0
80.0
FEMALES: Study Means, Nvo2max
00
j
s
(N
0
>
70.0 n
60.0 -
50.0 -
40.0 -
30.0 -
20.0 -
10.0 -
0.0 -
*
-------
MALES: Study Standard Deviations: Nvo2max
16.0 n
14.0 -
oj 12.0 -
1 10.0 -
•—^
* 8.0 -
6.0 -
4.0 -
2.0 -
0.0 -
(N
O
0.0
20.0
:
40.0
Age (Years)
60.0
80.0
FEMALES: Study Standard Deviations: Nvo2max
00
t
nvo2ma
14.0 n
12.0 -
10.0 -
8.0 -
6.0 -
4.0 -
2.0 -
o.o -
0
*
* • **
>^*J.
*
1 1 1 1 1 1 1
0 10.0 20.0 30.0 40.0 50.0 60.0 70.0
Age (Years)
Figure 7. Nvo2max Standard Deviations for Males and Females, Derived from Literature
Studies.
A-16
-------
4.2 Determining the NVo2max Distributions
Both the grouped mean and the individual datasets were evaluated for use in deriving the
nvo2max parameters.
The group means and SD were combined into single age/gender cohort values. The
combined means were calculated as mean of the group means, weighted by the number of
subjects. The group SD were calculated by transforming each group SD to a group
variance, calculating the mean variance (weighted by the number of subjects in each
study) and retransforming the variances to SDs. The combined group means and SDs are
given in Figures 8 and 9.
The combined group means were fairly well-behaved across age and gender cohorts (see
Figure 8), while the SD data (Figure 9) were noisier. These data may be appropriate for
use in the Physiology.txt file; however, it was noted that the group mean data, while
plentiful for children, were not very well represented in the adult (30+ years) age range
(especially for women). This is mainly due to the fact that very few investigators use
narrow age cohorts when studying adults, rather, it was far more common for broader age
groups to be used. These data were not included in the grouped mean analysis, as the
mean nvo2max for a broad age group cannot be assumed valid for the cohort represented
by the study age mean. Therefore, we opted to use the database of individual nvo2max
measurements to develop new distributions for the Physiology.txt file.
The individual nvo2max data were fit to several types of distributions (including normal,
beta, and lognormal distributions). It was determined that the normal distribution fit the
data best. The parameters (means and standard deviations) of the best-fit distributions
were obtained using the SAS PROC UNIVARIATE procedure. The SAS code used to fit
the data is given in Appendix B.
Both raw and smoothed nvo2max fits were calculated. Calculating 5-year running
averages did not smooth the data considerably. Therefore, the smoothed fits were
determined by choosing a best-fit functional form for the nv02max data. The data were
fit to functions as follows:
Mean (Age 0-20): Linear function
Mean (Age 21-100): Parabolic function
SD (Age 0-26): Linear function
SD (Age 27-100): Parabolic function
Fitting the data in this manner also allowed for all age/gender cohorts to be represented.
Since only cohorts having N>10 were fit to distributions, there were some cohorts for
which no parameters were calculated. The raw and smoothed fits for means are given in
Figure 10; analogous data for SD is given in Figure 11. The raw nvo2max parameters
were not as clean across ages as the body mass data (probably due to the much smaller
sample size), and thus the smoothed fits were selected for use in the attached
Physiology.txt file. As with body mass, the raw fits may be used at the direction of EPA.
A-17
-------
The results for the nvo2max means were in fact quite close to those in the current file.
However, the values exhibited much more consistent behavior across ages, and the values
for elderly persons were lower than previously. The SD values were also in the same
range as the current values, yet they no longer demonstrate nonsensical discontinuities
across ages.
The minimum and maximum nvo2max values were assumed to be the 1st and 99th
percentile of the best-fit lognormal distribution. (Note: this is different from the method
used for estimating the body mass limits. In that case, the samples were large enough
that the percentiles of the raw data were appropriate for use as minimum and maximum.
As the nvo2max data cohorts had much smaller N than the NHANES studies, the raw
percentiles were less appropriate.) The maximum and minimum values are shown in
Figures 12 and 13.
A-18
-------
MALES: Nvo2max, Combined Group Means
60 -,
™ 50 -
1 40 -
I 3° "
C 10 -
o -
0
20
40
Age (Years)
60
80
50 -,
j^ 40 -
^ 30 -
H 20 -
o
10 -
0
FEMALES: Nvo2max, Combined Group Means
0
10
60
70
20 30 40 50
Age (Years)
Figure 8. Combined Nvo2max Group Means for Males and Females
A-19
-------
MALES: Nvo2max, Combined Group SD
16 -
14 -
UJJ
^ 12 -
£ 10-
^j>
8 -
6 -
4 -
2 -
0 -
c
0
20
40
Age (Years)
60
80
10 -i
$ 8 -
I 6
1 4
o
« 2 H
FEMALES: Nvo2max, Combined Group SD
10
20
30 40
Age (Years)
50
60
70
Figure 9. Combined Nvo2max Group Standard Deviations.
A-20
-------
MALES: MEAN Nvo2max
x
cS
a so
-------
14 -
12 -
10 -
X
cS
P, 6-
2-
MALES: Nvo2max Standard Deviation
• Raw Values
• Fit Values
Current Physiology.txt
20
40 60
Age (Years)
100
FEMALES: Nvo2max Standard Deviation
Raw Values
Fit Values
Current Physiology.txt
20
40 60
Age (Years)
100
Figure 11. Nvo2max Normal Distribution Fits: Raw Fit Standard Deviations and Smoothed
Fits.
A-22
-------
oo
CN
60 -
50 -
40 -
30 -
20 -
10 -
0 -
0
MALES: Nvo2max Minimum
New Values
Current Physiology.txt
20
40 60
Age (Years)
80
100
45 n
40 -
oo 35 -
^ 25 "
I 20 -
-------
MALES: Nvo2max Maximum
80 -,
70 -
j) 60 -
H 50"
& 40 -
^ 30-
a 20 -
10 -
0 -
0
•New Values
• Current Physiology.txt
20
40 60
Age (Years)
80
100
o
I 20-
10 -
0
FEMALES: Nvo2max Maximum
0
•New Values
Current Physiology.txt
20
40 60
Age (Years)
80
100
..th
Figure 13. Nvo2max Maximums. 99 Percentile of the Best-fit Normal Distribution.
A-24
-------
5. DERIVATION OF NEW DISTRIBUTIONS FOR
HEMOGLOBIN CONTENT (HEMOGLOBIN DENSITY)
The new hemoglobin content values were derived from the combined NHANES 1999-
2000 and 2001-2002 datasets. As of December 2005, hemoglobin data had not yet been
released for the 2003-2004 study. The age data was provided in the Demographic
datasets (Demo.xpt and Demo_b.xpt, previously downloaded for the body mass analysis)
for the two survey periods, while hemoglobin content (in g/dL) was provided in the
Laboratory #25 (Complete Blood Count) datasets (Iab25.xpt and 125_b.xpt, which were
downloaded for this analysis). The dataset comprised 20,321 individuals; appropriate
sample weights were used for the combined 4-year (1999-2002) dataset as provided with
the NHANES 2001-2002 data release. Similarly to the body mass data, the hemoglobin
content values were analyzed in SAS. The age and hemoglobin datasets were merged
and fit to normal distributions using the SAS PROC UNIVARIATE procedure. The
FREQ option of the procedure was used to apply the sampling weights. The SAS code in
provided in the Appendix C.
Hemoglobin content statistics were estimated for single-year age and gender cohorts for
ages 1-19, as the behavior of the means were smooth in this age range. For persons 20
and over, the data were grouped in 5-year cohorts (20-24, 25-29, etc.) No blood count
data were available for subjects under 1 year of age or greater than 90. The age 0 mean
values were obtained by a linear regression of ages 1-20 (males) or 1 to 11 (females) back
to age 0. These were the ages for which the hemoglobin content demonstrated an
increase with age. The 91-95 and 96-100 mean values were obtained by a linear
regression of the 61-65 and older age groups. As the standard deviations did not appear
to behave as smoothly with age as did the mean values, the age 0 value was assumed
equal to the age 1 value, and the age 91-95 and 96-100 value was assumed equal to the
age 90-94 value.
The resulting means and standard deviations for the best-for normal distributions for
hemoglobin content are given in Figures 14 and 15. The current hemoglobin content
values are shown for comparison.
The main conclusions that can be made is that the current Physiology.txt input file
overestimates mean hemoglobin content in children and in older persons. The standard
deviation values in the current physiology.txt file are fairly close to those found in this
analysis. The new values are not very smooth over ages; EPA may elect to continue to
use the current values. It should be noted that the original reference for the current
hemoglobin statistics is unknown.
Note: In the current implementation of APEX, the hemoglobin content statistics affect
only the CO dose algorithm calculations.
A-25
-------
17 n
, , 16
5i is H
00
¥14"
I 13"
r^ 12
11
10
MALES: Mean Hemoglobin Content
* New Values
Current Physiology.txt
>»***+»++«, ^
rf/
20
40 60
Age (years)
80
100
17 n
16
^ 15 -
5 14 -
c
0^ -1 r-\
| H
O 12 ^
ffi 11
10
FEMALES: Mean Hemoglobin Content
» • * *
0 20 40 60 80 100
Age (years)
Figure 14. Mean Values of Hemoglobin Content as Derived from the 1999-2002 NHANES
Dataset, with Comparison to Current Physiology.txt Values
A-26
-------
MALES: Hemoglobin Content Standard Deviation
"S
O
O
ffi
2 n
1.5 -
1 <
0.5 -
n
» » »
r
^^"^ 2 ^ ^
4 New Values
20 40 60 80 100
Age (years)
FEMALES: Hemoglobin Content Standard Deviation
O
O
0.5 -
» + « » /—•-
20
40 60
Age (years)
80
100
Figure 15. Values of Hemoglobin Content Standard Deviation as Derived from the 1999-
2002 NHANES Dataset, with Comparison to Current Physiology.txt Values
A-27
-------
6. BLOOD VOLUME AS A FUNCTION OF HEIGHT AND
WEIGHT
In APEX, blood volume is estimated as a function of height and weight by the following
equation:
VUood = BVF*Weight+ K*Height3 - 30
where Vbiood is the blood volume (ml), Weight is in pounds, and height is in inches. BVF
is the blood volume factor that is read in from the physiology file, and K is a gender-
dependent constant (0.00683 for males, 0.00678 for females). This is a modification of
Allen's equation [168] to include the age/gender dependent BVF and adjusted for the
given units.
As previously mentioned, the data upon which the BVF values in the physiology file
were based could not be identified. The available documentation for pNEM documents a
non-age-dependent use of these equations.
In addition, no appropriate data were found for deriving new estimates for the BVF
variable as a function of age and gender for use with the Allen equations. It should be
noted however, that these equations were modified by Nadler [169]. These equations
seem to be used somewhat more often than the originals in the literature.
In addition, other (more recent) equations exist for estimation of blood volume from
height and weight specifically in children [170,171] or body surface area [172]. In
particular, Linderkamp et al. [170] derived prediction equations for blood volume as a
function of a number of physiological parameters for children in three different age
groups. It is recommended that further analysis of this study and others be undertaken.
However, inclusion of new blood volume equations in APEX would require changes
beyond the current physiology file (i.e. other, more intensive, code changes would be
needed). Thus, at the present time, no specific improvements to the current BVF values
in the physiology file can be made.
A-28
-------
REFERENCES
1. Astrand, I. 1960. Aerobic work capacity in men and women with special
reference to age. Acta. Phys. Scand. 49: Suppl. 169.
2. Astrand, PO and Rodhal K. 1977. Textbook of Work Physiology. 2nd addition,
McGraw-Hill, NY.
3. Heil DP, Freedson PS, Ahlquist LE, Price J, and Rippe JM. 1995. Nonexercise
regression models to estimate peak oxygen consumption. Med. Sci. Sports Exer
27:599-606.
4. Brainard J and Burmaster D. 1992. Bivariate distributions for height and weight
of men and women in the United States. Risk Analysis, 12(2):267-275.
5. Burmaster DE and Crouch E. 1997. Lognormal distributions of body weight as a
function of age for males and females in the United States, 1976-1980. Risk
Analysis, 17(4): 499-505.
6. Centers for Disease Control and Prevention (CDC). National Center for Health
Statistics (NCHS). National Health and Nutrition Examination Survey Data.
Hyattsville, MD: U.S. Department of Health and Human Services, Centers for
Disease Control and Prevention, 1976-1980.
.
7. U.S. Environmental Protection Agency (EPA). (2003) Handbook for use of data
from the national health and nutrition examination surveys (NHANES). National
Center for Environmental Assessment, Washington, DC: EPA/600/R-02/044.
Available from: National Technical Information Service, Springfield, VA, and
.
8. Centers for Disease Control and Prevention (CDC). National Center for Health
Statistics (NCHS). National Health and Nutrition Examination Survey Data.
Hyattsville, MD: U.S. Department of Health and Human Services, Centers for
Disease Control and Prevention, 1999-2004.
.
9. Center for Disease Control. National Center For Health Statistics. The National
Health and Nutriation Examination Survey (NHANES). Analytic and Reporting
Guidelines. December 2005. Available online at www.cdc.gov.
10. Andersen KL, Ilmarinen J, Rutenfranz J, Ottmann W, Berndt I, Kylian H. Leisure
time sport activities and maximal aerobic power during late adolescence. Eur J
Appl Physiol 1984; 52: 431-6.
11. Armstrong N, Kirby BJ, McManus AM, Welsman JR. Aerobic fitness of
prepubescent children. AnnHumBiol. 1995 Sep-Oct;22(5):427-41.
12. Armstrong N, Welsman JR. Assessment and interpretation of aerobic fitness in
children and adolescents.Exerc Sport Sci Rev. 1994;22:435-76.
13. Astrand I. Aerobic work capacity in men and women with special reference to
age. Acta Physiol Scand. 1960;49(Suppl 169): 1-92.
14. Eliakim A, Scheett T, Allmendinger N, Brasel JA, Cooper DM. Training, muscle
volume, and energy expenditure in nonobese American girls. J Appl Physiol.
2001 Jan;90(l):35-44.
A-29
-------
15. Falgairette, G. 1989. Evolution de la puissance maximal aerobie de 1'enfance a
1'age adulte: influence de 1'activite physique et al. sportive. Revue Des Sciences
et al. Techniques en Activites Physiques et al. Sportives. 10:43-58.
16. Fleg JL, Lakatta EG. Role of muscle loss in the age-associated reduction in VO2
max. J Appl Physiol. 1988 Sep;65(3): 1147-51.
17. Ghesquiere J, Reybrouck T, Faulkner JA, Cattaert A, Fagard R, Amery A.
Anaerobic threshold for long-term exercise and maximal exercise performance.
Ann ClinRes. 1982; 14 Suppl 34:37-41.
18. Kamon E, Pandolf KB. Maximal aerobic power during laddermill climbing,
uphill running, and cycling. J Appl Physiol. 1972 Apr;32(4):467-73.
19. Katch VL, Park MW. Minute-by-minute oxygen requirement and work efficiency
for constant-load exercise of increasing duration. Res Q. 1975 Mar;46(l):38-47.
20. McMiken DF. Oxygen deficit and repayment in submaximal exercise. Eur J Appl
Physiol Occup Physiol. 1976 Jun 21;35(2): 127-36.
21. Nevill A, Rowland T, Goff D, Mattel L, Ferrone L. Scaling or normalising
maximum oxygen uptake to predict 1-mile run time in boys. Eur J Appl Physiol.
2004 Jul;92(3):285-8. Epub 2004 Apr 9.
22. Pivarnik JM, Fulton JE, Taylor WC, Snider SA. Aerobic capacity in black
adolescent girls. Res Q Exerc Sport. 1993 Jun;64(2):202-7.
23. Posner JD, Gorman KM, Klein HS, Cline CJ. Ventilatory threshold:
measurement and variation with age. J Appl Physiol. 1987 Oct;63(4):1519-25.
24. Reybrouck T, Heigenhauser GF, Faulkner JA. Limitations to maximum oxygen
uptake in arms, leg, and combined arm-leg ergometry. J Appl Physiol. 1975
May;38(5):774-9.
25. Ribisl PM, Kachadorian WA. Maximal oxygen intake prediction in young and
middle-aged males. J Sports Med Phys Fitness. 1969 Mar;9(l): 17-22.
26. Schiller BC, Casas YG, Desouza CA, Seals DR. Maximal aerobic capacity across
age in healthy Hispanic and Caucasian women. J Appl Physiol. 2001
Sep;91(3): 1048-54.
27. Seals DR, Hagberg JM, Hurley BF, Ehsani AA, Holloszy JO. Endurance training
in older men and women. I. Cardiovascular responses to exercise. J Appl Physiol.
1984 Oct;57(4): 1024-9.
28. Sheel A, Seddon N, Knight A, McKenzie DC, R Warburton DE. Physiological
responses to indoor rock-climbing and their relationship to maximal cycle
ergometry. Med Sci Sports Exerc. 2003 Jul;35(7): 1225-31.
29. Talbot LA, Metter EJ, Fleg JL. Leisure-time physical activities and their
relationship to cardiorespiratory fitness in healthy men and women 18-95 years
old. Med Sci Sports Exerc. 2000 Feb;32(2):417-25.
30. Tanaka H, Desouza CA, Jones PP, Stevenson ET, Davy KP, Seals DR. Greater
rate of decline in maximal aerobic capacity with age in physically active vs.
sedentary healthy women. J Appl Physiol. 1997 Dec;83(6): 1947-53.
31. Adams. Personal Communication - T. McCurdy EPA.
32. Ahmad F, Kavey RE, Kveselis DA, Gaum WE, Smith FC. Responses of non-
obese white children to treadmill exercise. J Pediatr. 2001 Aug;139(2):284-90.
33. Andersen KL, Hermansen L. Aerobic work capacity in middle-aged Norwegian
men. J Appl Physiol. 1965 May;20(3):432-6.
A-30
-------
34. Andersen KL, Ilmarinen J, Rutenfranz J, Ottmann W, Berndt I, Kylian H, Ruppel
M. Leisure time sport activities and maximal aerobic power during late
adolescence. Eur J Appl Physiol Occup Physiol. 1984;52(4):431-6.
35. Andersen KL, Seliger V, Rutenfranz J, Mocellin R. Physical performance
capacity of children in Norway. Part I. Population parameters in a rural inland
community with regard to maximal aerobic power. Eur J Appl Physiol Occup
Physiol. 1974;33(3): 177-95.
36. Armstrong N and Welsman J. Daily physical activity estimated from continuous
heart rate monitoring and laboratory indices of aerobic fitness in pre-adolescent
children, (abs.) Res Q Exer Sport 64:A-24, 1993.
37. Armstrong N, Balding J, Gentle P, Kirby B. Estimation of coronary risk factors
in British schoolchildren: a preliminary report. Br J Sports Med. 1990
Mar;24(l):61-6.
38. Armstrong N, Balding J, Gentle P, Kirby B. Patterns of physical activity among
11 to 16 year old British children. BMJ. 1990 Jul 28;301(6745):203-5.
39. Armstrong N, Bray S. Physical activity patterns defined by continuous heart rate
monitoring. Arch Dis Child. 1991 Feb;66(2):245-7.
40. Armstrong N, Davies B, Heal M. The specificity of energy utilisation by trained
and untrained adolescent boys. Br J Sports Med. 1983 Sep; 17(3): 193-9.
41. Armstrong N, Kirby BJ, McManus AM, Welsman JR. Aerobic fitness of
prepubescent children. Ann Hum Biol. 1995 Sep-Oct;22(5):427-41. Erratum in:
Ann Hum Biol 1996 Mar-Apr;23(2):188.
42. Armstrong N, Welsman JR, Kirby BJ. Submaximal exercise and maturation in
12-year-olds. J Sports Sci. 1999Feb;17(2):107-14.
43. Armstrong N, Welsman JR, Kirby BJ. Peak oxygen uptake and maturation in 12-
yr olds. Med Sci Sports Exerc. 1998 Jan;30(l): 165-9.
44. Armstrong N, Williams J, Balding J, Gentle P, Kirby B. The peak oxygen uptake
of British children with reference to age, sex and sexual maturity. Eur J Appl
Physiol Occup Physiol. 1991;62(5):369-75.
45. Atomi Y, Fukunaga T, Yamamoto Y, Hatta H. Lactate threshold and vo2max of
trained and untrained boys relative to muscle mass and composition. In:
Rutenfranz J, Mocellin R, Klimt F, eds. children and exercise XII. International
series on sport sciences, Vol 16. Champaign, IL: Human Kinetics, 1986:227-232.
46. Atomi Y, Iwaoka K, Hatta H, Miyashita M, Yamamoto Y. Daily physical activity
levels in preadolescent boys related to VO2max and lactate threshold. Eur J Appl
Physiol Occup Physiol. 1986;55(2):156-61.
47. Bailey, D.A., Ross, W.D., Mirwald, R.L. and Weese, C. Size dissociation of
maximal aerobic power during growth in boys. Medicine in Sport 1978; 11: 140-
151.
48. Bale P. Pre- and post-adolescents' physiological response to exercise. Br J Sports
Med. 1981Dec;15(4):246-9.
49. Bar-Or O, Skinner JS, Bergsteinova V, Shearburn C, Royer D, Bell W, Haas J,
Buskirk ER. Maximal aerobic capacity of 6—15 year-old girls and boys with
subnormal intelligence quotients. ActaPaediatr Scand Suppl. 1971;217:108-13.
50. Becker DM, Vaccaro P. Anaerobic threshold alterations caused by endurance
training in young children. J Sports Med Phys Fitness. 1983 Dec;23(4):445-9.
A-31
-------
51. Binkhorst 1984 (Ref??) From T. McCurdy spreadsheet, dated Nov 15, 2000.
52. Binkhorst RA, Saris WHM, Noordeloos AM, et al. Maximal oxygen consumption
of children (6 to 18 years) predicted from maximal and submaximal values in
treadmill and bicycle tests. In: Rutenfranz J, Mocellin R, Klimt F, eds. children
and exercise XII. International series on sport sciences, Vol 16. Champaign, IL:
Human Kinetics, 1986:227-232.
53. Boileau RA, Bonen A, Heyward VH, Massey BH. Maximal aerobic capacity on
the treadmill and bicycle ergometer of boys 11-14 years of age. J Sports Med
Phys Fitness. 1977 Jun; 17(2): 153-62.
54. Bonen A, Heyward VH, Cureton KJ, Boileau RA, Massey BH. Prediction of
maximal oxygen uptake in boys, ages 7-15 years. Med Sci Sports. 1979
Spring; ll(l):24-9.
55. Brown, J.D., A.D. Mahon and D.M. Plank. Attainment of maximal exercise
criteria in boys and men. Journal of Sports Medicine and Physical Fitness, 42:
135-140, 2002.
56. Buskirk E, Taylor HL. Maximal oxygen intake and its relation to body
composition, with special reference to chronic physical activity and obesity. J
ApplPhysiol. 1957 Jul;ll(l):72-8.
57. Chausow 1984 (Ref??) In T. McCurdy spreadsheet, dated Nov 15, 2000.
58. Cooper DM, Berry C, Lamarra N, Wasserman K. Kinetics of oxygen uptake and
heart rate at onset of exercise in children. J Appl Physiol. 1985 Jul;59(l):211-7.
59. Cooper DM, Weiler-Ravell D, Whipp BJ, Wasserman K. Aerobic parameters of
exercise as a function of body size during growth in children.! Appl Physiol. 1984
Mar;56(3):628-34.
60. Cumming GR, Borysyk LM. Criteria for maximum oxygen uptake in men over
40 in a population survey. Med Sci Sports. 1972 Spring;4(l): 18-22.
61. Cumming GR, Everatt D, Hastman L. Bruce treadmill test in children: normal
values in a clinic population. Am J Cardiol. 1978 Jan;41(l):69-75.
62. Cumming GR, Goulding D, Baggley G. Failure of school physical education to
improve cardiorespiratory fitness. Can Med Assoc J. 1969 Jul 26;101(2):69-73.
63. Cunningham DA, Paterson DH, Blimkie CJ, Donner AP. Development of
cardiorespiratory function in circumpubertal boys: a longitudinal study. J Appl
Physiol. 1984Feb;56(2):302-7.
64. Cunningham DA, Stapleton JJ, MacDonald 1C, Paterson DH. Daily energy
expenditure of young boys as related to maximal aerobic power. Can J Appl Sport
Sci. 1981Dec;6(4):207-ll.
65. Cunningham DA, Telford P, Swart GT. The cardiopulmonary capacities of young
hockey players: age 10. Med Sci Sports. 1976 Spring;8(l):23-5.
66. Cunningham DA, van Waterschoot BM, Paterson DH, Lefcoe M, Sangal SP.
Reliability and reproducibility of maximal oxygen uptake measurement in
children. Med Sci Sports. 1977 Summer;9(2): 104-8.
67. Cureton KJ, Sloniger MA, Black DM, McCormack WP, Rowe DA. Metabolic
determinants of the age-related improvement in one-mile run/walk performance in
youth. Med Sci Sports Exerc. 1997 Feb;29(2):259-67.
A-32
-------
68. Cureton KJ, Sloniger MA, O'Bannon JP, Black DM, McCormack WP. A
generalized equation for prediction of VO2peak from 1-mile run/walk
performance. Med Sci Sports Exerc. 1995 Mar;27(3):445-51.
69. Davies CT. Limitations to the prediction of maximum oxygen intake from
cardiac frequency measurements. J Appl Physiol. 1968 May;24(5):700-6.
70. Davies CT. Metabolic cost of exercise and physical performance in children with
some observations on external loading. Eur J Appl Physiol Occup Physiol.
1980;45(2-3):95-102.
71. Dehn MM, Bruce RA. Longitudinal variations in maximal oxygen intake with
age and activity. J Appl Physiol. 1972 Dec;33(6):805-7.
72. Docherty D, Wenger HA, Collis ML. The effects of resistance training on aerobic
and anaerobic power of young boys.Med Sci Sports Exerc. 1987 Aug;19(4):389-
92.
73. Eiberg S, Hasselstrom H, Gronfeldt V, Froberg K, Svensson J, Andersen LB.
Maximum oxygen uptake and objectively measured physical activity in Danish
children 6-7 years of age: the Copenhagen school child intervention study. Br J
Sports Med. 2005 Oct;39(10):725-30.
74. Eisenman PA, Golding LA. Comparison of effects of training on Vo2max in girls
and young women. Med Sci Sports. 1975 Summer;7(2): 136-8.
75. Fawkner S, Armstrong N. Modeling the vo2 kinetic response to moderate
intensity exercise in children. Acta Kinesiol Universit Tartu 7:80-84, 2002.
76. Fox EL, Bartels RL, Billings CE, O'Brien R, Bason RM, Mathews DK.
Frequency and duration of interval training programs and changes in aerobic
power. J Appl Physiol 1975; 38:481-84
77. Fredriksen PM, Ingjer F, Nystad W, Thaulow E. A comparison of VO2(peak)
between patients with congenital heart disease and healthy subjects, all aged 8-17
years. Eur J Appl Physiol Occup Physiol. 1999 Oct;80(5):409-16.
78. Gastin PB, Costill DL, Lawson DL, Krzeminski K, McConell GK. Accumulated
oxygen deficit during supramaximal all-out and constant intensity exercise. Med
Sci Sports Exerc. 1995 Feb;27(2):255-63.
79. George JD. Alternative approach to maximal exercise testing and VO2 max
prediction in college students. Res Q Exerc Sport. 1996 Dec;67(4):452-7.
80. Gilliam TB, Sady S, Thorland WG, Weltman AL. Comparison of peak
performance measures in children ages 6 to 8, 9 to 10, and 11 to 13 years. Res Q.
1977Dec;48(4):695-702.
81. Gilliam, T. & Freedson, P. Effects of a twelve-week school physical fitness
program on peak VO2, body composition, and blood lipids in 7 to 9 year old
children. Int J of Sports Med 1980; 1:73-78.
82. Gutin B, Barbeau P, Owens S, Lemmon CR, Bauman M, Allison J, Kang HS,
Litaker MS. Effects of exercise intensity on cardiovascular fitness, total body
composition, and visceral adiposity of obese adolescents. Am J Clin Nutr. 2002
May;75(5):818-26.
83. Hansen, H. S., K. Froberk, J. R. Nielsen, and N. Hyldebrandt. A new approach to
assessing maximal aerobic power in children: the Odense School Child Study.
Euro. J. Appl. Physiol. 58: 618-624, 1989.
A-33
-------
84. Harms C, Cordain L, Stager J, Sockler J, and Harris M. Body fat mass affects
postexercise metabolism in males of similar lean body mass. Med Exer Nutr
Health 4:33-39, 1995.
85. Hebestreit H, Bar-Or O, McKinty C, Riddell M, Zehr P. Climate-related
corrections for improved estimation of energy expenditure from heart rate in
children. J Appl Physiol. 1995 Jul;79(l):47-54.
86. Hebestreit H, Kriemler S, Hughson RL, Bar-Or O. Kinetics of oxygen uptake at
the onset of exercise in boys and men. J Appl Physiol. 1998 Nov;85(5):1833-41.
87. Hebestreit H, Mimura K, Bar-Or O. Recovery of muscle power after high-
intensity short-term exercise: comparing boys and men. J Appl Physiol. 1993
Jun;74(6):2875-80.
88. Hermansen L, Andersen KL. Aerobic work capacity in young Norwegian men
and women. J Appl Physiol. 1965 May;20(3):425-31.
89. Hossack KF, Bruce RA. Maximal cardiac function in sedentary normal men and
women: comparison of age-related changes. J Appl Physiol. 1982 Oct;53(4):799-
804.
90. Humphrey LD, Falls HB. Circulatory and metabolic responses of young women
at HR = 180 and HR = max. Res Q. 1975 Dec;46(4):498-504.
91. Hunt B. R, George J.D., Vehrs P.R., Fisher A. G., Fellingham G. W. (2000),
Validity of a submaximal 1-mile track jog test in predicting VO2max in fit
teenagers. Pediatric exercise science 12(1):80-90.
92. Ikai M, Kitagawa K. Maximum oxygen uptake of Japanese related to sex and
age. Med Sci Sports. 1972 Fall;4(3): 127-31.
93. Kaminsky L, and Whaley M. Effect of interval-type exercise on excess
postexercise oxygen consumption in obese and normal-weight women. Med Exer
Nutr Health 2:106-111,1993.
94. Kaminsky LA, Wehrli KW, Mahon AD, Robbins GC, Powers DL, Whaley MH.
Evaluation of a shallow water running test for the estimation of peak aerobic
power. Med Sci Sports Exerc. 1993 Nov;25(ll): 1287-92.
95. Kanaley JA, Boileau RA. The onset of the anaerobic threshold at three stages of
physical maturity. J Sports Med Phys Fitness. 1988 Dec;28(4):367-74.
96. Katch FI, McArdle WD, Pechar GS. Relationship of maximal leg force and leg
composition to treadmill and bicycle ergometer maximum oxygen uptake. Med
Sci Sports. 1974 Spring;6(l):38-43.
97. Kemper H and Vershuur R. Maximal aerobic power in 13 and 14 year old
teenagers in relation to biologic age. Int J Sports Med. 2:97-100, 1981.
98. Koch G, Eriksson BO. Effect of physical training on pulmonary ventilation and
gas exchange during submaximal and maximal work in boys aged 11 to 13 years.
Scand J Clin Lab Invest. 1973 Jan;31(l):87-94.
99. Kohrt WM, Malley MT, Coggan AR, Spina RJ, Ogawa T, Ehsani AA, Bourey
RE, Martin WH 3rd, Holloszy JO. Effects of gender, age, and fitness level on
response of VO2max to training in 60-71 yr olds. J Appl Physiol. 1991
Nov;71(5):2004-ll.
100. Krahenbuhl GS, Pangrazi RP, Burkett LN, Schneider MJ, Petersen G. Field
estimation of VO2 max in children eight years of age. Med Sci Sports. 1977
Spring;9(l):37-40.
A-34
-------
101. Krahenbuhl GS, Pangrazi RP, Petersen GW, Burkett LN, Schneider MJ. Field
testing of cardiorespiratory fitness in primary school children. Med Sci Sports.
1978Fall;10(3):208-13.
102. Krahenbuhl GS, Pangrazi RP. Characteristics associated with running
performance in young boys. Med Sci Sports Exerc. 1983;15(6):486-90.
103. Loftin M, Strikmiller P, Warren B, Myers L, Schroth L, Pittman J, Harsha D, and
Sothern M. Comparison and relationship of Vo2peak and physical activity
patterns in elementary and high school females. Fed Exer Sci. 10:153-163, 1998.
104. Macek M and Vavra J. The adjustment of oxygen uptake at the onset of exercise:
a comparison between prepubertal boys and young adults. In J Sports Med 1:70-
72, 1980.
105. Macek M, Vavra J, Novosadova J. Prolonged exercise in prepubertal boys. I.
Cardiovascular and metabolic adjustment. Eur J Appl Physiol Occup Physiol.
1976Sep23;35(4):291-8.
106. Macek M, Vavra J. Cardiopulmonary and metabolic changes during exercise in
children 6-14 years old. J Appl Physiol. 1971 Feb;30(2):200-4.
107. Mahler DA, Mejia-Alfaro R, Ward J, Baird JC. Continuous measurement of
breathlessness during exercise: validity, reliability, and responsiveness. J Appl
Physiol. 2001 Jun;90(6):2188-96.
108. Mahon AD, Anderson CS, Hipp MJ, Hunt KA. Heart rate recovery from
submaximal exercise in boys and girls. Med Sci Sports Exerc. 2003
Dec;35(12):2093-7.
109. Mahon AD, Gay JA, Stolen KQ. Differentiated ratings of perceived exertion at
ventilatory threshold in children and adults. Eur J Appl Physiol Occup Physiol.
1998 Jul;78(2): 115-20.
110. Mahon AD, Vaccaro P. Ventilatory threshold and VO2max changes in children
following endurance training. Med Sci Sports Exerc. 1989 Aug;21(4):425-31.
111. Mahon, A.D., C.C. Cheatham, K.Q. Kelsey, and J.D. Brown. Plasma fibrinogen,
physical activity and aerobic fitness in children. In: Children and Exercise XIX:
19th International Seminar on Paediatric Work Physiology, N. Armstrong, B.
Kirby, and J. Wellsman, Editors. London: Chapman & Hall Ltd., 1997. rs.
London: Chapman & Hall Ltd., 997
112. Maksud MG, Coutts KD, Hamilton LH. Oxygen uptake, ventilation, and heart
rate. Study in Negro children during strenuous exercise. Arch Environ Health.
1971 Jul;23(l):23-8.
113. Mandigout S, Lecoq AM, Courteix D, Guenon P, Obert P. Effect of gender in
response to an aerobic training programme in prepubertal children. Acta Paediatr.
2001 Jan;90(l):9-15.
114. Massicotte DR, Macnab RB. Cardiorespiratory adaptations to training at
specified intensities in children. Med Sci Sports. 1974 Winter;6(4):242-6.
115. Matsui H, Miyashita M, Miura M, Kobayashi K, Hoshikawa T. Maximum oxygen
intake and its relationship to body weight of Japanese adolescents.Med Sci Sports.
1972 Spring;4(l):27-32.
116. McArdle WD and Magel JR. Physical work capacity and maximum oxygen
uptake in treadmill and bicycle exercise. Med Sci Sports 2: 118-126, 1971
McCardle et al. 1972
A-35
-------
117. McDonough JR, Kusumi F, Bruce RA. Variations in maximal oxygen intake with
physical activity in middle-aged men. Circulation. 1970 May;41(5):743-51.
118. McManus AM, Armstrong N, Williams CA. Effect of training on the aerobic
power and anaerobic performance of prepubertal girls. Acta Paediatr. 1997
May;86(5):456-9.
119. McMurray RG, Guion WK, Ainsworth BE, Harrell JS. Predicting aerobic power
in children. A comparison of two methods. J Sports Med Phys Fitness. 1998
Sep;38(3):227-33.
120. Mercier J, Varray A, Ramonatxo M, Mercier B, Prefaut C. Influence of
anthropometric characteristics on changes in maximal exercise ventilation and
breathing pattern during growth in boys. Eur J Appl Physiol Occup Physiol.
1991;63(3-4):235-41.
121. Murray TD, Walker JL, Jackson AS, Morrow JR Jr, Eldridge JA, Rainey DL.
Validation of a 20-minute steady-state jog as an estimate of peak oxygen uptake
in adolescents. Res Q Exerc Sport. 1993 Mar;64(l):75-82.
122. Nakagawa A, Ishiko T. Assessment of aerobic capacity with special reference to
sex and age of junior and senior highschool students in Japan.
123. Palgi Y, Gutin B, Young J, Alejandro D. Physiologic and anthropometric factors
underlying endurance performance in children. Int J Sports Med. 1984
Apr;5(2):67-73.
124. Paterson DH, Cunningham DA, Bumstead LA. Recovery O2 and blood lactic
acid: longitudinal analysis in boys aged 11 to 15 years. Eur J Appl Physiol Occup
Physiol. 1986;55(l):93-9.
125. Paterson DH, Cunningham DA, Donner A. The effect of different treadmill
speeds on the variability of VO2 max in children. Eur J Appl Physiol Occup
Physiol. 1981;47(2): 113-22.
126. Paterson DH, McLellan TM, Stella RS, Cunningham DA. Longitudinal study of
ventilation threshold and maximal O2 uptake in athletic boys. J Appl Physiol.
1987 May;62(5):2051-7.
127. Petzl DH, Haber P, Schuster E, Popow C, Haschke F. Reliability of estimation of
maximum performance capacity on the basis of submaximum ergometric stress
tests in children 10-14 years old. Eur J Pediatr. 1988 Feb; 147(2): 174-8.
128. Pivarnik JM, Bray MS, Hergenroeder AC, Hill RB, Wong WW. Ethnicity affects
aerobic fitness in US adolescent girls. Med Sci Sports Exerc. 1995
Dec;27(12): 1635-8.
129. Pivarnik JM, Dwyer MC, Lauderdale MA. The reliability of aerobic capacity
(VO2max) testing in adolescent girls. Res Q Exerc Sport. 1996 Sep;67(3):345-8.
130. Pivarnik, J.M., W.C. Taylor, and S.S. Cummings. Longitudinal assessment of
aerobic fitness in middle-school African American girls. Pediatr Exerc Sci 10: 21-
27, 1998.
131. Reybrouck T, Ghesquiere J, Weymans M, Amery A. Ventilatory threshold
measurement to evaluate maximal endurance performance. Int J Sports Med. 1986
Feb;7(l):26-9.
132. Rivera-Brown, A.M. and Frontera, W.R. Achievement of plateau and reliability of
VO2max in active adolescents: Effects of testing ergometer. Fed. Exer. Sci.,
10:164-175, 1998.
A-36
-------
133. Rivera-Brown, A.M., Alvarez ,A., Rodriguez-Santana, J., Benetti, P. Anaerobic
power and achievement of VO2 plateau in pre-pubertal boys. Int. J. Sports Med.
22:111-115,2001.
134. Rivera-Brown, AM, Rivera, MA, Frontera, WR (1992) Applicability of criteria
for VO2max in active children Pediatr Exerc Sci 4,331-339.
135. Rogers DM, Olson BL, Wilmore JH. Scaling for the VO2-to-body size
relationship among children and adults. J Appl Physiol. 1995 Sep;79(3):958-67.
136. Rogers, DM, Turley, KR, Kujawa, KI, Harper, KM, Wilmore, J. (1995)
Allometric scaling factors for oxygen uptake during exercise in children Pediatr
Exerc Sci 7,12-25.
137. Rowland T, Kline G, Goff D, Martel L, and Ferrone L. Physiological
determinants of maximal aerobic power in healthy 12-year-old boys. Fed Exer
Sci 11:317-326, 1999.
138. Rowland TW, Green GM. Physiological responses to treadmill exercise in
females: adult-child differences. Med Sci Sports Exerc. 1988 Oct;20(5):474-8.
139. Rowland TW, Martel L, Vanderburgh P, Manos T, Charkoudian N. The
influence of short-term aerobic training on blood lipids in healthy 10-12 year old
children. Int J Sports Med. 1996 Oct;17(7):487-92.
140. Rowland, T, Miller K, Vanderburgh P, Goff D, Martel L, and Ferrone L.
Cardiovascular fitness in premenarcheal girls and young women. Int J Sports Med
20: 1-5, 1999.
141. Rutenfranz J, Macek M, Lange Andersen K, Bell RD, Vavra J, Radvansky J,
Klimmer F, Kylian H. The relationship between changing body height and
growth related changes in maximal aerobic power. Eur J Appl Physiol Occup
Physiol. 1990;60(4):282-7.
142. Sady, S. Transient oxygen uptake and heart rate responses at the onset of relative
endurance exercise in prepubertal boys and adult men. Int. J. Sports Med. 2: 240-
244, 1981.
143. Sargent 1985 (Ref??) From T. McCurdy spreadsheet, dated Nov 15, 2000.
144. Saris WH, Noordeloos AM, Ringnalda BEM, et al. Binkhorst RA, Kemper HCG,
Saris WHM, editors. Children and Exercise XI. Champaign: Human Kinetics
Publisher; 1985; Reference values for aerobic power of healthy 4- to 18-year-old
Dutch children: Preliminary results, p. 151-60.
145. Savage MP, Petratis MM, Thomson WH, Berg K, Smith JL, Sady SP. Exercise
training effects on serum lipids of prepubescent boys and adult men. Med Sci
Sports Exerc. 1986 Apr; 18(2): 197-204.
146. Shuleva K, Hunter G, Hester D, Dunaway D. Exercise oxygen uptake in 2 to 6
year olds. Fed Exer Sci 2:130-139, 1990.
147. Sobolova V, Seliger V, Grussova D, Machovcova J, Zelenka V. The influence of
age and sports training in swimming on physical fitness. Acta Paediatr Scand
Suppl. 1971;217:63-7.
148. Spurr, G.B. & Reina, J.C. (1990). Daily pattern of %VO2 max and heart rates in
normal and undernourished school children. Med. Sci. Sports Exerc. 22:643-652.
149. Stewart KJ, Gutin B. Effects of physical training on cardiorespiratory fitness in
children. Res Q. 1976 Mar;47(l): 110-20.
A-37
-------
150. Stoedefalke K, Armstrong N, Kirby BJ, Welsman JR. Effect of training on peak
oxygen uptake and blood lipids in 13 to 14-year-old girls. Acta Paediatr. 2000
Nov;89(ll): 1290-4.
151. Sundberg S, Elovainio R. Cardiorespiratory function in competitive endurance
runners aged 12-16 years compared with ordinary boys. Acta Paediatr Scand.
1982Nov;71(6):987-92.
152. Sunnegardh J, Bratteby LE. Maximal oxygen uptake, anthropometry and physical
activity in a randomly selected sample of 8 and 13 year old children in Sweden.
Eur J Appl Physiol Occup Physiol. 1987;56(3):266-72.
153. Tolfrey K, Campbell IG, Batterham AM. Exercise training induced alterations in
prepubertal children's lipid-lipoprotein profile. Med Sci Sports Exerc. 1998
Dec;30(12): 1684-92.
154. W Bell (1997) Aerobic fitness and fat patterning in adolescent boys. In: Children
and Exercise (Eds N Armstrong, B Kirby, J Welsman) pp 63-68. London: E & F
Spon.
155. W.D. McArdle et al. (1972) Reliability and interrelationships between maximal
oxygen uptake, physical work capacity and step test scores in college women.
Medicine and Science in Sports, Vol 4, p!82-186.
156. Washington RL, van Gundy JC, Cohen C, Sondheimer HM, Wolfe RR. Normal
aerobic and anaerobic exercise data for North American school-age children. J
Pediatr. 1988 Feb;112(2):223-33.
157. Welsman JR, Armstrong N, Nevill AM, Winter EM, Kirby B J. Scaling peak VO2
for differences in body size. Med Sci Sports Exerc. 1996 Feb;28(2):259-65.
158. Welsman JR, Armstrong N, Withers S. Responses of young girls to two modes of
aerobic training. Br J Sports Med. 1997 Jun;31(2): 139-42.
159. Welsman JR, Armstrong N. Daily physical activity and blood lactate indices of
aerobic fitness in children. Br J Sports Med. 1992 Dec;26(4):228-32.
160. Williams CA, Armstrong N, Powell J. Aerobic responses of prepubertal boys to
two modes of training. Br J Sports Med. 2000 Jun;34(3): 168-73.
161. Williford FIN, Blessing DL, Duey WJ, Barksdale JM, Wang N, Olson MS, Teel S.
Exercise training in black adolescents: changes in blood lipids and Vo2max. Ethn
Dis. 1996Autumn;6(3-4):279-85.
162. Wilmore JH, Sigerseth PO. Physical work capacity of young girls, 7-13 years of
age. J Appl Physiol. 1967 May;22(5):923-8.
163. Yamaji K, Miyashita M, Shephard RJ. Relationship between heart rate and
relative oxygen intake in male subjects aged 10 to 27 years. J Hum Ergol (Tokyo).
1978 Sep;7(l):29-39.
164. Yoshizawa S, Ishizaki T, and Honda H. Aerobic power and endurance running
in young children. In: Rutenfranz J, Mocellin R, Klimt F, eds. children and
exercise XII. International series on sport sciences, Vol 16. Champaign, IL:
Human Kinetics, 1986:227-232.
165. Yoshizawa S, Honda H, NakamuraN. Effects of an 18-month endurance run
program on maximal aerobic power in 4- to 6- year old girls. Pediatr Exer Sci
9:33-43, 1997.
A-38
-------
166. Yoshizawa S, Honda H, Urushibara M, Nakamura N. Effects of endurance run on
circulorespiratory system in young children. J Hum Ergol (Tokyo). 1990
Jun;19(l):41-52.
167. Yoshizawa S, Ishizaki T, Honda H. Physical fitness of children aged 5 and 6
years. J Hum Ergol (Tokyo). 1977 Sep;6(l):41-51.
168. Allen TH, Peng MT, et al. Prediction of blood volume and adiposity in man from
body weight and cube of height. Metabolism. 1956; 5: 328.
169. Nadler SB, Hidalgo JU, Bloch T. Prediction of blood volume in normal human
adults. Surgery. 1962;51:224-232.
170. Linderkamp O, Versmold HT, et al. Estimation and prediction of blood volume in
infants and children. Europ J Pediat. 1977; 125: 227-234.
171. Cropp GJA. Changes in blood and plasma volumes during growth. J Pediatrics.
1971; 78: 220-229.
172. Smetannikov Y, Hopkins D. Intraoperative bleeding: A mathematical model for
minimizing hemoglobin loss. Transfusion. 1996; 36: 832-835.
A-39
-------
Appendix A. SAS Code for Estimating the Body Mass Distributions
K K Isaacs 10/2005
Alion Science and Technology
Data weight;
merge Demo Demo b Demo c Bmx Bmx b r Bmx c;
by SEQN;
mass=BMXWT;
gen=RIAGENDR;
ageyrs=RIDAGEYR;
agemonths=RIDAGEEX;
wt = (2/3)*WTMEC4YR;
if (SEQN>21004) THEN wt=(1/3)*WTMEC2YR;
if agemonths<12 and agemonths>0 THEN ageyrs=0;
keep SEQN mass gen ageyrs agemonths wt;
run;
proc sort data=weight;
by gen ageyrs;
run;
Proc univariate data=weight;
by gen ageyrs;
var mass;
freg wt;
histogram mass / lognormal;
run;
A-40
-------
Appendix B. SAS Code for Estimating the Normalized Vo2max
Distributions
Adams experimental data provided in Excel form by Stephen Graham and Tom McCurdy, EPA
Other data collected by Alion Science and Tech.
This work was performed for WA 10, APEX/SHEDS Physiology File Update
K. K. Isaacs October 2005
proc sort data=alldata;
by gender age;
run;
Proc univariate data=alldata;
by gender age;
var nvo2max;
histogram nvo2max / normal;
output out=outputdatal N=samplesize mean=Mean
std=StdDeviation ProbN=NormalFit;
run;
Proc export data=outputdatal outfile="H:\kki-05-PHYSIOLOGY_10\Alldata_vo2max.csv"
replace;
run;
A-41
-------
Appendix C. SAS Code for Estimating the Hemoglobin Content
Data
I* This program calculates best fit normal distributions for hemoglobin content
from the NHANES 1999-2000 and 2001-2002 datasets.
Distributions are derived from hemoglobin content and age data downloaded from the CDC
site at
http://www.cdc.gov/nchs/about/maj or/nhanes/nhanes 9 9-00.htm
and
http://www.cdc.gov/nchs/about/maj or/nhanes/nhanes01-02.htm
2001-2002
125_b.xpt (NHANES Lab dataset #25)
Demo_b.xpt (NHANES Demographic Data, contains age in years or months)
*/
Data Hb;
merge Demo Lab25 Demo b L25 b;
by SEQN; * Sample number;
Hb=LBXHGB; * Hb content g/dL;
gen=RIAGENDR; * Gender;
ageyrs=RIDAGEYR; * Age in years;
agemonths=RIDAGEEX; * Age in months;
wt = WTMEC4YR; * 4-year sample weights;
if agemonths<12 and agemonths>0 THEN ageyrs=0; * Age 0;
if ageyrs>20 then ageyrs=(floor(ageyrs/5)+1)*5; * Bin in 5-year incs;
keep SEQN Hb gen ageyrs agemonths wt;
run;
proc sort data=Hb;
by gen ageyrs;
run;
Proc univariate data=Hb;
by gen ageyrs;
var Hb;
reg wt;
histogram Hb / normal;
output out=outputs N=samplesize mean=Mean
std=StdDeviation ProbN=NormalFit;
run;
Proc export data=outputs outfile="H:\kki-05-PHYSIOLOGY_10\Hemoglobin\HbFitswt.csv"
replace;
run;
A-42
-------
Appendix D. The New Physiology.Txt File
Note: The values contained in the file conform to the current APEX read formats. That
is, the number of decimal places for each parameter is dictated by the APEX code. It is
likely that this will change in the future, at which point more significant digits could be
added to the Physiology.txt file.
Males
age 0-100, then females age
NVO2max distribution
0-100 (last revised 12-20-05)
Age
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
Source Distr
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Mean
48
48
48
49
49
49
50
50
50
51
51
51
52
52
52
53
53
53
53
54
54
54
53
52
51
51
50
49
48
48
47
46
46
45
44
44
43
42
42
41
40
40
39
39
38
38
37
36
36
35
35
34
34
34
33
33
32
32
31
.3
.6
.9
.2
.5
.8
.1
.4
.8
.1
.4
.7
.0
.3
.6
.0
.3
.6
.9
.2
.5
.2
.4
.6
.8
.1
.3
.6
.8
.1
.4
.7
.0
.3
.6
.0
.3
.7
.1
.4
.8
.2
.7
.1
.5
.0
.4
.9
.4
.9
.4
.9
.5
.0
.6
.1
.7
.3
.9
SD
1.
2 .
2 .
2 .
3 .
3 .
3 .
4 .
4 .
4 .
5.
5.
5.
5.
6 .
6 .
6 .
7 .
7.
7.
8 .
8 .
8 .
9.
9.
9.
10
10
10
10
9.
9.
9.
9.
9.
9.
8 .
8 .
8 .
7.
5.
5.
5.
5.
5.
5.
5.
5.
5.
5.
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
7
0
4
7
0
3
7
0
3
6
0
3
6
9
2
6
9
2
5
9
2
5
8
2
5
8
.7
.5
.3
.1
9
7
6
4
2
0
9
7
6
3
5
5
5
5
5
5
5
5
5
5
9
9
9
9
9
9
9
9
9
Lower
44
43
43
43
42
42
41
41
40
40
39
39
39
38
38
37
37
36
36
35
35
34
32
31
29
28
25
25
24
24
24
24
23
23
23
23
22
22
22
25
28
28
28
28
28
28
28
28
28
28
23
23
23
23
23
23
23
23
23
.3
.8
.4
.0
.5
.1
.6
.2
.8
.3
.9
.4
.0
.6
.1
.7
.3
.8
.4
.9
.5
.5
.9
.4
.8
.3
.5
.2
.9
.6
.3
.0
.8
.5
.2
.0
.7
.4
.2
.5
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.5
.5
.5
.5
.5
.5
.5
.5
.5
Upper
52
53
54
55
56
57
58
59
60
61
62
64
65
66
67
68
69
70
71
72
73
74
74
73
73
73
75
74
72
71
70
69
68
67
66
65
64
62
61
54
50
50
50
50
50
50
50
50
50
50
42
42
42
42
42
42
42
42
42
.2
.3
.4
.4
.5
.6
.6
.7
.8
.8
.9
.0
.0
.1
.2
.2
.3
.4
.4
.5
.6
.0
.0
.9
.9
.9
.2
.0
.8
.6
.4
.3
.2
.1
.0
.0
.0
.9
.9
.1
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.7
.7
.7
.7
.7
.7
.7
.7
.7
Assumptions
A-43
-------
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
31
31
30
30
30
29
29
29
28
28
28
27
27
27
27
27
26
26
26
26
26
25
25
25
25
25
25
25
25
25
24
24
24
24
24
24
24
24
24
25
25
25
35
36
36
36
37
37
37
38
38
38
39
39
39
40
40
40
41
41
41
42
42
42
41
40
40
39
39
38
37
37
36
36
35
34
34
33
33
32
32
.5
.1
.7
.4
.0
.7
.4
.1
.8
.5
.2
.9
.7
.4
.2
.0
.7
.5
.4
.2
.0
.8
.7
.6
.4
.3
.2
.1
.1
.0
.9
.9
.9
.8
.8
.8
.8
.9
.9
.0
.0
.1
.9
.2
.5
.9
.2
.5
.9
.2
.5
.9
.2
.5
.9
.2
.5
.9
.2
.5
.8
.2
.5
.1
.5
.8
.2
.6
.0
.4
.8
.2
.6
.0
.5
.9
.4
.9
.4
.9
.4
4
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
6
6
6
6
6
6
6
6
6
6
7
7
7
7
7
7
7
7
7
7
7
8
8
8
8
8
8
8
7
7
7
7
7
7
6
6
6
6
.9
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.9
.0
.1
.2
.3
.4
.5
.6
.7
.8
.9
.0
.0
.1
.2
.3
.4
.5
.6
.7
.8
.9
.0
.1
.2
.3
.4
.4
.1
.9
.7
.6
.4
.2
.0
.8
.7
.5
.4
23
21
21
21
21
21
21
21
21
21
21
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
22
22
22
22
22
22
22
22
23
23
23
23
23
23
23
23
23
24
24
24
24
23
22
22
21
20
19
18
18
18
18
18
18
18
18
18
17
17
17
.5
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.2
.3
.4
.5
.6
.7
.8
.9
.0
.1
.3
.4
.5
.6
.7
.8
.9
.0
.1
.2
.4
.7
.9
.0
.1
.3
.5
.9
.8
.7
.6
.5
.4
.2
.1
.0
.8
.7
.6
42
41
41
41
41
41
41
41
41
41
41
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
38
49
50
50
51
51
52
52
53
54
54
55
55
56
56
57
57
58
59
59
60
60
60
60
59
59
58
58
57
56
55
54
53
52
51
50
49
48
48
47
.7
.8
.8
.8
.8
.8
.8
.8
.8
.8
.8
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.6
.2
.7
.3
.8
.4
.9
.5
.0
.6
.1
.7
.2
.8
.3
.9
.5
.0
.6
.1
.7
.5
.1
.6
.2
.8
.4
.8
.7
.6
.6
.6
.6
.7
.7
.8
.9
.0
.2
A-44
-------
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
Males
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
age 0-100,
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
Normal
31.
31.
31.
30.
30.
29.
29.
28 .
28 .
28 .
27.
27.
26 .
26 .
26 .
25.
25.
25.
24 .
24 .
24 .
24 .
23 .
23 .
23 .
23 .
22 .
22 .
22 .
22 .
21.
21.
21.
21.
21.
21.
21.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
20.
then females
9
4
0
5
1
6
2
8
4
0
6
2
8
5
1
8
5
2
9
6
3
0
7
5
2
0
7
5
3
1
9
7
6
4
3
1
0
9
8
7
6
5
4
3
3
3
2
2
2
2
2
2
2
3
3
4
4
5
6
7
8
9
age
Body mass distribution,
Age
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Source
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
Distr
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
GM
7.8
11.
13 .
16 .
18 .
21.
23 .
27.
31.
34 .
38 .
44 .
48 .
55.
62 .
67.
4
9
0
5
6
1
1
7
7
3
1
0
4
8
7
6 .
6 .
6 .
5.
5.
5.
5.
5.
5.
5.
5.
5.
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
4 .
0-
kg
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
2
1
0
8
7
6
5
4
3
2
1
0
9
8
8
7
7
6
6
5
5
5
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
100
GSD
301
143
146
154
165
234
213
216
302
265
280
308
315
340
293
255
17.4
17.3
17.1
17.0
16 .8
16 .6
16 .5
16 .3
16 .1
16 .0
15.8
15.6
15.4
15.2
15.1
14 .9
14 .7
14 .5
14 .3
14 .1
13 .9
13 .6
13 .4
13 .2
13 .0
12 .8
12 .5
12 .3
12 .1
11.9
11.7
11.5
11.4
11.2
11.1
10.9
10.8
10.7
10.6
10.4
10.4
10.3
10.2
10.1
10.1
10.0
10.0
10.0
10.0
10.0
10.0
10.0
10.0
10.1
10.1
10.2
10.2
10.3
10.4
10.5
10.6
10.7
46 .4
45.6
44 .8
44 .0
43 .3
42 .6
41.9
41.2
40.6
40.0
39.4
38 .8
38 .3
37.7
37.2
36 .7
36 .3
35.9
35.4
35.1
34 .7
34 .3
34 .0
33 .7
33 .4
33 .2
33 .0
32 .7
32 .5
32 .3
32 .1
32 .0
31.8
31.6
31.5
31.3
31.2
31.1
31.0
30.9
30.8
30.7
30.6
30.6
30.5
30.5
30.4
30.4
30.4
30.4
30.4
30.4
30.4
30.5
30.5
30.6
30.6
30.7
30.8
30.9
31.0
31.1
(last revised 12-20-05)
Lower
3 .6
8 .2
9.8
11.7
11.1
13 .7
16 .1
19.3
19.1
24 .0
24 .3
26 .2
27.7
27 .7
35.7
41.5
A-45
Upper Assumptions
11.8
16 .1
20.9
23 .7
28 .1
42 .4
41.1
46 .8
66 .2
69.9
72 .9
83 .8
94 .8
106 .6
121.0
117.9
-------
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
72
73
75
77
78
78
83
80
81
84
81
85
84
82
81
81
84
88
81
87
83
85
84
84
90
87
88
88
88
87
88
86
84
86
84
88
89
89
90
88
84
87
85
84
87
89
84
89
90
89
86
86
85
87
82
79
82
85
83
84
78
79
79
77
79
75
76
74
75
71
74
73
72
72
71
70
70
69
69
68
67
.5
.1
.1
.2
.0
.2
.8
.6
.7
.8
.8
.2
.3
.1
.6
.3
.7
.2
.2
.2
.4
.8
.1
.6
.1
.4
.3
.4
.5
.1
.2
.5
.8
.2
.7
.0
.9
.0
.1
.3
.8
.5
.1
.2
.0
.0
.8
.1
.0
.9
.8
.2
.2
.1
.8
.6
.0
.6
.0
.5
.7
.4
.9
.6
.9
.4
.8
.6
.3
.8
.0
.4
.7
.1
.5
.9
.3
.6
.0
.4
.8
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
.267
.248
.243
.245
.250
.297
.292
.222
.251
.206
.273
.249
.272
.236
.262
.249
.235
.231
.221
.251
.228
.241
.260
.196
.246
.173
.205
.233
.200
.205
.243
.229
.186
.240
.179
.208
.216
.228
.216
.222
.195
.253
.266
.182
.232
.207
.228
.262
.193
.215
.228
.207
.191
.222
.210
.240
.204
.196
.217
.185
.207
.170
.195
.155
.174
.157
.180
.158
.205
.191
.170
.170
.160
.160
.160
.160
.160
.150
.150
.150
.150
45
49
51
52
50
46
53
50
50
50
48
50
51
50
52
48
49
64
53
61
45
59
52
61
58
61
62
54
56
60
54
49
56
47
53
57
55
58
64
55
45
58
51
58
57
49
56
56
59
58
54
43
61
50
46
51
51
56
53
56
55
58
41
56
56
55
54
53
41
46
50
50
50
50
49
49
49
49
49
48
48
.8
.9
.2
.6
.5
.8
.3
.5
.6
.2
.9
.0
.0
.6
.5
.8
.7
.8
.1
.0
.8
.3
.8
.2
.5
.3
.2
.0
.6
.6
.2
.9
.3
.0
.4
.9
.2
.2
.1
.1
.0
.3
.6
.7
.3
.9
.0
.3
.1
.1
.0
.1
.2
.7
.5
.0
.9
.2
.3
.5
.9
.7
.1
.4
.0
.8
.4
.2
.5
.9
.6
.4
.2
.0
.8
.6
.4
.3
.1
.9
.7
139.1
136 .6
144 .2
134 .5
130.0
199.2
155.4
137.6
132 .6
136 .1
164 .5
153 .9
167.2
147.2
139.0
170.6
135.8
146 .3
136 .9
193 .3
140.5
150.9
149.7
140.6
154 .0
117.7
144 .0
145.3
128 .9
160.2
154 .3
188 .3
128 .3
171.3
124 .4
143 .6
144 .9
143 .3
155.2
138 .6
110.3
160.0
179.0
112 .4
141.7
162 .8
152 .1
171.6
119.0
126 .3
150.1
127.5
163 .2
127.2
125.5
122 .8
132 .7
128 .3
120.0
133 .5
121.1
109.3
115.1
107.8
111.9
111.9
111.8
107.0
109.5
105.8
101.1
99.1
97.2
95.2
93 .2
91.3
89.3
87.4
85.4
83 .4
81.5
A-46
-------
97
98
99
100
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
67
66
65
65
7.
11
13
15
18
20
22
26
30
35
40
46
50
56
57
60
61
61
64
66
67
67
66
69
70
66
73
70
74
69
70
73
72
72
69
73
73
70
75
72
72
73
73
73
75
76
77
72
74
72
75
72
74
74
72
76
77
72
74
80
75
77
73
72
75
72
73
75
73
74
69
69
69
71
70
70
69
.1
.5
.9
.3
4
.1
.3
.6
.0
.4
.5
.5
.5
.2
.6
.6
.7
.6
.2
.1
.6
.2
.6
.2
.0
.2
.8
.7
.3
.3
.0
.6
.4
.1
.6
.0
.9
.7
.8
.0
.5
.0
.6
.3
.9
.4
.7
.4
.7
.8
.5
.8
.6
.8
.2
.9
.5
.7
.4
.0
.3
.4
.5
.6
.8
.1
.3
.3
.4
.9
.1
.8
.2
.4
.0
.1
.9
.4
.4
.5
.5
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
.140
.140
.140
.140
.304
.163
.158
.160
.171
.229
.194
.239
.315
.271
.304
.302
.274
.275
.248
.249
.255
.248
.281
.274
.262
.262
.273
.304
.289
.283
.281
.281
.312
.250
.305
.278
.281
.307
.230
.306
.289
.284
.295
.251
.289
.268
.270
.314
.266
.308
.304
.298
.303
.261
.292
.240
.283
.259
.281
.231
.315
.252
.267
.277
.260
.240
.198
.238
.281
.254
.242
.266
.250
.225
.188
.232
.240
.240
.277
.216
.199
48
48
48
47
3 .
7.
10
11
12
12
15
16
19
20
22
27
27
33
37
34
40
41
42
41
41
39
42
40
47
44
45
41
44
39
42
43
41
44
46
44
44
48
43
41
45
50
47
45
49
41
46
47
44
45
48
42
45
46
44
53
45
48
45
50
51
50
49
46
41
35
48
47
39
48
45
45
40
47
37
46
48
.5
.3
.1
.9
7
4
.1
.0
.8
.6
.9
.9
.8
.3
.7
.7
.8
.4
.7
.9
.9
.5
.4
.6
.5
.7
.0
.3
.5
.8
.3
.4
.3
.3
.1
.7
.5
.9
.6
.2
.6
.1
.7
.6
.5
.5
.1
.6
.5
.6
.6
.8
.2
.1
.4
.5
.7
.2
.3
.6
.6
.6
.0
.9
.3
.7
.7
.9
.1
.9
.4
.2
.3
.0
.9
.5
.7
.4
.4
.8
.8
79.5
77.6
75.6
73 .6
12 .1
15.3
20.4
27.9
29.1
40.4
36 .7
51.0
60.8
58 .6
71.2
84 .6
93 .3
99.5
110.0
108 .4
113 .8
133 .1
123 .6
118 .5
122 .6
123 .7
123 .5
143 .0
144 .5
131.8
128 .9
140.9
142 .1
116 .3
151.5
125.9
139.7
135.2
115.3
138 .4
150.1
152 .1
151.7
123 .1
137.4
156 .9
146 .1
159.5
153 .0
141.5
145.8
130.6
166 .0
125.5
175.7
120.2
146 .6
176 .6
123 .1
125.6
134 .9
122 .6
117.7
133 .0
128 .3
125.6
121.1
119.9
132 .5
113 .7
113 .3
123 .8
120.7
118 .0
102 .8
108 .1
103 .8
127.6
106 .4
117.4
101.7
A-47
-------
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
Males
Age
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
CDC
age 0-100
Source
R47g
R47g
R47g
R47h
R47h
R47h
R47h
R47h
R47h
R47h
R47i
R47i
R47i
R47i
R47i
R47i
R47i
R47i
R47J
R47J
R47J
R47J
R47J
R47J
R47J
R47J
R47J
R47J
R47J
R47J
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
LN
70.1
66 .4
67.8
62 .2
65.4
64 .8
62 .9
62 .2
61.5
62 .4
61.8
61.3
60.7
60.2
59.6
59.1
58 .5
58 .0
57.4
56 .9
56 .3
55.8
55.2
54 .7
then females age
Regression
DV
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
0
equation
IV
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
1.240
1.211
1.200
1.255
1.184
1.260
1.196
1.216
1.209
1.210
1.210
1.210
1.210
1.210
1.200
1.200
1.200
1.200
1.200
1.200
1.200
1.190
1.190
1.190
-100
40.3
44 .1
46 .2
41.2
42 .7
40.6
44 .7
43 .5
42 .3
41.9
41.7
41.5
41.3
41.1
40.9
40.7
40.5
40.3
40.1
39.9
39.7
39.5
39.3
39.1
119.8
109.8
98 .4
121.4
91.4
120.0
101.2
108 .4
93 .2
101.2
100.3
99.4
98 .4
97.5
96 .6
95.7
94 .8
93 .9
93 .0
92 .1
91.2
90.3
89.4
88 .5
(last revised 6-
Estimate for
Slope
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
.244
.244
.244
.095
.095
.095
.095
.095
.095
.095
.074
.074
.074
.074
.074
.074
.074
.074
.063
.063
.063
.063
.063
.063
.063
.063
.063
.063
.063
.063
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
Interc
-0.127
-0.127
-0.127
2 .110
2 .110
2 .110
2 .110
2 .110
2 .110
2 .110
2 .754
2 .754
2 .754
2 .754
2 .754
2 .754
2 .754
2 .754
2 .896
2 .896
2 .896
2 .896
2 .896
2 .896
2 .896
2 .896
2 .896
2 .896
2 .896
2 .896
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
3 .653
A-48
RMR
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
SE
290
290
280
280
280
280
280
280
280
280
440
440
440
440
440
440
440
440
640
640
640
640
640
640
640
640
640
640
640
640
700
700
700
700
700
700
700
700
700
700
700
700
700
700
700
700
700
700
700
700
700
700
700
700
11-98)
Units med.
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
wgt
2
2
3
3
3
4
4
4
4
5
5
5
6
6
6
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
.1
.7
.2
.6
.8
.0
.3
.5
.8
.0
.4
.7
.0
.3
.9
.2
.7
.6
.3
.4
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
-------
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R47k
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R471
R47a
R47a
R47a
R47b
R47b
R47b
R47b
R47b
R47b
R47b
R47c
R47C
R47C
R47C
R47C
R47C
R47C
R47C
R47d
R47d
R47d
R47d
R47d
R47d
R47d
R47d
R47d
R47d
R47d
R47d
R47e
R47e
R47e
R47e
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.048
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.049
.244
.244
.244
.085
.085
.085
.085
.085
.085
.085
.056
.056
.056
.056
.056
.056
.056
.056
.062
.062
.062
.062
.062
.062
.062
.062
.062
.062
.062
.062
.034
.034
.034
.034
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
-
-
-
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
3
3
3
3
.653
.653
.653
.653
.653
.653
.653
.653
.653
.653
.653
.653
.653
.653
.653
.653
.653
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
.459
0.130
0.130
0.130
.033
.033
.033
.033
.033
.033
.033
.898
.898
.898
.898
.898
.898
.898
.898
.036
.036
.036
.036
.036
.036
.036
.036
.036
.036
.036
.036
.538
.538
.538
.538
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
.700
.700
.700
.700
.700
.700
.700
.700
.700
.700
.700
.700
.700
.700
.700
.700
.700
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.690
.250
.250
.250
.290
.290
.290
.290
.290
.290
.290
.470
.470
.470
.470
.470
.470
.470
.470
.500
.500
.500
.500
.500
.500
.500
.500
.500
.500
.500
.500
.470
.470
.470
.470
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
2
2
3
3
3
3
3
4
4
4
4
5
5
5
5
6
6
6
5
5
6
6
6
6
6
6
6
6
6
6
5
5
5
5
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.0
.5
.0
.3
.5
.7
.9
.1
.4
.7
.9
.2
.5
.7
.9
.0
.1
.2
.7
.8
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.7
.7
.7
.7
A-49
-------
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
Males
Blood
Age
0
1
2
3
4
5
6
7
8
9
10
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47e
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
R47f
age 0-100
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BMR
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
BM
then females age
Volume factor and
BLDFAC
17.0
17.0
17.0
17.0
17.0
17.0
17.0
17.0
17.0
17.0
17.0
HGMN
11.9
12 .2
12 .4
12 .7
12 .8
13 .0
13 .2
13 .5
13 .4
13 .6
13 .6
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0-
Hemoglobin
HGSTD
1.0
1.0
0.8
0.8
0.8
0.9
0.9
0.8
0.8
1.0
0.9
034
034
034
034
034
034
034
034
034
034
034
034
034
034
034
034
034
034
034
034
034
034
034
034
034
034
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
038
100
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
3 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
2 .
(HG
538
538
538
538
538
538
538
538
538
538
538
538
538
538
538
538
538
538
538
538
538
538
538
538
538
538
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
755
last
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.470
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
.450
revised
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
MJ/day
12-20-05)
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.7
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
5.2
content
A-50
-------
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
17
17
17
17
17
17
17
17
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
.0
.0
.0
.0
.0
.0
.0
.0
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
13
14
14
14
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
13
.7
.0
.3
.7
.1
.4
.5
.7
.8
.8
.7
.7
.7
.7
.7
.7
.7
.7
.7
.7
.6
.6
.6
.6
.6
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.4
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.1
.1
.1
.1
.1
.0
.0
.0
.0
.0
.7
.7
.7
.7
.7
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.0
.0
.0
.0
.0
.8
0
1
1
1
1
1
1
1
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
.7
.0
.0
.0
.0
.0
.0
.0
.8
.9
.9
.9
.9
.9
.9
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.0
.0
.0
.0
.0
.2
.2
.2
.2
.2
.2
.2
.2
.2
.2
.4
.4
.4
.4
.4
.5
.5
.5
.5
.5
.4
.4
.4
.4
.4
.8
.8
.8
.8
.8
.8
A-51
-------
92
93
94
95
96
97
98
99
100
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
20
20
20
20
20
20
20
20
20
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
.4
.4
.4
.4
.4
.4
.4
.4
.4
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.0
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
13
13
13
13
13
13
13
13
13
12
12
12
12
12
12
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
.8
.8
.8
.8
.5
.5
.5
.5
.5
.2
.3
.6
.5
.8
.9
.0
.1
.3
.4
.6
.5
.6
.5
.6
.5
.5
.5
.5
.4
.5
.5
.5
.5
.5
.5
.3
.3
.3
.3
.3
.3
.3
.3
.3
.3
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.6
.6
.6
.6
.6
.7
.7
.7
.7
.7
.8
.8
.8
.8
.8
.8
.8
.8
.8
.8
.8
.8
.8
.8
.8
.8
1
1
1
1
1
1
1
1
1
0
0
0
1
0
1
0
0
0
0
1
0
0
1
1
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
.8
.8
.8
.8
.8
.8
.8
.8
.8
.7
.7
.8
.0
.8
.0
.8
.8
.8
.8
.0
.9
.9
.0
.0
.9
.1
.1
.2
.1
.1
.2
.2
.2
.2
.2
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.2
.2
.2
.2
.2
.3
.3
.3
.3
.3
.2
.2
.2
.2
.2
.1
.1
.1
.1
.1
.2
.2
.2
.2
.2
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
.1
A-52
-------
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
14
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
13
.8
.8
.8
.8
.8
.8
.8
.8
.8
.6
.6
.6
.6
.6
.4
.4
.4
.4
.4
.2
.2
.2
.2
.2
.0
.0
.0
.0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
.1
.1
.1
.1
.3
.3
.3
.3
.3
.2
.2
.2
.2
.2
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
.6
100
14 .6
13 .0
1.6
A-53
-------
Appendix E. All Derived Physiological Parameters
Table 1. Nv02max Values for Males: Raw and Smoothed Fits.
Age
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
18.00
19.00
20.00
21.00
22.00
23.00
24.00
25.00
26.00
27.00
28.00
29.00
30.00
31.00
32.00
33.00
34.00
35.00
36.00
37.00
38.00
39.00
40.00
MEAN
Raw
Fit Values
51.37
53.46
51.10
51.28
50.13
50.70
52.74
52.93
53.18
49.46
49.77
51.98
59.88
56.80
54.60
54.61
53.76
57.23
50.90
50.06
46.38
48.32
51.02
45.59
45.86
46.90
42.08
44.48
38.63
42.63
40.41
39.70
40.62
39.02
MEAN
Smoothed
Fit Values
48.25
48.56
48.88
49.19
49.50
49.82
50.13
50.44
50.76
51.07
51.39
51.70
52.01
52.33
52.64
52.95
53.27
53.58
53.90
54.21
54.52
54.23
53.42
52.63
51.84
51.07
50.31
49.56
48.82
48.10
47.38
46.67
45.98
45.30
44.63
43.97
43.32
42.68
42.05
41.44
40.83
MALES
SD SD
Raw Smoothed
Fit Values Fit Values
2.86
2.86
6.26
5.87
6.04
7.13
5.13
4.72
5.57
6.06
6.93
7.48
9.65
9.31
8.17
8.40
9.60
10.44
10.63
9.66
8.95
10.47
12.31
9.91
10.14
11.03
9.08
8.95
10.10
7.11
8.81
6.22
8.01
8.28
1.71
2.04
2.36
2.68
3.01
3.33
3.65
3.98
4.30
4.62
4.95
5.27
5.59
5.92
6.24
6.56
6.89
7.21
7.53
7.86
8.18
8.50
8.83
9.15
9.47
9.80
10.69
10.49
10.29
10.10
9.92
9.73
9.55
9.38
9.20
9.03
8.87
8.71
8.55
8.40
8.25
MIN
(IstPctl)
44.26
43.82
43.39
42.95
42.51
42.07
41.63
41.19
40.76
40.32
39.88
39.44
39.00
38.56
38.13
37.69
37.25
36.81
36.37
35.93
35.50
34.45
32.89
31.35
29.81
28.29
25.45
25.16
24.88
24.60
24.32
24.04
23.76
23.49
23.22
22.95
22.69
22.42
22.16
21.90
21.64
MAX
(99th Pctl)
52.24
53.30
54.37
55.43
56.50
57.56
58.63
59.70
60.76
61.83
62.89
63.96
65.02
66.09
67.16
68.22
69.29
70.35
71.42
72.48
73.55
74.01
73.95
73.91
73.88
73.86
75.17
73.96
72.77
71.59
70.44
69.31
68.20
67.10
66.03
64.98
63.95
62.94
61.94
60.97
60.02
A-54
-------
Age
41.00
42.00
43.00
44.00
45.00
46.00
47.00
48.00
49.00
50.00
51.00
52.00
53.00
54.00
55.00
56.00
57.00
58.00
59.00
60.00
61.00
62.00
63.00
64.00
65.00
66.00
67.00
68.00
69.00
70.00
71.00
72.00
73.00
74.00
75.00
76.00
77.00
78.00
79.00
80.00
81.00
82.00
83.00
84.00
85.00
86.00
MEAN
Raw
Fit Values
39.72
35.58
39.98
38.65
40.15
40.67
41.51
38.92
34.65
33.85
32.52
36.31
36.23
33.91
33.40
31.68
32.47
33.24
33.05
29.02
31.68
29.72
30.90
30.65
29.86
28.60
29.47
28.95
31.13
27.12
28.56
27.62
27.84
25.05
23.74
23.68
MEAN
Smoothed
Fit Values
40.24
39.66
39.09
38.53
37.98
37.44
36.92
36.40
35.90
35.41
34.92
34.45
34.00
33.55
33.11
32.69
32.27
31.87
31.48
31.10
30.73
30.37
30.02
29.69
29.36
29.05
28.75
28.46
28.18
27.91
27.65
27.41
27.17
26.95
26.74
26.54
26.35
26.17
26.00
25.84
25.70
25.57
25.44
25.33
25.23
25.14
MALES
SD SD
Raw Smoothed
Fit Values Fit Values
9.96
9.85
6.46
7.60
6.59
7.89
9.68
10.52
7.68
6.49
4.51
7.08
7.31
5.29
5.08
6.52
6.33
6.32
6.45
3.59
6.95
5.09
8.06
5.32
6.90
5.51
5.25
5.63
6.43
3.44
5.71
5.03
6.27
6.68
4.99
5.88
8.10
7.96
7.82
7.69
7.56
7.43
7.31
7.19
7.07
6.96
6.86
6.75
6.65
6.56
6.46
6.37
6.29
6.21
6.13
6.06
5.99
5.92
5.86
5.80
5.75
5.70
5.65
5.61
5.57
5.53
5.50
5.47
5.45
5.43
5.41
5.40
5.39
5.38
5.38
5.39
5.39
5.39
5.39
5.39
5.39
5.39
MIN
(IstPctl)
21.39
21.14
20.89
20.64
20.40
20.16
19.91
19.68
19.44
19.21
18.98
18.75
18.52
18.30
18.08
17.86
17.64
17.43
17.22
17.01
16.80
16.60
16.40
16.20
16.00
15.80
15.61
15.42
15.23
15.05
14.86
14.68
14.50
14.33
14.15
13.98
13.81
13.65
13.48
13.32
13.17
13.04
12.92
12.81
12.70
12.62
MAX
(99th Pctl)
59.09
58.18
57.28
56.41
55.56
54.73
53.92
53.12
52.35
51.60
50.87
50.16
49.47
48.79
48.14
47.51
46.90
46.31
45.74
45.19
44.66
44.14
43.65
43.18
42.73
42.30
41.89
41.50
41.13
40.78
40.45
40.13
39.84
39.57
39.32
39.09
38.88
38.69
38.52
38.37
38.22
38.09
37.97
37.86
37.76
37.67
A-55
-------
Age
87.00
88.00
89.00
90.00
91.00
92.00
93.00
94.00
95.00
96.00
97.00
98.00
99.00
100.00
MEAN MEAN
Raw Smoothed
Fit Values Fit Values
25.06
25.00
24.94
24.90
24.86
24.84
24.83
24.83
24.84
24.87
24.90
24.95
25.00
25.07
MALES
SD SD
Raw Smoothed
Fit Values Fit Values
5.39
5.39
5.39
5.39
5.39
5.39
5.39
5.39
5.39
5.39
5.39
5.39
5.39
5.39
MIN
(IstPctl)
12.54
12.47
12.42
12.37
12.34
12.32
12.31
12.31
12.32
12.34
12.37
12.42
12.48
12.54
MAX
(99th Pctl)
37.59
37.52
37.47
37.42
37.39
37.37
37.36
37.36
37.37
37.39
37.43
37.47
37.53
37.60
Table 2. Nv02max Values for Females: Raw and Smoothed Fits
Age
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
18.00
19.00
MEAN
Raw
Fit Values
30.56
45.53
43.88
43.03
42.00
37.57
39.57
35.51
38.22
45.67
43.87
MEAN
Smoothed
Fit
Values
35.88
36.21
36.54
36.87
37.20
37.54
37.87
38.20
38.53
38.86
39.19
39.52
39.85
40.18
40.51
40.85
41.18
41.51
41.84
42.17
FEMALES
SD SD
Raw Smoothed
Fit Fit
Values Values
9.90
6.27
5.26
6.88
7.48
6.79
5.43
5.36
8.86
8.53
7.83
5.90
6.00
6.09
6.19
6.28
6.38
6.47
6.57
6.66
6.76
6.85
6.95
7.04
7.14
7.23
7.33
7.42
7.52
7.61
7.71
MIN
(1st
Pctl)
22.15
22.26
22.37
22.48
22.59
22.70
22.81
22.92
23.03
23.14
23.25
23.36
23.47
23.58
23.69
23.80
23.91
24.02
24.13
24.24
MAX
(99th Pctl)
49.61
50.17
50.72
51.27
51.82
52.37
52.93
53.48
54.03
54.58
55.13
55.69
56.24
56.79
57.34
57.89
58.45
59.00
59.55
60.10
A-56
-------
Age
20.00
21.00
22.00
23.00
24.00
25.00
26.00
27.00
28.00
29.00
30.00
31.00
32.00
33.00
34.00
35.00
36.00
37.00
38.00
39.00
40.00
41.00
42.00
43.00
44.00
45.00
46.00
47.00
48.00
49.00
50.00
51.00
52.00
53.00
54.00
55.00
56.00
57.00
58.00
59.00
60.00
61.00
62.00
63.00
64.00
MEAN
Raw
Fit Values
42.52
43.45
43.22
43.87
41.14
38.20
38.98
34.94
38.08
35.13
35.79
35.22
36.06
34.95
38.13
32.63
33.59
31.11
33.12
28.80
29.06
29.54
30.90
27.60
29.33
28.53
29.41
30.49
27.92
26.48
29.80
27.49
28.95
23.77
25.34
26.05
26.30
26.06
23.67
24.70
21.63
26.64
23.84
MEAN
Smoothed
Fit
Values
42.50
42.10
41.45
40.81
40.18
39.56
38.95
38.35
37.75
37.17
36.60
36.04
35.48
34.94
34.41
33.88
33.37
32.87
32.37
31.89
31.42
30.95
30.50
30.05
29.62
29.19
28.78
28.37
27.97
27.59
27.21
26.84
26.49
26.14
25.80
25.48
25.16
24.85
24.55
24.27
23.99
23.72
23.46
23.21
22.97
FEMALES
SD SD
Raw Smoothed
Fit Fit
Values Values
7.69
8.51
7.59
10.13
8.22
7.09
11.12
8.02
9.80
6.30
9.10
7.89
6.93
9.51
7.08
4.88
6.17
5.13
3.76
5.14
5.74
8.00
6.82
4.32
4.17
4.90
6.00
7.15
6.05
5.36
5.13
3.66
5.83
3.56
4.61
4.29
4.91
4.07
4.81
4.65
4.99
7.38
3.77
7.80
7.90
7.99
8.09
8.18
8.28
8.37
8.35
8.14
7.94
7.74
7.55
7.37
7.19
7.01
6.84
6.68
6.52
6.37
6.22
6.08
5.95
5.82
5.70
5.58
5.47
5.36
5.26
5.16
5.07
4.99
4.91
4.83
4.77
4.70
4.65
4.60
4.55
4.51
4.48
4.45
4.43
4.41
4.40
4.39
MIN
(1st
Pctl)
24.35
23.73
22.86
21.99
21.14
20.30
19.47
18.93
18.82
18.71
18.59
18.47
18.35
18.23
18.10
17.97
17.83
17.70
17.55
17.41
17.26
17.11
16.96
16.80
16.64
16.48
16.31
16.14
15.97
15.79
15.61
15.43
15.24
15.06
14.86
14.67
14.47
14.27
14.06
13.85
13.64
13.43
13.21
12.99
12.76
MAX
(99th Pctl)
60.65
60.48
60.05
59.63
59.22
58.82
58.43
57.76
56.69
55.64
54.61
53.60
52.62
51.66
50.72
49.80
48.91
48.04
47.19
46.37
45.57
44.79
44.03
43.30
42.59
41.90
41.24
40.60
39.98
39.38
38.81
38.26
37.73
37.23
36.74
36.29
35.85
35.44
35.05
34.68
34.33
34.01
33.71
33.44
33.18
A-57
-------
Age
65.00
66.00
67.00
68.00
69.00
70.00
71.00
72.00
73.00
74.00
75.00
76.00
77.00
78.00
79.00
80.00
81.00
82.00
83.00
84.00
85.00
86.00
87.00
88.00
89.00
90.00
91.00
92.00
93.00
94.00
95.00
96.00
97.00
98.00
99.00
100.00
MEAN
Raw
Fit Values
20.26
20.38
20.49
22.05
21.92
20.38
25.30
21.21
20.46
20.63
20.60
20.91
22.27
19.93
22.80
23.19
19.29
13.44
28.03
17.00
18.69
18.18
27.15
18.18
MEAN
Smoothed
Fit
Values
22.74
22.52
22.31
22.11
21.92
21.74
21.57
21.41
21.26
21.12
20.99
20.87
20.76
20.65
20.56
20.48
20.41
20.34
20.29
20.25
20.21
20.19
20.18
20.17
20.18
20.20
20.22
20.26
20.30
20.36
20.42
20.50
20.58
20.67
20.78
20.89
FEMALES
SD SD
Raw Smoothed
Fit Fit
Values Values
3.83 4.39
4.39
4.39
3.90 4.39
4.56 4.39
4.15 4.39
4.39
4.39
4.59 4.39
4.39
3.80 4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
4.39
MIN
(1st
Pctl)
12.53
12.31
12.10
11.90
11.71
11.53
11.36
11.20
11.05
10.91
10.78
10.66
10.55
10.44
10.35
10.27
10.20
10.13
10.08
10.04
10.00
9.98
9.97
9.96
9.97
9.98
10.01
10.05
10.09
10.15
10.21
10.28
10.37
10.46
10.57
10.68
MAX
(99th Pctl)
32.95
32.73
32.52
32.32
32.13
31.95
31.78
31.62
31.47
31.33
31.20
31.08
30.97
30.86
30.77
30.69
30.62
30.55
30.50
30.46
30.42
30.40
30.39
30.38
30.39
30.41
30.43
30.47
30.51
30.57
30.63
30.71
30.79
30.88
30.99
31.10
A-58
-------
Table 3. Body Mass Raw Fits.
Age
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
18.00
19.00
20.00
21.00
22.00
23.00
24.00
25.00
26.00
27.00
28.00
29.00
30.00
31.00
32.00
33.00
34.00
35.00
36.00
37.00
38.00
39.00
40.00
41.00
42.00
43.00
Geometric
Mean
7.767
11.440
13.932
15.967
18.475
21.618
23.142
27.072
31.651
34.656
38.329
44.149
47.988
55.364
62.832
67.650
72.460
73.081
75.060
77.182
77.952
78.239
83.845
80.607
81.706
84.818
81.812
85.166
84.321
82.144
81.581
81.275
84.715
88.188
81.163
87.192
83.404
85.759
84.132
84.611
90.071
87.425
88.290
88.423
MALES
GSD
1.301
1.143
1.146
1.154
1.165
1.234
1.213
1.216
1.302
1.265
1.280
1.308
1.315
1.340
1.293
1.255
1.267
1.248
1.243
1.245
1.250
1.297
1.292
1.222
1.251
1.206
1.273
1.249
1.272
1.236
1.262
1.249
1.235
1.231
1.221
1.251
1.228
1.241
1.260
1.196
1.246
1.173
1.205
1.233
Min
3.6
8.2
9.8
11.7
11.1
13.7
16.1
19.3
19.1
24
24.3
26.2
27.7
27.7
35.7
41.5
45.8
49.9
51.2
52.6
50.5
46.8
53.3
50.5
50.6
50.2
48.9
50
51
50.6
52.5
48.8
49.7
64.8
53.1
61
45.8
59.3
52.8
61.2
58.5
61.3
62.2
54
Max
11.8
16.1
20.9
23.7
28.1
42.4
41.1
46.8
66.2
69.9
72.9
83.8
94.8
106.6
121
117.9
139.1
136.6
144.2
134.5
130
199.2
155.4
137.6
132.6
136.1
164.5
153.9
167.2
147.2
139
170.6
135.8
146.3
136.9
193.3
140.5
150.9
149.7
140.6
154
117.7
144
145.3
Geometric
Mean
7
11
13
15
18
20
22
26
30
35
40
46
50
56
57
60
61
61
64
66
66
67
66
69
70
66
72
70
74
69
70
73
72
72
69
73
73
70
75
72
72
73
73
73
429
119
258
587
005
353
454
483
534
235
550
579
673
649
214
091
582
229
591
156
981
218
823
721
284
300
973
604
363
110
616
039
938
710
773
044
547
019
587
295
888
363
697
438
FEMALES
GSD Min
1.304
1.163
1.158
1.160
1.171
1.229
1.194
1.239
1.315
1.271
1.304
1.302
1.274
1.275
1.248
1.249
1.255
1.248
1.281
1.274
1.262
1.262
1.273
1.304
1.289
1.283
1.281
1.281
1.312
1.250
1.305
1.278
1.281
1.307
1.230
1.306
1.289
1.284
1.295
1.251
1.289
1.268
1.270
1.314
3.7
7.4
10.1
11
12.8
12.6
15.9
16.9
19.8
20.3
22.7
27.7
27.8
33.4
37.7
34.9
40.9
41.5
42.4
41.6
41.5
39.7
42
40.3
47.5
44.8
45.3
41.4
44.3
39.3
42.1
43.7
41.5
44.9
46.6
44.2
44.6
48.1
43.7
41.6
45.5
50.5
47.1
45.6
Max
12.1
15.3
20.4
27.9
29.1
40.4
36.7
51
60.8
58.6
71.2
84.6
93.3
99.5
110
108.4
113.8
133.1
123.6
118.5
122.6
123.7
123.5
143
144.5
131.8
128.9
140.9
142.1
116.3
151.5
125.9
139.7
135.2
115.3
138.4
150.1
152.1
151.7
123.1
137.4
156.9
146.1
159.5
A-59
-------
Age
44.00
45.00
46.00
47.00
48.00
49.00
50.00
51.00
52.00
53.00
54.00
55.00
56.00
57.00
58.00
59.00
60.00
61.00
62.00
63.00
64.00
65.00
66.00
67.00
68.00
69.00
70.00
71.00
72.00
73.00
74.00
75.00
76.00
77.00
78.00
79.00
80.00
81.00
82.00
83.00
84.00
85.00
86.00
87.00
88.00
89.00
Geometric
Mean
88.528
87.102
88.157
86.547
84.793
86.235
84.659
87.975
89.886
89.012
90.098
88.268
84.796
87.501
85.116
84.190
87.044
89.007
84.788
89.137
89.974
89.891
86.814
86.207
85.172
87.116
82.775
79.630
82.011
85.590
83.001
84.465
78.733
79.376
79.909
77.629
79.866
75.405
76.798
74.611
75.325
71.776
73.986494
73.364276
72.742058
72.11984
MALES
GSD Min
1.200
1.205
1.243
1.229
1.186
1.240
1.179
1.208
1.216
1.228
1.216
1.222
1.195
1.253
1.266
1.182
1.232
1.207
1.228
1.262
1.193
1.215
1.228
1.207
1.191
1.222
1.210
1.240
1.204
1.196
1.217
1.185
1.207
1.170
1.195
1.155
1.174
1.157
1.180
1.158
1.205
1.191
1.17
1.17
1.16
1.16
56.6
60.6
54.2
49.9
56.3
47
53.4
57.9
55.2
58.2
64.1
55.1
45
58.3
51.6
58.7
57.3
49.9
56.04
56.3
59.1
58.1
54
43.1
61.2
50.7
46.5
51
51.9
56.2
53.3
56.5
55.9
58.7
41.1
56.4
56
55.8
54.4
53.2
41.5
46.9
50.57
50.38
50.19
50
Max
128.9
160.2
154.3
188.3
128.3
171.3
124.4
143.6
144.9
143.3
155.2
138.6
110.3
160
179
112.4
141.7
162.8
152.1
171.6
119
126.3
150.1
127.5
163.2
127.2
125.5
122.8
132.7
128.3
120
133.5
121.1
109.3
115.1
107.8
111.9
111.9
111.8
107
109.5
105.8
101.07
99.113
97.154
95.194
Geometric
Mean
75.742
76.795
77.544
72.849
74.646
72.844
75.217
72.941
74.472
74.733
72.413
75.951
77.322
72.378
74.548
80.638
75.777
77.121
73.347
72.308
75.440
72.910
73.101
75.835
73.207
74.368
68.977
69.083
69.898
71.360
70.410
70.526
69.549
70.128
66.375
67.780
62.214
65.397
64.755
62.886
62.215
61.453
62.400356
61.847614
61.294872
60.74213
FEMALES
GSD Min
1.266
1.308
1.304
1.298
1.303
1.261
1.292
1.240
1.283
1.259
1.281
1.231
1.315
1.252
1.267
1.277
1.260
1.240
1.198
1.238
1.281
1.254
1.242
1.266
1.250
1.225
1.188
1.232
1.240
1.240
1.277
1.216
1.199
1.240
1.211
1.200
1.255
1.184
1.260
1.196
1.216
1.209
1.21
1.21
1.21
1.21
49.5
41.6
46.6
47.8
44.2
45.1
48.4
42.5
45.7
46.2
44.3
53.6
45.6
48.6
45
50.9
51.3
50.7
49.7
46.9
41.1
35.9
48.4
47.2
39.3
48
45.9
45.5
40.7
47.4
37.4
46.8
48.8
40.3
44.1
46.2
41.2
42.7
40.6
44.7
43.5
42.3
41.85
41.66
41.47
41.27
Max
153
141.5
145.8
130.6
166
125.54
175.7
120.2
146.6
176.6
123.1
125.6
134.9
122.6
117.7
133
128.3
125.6
121.1
119.9
132.5
113.7
113.3
123.8
120.7
118
102.8
108.1
103.8
127.6
106.4
117.4
101.7
119.8
109.8
98.4
121.4
91.4
120
101.2
108.4
93.2
101.16
100.26
99.351
98.445
A-60
-------
Age
90.00
91.00
92.00
93.00
94.00
95.00
96.00
97.00
98.00
99.00
100.00
MALES
Geometric
Mean
71 .497622
70.875404
70.253186
69.630968
69.00875
68.386532
67.764314
67.142096
66.519878
65.89766
65.275442
GSD
1.16
1.16
1.16
1.15
1.15
1.15
1.15
1.14
1.14
1.14
1.14
Min
49.81
49.62
49.44
49.25
49.06
48.87
48.68
48.49
48.3
48.11
47.92
Max
93.235
91.276
89.317
87.358
85.399
83.44
81.481
79.522
77.563
75.604
73.645
FEMALES
Geometric
Mean
60.189388
59.636646
59.083904
58.531162
57.97842
57.425678
56.872936
56.320194
55.767452
55.21471
54.661968
GSD
1.21
1.2
1.2
1.2
1.2
1.2
1.2
1.2
1.19
1.19
1.19
Min
41.08
40.88
40.69
40.49
40.3
40.1
39.91
39.71
39.52
39.32
39.13
Max
97.538
96.632
95.726
94.82
93.914
93.008
92.102
91.195
90.289
89.383
88.477
**
Dark shading (age 86+) designates linear forecast.
Table 4. Body Mass Smoothed Fits (5-Year Running Averages).
Age
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
18.00
19.00
20.00
21.00
22.00
Geometric
Mean
7.767209794
1 1 .44008024
13.93227373
15.96664726
18.47458493
21.61756114
23.14243627
27.07246068
31.6505017
34.65600448
38.32939135
44.14863459
47.98795299
55.36374737
62.83159173
67.65031426
72.45980541
73.08089659
75.06031573
77.18236513
77.95205826
78.45564692
79.56489519
MALES
GSD Min
1.300901
1.143324
1.145566
1.153689
1.164972
1 .233822
1.213499
1.215834
1.301873
1.265317
1.279707
1.30753
1.314848
1 .33952
1.292533
1.254999
1 .267468
1 .248405
1 .243204
1 .244928
1 .250326
1.265585
1.261251
3.6
8.2
9.8
11.7
11.1
13.7
16.1
19.3
19.1
24
24.3
26.2
27.7
27.7
35.7
41.5
45.8
49.9
51.2
52.6
50.5
50.88
50.74
Max
11.8
16.1
20.9
23.7
28.1
42.4
41.1
46.8
66.2
69.9
72.9
83.8
94.8
106.6
121
117.9
139.1
136.6
144.2
134.5
130
152.66
151.34
Geometric
Mean
7.428916349
11.11947416
13.25797158
15.58684049
18.00506307
20.35285099
22.45431948
26.48323788
30.53391399
35.23472141
40.54996835
46.57910267
50.67329267
56.64881107
57.21362103
60.09135575
61.58214656
61.22931022
64.59054256
66.15556407
66.98146906
66.35375002
67.37976393
FEMALES
GSD Min
1 .304229
1.162608
1.158434
1.159883
1.17108
1 .229237
1.194119
1 .23892
1.315137
1.271364
1 .303997
1.302182
1 .273946
1 .275455
1 .24795
1 .24897
1.255162
1 .248057
1.281298
1 .274083
1.261822
1 .270386
1 .274844
3.7
7.4
10.1
11
12.8
12.6
15.9
16.9
19.8
20.3
22.7
27.7
27.8
33.4
37.7
34.9
40.9
41.5
42.4
41.6
41.5
41.44
41.02
Max
12.1
15.3
20.4
27.9
29.1
40.4
36.7
51
60.8
58.6
71.2
84.6
93.3
99.5
110
108.4
113.8
133.1
123.6
118.5
122.6
122.38
126.26
A-61
-------
Age
23.00
24.00
25.00
26.00
27.00
28.00
29.00
30.00
31.00
32.00
33.00
34.00
35.00
36.00
37.00
38.00
39.00
40.00
41.00
42.00
43.00
44.00
45.00
46.00
47.00
48.00
49.00
50.00
51.00
52.00
53.00
54.00
55.00
56.00
57.00
58.00
59.00
60.00
61.00
62.00
63.00
64.00
65.00
66.00
67.00
68.00
Geometric
Mean
80.46958232
81 .84267254
82.55729313
82.82151847
83.56439112
83.65195203
83.00459482
82.89721864
82.80701235
83.58034187
83.38418057
84.50647805
84.9321819
85.14102649
84.32994666
85.01958212
85.59524544
86.39949423
86.90564401
87.76379051
88.54719729
87.95342484
88.09985934
87.751282
87.02523405
86.56661258
86.07815707
86.04175058
86.70964624
87.55345712
88.32616726
89.04784314
88.4120991
87.93495739
87.15584772
85.97418819
85.72952642
86.57173577
86.0292098
86.83331368
87.99005122
88.55927286
88.12051692
88.40439667
87.61146369
87.03986775
MALES
GSD Min
1 .262527
1.253588
1 .248802
1 .240222
1.250399
1 .247428
1.258753
1.253937
1.251132
1 .242848
1.239735
1.237533
1.233184
1 .234298
1.240177
1.235131
1.233983
1 .223065
1.215924
1.210495
1.211458
1.203416
1.217379
1.222211
1.212835
1 .220669
1.215489
1.208607
1.206031
1.2144
1 .209663
1.218268
1.215526
1 .222906
1.230391
1.223617
1 .22558
1 .228074
1 .222944
1 .22222
1 .224363
1 .220869
1 .225034
1.220898
1.206714
1.212565
50.34
50.28
50.7
50.04
50.14
50.14
50.6
50.58
50.52
53.28
53.78
55.48
54.88
56.8
54.4
56.02
55.52
58.62
59.2
59.44
58.52
58.94
57.52
55.06
55.52
53.6
52.16
52.9
53.96
54.34
57.76
58.1
55.52
56.14
54.82
53.74
54.18
55.16
54.708
55.648
55.728
55.888
56.708
54.12
55.1
53.42
Max
150.96
152.18
145.24
144.94
150.86
153.78
154.36
155.58
151.96
147.78
145.72
156.58
150.56
153.58
154.26
155
147.14
142.58
141.2
140.32
137.98
139.22
146.54
155.4
152
160.48
153.32
151.18
142.5
145.5
142.28
145.12
138.46
141.48
148.62
140.06
140.68
151.18
149.6
148.12
149.44
146.36
143.82
138.9
137.22
138.86
Geometric
Mean
68.20537834
68.06901959
69.21992781
69.97607936
70.90453453
70.66975978
71.53295767
71.54621552
72.01313142
71 .6826276
71.81523165
72.30094254
72.40264379
71.81884258
72.3941641
72.89859355
72.86733489
72.830387
73.56585153
73.13604869
73.82543503
74.60684165
75.44302619
75.27348935
75.51517243
74.93569966
74.62001355
73.69947055
74.02400492
74.04127315
73.95491798
74.10188224
74.97813364
74.55937637
74.52242942
76.16748501
76.13252691
76.09221736
76.28599922
75.83796229
74.79845832
74.22522224
73.42130739
73.91902253
74.09892054
73.88448789
FEMALES
GSD Min
1.277813
1.282127
1 .285979
1 .287735
1.289413
1.28161
1 .285847
1.285108
1 .28495
1.283915
1 .280002
1 .280205
1 .282492
1.283151
1 .280709
1 .284821
1.281407
1.277165
1 .274483
1 .278433
1.281514
1 .285327
1 .292445
1.29795
1 .295658
1 .29459
1.291492
1 .278764
1 .27574
1 .266893
1 .270898
1.25873
1 .273724
1 .267547
1 .269258
1 .268509
1 .274205
1.259138
1.248419
1 .242552
1 .243365
1 .242246
1 .242692
1 .256261
1 .258602
1 .247351
42.2
42.86
43.98
43.86
44.66
43.02
42.48
42.16
42.18
42.3
43.76
44.18
44.36
45.68
45.44
44.44
44.7
45.88
45.68
46.06
47.64
46.86
46.08
46.22
45.94
45.06
46.42
45.6
45.18
45.58
45.42
46.46
47.08
47.66
47.42
48.74
48.28
49.3
49.52
49.9
47.94
44.86
44.4
43.9
42.38
43.76
Max
131.46
133.3
134.34
137.82
137.64
132
135.94
135.34
135.1
133.72
133.52
130.9
135.74
138.22
141.52
143.08
142.88
144.24
143.04
144.6
150.58
151.4
149.18
146.08
147.38
141.888
148.728
143.608
146.808
148.928
148.44
138.42
141.36
136.56
124.78
126.76
127.3
125.44
125.14
125.58
125.48
122.56
120.1
120.64
120.8
117.9
A-62
-------
Age
69.00
70.00
71.00
72.00
73.00
74.00
75.00
76.00
77.00
78.00
79.00
80.00
81.00
82.00
83.00
84.00
85.00
86.00
87.00
88.00
89.00
90.00
91.00
92.00
93.00
94.00
95.00
96.00
97.00
98.00
99.00
100.00
Geometric
Mean
85.61667034
84.17987726
83.34052655
83.42413149
82.60108145
82.93914453
82.75981666
82.23282472
81.09670348
80.02242628
79.10265965
78.43698141
77.92142176
76.86173441
76.40090269
74.7828307
74.4992137
73.81250877
73.43871959
72.79767378
72.74205786
72.11983988
71.4976219
70.87540393
70.25318595
69.63096798
69.00875
68.38653203
67.76431405
67.14209607
66.5198781
66.20876911
MALES
GSD Min
1.211601
1.213949
1.213444
1.214423
1.213508
1.208433
1.201809
1.19492
1.194698
1.182282
1.180128
1.170309
1.17229
1.164819
1.174796
1.17822
1.180574
1.177977
1.179377
1.170756
1.164535
1.162177
1.159818
1.157459
1.1551
1.152742
1.150383
1.148024
1.145665
1.143307
1.140948
1.139769
51.1
50.5
52.26
51.26
51.78
53.78
54.76
56.12
53.1
53.72
53.62
53.6
52.74
55.16
52.18
50.36
49.31362
48.50949
47.90759
49.60794
50.19052
50.00172
49.81292
49.62412
49.43532
49.24652
49.05772
48.86892
48.68012
48.49132
48.30252
48.20812
Max
138.7
133.24
134.28
127.3
125.86
127.46
127.12
122.44
119.8
117.36
113.04
111.2
111.7
110.08
110.42
109.2
107.0343
104.4969
102.5276
99.66646
97.15354
95.19446
93.23538
91.27631
89.31723
87.35815
85.39908
83.44
81 .48092
79.52185
77.56277
76.58323
Geometric
Mean
73.09783811
72.29417333
71.10679374
70.73734313
69.94568967
70.25540111
70.34868617
70.39465694
69.39757689
68.87151054
67.20924709
66.37891246
65.30424541
64.60647334
63.49351577
63.34131978
62.74189138
62.16041309
61 .84223228
61.5476723
61.29487188
60.74212987
60.18938785
59.63664584
59.08390383
58.53116181
57.9784198
57.42567778
56.87293577
56.32019375
55.76745174
55.49108073
FEMALES
GSD Min
1 .234088
1.23216
1 .227009
1.225132
1 .235452
1.241
1 .23444
1 .234265
1.228511
1.213229
1.221048
1.217987
1 .222093
1.219074
1 .22226
1.213219
1.218822
1 .208932
1.211505
1.209819
1.209189
1 .207753
1.206317
1 .20488
1 .203444
1 .202008
1.200571
1.199135
1.197699
1.196263
1.194826
1.194108
45.76
45.18
43.88
45.5
43.38
43.56
44.22
44.14
43.48
45.24
44.12
42.9
42.96
43.08
42.54
42.76
42.59095
42.80295
42.15598
41.71005
41.46516
41.27036
41.07556
40.88075
40.68595
40.49115
40.29634
40.10154
39.90674
39.71193
39.51713
39.41973
Max
115.72
114.68
110.68
112.06
109.74
112.66
111.38
114.58
111.02
109.42
110.22
108.16
108.2
106.48
108.48
102.84
104.7926
100.844
100.4742
98.48308
99.35077
98.44462
97.53846
96.63231
95.72615
94.82
93.91385
93.00769
92.10154
91.19538
90.28923
89.83615
A-63
-------
Table 5. Hemoglobin Content.
Age
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
MALES
MEAN STD
11.927 0.993545
12.20959 1.013091
12.42075 0.823171
12.69015 0.83159
12.8006 0.80152
12.95822 0.878515
13.19574 0.893008
13.46198 0.836639
13.35161 0.833121
13.59742 0.971019
13.63062 0.906785
13.66 0.726155
13.9727 0.955869
14.28293 1.036749
14.70654 1.020254
15.13583 1.04546
15.36442 1.021623
15.45945 0.979296
15.7487 1.02514
15.76812 0.831813
15.79371 0.880956
15.71703 0.91072
15.70837 1.045808
15.55635 0.959964
15.43525 1.021741
15.44038 1.105939
15.41492 1.096952
15.31983 1.123792
15.27653 0.97796
15.07274 1.192645
14.96193 1.24457
14.72786 1.418355
14.51 1.476879
14.52915 1.352814
13.97647 1.757686
13.801 1.757686
13.534 1.757686
FEMALES
MEAN STD
12.209 0.729499905
12.27307 0.719158646
12.55018 0.843436666
12.4519 0.965868504
12.83442 0.773409545
12.87154 0.969254536
13.01866 0.828912341
13.09899 0.754370806
13.25291 0.826349227
13.36671 0.808377267
13.58919 1.034306588
13.52681 0.90041802
13.6273 0.884271668
13.46986 0.97623121
13.58878 1.034527514
13.47154 0.856131982
13.50562 1.088863466
13.49842 1.117860417
13.46091 1.18250671
13.35445 1.090493585
13.5016 1.072791517
13.47168 1.170602542
13.2967 1.145254677
13.34583 1.134192006
13.4881 1.163867696
13.48617 1.348669176
13.61113 1.193756618
13.67737 1.106237392
13.83717 1.237714453
13.76529 1.093354796
13.81911 1.093565513
13.79013 1.056812752
13.84426 1.30818261
13.57546 1.238910845
13.43767 1.552685662
13.2085 1.552685662
13.005 1.552685662
A-64
-------
Appendix B
COHb Module for APEX4.3
This appendix describes the probabilistic carboxyhemoglobin (COHb) module used in the
current APEX4.3 model. The approach described here is based primarily on the COHb module
originally described by Biller and Richmond (1982) and contained within two reports (Johnson
et al., 1992; Johnson et al., 2000) and used in EPA probabilistic NAAQS exposure model for CO
(pNEM/CO), a predecessor of APEX4.3. This appendix also describes the principal changes
made to the COHb module when it was incorporated into APEX4.3, including a change in the
method used to solve the Coburn-Forster-Kane (CFK) equation (Coburn et al., 1965). Lastly,
section B.6 summarizes outputs that were generated in performing event-level model
simulations.
B.I Base Physiological Model for Computing COHb Levels
Using time/activity data obtained from various diary studies, APEX constructs a composite diary
for each simulated person in the specified population at risk. The composite diary consists of a
sequence of events spanning the specified period of the exposure assessment (typically one
calendar year). Each event is defined by a start time, a duration, a geographic location, a
microenvironment, and an activity. Using various algorithms described in Section 4 of the
Quantitative Risk and Exposure Assessment for Carbon Monoxide (CO REA), APEX4.3
provides estimates of CO concentration and alveolar ventilation rate for each event in the
composite diary. APEX4.3 then uses these data, together with estimates of various physiological
parameters specific to the simulated individual, to estimate the percent COHb in the blood
(%COHb) as an average %COHb value over the duration of each exposure event and as an
instantaneous %COHb level at the end of each event.
The %COHb calculation is based on the solution to the non-linear CFK equation,
previously described in Appendix E of Johnson et al. (2000). The CFK model describes the rate
of change of COHb blood levels as a function of the following quantities:
1. Inspired CO pressure
2. COHb level
3. Oxyhemoglobin (O2Hb) level
4. Hemoglobin (Hb) content of blood
5. Blood volume
6. Alveolar ventilation rate
7. Endogenous CO production rate
8. Mean pulmonary capillary oxygen pressure
9. Pulmonary diffusion rate of CO
10. Haldane coefficient (M)
11. Barometric pressure
12. Vapor pressure of water at body temperature (i.e., 47 torr).
B-l
-------
If all of the listed quantities except COHb level are constant over some time interval, the
CFK equation has a linear form over the interval and is readily integrated. The solution to the
linear form gives reasonably accurate results for lower levels of COHb. However, CO and
oxygen compete for the available hemoglobin and are, therefore, not independent of each other.
If this dependency is taken into account, the resulting differential equation is no longer linear.
Peterson and Stewart (1975) proposed a heuristic approach to account for this dependency which
assumed the linear form and then adjusted the O2Hb level iteratively based on the assumption of
a linear relationship between COHb and O2Hb. This approach was used in the COHb module of
the original CO-NEM exposure model (Biller and Richmond, 1982; Johnson and Paul, 1983).
Alternatively, it is possible to determine COHb at any time by numerical integration of
the nonlinear CFK equation (e.g., by use of the Runge-Kutta method) if one assumes a particular
relationship between COHb and O2Hb. Muller and Barton (1987) demonstrated that assuming a
linear relationship between COHb and O2Hb leads to a form of the CFK equation equivalent to
the Michaelis-Menten kinetic model which can be analytically integrated. However, the
analytical solution in this case cannot be solved explicitly for COHb. Muller and Barton (1987)
demonstrated a binary search method for determining the COHb value.
The COHb module used in pNEM/CO employed a linear relationship between COHb and
O2Hb which was consistent with the basic assumptions of the CFK model. The approach
differed from the linear forms used by other modelers in that the Muller and Barton (1987)
solution was employed. However, instead of the simple binary search described by Muller and
Barton (1987), a combination of the binary search and Newton-Raphson root-finding methods
was used to solve for COHb (Press et al., 1986). Using the Muller and Barton (1987) solution
increased computation time compared to the Peterson and Stewart (1975) method but was shown
to be faster than fourth-order Runge-Kutta numerical integration.
APEX4.3 employs a different approach in which the CFK equation is solved using a
fourth-order Taylor's series expansion with subintervals. This method, first incorporated in
Version 3 of APEX, is described in Section B.2 of this appendix. A more detailed description
can be found in the Programmer's Guide for the APEX3 model (Glen, 2002).
B.2 CFK Model for Estimation of Carboxyhemoglobin
Table B-l defines the variables which appear in the equations of this section. Coburn,
Forster, and Kane (1965) derived the following differential equation governing COHb levels in
the blood upon exposure to CO.
d[COHb} = Vco } PIco Pc2
dt Vb BVb MBVb[O2Hb} l q' " '
where,
1 P -P
B = —— + B . H2° (Eq. B-2)
n TT~ \ ~i /
DLco VA
B-2
-------
Table B-l. Definitions of CFK model variables.
Variable
t
[COHb]
[02Hb]
[RHb]
[COHb]0
[THb]0
%[COHb]
%[O2Hb]
%[COHb]0
%[COHB]«
ICQ
PCco
PC02
PB
PRO
VA
Vco
DL
CO
M
k
vb
Hb
%MetHb
Definition
Time from start of an exposure event
Concentration of carboxyhemoglobin (COHb) in blood at
time t
Concentration of oxyhemoglobin (O2Hb) in blood at time t
Concentration of reduced hemoglobin in blood
[COHb] at t = 0
[RHb] + [COHb] + [O2Hb]
[COHb] expressed as percent of [RHb]0
[O2Hb] expressed as percent of [RHb]0
[COHb] at t = 0
[COHb] at t = -
Pressure of inspired CO in air saturated with water vapor
at body temperature
Mean pulmonary capillary CO pressure
Mean pulmonary capillary O2 pressure
Barometric pressure
Vapor pressure of water at body temperature, or 47
Alveolar ventilation rate
Endogenous CO production rate
Pulmonary CO diffusion rate
Haldane coefficient
Equilibrium constant for reaction O2 + RHb = O2Hb
Blood volume
Total hemoglobin in blood
Methemoglobin as weight percent of Hb
Units
minutes
ml CO per ml blood
at STPD
ml O2 per ml blood at STPD
equivalent ml CO per ml of
blood at STPD
ml CO per ml blood
at STPD
%
%
%
%
torr
torr
torr
torr
torr
ml/min STPD
ml/min STPD
ml/min/torr, STPD
ml
g/100ml
%
Notes:
1 Standard Temperature Pressure, and Dry (STPD)
-------
If the only quantity in equation (B-l) that varies with time is [COHb], the CFK equation
is linear and can be readily integrated. However, since oxygen (02) and CO compete for the
available Hb, [COHb] and [O2Hb] must be related. Increasing [COHb] will result in decreasing
[O2Hb]. Thus the CFK equation is not linear and requires the relationship between the two
quantities to be known if it is to be accurately integrated over a wide range of COHb levels.
Various linear relationships between [COHb] and [O2Hb] have been used (see Marcus,
1980; McCartney, 1990; Muller and Barton, 1987; and Tikuisis et al., 1987). A relationship not
previously used follows directly from the basic assumptions of the CFK model. The CFK model
employs the Haldane coefficient, which is the equilibrium constant associated with the following
reaction representing the replacement of O2 in O2Hb by CO:
CO + O2Hb ^O2 + COHb (Eq. B-3)
Equation B-4, the Haldane relationship, applies approximately at equilibrium conditions:
Pc0 [COHb]
Pcco[02Hb]
= M (Eq. B-4)
The Haldane coefficient, M, is the chemical equilibrium constant for reaction (B-3). The
above reaction can also be viewed as the difference between two competing chemical reactions:
CO + RHb ^ COHb (Eq. B-5)
O2 + RHb ^ O2Hb (Eq. B-6)
Subtracting (B-6) from (B-5) yields (B-3). If (B-3) is in equilibrium, then (B-5) and (B-
6) are in equilibrium. If k represents the equilibrium constant for (B-6) then:
-k (Eq.B-7)
Pc0\RHV\
It is known that an individual breathing air free of CO for an extended period will have
about 97% of their reactive Hb bound with oxygen (O2Hb) and the remainder (3%) as the
reduced form (RHb). It is also known that at one atmosphere barometric pressure, the mean
pulmonary capillary oxygen pressure is approximately 100 torr. Substituting into (B-7) yields
0.32 as the approximate value of k at body temperature. From mass balance considerations:
[O2Hb] + [COHb] + [RHb] = [THb]0 (Eq. B-8)
Eliminating [RHb] between (B-7) and (B-8) and solving for [O2Hb] yields:
kPc
[02Hb] = °2 ([THb]0 - [COHb]) (Eq. B-9)
B-4
-------
This equation represents the aforementioned linear form of the CFK equation. It has the
same form as a relationship given by McCartney (1990), but replaces the constant in the
McCartney equation by the term in (B-9) involving the mean pulmonary capillary oxygen
pressure and the equilibrium constant k. Substituting (B-9) into (B-l) yields a CFK equation free
of [O2Hb] and fully consistent with Coburn, Forster, and Kane's original derivation.
dt Vb BVb [THb]0 -[COHb] kMBV
In working with the CFK model it is convenient to express COHb as a percent of [RHb]o.
Multiplying (B-10) by 100 and dividing by [RHb]o yields the expression
d%\COHb\= 100 Vco | PIco %[COHb] 100(1 + ^C02) B
dt [THb]0 Vb BVb WO-%[COHb] k[RHb]0MBVb
Equation (B-l 1) can be written in the form suggested by Muller and Barton (1987):
dt WO-%[COHb]
where,
BV
(Eq.B-13)
= - - _ B
k[THb}0MBVb
Given values for the atmospheric pressure and the physiological variables in equations
(B-l 2) through (B-l 4), the value of %[COHb] at time t can be found by numerical integration
using such techniques as the fourth-order Runge-Kutta method (Press et al., 1986). Muller and
Barton (1987) demonstrated that an equation of the form of (B-12) is equivalent to a Michaelis-
Menten kinetics model which can be integrated. The integration yields:
_(C0 +Cl)t + %[COHb}, -%[COHb}0 -(100-%[CO/fl] Jin - = 0 (Eq. B-15)
01 ' ° ™ %[COHb]a-%[COHb]0
The equation for %[COHb]oo is obtained by setting equation (B-12) equal to zero and
solving for %[COHb], which is now equal to %[COHb]oo:
%[COHb]a> = 10°C° (Eq. B- 1 6)
(C0+Q)
B-5
-------
Equation (B-15) cannot be solved explicitly for %[COHb]. Muller and Barton (1987)
suggest the binary search method as one way to find the value of %[COHb]. Press et al. (1986)
contend a combination of the binary search and Newton-Raphson methods is faster on average.
Consequently, the pNEM/CO version of the COHb module used a combination of the binary
search and Newton-Raphson root finding methods to solve for COHb (Press et al., 1986). Using
the Muller and Barton (1987) solution increased the computation time when compared with the
Peterson and Stewart (1975) method, however it was still shown to be faster than the fourth-
order Runge-Kutta numerical integration.
The current version of APEX (APEX4.3) employs an alternative approach in which the
CFK equation is solved using a fourth-order Taylor's series expansion with subintervals. This
method, first incorporated in Version 3 of APEX, is described in detail in the Programmer's
Guide for the APEX3 Model by Glen (2002). This reference also includes the results of various
tests conducted on 10 candidate methods for solving the CFK equation. The selected method
(fourth-order Taylor series with subintervals) was chosen because of its simplicity, fast execution
speed, and ability to produce relatively accurate estimates of %COHb at both low and high levels
of CO exposure. Additional information concerning the %COHb calculation method and its
theoretical basis can be found in Section 10.2 of US EPA (2008a).
In developing the fourth-order Taylor Series expansion approach, Glen (2002) began by
defining N(t) as the %COHb level in the blood at time t, a quantity that is mathematically
restricted to range between 0 and 100 (percent). N(t) satisfies the following differential equation:
N'(t) = Co - Ci N(t) / (100 - N(t) ) (Eq. B-17)
where Co and Ci are constants (at least over the duration of one event) that depend on physical
and physiological parameters and on the CO concentration in the air. Equation (B-17) is
equivalent to (B-12) above, except that (B-12) uses the symbol %[COHb] instead of N(t).
The task of expanding N(t) in a Taylor's series becomes simpler if the following new
variables are defined:
D0= 1-N(0)/100 (Eq. B-18)
A0= C0/(C0 + Ci) (Eq.B-19)
Ai = Ci / (Co + Ci) (Eq. B-20)
D = Do-Ai (Eq. B-21)
z = (Co + Ci) t / (lOOxDo2) (Eq. B-22)
The z variable is a re-scaled time variable that is dimensionless. It is used as the
independent variable for the Taylor's series expansion. In equations expressed as functions of z
rather than t, any primes will indicate the derivatives with respect to z.
-------
Expressing (B-17) as a function of z yields the expression
]ST(z) = 100 D02 Ao - 100 D02 AI N(z) / ( 100 - N(z) ) (Eq. B-23)
The Taylor's series about the origin (z = 0) for N(z) is given by
N(z) = N(0) + 1ST (0) z + N''(0) z2 / 2 + N^(0) z3 / 6 + Niv(0) z4/ 24 + ... (Eq. B-24)
Through a series of algebraic substitutions, Glen (2002) shows that the Taylor series
expansion of N(z) truncated to the fourth order can be represented by
T4(z) = T3(z) - 100 AI Do D (Ai2 - 8 D AI + 6 D 2) z4 / 24 (Eq. B-25)
where
T3(z) = N(0) +100 Do D z -100 AI D0 D z2 / 2 + 100 AI D0 D (Ai - 2D) z3 / 6 (Eq. B-26)
Tests showed that the fourth-order Taylor series expansion (B-25) provided greater
accuracy than the third-order expansion for z values close to one. Glen (2002) found that z
values below one generally correspond to N(0) values or %COHb below forty to fifty percent for
one-hour exposure events.
The z value for a given event depends on the event duration, the initial %COHb level
N(0), and on the physiological parameters, and can be directly evaluated at the start of each
event. For events with a z value above some threshold, it is possible to improve the performance
of (B-25) by dividing the event into smaller events ("subintervals"), each with a shorter duration
and hence smaller z value. As the subinterval duration decreases, accuracy increases at the
expense of program execution time. APEX4.3 enables the user to select a limit on z which in
turn determines the number of subintervals to be used in applying the fourth-order Taylor
expansion. Glen (2002) recommends that the limit on z be set at 0.4 or 0.5.
B.3 Application of the COHb Model in APEX4.3
Description of APEX4.3 for CO
APEX4.3 follows the daily activities over an extended period of a finite set of simulated
individuals residing within a given geographic area. The period may be a single season or a
calendar year. Each simulated individual is defined by a set of general demographic
characteristics that includes age, gender, and body weight. The values of these factors are used
to derive values for blood volume, menstrual phase, endogenous CO production rate, and other
factors required by the COHb module (see Section B.4). The exposure of each individual is
represented by a continuous sequence of exposure events which span the time period of interest.
Each exposure event represents a time interval of 60 minutes or less during which the individual
resides in a single environment and engages in a single activity. To permit calculation of hourly
average exposures, exposure events are not permitted to fall in more than one clock hour.
Consequently, the passage from one exposure event to the next is indicated by a change in
microenvironment, activity, or clock hour. Algorithms within APEX4.3 calculate an average CO
B-7
-------
concentration for each exposure event according to the time, district, and microenvironment
specified for the event. As the exposure events for a simulated individual are contiguous, the
model can combine these concentrations to generate output distributions of one-hour and running
eight-hour exposures for each individual. The exposures calculated for the simulated individuals
can then be population-weighted to produce exposure distributions for population groups of
particular interest (e.g., people with coronary heart disease).
APEX4.3 constructs a year-long time/activity pattern for each simulated individual by
sampling 24-hour activity patterns from the Consolidated Human Activity Data Base (CHAD)
(McCurdy et al., 2000; US EPA, 2002), which is described in Section 4.4.3 of the CO REA. The
sampling approach attempts to match the 24-hour activity patterns to the simulated individual
and exposure period according to the demographic characteristics of the individual and the
season, day type (weekday/weekend), and maximum temperature of each day in the specified
exposure period.
The COHb Module
The COHb module in APEX4.3 currently employs the version of the CFK model
represented by equations (B-12) through (B-14) to compute an average COHb value over the
duration of each exposure event and an instantaneous COHb level at the end of each event. To
perform these computations, the COHb module requires information on each of the quantities
listed in the section describing the CFK model. In addition, the COHb level at the beginning of
the exposure event must be known. This latter quantity is usually the COHb level computed at
the end of the previous contiguous exposure event. To obtain the initial COHb at the start of the
exposure period, the computation is started one day before the beginning of the period. The
effect of the initial COHb value on the end value is negligible after about 15 hours. The program
stores the calculated COHb values for each exposure event and outputs distributions of COHb
levels by population group for averaging times ranging from one hour to one day.
Assignment of CFK Model Input Data for an Exposure Event
Section B.4 describes the equations and procedures used by the APEX4.3 COHb module
to obtain the values of the input variables for equations (B-2) and (B-13) through (B-16). A brief
overview is given here.
The actual inspired CO level can change significantly during an exposure event. The
model supplies an average exposure concentration for the event, which is used as the CO input.
The time constant for the change in COHb is sufficiently large that the use of concentrations
based on averaging times up to one hour can be used in place of the instantaneous concentrations
over the averaging time period with little loss of accuracy in estimating the COHb level at the
end of the exposure event. Furthermore, applying the average concentrations to a contiguous
sequence of exposure events does not cause an accumulation of error.
The COHb model presently used in APEX4.3 does not account for changing barometric
pressure. It uses a constant barometric pressure calculated for each study area as a function of
the average elevation of the area above sea level. The pressure at sea level is taken to be 760
torr.
The remaining input variables to the CFK model are all physiological parameters. While
-------
the Haldane coefficient, the equilibrium constant k, and average pulmonary capillary oxygen
pressure are treated as having the same constant values for all individuals, the remaining
physiological input variables will vary among individuals. The next section describes the
methods used to generate the various physiological input variables for each combination of
individual and calendar day processed by APEX.
B.4 Computation of Input Data for the COHb Module
As discussed in the previous section and in Sections 4.4.5 and 4.4.7 of the CO REA, the
algorithms used to estimate VE and COHb require values for various physiological parameters
such as body mass, blood volume, and pulmonary diffusion rate. Table B-2 provides a list and
description of the principal parameters; additional parameters are listed and described in Chapter
5 of US EPA (2008a). An algorithm within APEX4.3 probabilistically generates a value for
each parameter on the list (collectively referred to as a physiological profile) for each simulated
individual. Figure B-l is a flow diagram showing the process by which each physiological
profile is generated. Each of the generated physiological profiles is internally consistent, in that
the functional relationships among the various parameters are maintained. For example, blood
volume is determined as a function of weight and height, where height is estimated as a function
of weight. Weight in turn is selected from a distribution specific to gender and age.
For each simulated individual, APEX4.3 computes exposure for a contiguous sequence of
exposure events spanning the total time period of the computation. This multi-day sequence of
exposure events is determined by random sampling day-long event sequences from a set of pools
of 24-hour activity patterns. An individual 24-hour pattern in one of these pools is referred to
here as a unit exposure sequence (UES). Each pool consists of a collection of UESs that are
specific to selected demographic characteristics of the individual (e.g., age and gender), season,
day type (weekday/weekend), and maximum daily temperature.
A UES is a contiguous set of exposure events spanning 24 hours. Each event is
characterized by start time, duration, home/work status, microenvironment, and activity. All
exposure events are constrained to occur entirely within a clock hour. The CFK model within
the COHb module is called for each exposure event. Each event requires the following data.
Time duration of event (minutes)
Inspired CO partial pressure averaged over the event (torr)
Percent COHb at the start of the event (%)
Alveolar ventilation rate (ml/min STPD)
Average pulmonary capillary oxygen pressure (torr)
Haldane Coefficient (unitless)
Equilibrium constant for the reaction of Q^
Atmospheric pressure (torr)
Blood volume (ml)
Total potential reduced hemoglobin content of blood (ml CO/ml STPD)
Pulmonary CO diffusion rate (ml/min/torr STPD)
Endogenous CO production rate (ml/min STPD)
B-9
-------
Table B-2. Principal parameters included in the physiological profile for adults for applications of APEX4.3.
Parameter
Algorithm(s)
Containing
Parameter
Other Parameters
Required for
Calculating
Parameter
Method Used to Estimate Parameter Value
Age
COHb
Ventilation
rate
Demographic group
Randomly selected from population-weighted distribution specific to demographic
group
Gender
COHb
Ventilation
rate
Demographic group
Randomly selected from population-weighted distribution specific to demographic
group
Body Weight
COHb
Ventilation
rate
Gender
Age
Randomly selected from population-weighted lognormal distributions with age- and
gender-specific geometric mean (GM) and geometric standard deviation (GSD)
derived from data from the National Health and Nutrition Examination Survey
(NHANES), for the years 1999-2004 (Isaacs and Smith, 2005)
Height
COHb
Weight
Gender
Estimated using equations developed by Johnson (1998) using height and weight
data provided by Brainard and Burmaster (1992).
Males: height = 34.43 inches + (6.67)[ln(weight)] + (2.38 inches)(z)
Females: height = 48.07 inches + (3.07)[ln(weight)] + (2.48 inches)(z)
The z term is randomly selected from a unit normal [N(0,1)] distribution. Units:
height (inches), weight (Ibs).
Menstrual phase
COHb
Gender
Age
If gender = female, menstrual phase was randomly assigned in alternating 14-day
cycles according to the following age-specific probabilities.
Age < 12 or >50: 100% premenstrual
Age 12 through 50: 50% premenstrual, 50% postmenstrual.
B-10
-------
Parameter
Algorithm(s)
Containing
Parameter
Other Parameters
Required for
Calculating
Parameter
Method Used to Estimate Parameter Value
Blood volume
COHb
Gender
Weight
Height
Blood volume (Vb) was determined according to gender by the following equations
developed from Allen et al. (1956) which were modified to accept the units used for
height and weight.
Males: Vb = (20.4)(weight) + (0.00683)(H3) - 30
Females: Vb = (14.6)(weight) + (0.00678)(H3) - 30
Units: blood volume (ml), weight (Ibs), height (inches).
Hemoglobin
content of the
blood, Hb
COHb
Gender
Age
Randomly selected from normal distribution with arithmetic mean (AM) and
arithmetic standard deviation (ASD) determined by gender and age based on data
obtained from the National Health and Nutrition Examination Survey (NHANES),
for the years 1999-2004 (see Isaacs and Smith (2005) in the CO REA, Appendix A)
Units: grams of Hb per deciliter of blood
B-ll
-------
Parameter
Algorithm(s)
Containing
Parameter
Other Parameters
Required for
Calculating
Parameter
Method Used to Estimate Parameter Value
Pulmonary CO
diffusion rate,
DL
COHb
Gender
Height
Age
Pulmonary CO diffusion rate (DL) was determined according to gender, height, and
age according to the following equations obtained from a paper by Salorinne (1976)
and modified to conform to the units used in the COHb module.
Males:
= (0.361)(height) - (0.232)(age) + 16.3 ml/min/torr
Females:
DLCQ = (0.556)(height) - (0.115)(age) - 5.97 ml/min/torr
Units:
DL (ml/min/torr), height (inches), age (years).
Given the alveolar ventilation rate for the exposure event the associated adjusted
pulmonary diffusion rate is calculated as:
DL (Adjusted) = DL (Base) + 0.000845F, - 5.7
Endogenous CO
production rate
COHb
Gender
Age
Menstrual phase
Endogenous CO production rate was randomly selected from a lognormal
distribution with geometric mean (GM) and geometric standard deviation (GSD)
determined according to the following equations specific to age, gender, and
menstrual phase groupings (see data in Table B-3).
Males, 18+: GM = 0.473, GSD = 1.316
Females, 18+, premenstrual: GM = 0.497, GSD = 1.459
Females, 18+, postmenstrual: GM = 0.311, GSD = 1.459
Units: GM (ml/hr), GSD (dimensionless).
B-12
-------
Parameter
Resting
metabolic rate
(RMR)
Energy
conversion factor
(ECF)
NV02max
V02max
Algorithm(s)
Containing
Parameter
Ventilation
rate
Ventilation
rate
Ventilation
rate
Ventilation
rate
Other Parameters
Required for
Calculating
Parameter
Gender
Age
Body Weight
Gender
Gender
Age
NV02max
Body Weight
Method Used to Estimate Parameter Value
See Section 4.4.5 of CO REA and Chapter 5 of US EPA (2008a).
See Section 4.4.5 of CO REA and Chapter 5 of US EPA (2008a).
See Section 4.4.5 of CO REA and Chapter 5 of US EPA (2008a).
See Section 4.4.5 of CO REA and Chapter 5 of US EPA (2008a).
B-13
-------
Gender
Demographic Profile of
Simulated Individual
Weight
Height
Menstrual Phase
Endogenous CO
Production Rate
Blood Volume
Total Hemoglobin
Content
of the Blood
Pulmonary CO
Diffusion Rate
Resting Metabolic Rate
(RMR)
Energy Conversion Factor
(ECF)
Maximum Normalized
Oxygen Uptake Rate
(NV02max)
Maximum Oxygen Uptake
Rate(V02max)
Age
Figure B-l. Flow diagram for physiological profile generator. Input data is supplied at
the start of the APEX4.3 computation.
B-14
-------
Given these data as inputs, the module computes the percent COHb at the end of the
exposure event. This value is used by the module as the initial percent COHb for the next
contiguous exposure event. The module also computes the average percent COHb value for each
exposure event. The main program retains these values and uses them to calculate percent
COHb values for averaging times ranging from one hour to one day.
Some of the above data do not change during an APEX4.3 computer run and, therefore,
need to be supplied to the computer program only once at the start. Some of the data vary with
the individual and therefore need to be supplied at the beginning of each activity day. Other data
tend to change with the exposure event and therefore need to be supplied for each new exposure
event.
Barometric Pressure
A constant barometric pressure is assumed for the study area based on the average height
above sea level:
PB = 760 x exp(-0.0000386 x Altitude) (Eq. B-27)
where altitude is the average height (in feet) of the study area above sea level (US EPA, 1978).
The altitude was set at 5,183 feet for Denver and 328 feet for Los Angeles.
Average Pulmonary Capillary Oxygen Pressure
The equation employed is based on an approximation used by Peterson and Stewart
(1975) in which 49 torr is subtracted from the partial pressure of inspired oxygen. This leads to
the following approximate relationship:
Pc02 = 0.209(PB - 47) - 49 (Eq. B-28)
where 0.209 is the mole fraction of O2 in dry air and 47 is the vapor pressure of water at body
temperature. This expression was used in an investigation of the CFK equation by Tikuisis et al.
(1987). A value of 100 torr is commonly used since Equation (B-28) generates this value for a
barometric pressure equivalent to 760 torr.
Haldane Coefficient
The value of 218 was used for the Haldane coefficient. While measured values in the
range 210 to 270 have been reported in the extant literature, most researchers use values within
the range of 210 to 240. In the early 1980's, the Clean Air Scientific Advisory Committee
(CASAC) expressed the opinion to EPA (Friedlander, 1982) that the most careful work done in
this area was that by Rodkey et al. (1969), who determined a value of 218. This value was
selected for use in the COHb module of the earlier CO-NEM exposure model. Other researchers
using values in the range 218 to 220 include Peterson and Stewart (1970), Marcus (1980), Collier
and Goldsmith (1983), and Muller and Barton (1987). As the value 218 falls within the range
currently used by researchers, we have elected to continue using this value in APEX4.3.
Equilibrium Constant for the Reaction of O? and RHb
This quantity was estimated in Section B.2 to have the value 0.32 based on the
B-15
-------
observation that %[RHb] is about 3% in individuals breathing air which is free of CO and a
value of 100 torr for Pco^ .
Total Reduced Hemoglobin in the Absence of O? and CO
The quantity [THb]o is expressed as equivalent milliliters of Q^ or CO at STPD per
milliliter of blood. Total Hb blood levels are customarily expressed as grams per deciliter of
blood. The total Hb level in the absence of COHb and O2Hb would consist principally of RHb
which can react with O2 or CO and MetHb which cannot. Total Hb blood levels also tend to be
higher in people living at higher altitudes. To relate [THb]o to Hb, it is therefore necessary to
correct for the MetHb present, adjust for the effect of altitude, and convert to equivalent
milliliters of CO at STPD. The later conversion is based on the observation that a gram of
reduced Hb can react with a maximum of 1 .39 ml of Oi or CO at STPD. The application of
these three factors yields the equation:
\RHb}0 = 1 .39 x #6(100 - %MetHb) x 1 + (Eq B-29)
where HbAlt is the percent increase in Hb due to exposure to altitude and is given by (US EPA,
1978):
HbAlt = 2.76e°-OOOU49A""ude
Hb in equation (B-29) is a sea level value. Hb level in a human population is normally
distributed with the mean Hb and standard deviation both dependent on gender and age class (see
entry in Table B-2 for the distributions of Hb by age and gender). Given the hemoglobin content
of the blood based on the distributions listed in Table B-2, [THb]o is calculated using equation
(B-29). The weight percent MetHB, %MetHB, is taken to be 0.5% of the weight of Hb (Muller
and Barton, 1987).
Determination of Weight
Body mass or weight (in kg) was determined by fitting lognormal distributions to data
organized by age and gender from the National Health and Nutrition Examination Survey for the
years 1999-2004 (Isaacs and Smith, 2005). Distribution parameters were estimated for single-
year age cohorts for both genders for ages 0-85. As the NHANES 1999-2004 studies only
covered persons up to age 85, linear forecasts for the parameters were made for ages 86-100, as
based on the data for ages 60 and greater.
Determination of Height
The following equations were used to estimate height as a function of gender and weight.
Equations B-30 and B-3 1 were derived by Johnson (1998) using height and weight data provided
by Brainard and Burmaster (1992).
males: height = 34.43 inches + (6.67)[ln(weight)] + (2.38 inches)(z) (Eq. B-30)
females: height = 48.07 inches + (3.07)[ln(weight)] + (2.48 inches)(z) (Eq. B-3 1)
where the z term was randomly selected from a unit normal [N(0,l)] distribution.
B-16
-------
Base Pulmonary Diffusion Rate of CO
A base lung diffusivity of CO for the individual is calculated as follows:
Men: DLm = 0.361 xheight-0.232xage +16.3 (Eq. B-32)
Women: DLm = 0.556 x height-O.I \5xage-5.97 (Eq. B-33)
where height is in inches and age is in years.
The regression equations were obtained from a paper by Salorinne (1976) and modified
to conform to the units used in the COHb module. The Salorinne data were obtained for non-
exercising individuals. Tikuisis et al. (1992), working with eleven male subjects at various
exercise levels, showed significant increase in lung diffusivity of CO with increasing alveolar
ventilation rate. Regression analyses of data provided by Tikuisis et al. (1992) for the individual
subjects in the study showed the relationship to be linear. From this relationship and the heights
and ages of the subjects in the Tikuisis et al. (1992) study, it was determined that the Salorinne
(1976) equations for male subjects correspond to an alveolar ventilation rate of 6.69 1/min STPD.
In the absence of other data it is assumed that this same value applies to women. Thus, for each
twenty-four hour period equations B-32 and B-33 are used to compute lung diffusion rates of CO
for a base case alveolar ventilation rate of 6.69 1/min STPD. As will be seen, this value is
adjusted to account for the actual ventilation rate experienced by the simulated individual during
each individual exposure event.
Endogenous Rate of CO Production
The endogenous CO production rates taken from a number of sources show the rate to be
distributed lognormally in the population (see Table B-3 for data and sources). The distribution
is different for men and women. For a woman there is a further difference depending on whether
she is in her premenstrual or postmenstrual phase. Table B-2 presents these distributions
classified by class, gender, and menstrual phase.
For each male individual, APEX4.3 specifies a single value for endogenous CO
production rate and uses it for all days of the year. For each female individual between 18 and
64 years of age, APEX4.3 specifies one value of endogenous CO production rate to represent
premenstrual days and one value to represent postmenstrual days. Female individuals under 12
years and older than 50 are assumed to be premenstrual; consequently, APEX4.3 specifies a
single value for endogenous CO production rate to be used for all days of the year. The specified
values are randomly selected from the appropriate distributions presented in Table B-2. A
random number, z, is sampled from the standardized normal distribution, N(0,l) to make each
selection. The appropriate endogenous CO production rate is then obtained from:
Vco = 0.01667 x (geom.mean) x (geom.S.D.y (Eq. B-34)
The constant term converts ml/hr to ml/min.
A probabilistic algorithm within APEX4.3 assigns a menstrual phase to each day of the
year for female individuals aged 12 to 50 years. The algorithm randomly assigns a number
B-17
-------
between 1 and 28 to January 1. The number is increased by one for each successive day until
number 28 is reached. The next day is numbered 1 and the 28-day numbering cycle is repeated
until each day of the year has been assigned a number between 1 and 28. Days numbered 1
through 14 are identified as post-menstrual days; days numbered 15 through 28 are identified as
pre-menstrual days.
B-18
-------
Table B-3. Literature data used to derive lognormal distributions used to estimate endogenous CO production rate.
Reference
Berketal. (1974)
Brouillard et al. (1975)
Coburnetal. (1963)
Coltman and Dudley,
(1969)
Delivoria-Papadoppulos
etal. (1974)
Luomanmaki and
Coburn (1969)
Lynch and Moede
(1972)
Merke et al. (1975)
Werner and Lindahl
(1980)
Gender
Male
Male
Male
Male
Male
Female
Male
Male
Female
Female
Female
Male
Menstrual
Cycle
(pre/post)
NA
NA
NA
NA
NA
Pre
Post
NA
NA
Pre
Post
Pre
Post
NA
Endogenous CO Production Rate (ml/hr
0.43
0.81
0.35
0.45
0.58
0.33
0.46
0.57
0.23
0.38
0.40
0.72
0.48
0.64
0.40
0.54
0.58
0.57
0.40
0.38
0.45
0.26
0.54
0.51
0.42
0.81
0.37
0.23
0.86
0.47
0.76
0.52
0.33
0.39
0.51
0.36
0.60
0.72
0.34
0.41
0.26
0.23
0.25
0.35
0.23
0.48
0.59
0.70
0.43
0.55
0.45
0.99
0.41
0.54
0.65
0.33
0.20
0.52
0.24
0.31
0.80
0.35
0.37
0.39
0.48
0.26
0.38
0.55
0.42
0.22
0.80
0.55
0.69
0.72
0.51
0.49
0.40
0.53
0.16
0.62
0.44
0.15
0.54
0.32
0.70
0.42
0.45
0.43
0.30
0.44
0.29
0.21
0.68
0.43
0.36
0.57
0.50
0.48
0.28
0.35
0.65
B-19
-------
B.5 Input Data Supplied By APEX 4.3 with Each Exposure Event
Duration of Exposure Event
The duration of the exposure event in minutes is supplied by the main program to the
COHb module.
Partial Pressure of Inspired Carbon Monoxide
The main program supplies the inspired CO concentration averaged over the duration of
the exposure expressed as ppm. This quantity is converted to pressure via:
Plco = (CO) x (Pb - 47) x 1(T6 (Eq. B-35)
Initial Percent COHb Level at Start of Exposure Event
The program retains the percent COHb computed at the end of the previous exposure
event and uses this value as the initial percent COHb for the present event. The starting COHb at
the beginning of an activity day is the final COHb level at the end of the preceding activity day.
This latter procedure is used for the first activity day of the overall computation since the
program starts the day before the overall period covered by the APEX4.3 computation.
Alveolar Ventilation Rate
The main program supplies the COHb module with ventilation rate derived from the
algorithm discussed in Section 4.4.5 of this report.
Adjusted Pulmonary Diffusion Rate of CO
Given the alveolar ventilation rate for the exposure event the associated adjusted
pulmonary diffusion rate can be calculated from:
DL (Adjusted) = DL (Base)+ 0.000845VA -5.7 (Eq. B-36)
B.6 Analysis of Selected APEX COHb Outputs
This section provides analysis of APEX outputs by using the event-level APEX
simulations. The APEX events and hourly output files can provide event-level and hourly-level
exposure and dose profiles for all simulated individuals (US EPA, 2008a, 2008b). We generated
the hourly time-series of dose for approximately 400 CHD simulated individuals, with the CHD
population defined as described in section 5.5.1 of the CO REA.1 Both the Los Angeles and
Denver study areas were evaluated using two air quality scenarios: as is 2006 ambient air quality
and historical2 air quality adjusted to just meet the current 8-hour standard. Two model
1 For each of these particular model runs, APEX generated the complete time-series of exposure and dose
for 5,000 persons. Of the 5,000 persons simulated, APEX generated 438 CHD persons in Denver and 394 CHD
persons in Los Angeles based on the NHIS CHD prevalence rates.
21997 ambient monitoring data was used in Los Angeles, 1995 data was used in Denver. An additional air
quality/exposure scenario included air quality adjusted to just meet a 99th percentile 8-hour daily maximum of 5.0
ppm, though because these endogenous CO production runs were based on the same historical years of data in each
B-20
-------
simulations were run for each scenario and study area; the first generated total COHb levels (i.e.,
COHb dose resulting from both ambient CO exposure and endogenous CO production) and the
second simulation generated COHb levels with all exposure concentrations set to zero (i.e., to
estimate COHb resulting from endogenous CO production alone). See CO REA chapters 4 and 5
for details regarding all other relevant APEX model simulation settings.
First, an analysis of the contribution of endogenous CO production to COHb levels is
provided. This is followed by an analysis of the ambient concentration profile and dose time
series for a few selected individuals. Finally, the total maximum end-of-hour COHb levels are
evaluated with respect to the maximum COHb contribution from ambient exposure and
maximum COHb contribution from endogenous CO production.
B.6.1 Contribution of Endogenous CO Production to COHb Level
We were interested in determining the relative contribution of endogenous CO
production to an individual's total COHb level. In this first analysis, we generated the hourly
time-series of dose for the approximately 400 CHD simulated individuals in the absence of CO
exposure. When an APEX model simulation is performed with concentration input values equal
to zero, endogenous CO production will be entirely responsible for the calculated COHb levels
for each simulated person.
We note here that while the ambient CO concentration is set to zero for these runs, the
temperature data associated with each simulation time period will affect the activity diaries
sampled from CHAD used to construct each individual's longitudinal activity pattern profile.
This will have a small impact on the estimated COHb doses due to, for example, differences in
each simulated individual's ventilation rates that are linked to the activity pattern profile.
Therefore, we expect there to be small differences in the contribution of endogenous CO
production to COHb level when comparing the two exposure scenarios within each study area.
Furthermore, we expect the contribution of endogenous COHb production to be greater in
Denver when compared with that of simulated persons in Los Angeles, given the large
differences in altitude between the two study areas.
First, we generated descriptive statistics for each individual's 8,760 hourly average
COHb dose values resulting from endogenous CO production alone. Then annual means were
calculated for each individual and used to generate a population distribution of annual means for
each study area and air quality scenario. These population statistics are summarized in Table B-
4. The annual average hourly COHb level for the population was 0.32% for the simulated CHD
population in Denver and about 0.27% for the simulated CHD population in Los Angeles. There
was variability in the population means as indicated by the range of individual annual mean
values that extended from a low of about 0.11% COHb, to a high of about 1.1% COHb. Overall
the population variability in annual average COHb, as measured by the coefficient of variability
(COV), was about 42% for both study areas and for both exposure scenarios.
The variability in each individual's annual mean COHb, as measured by an average of
each person's respective COV for the year, was also calculated and found generally consistent
for each simulated individual, and when considering the two study areas and both exposure
study area, the generated time-series data for the endogenous CO contribution would be identical. These exposure
scenarios are described more fully in section 5.6 of the CO REA.
B-21
-------
scenarios. On average, an individual's hourly COHb was estimated to vary by about 22-23% of
their annual average hourly COHb. Note the standard deviation of the individual-based COVs
equaled about 5%. When compared to the population variability statistics, it can be seen that
there is greater inter-personal variability in the contribution of endogenous CO production to
COHb levels than intra-personal variability. This finding is generally as expected, given the
wide range of interpersonal attributes that might influence the COHb level including age, gender,
body mass, blood hemoglobin content, endogenous CO production rate, etc., compared with the
limited number of intra-personal attributes that might influence COHb level, such as alveolar
ventilation rate. Note that this analysis does not include variability in COHb due to diseases and
other medical conditions, some of which may increase or decrease the endogenous CO
production rate (e.g., hemolytic anemia and infection) (ISA, Section 4.5).
B.6.2 Time-Series of COHb Levels In Individuals and Associated Ambient Concentrations
The time-series of hourly average COHb levels was plotted along with the hourly
ambient CO concentration for three simulated CHD persons in Los Angeles. The purpose was to
illustrate the variation in COHb levels occurring over time, with respect to endogenous CO
production and ambient CO concentrations3 (which of course would ultimately be related to
ambient CO exposure). We purposefully selected these three simulated persons to represent a
low, mid, and high level COHb time-series. A full week of data was extracted from the air
quality scenario of just meeting the current standard and included the hours where the ambient
CO concentration approached the 8-hour CO standard of 9.4 ppm. Note again that the total
COHb in this analysis refers to the COHb dose resulting from both ambient CO exposure and
endogenous CO production. The time-series for all three persons was for the 1-week period
from December 17 through December 23 (Figure B-2), though the individuals have entirely
different ambient CO concentration profiles and endogenous CO production rates, and therefore
different total COHb levels over the time period of interest.
The person designated as 'low' had a mean total COHb dose level of about 0.6% across
the time period, with hourly total COHb levels ranging from about 0.4 to 0.7% (Figure B-2, top).
Hourly ambient CO concentrations used in calculating this person's exposure were generally low
(i.e., less than 1 ppm) though, on occasion, ranged upwards to 4 ppm. This individual, while
experiencing low total COHb levels across this time period, actually has a relatively high
contribution from endogenous CO production (on average contributing to a COHb level of about
0.4 % over the 1-week period) with limited COHb contribution from ambient CO exposure (on
average contributing to a COHb level of about 0.2%).
The person designated as 'mid' also had a mean COHb dose level of about 0.6% (Figure
B-2, middle) though with greater variability in hourly COHb level when compared with the
'low' person. The COHb profile for the 'mid' person extended upwards to a peak COHb level of
about 1.2% on several hourly excursions, largely in response to exposure to ambient CO
concentrations. The estimated contribution to COHb levels resulting from endogenous CO
production for the 'mid' person was within the lower range of the population average (Table B-
4), having an average hourly COHb level from endogenous CO of less than 0.2% for this time
period.
3 This is the ambient concentration used as input to APEX, only adjusted for the air quality scenario. The
1997 Los Angeles monitoring data were adjusted by a factor of 0.627 to just meet the current 8-hour standard.
B-22
-------
The response to ambient CO concentrations is more notable when observing the profile
of the designated 'high' person (Figure B-2, bottom). On average, this person had a total COHb
level of 1.0% across the illustrated 1-week period, though 1-hour total COHb levels peaked just
above 2.9% following the upwards spiking of associated hourly ambient CO concentrations of
between 10 and 12 ppm. Note that COHb levels resulting from endogenous CO production
alone for this individual were similar to the population average (Table B-4), at about 0.3%
COHb for this time period.
This analysis suggests that moderate to high total COHb levels result from ambient
exposure rather than endogenous CO production, while individuals with lower total COHb levels
have a greater contribution to total COHb from endogenous CO.
B.6.3 Evaluation of Maximum End-of-Hour COHb Levels with Respect to Contribution
from Ambient CO Exposure and Endogenous CO Production
In this third analysis, we calculated three dose metrics using the event-level output files
generated for the CHD population in each study area and for the two air quality scenarios: as is
air quality and just meeting the current 8-hour standard. Consistent with the standard output
generated for the CO REA, we generated a single maximum end-of-hour COHb level for each
simulated CHD person, resulting from both ambient CO exposure and endogenous CO
production (and termed total COHb in this document). Next, we used the COHb levels
generated from APEX simulations employing zero exposure concentrations, to estimate a
maximum COHb level resulting from endogenous CO production alone. And finally, the time-
series of COHb levels resulting from endogenous CO production were subtracted from the time-
series of total COHb (i.e., from endogenous production and ambient exposure), to estimate the
maximum COHb level resulting from ambient CO exposure alone. In considering the results, we
note that the focus is on ambient CO contribution to exposure, although in some cases, other CO
sources can play a more important role in COHb levels (as described in section 2.3 of CO REA).
The percentiles of the distribution of each of these metrics were calculated in 0.5
percentile increments, the results of which are illustrated in Figure B-3. Selected percentiles of
each distribution are provided in Table B-5. As noted above for the distribution of annual mean
COHb levels resulting from endogenous CO production, the maximum contribution is nearly
identical when comparing the air quality scenarios within the study areas. In general, over 98%
of the simulated CHD population has a maximum end-of-hour COHb level resulting from
endogenous CO production of less than 1.0%. There is a difference in the highest maximum
end-of-hour COHb level contributed by endogenous CO production when comparing Denver to
Los Angeles; the maximum extends upwards to a COHb level of about 1.5% in Denver, while in
Los Angeles the highest maximum end-of-hour COHb level is about 1.0%.
The distribution of the maximum end-of-hour COHb levels resulting from ambient
exposure alone is consistently higher than that resulting from endogenous CO production,
indicating the relative importance of ambient CO exposure; however, it is not always the case
that the maximum end-of-hour total COHb level occurred simultaneously with the maximum
contribution from ambient exposure. About half of the time, this was the case (Table B-6),
though it varied based on the study area and air quality scenario considered. Even when a
person's maximum total COHb did not occur at the same time as the maximum ambient
contribution, the ambient contribution, on average was between 51-66% of a person's maximum
total COHb.
B-23
-------
Table B-4. Descriptive statistics for annual average hourly COHb levels resulting from endogenous CO production alone and
using an APEX simulated CHD population.
Study
Area
Denver
Los
Angeles
Air
Quality
Scenario
As Is
Current
Standard
As Is
Current
Standard
CHD Population Annual Averag
Mean
0.315
0.315
0.268
0.268
stdev
0.132
0.133
0.111
0.111
min
0.121
0.122
0.095
0.095
P1
0.136
0.136
0.105
0.106
P5
0.161
0.161
0.139
0.139
e Hourly COHb Levels (%)a
p50
0.285
0.285
0.246
0.245
p95
0.573
0.576
0.506
0.504
p99
0.799
0.804
0.675
0.665
max
1.074
1.087
0.695
0.695
COV
42.1
42.3
41.5
41.5
Individual Variability
in Hourly COHb
Levels (%)b
Mean of
COVs
20.2
20.1
22.8
22.8
Stdev of
COVs
5.2
5.2
5.6
5.6
Notes:
a An annual hourly average COHb (i.e., the 8,760 hourly values) for each simulated individual was first calculated. Then the mean, standard
deviation, and selected percentiles of this annual hourly average were calculated for the simulated population (i.e., 438 persons in Denver,
394 persons in Los Angeles).
b The COV (mean/standard deviation) was calculated for each individual using each person's 8,760 hourly values. Then the mean and
standard deviation of these individual COVs was calculated for each simulated population
Table B-5. Selected percentiles of the maximum end-of-hour COHb levels using an APEX simulated CHD population,
considering the maximum contribution from endogenous CO production, the maximum contribution from
ambient CO exposure and the maximum total COHb level.
Study
Area
Denver
Los
Angeles
Air
Quality
Scenario
As Is
Current
Standard
As Is
Current
Standard
Maximum End-of-Hour COHb Level (%)a
Endogenous Contribution
min
0.19
0.19
0.14
0.14
med
0.43
0.43
0.38
0.37
max
1.55
1.54
1.04
1.04
Ambient Contribution
min
0.24
0.43
0.24
0.30
med
0.50
0.94
0.60
0.72
max
1.70
2.74
1.70
2.77
Total
min
0.46
0.64
0.46
0.48
med
0.80
1.21
0.84
0.97
max
1.85
2.95
1.92
3.04
Notes:
3 Selected percentiles correspond to COHb levels in Figure B-3.
B-24
-------
3.0
2.5-
_2.0-
£
| 1.5
O
0.5-
0.0
3.0
Person 1:
Low COHb
COHb contribution from
endogenous CO
COHb total (ambient CO
exposure + endogenous CO)
- - - Ambient monitor concentration
2.5-
_2.0-
Person 2:
Mid COHb
COHb contribution from
endogenous CO
COHb total (ambient CO
exposure + endogenous CO)
12
-- 10
8 f
Q.
a.
O
6 ^
'c
Ol
!5
+ 4 I
2
12
-- 10
2.5-
1.5-
O
O
1.0-
0.5-
0.0
Person 3:
High COHb
COHb contribution from
endogenous CO
COHb total (ambient CO
exposure + endogenous CO)
- - - Ambient monitor concentration
-- 10
-- 8
-- 6
a.
O
4 I
2
0
8400 8424 8448 8472 8496 8520 8544 8568
Hour of Day in Year
Figure B-2. Time-series profile of COHb levels and ambient CO for three simulated
CHD persons in Los Angeles - air quality just meeting the current standard.
B-25
-------
Denver: As Is Air Quality
Los Angeles: As Is Air Quality
100
90
80
70
60
50
40
30
20
10
0
a.
o
Q.
O
Q.
Maximum Contribution from Ambient CO Exposure Alone
- - - Maximum Contribution from Endogenous CO Production Alone
Maximum (Ambient Exposure + Endogenous CO Production)
Maximum Contribution from Ambient CO Exposure Alone
- - - Maximum Contribution from Endogenous CO Production Alone
Maximum (Ambient Exposure + Endogenous CO Production)
Denver: Just Meeting the Current Standard
Los Angeles: Just Meeting the Current Standard
100 -F
90 :-
80 :-
I ^o;
| 60 :
Q.
O
"~r -/^- •$>**—
/ /s\
h ' S
/ f
'-! /
/
/
/
, • i
Maximum Contribution from Ambient CO Exposure Alone
• • • Maximum Contribution from Endogenous CO Production Alone
Maximum (Ambient Exposure + Endogenous CO Production)
50 :-
40 :-
Q.
O
Q_
30 :-
20 f- ----i /--
10 i ' /
V ^
Maximum Contribution from Ambient CO Exposure Alone
- - - Maximum Contribution from Endogenous CO Production Alone
Maximum (Ambient Exposure + Endogenous CO Production)
0.0 0.5 1.0 1.5 2.0 2.5
Maximum End-of-Hour COHb (%)
3.0
3.5 0.0
1.0 1.5 2.0 2.5
Maximum End-of-Hour COHb (%)
3.0
3.5
Figure B-3. Maximum end-of-hour COHb levels using an APEX simulated CHD population in Denver (left) and Los Angeles
(right), considering as is air quality (top) and air quality adjusted to just meet the current 8-hour standard
(bottom).
B-26
-------
Table B-6. Evaluation of the ambient contribution occurring simultaneously with a
CHD person's maximum total end-of-hour COHb in Denver and Los
Angeles.
Study
Area
Denver
Los
Angeles
Air
Quality
Scenario
As Is
Current
Standard
As Is
Current
Standard
Maximum Ambient COHb
Contribution COHb Occurred
Simultaneously with
Maximum Total COHb
Yes
No
Yes
No
Yes
No
Yes
No
Persons
(n)
183
255
320
118
213
181
229
165
(%)
41.8
58.2
73.0
27.9
54.1
45.2
58.1
41.9
Percent of Ambient COHb
Contribution to Maximum
Total COHb3
min
32.5
7.5
57.6
23.4
43.5
20.1
43.5
26.2
mean
73.5
50.6
79.9
65.9
77.0
57.7
79.6
63.6
max
95.6
86.2
97.6
89.9
93.7
90.1
95.2
90.5
Notes:
a The ambient COHb contribution that occurred at the same time as the persons maximum total COHb
was used for this calculation.
Each simulated CHD person's maximum end-of-hour total COHb level was also plotted
against their respective ambient contribution to that maximum COHb level. We separated each
of the air quality scenarios, the study areas, and whether the maximum ambient contribution
coincided with the maximum total COHb level or did not. The results of this analysis are
illustrated in Figures B-4 and B-5 for Denver and Los Angeles, respectively.
As expected, strong linear relationships exist when the maximum ambient contribution
occurred at the same time as the maximum total COHb level, regardless of air quality scenario or
study area. The coefficient of determination (R2 values) ranged from about 0.80 to 0.92, with
linear regression slopes very close to unity, supporting the importance of the ambient
contribution to these persons' maximum total COHb level. Regression intercepts were also
similar to one another, ranging from a value about 0.21 to 0.31%, and generally approximate the
average endogenous contribution to COHb.
Relatively weaker relationships were exhibited when the maximum total COHb level did
not correspond with the maximum ambient contribution. The coefficient of determination
ranged from about 0.11 to 0.48 with regression slopes ranging from 0.4 to 0.7. The relative
greater importance of the endogenous contribution to COHb for these persons is indicated by the
higher regression intercepts, estimated to range from about 0.51 to 0.65% COHb.
The most important message conveyed by these figures illustrating simulations for which
exposure included only contribution from ambient CO sources is that for all simulated persons
having a maximum end-of-hour COHb level at or above 2.0%, and for most persons having a
maximum end-of-hour COHb level at or above 1.5%, the ambient contribution is a more
important factor than endogenous CO production. This is indicated by the relatively few data
points that greatly deviate from the bounded regression lines at or above these selected COHb
levels, regardless of whether the maximum total COHb occurred at the same time as the
maximum ambient contribution.
B-27
-------
Denver: As Is Air Quality
Max Ambient Contribution Did Not Occur at Same Time as Max Total COHb
Denver: Just Meeting the Current Standard
Max Ambient Contribution Did Not Occur at Same Time as Max Total COHb
Denver: As Is Air Quality
Max Ambient Contribution Occurred at Same Time as Max Total COHb
Denver: Just Meeting the Current Standard
Max Ambient Contribution Occurred at Same Time as Max Total COHb
0.5 1.0 1.5 2.0 2.5 3.0
Ambient Contribution to Maximum End-of-hour COHb (%)
3.5 0.0
0.5 1.0 1.5 2.0 2.5 3.0
Ambient Contribution to Maximum End-of-hour COHb (%)
3.5
Figure B-4. Relationship between ambient contribution to COHb level and the corresponding maximum total end-of-hour
COHb level in Denver. Air quality as is (left) and just meeting the current standard (right). Maximum ambient
contribution did not occur with maximum total COHb (top) and maximum ambient contribution did occur with
maximum total COHb (bottom).
B-28
-------
3.5
i3.0
.a
O 9^
o 2-5
^ 2.0;
5,
Los Angeles: As Is Air Quality
Max Ambient Contribution Did Not Occur at Same Time as Max Total COHb
Los Angeles: Just Meeting the Current Standard
Max Ambient Contribution Did Not Occur at Same Time as Max Total COHb
.5
1.1
.0 :
S
0.5 --
0.0
y = 0.5914x+0.5396
R2 = 0.2344
y = 0.7481x +0.5052
R2 = 0.4342
3.5
Los Angeles: As Is Air Quality
Max Ambient Contribution Occurred at Same Time as Max Total COHb
Los Angeles: Just Meeting the Current Standard
Max Ambient Contribution Occurred at Same Time as Max Total COHb
£3.0
£l
O 9^
o 2-5
2.0 :-
LU
ro
5
0.0
1.0--- --.-r
y = 0.9938X + 0.2099
R2 = 0.9162
0.0
3.5
0.5 1.0 1.5 2.0 2.5 3.0 3.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Ambient Contribution to Maximum End-of-hour COHb (%) Ambient Contribution to Maximum End-of-hour COHb (%)
Figure B-5. Relationship between ambient contribution to COHb level and the corresponding maximum total end-of-hour
COHb level in Los Angeles. Air quality as is (left) and just meeting the current standard (right). Maximum
ambient contribution did not occur with maximum total COHb (top) and maximum ambient contribution did
occur with maximum total COHb (bottom).
B-29
-------
B.7 References
Allen TH, Peng MT, Chen KP, Huang TF, Chang C, Fang HS. (1956). Prediction of blood volume and adiposity in
man from body weight and cube of height. Metabolism. 5:328-345.
Biller WF and Richmond HM. (1982). Sensitivity Analysis on Coburn Model Predictions of COHb Levels
Associated with Alternative CO Standards. Report to Strategies and Air Standards Division of the Office
of Air quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park,
NC, November, 1982.
Brouillard RP, Contrad ME, Bensinger TA. (1975). Effect of blood in the gut on measurements of endogenous
carbon monoxide production. Blood. 45:67-69.
Brainard J and Burmaster D. (1992). Bivariate distributions for height and weight of men and women in the United
States. RiskAnalysis. 12(2):267-275.
Berk PD, Rodkey FL, Blaschke TF, Collison HA, Waggoner JG. (1974). Comparison of plasma bilirubin turnover
and carbon monoxide production in man. J Lab ClinMed. 83:29-37.
Coburn RF, Forster RE, Kane RB. (1965). Considerations of the physiology and variables that determine blood
carboxyhemoglobin concentration in man. J Clin Invest. 44:1899-1910.
Coburn RF, Blakemore WS, Forster RE. (1963). Endogenous carbon monoxide production in man. J Clin Invest.
42:1172-1178.
Collier C and Goldsmith JR. (1983). Interactions of carbon monoxide and hemoglobin at high altitude. Atmos
Environ. 17:723-728.
Coltman CA and Dudley III GM. 1969. The relationship between endogenous carbon monoxide and total heme
mass in normal and abnormal subjects. AmJMedSci. 258:374-385.
Delivoria-Papadoppulos M, Coburn RF, Forster RE. (1974). Cyclic variation of rate of carbon monoxide
production in normal women. JAppl Physiol. 36:49-71.
Friedlander SK. (1982). Letter from the Clean Air Scientific Advisory Committee to US EPA Administrator.
August 31, 1982.
Glen G. (2002). Programmer's Guide for the APEX3 Model. Prepared by ManTech Environmental Technology,
Inc., for the U.S. Environmental Protection Agency, Research Triangle Park, NC. April 30.
Issacs K and Smith L. (2005). New Values for Physiological Parameters for the Exposure Model Input File
Physiology.txt. Technical memorandum to Tom McCurdy, NERL WA10. December 20, 2005. Provided
in Appendix A of the CO REA.
Johnson T and Paul RA. (1983). The NAAQS Model (NEM) Applied to Carbon Monoxide. Report no. EPA
450/5-83-003. U.S. Environmental Protection Agency, Research Triangle Park, NC.
Johnson T. (1998). Memo No. 5: Equations for Converting Weight to Height Proposed for the 1998 Version of
pNEM/CO. Memorandum Submitted to U.S. Environmental Protection Agency. TRJ Environmental, Inc.,
713 Shadylawn Road, Chapel Hill, North Carolina 27514.
Johnson T, Capel J, Olaguer E, Wijnberg L. (1992). Estimation of Carbon Monoxide Exposures and Associated
Carboxyhemoglobin Levels in Denver Residents Using a Probabilistic Version of NEM. Prepared by TRJ
Environmental, Inc., for the Office of Air Quality Planning and Standards, U.S. Environmental Protection
Agency, Research Triangle Park, NC.
B-30
-------
Johnson T, Mihlan G, LaPointe J, et al. (2000). Estimation of Carbon Monoxide Exposures and Associated
Carboxyhemoglobin Levels in Residents of Denver and Los Angeles Using pNEM/CO (Version 2.1).
Prepared by TRJ Environmental, Inc., for the Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, Research Triangle Park, NC.
Luomanmaki K and Coburn RF. (1969). Effects of metabolism and distribution of carbon monoxide on blood and
body stores. AmJPhysiol. 217(2):354-362.
Lynch SR and Moede AL. (1972). Variation in the rate of endogenous carbon monoxide production in normal
human beings. JLabClinMed. 79:85-95.
Marcus AH. (1980). Mathematical models for carboxyhemoglobin. Atmos Environ. 14:841-844.
McCartney ML. (1990). Sensitivity analysis applied to Coburn-Forster-Kane models of carboxhemoglobin
rormation. Am IndHygAssoc J. 51(3):169-177.
McCurdy T, Glen G, Smith L, and Lakkadi Y. (2000). The National Exposure Research Laboratory's Consolidated
Human Activity Database. J Expos Anal and Environ Epidemiol. 10:566-578.
Merke C, Cavallin-Stahl E, Lundh B. (1975). Carbon monoxide production and reticulocyte count in normal
women: effect of contraceptive drugs and smoking. ActaMedScan. 198:155-160.
Muller KE and Barton CN. (1987). A nonlinear version of the Coburn, Forster and Kane model of blood
carboxyhemoglobin. Atmos Environ. 21:1963-1967.
Peterson JE and Stewart RD. (1970). Absorption and elimination of carbon monoxide by inactive young men.
Arch Environ Health. 21:165-171.
Peterson JE and Stewart RD. (1975). Predicting the carboxhemoglobin levels resulting from carbon monoxide
exposures. JApplPhysiol. 39:633-638.
Press WH, Flannery BP, Teukolsky SA, Vettering WT. (1986). Numerical Recipes: The Art of Scientific
Computing. 1st edition. Cambridge University Press, Cambridge MA.
Rodkey FL, O'Neal JD, Collison HA. (1969). Oxygen and carbon monoxide equilibria of human adult hemoglobin
at atmospheric and elevated pressure. Blood. 33:57-65.
Salorinne Y. (1976). Single-breath pulmonary diffusing capacity. ScandJResp Diseases. Supplementum 96.
Tikuisis P, Madill HD, Gill BJ, Lewis WF, Cox KM, Kane DM. (1987). A critical analysis of the use of the CFK
equation in predicting COHb formation. Am Ind Hyg Assoc J'. 48(3):208-213.
Tikuisis P, Kane DM, McLellan TM, Buick F, Fairburn SM. (1992). Rate of formation of carboxyhemoglobin in
exercising humans exposed to carbon monoxide. JAppl Physiol. 72(4): 1311-1319.
US EPA. (1978). Altitude as a Factor in Air Pollution. Report no. EPA-600/9-78-015. U.S. Environmental
Protection Agency, Research Triangle Park, NC.
US EPA. (2002). EPA's Consolidated Human Activities Database. Data and associated documentation available
at: http://www.epa.gov/chad/.
US EPA. (2008a). Total Risk Integrated Methodology (TRIM) Air Pollutants Exposure Model Documentation
(TRIM.Expo/APEX, Version 4.3). Volume 2: Technical Support Document. Report no. EPA-452/B-08-
00Ib. Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research
Triangle Park, NC. Available at: http://www.epa.gov/ttn/fera/human apex.html.
B-31
-------
US EPA. (2008b). Total Risk Integrated Methodology (TRIM) Air Pollutants Exposure Model Documentation
(TRIM.Expo/APEX, Version 4.3). Volume 1: Users Guide. Report no. EPA-452/B-08-00la. Office of
Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC.
Available at: http://www.epa.gov/ttn/fera/human apex.html.
Werner B and Lindahl J. (1980). Endogenous carbon monoxide production after bicycle exercise in healthy
subjects and in patients with hereditary spherocytosis. Scan J Lab Invest. 40:319-324.
B-32
-------
Appendix C
Isaacs et al. (2009) Reference Used in Developing D and A
Statistics Input to APEX Model
The following presents a reformatted version of the Isaacs et al. (2009) presentation to allow for
easier reading. The poster, included at the end of this Appendix in its entirety, was originally
presented at the American Time Use Research Conference, June 25-26, 2009, University of
Maryland, College Park, MD.
C-l
-------
Statistical Properties of Longitudinal Time-Activity Data for Use in EPA Exposure Models
Kristin Isaacs1, Thomas McCurdy2, April Errickson3, Susan Forbes3, Graham Glen1, Stephen
Graham4, Lisa McCurdy5, Melissa Nysewander1, Luther Smith1, Nicolle Tulve2, and Daniel
Vallero2
1 Alion Science and Technology, Research Triangle Park, NC, 2Human Exposure and
Atmospheric Sciences Division, National Exposure Research Laboratory, US Environmental
Protection Agency, Research Triangle Park, NC, 3School of Information and Library Science,
University of North Carolina at Chapel Hill, Chapel Hill, NC, 4Office of Air Quality Planning
and Standards, US Environmental Protection Agency, Research Triangle Park, NC,
5Homemaker, Durham, NC.
ABSTRACT
Realistic simulation of longitudinal activity patterns is necessary for appropriately
reproducing the frequency and duration of pollutant exposures in human exposure models. In
EPA's exposure models, longitudinal activity diaries for simulated persons are constructed from
the 1-day cross sectional activity diaries in the Consolidated Human Activity Database (CHAD).
Recently, new algorithms have been developed to construct longitudinal diaries from CHAD
diaries based on realistic variance and autocorrelation properties of diary characteristics relevant
to pollutant exposure. Characteristics of particular interest include time spent in particular
microenvironments and time spent in activities that produce high ventilation rates. However,
few multi-day data are currently available for estimating accurate statistical properties for these
quantities. Results from a recent time-activity study of 10 adults and one newborn child are
presented here. The participants recorded their personal location and activity for two-week
periods in each of four seasons in 2006 and 2007. The data were recorded 24 hours a day, in
increments as small as one minute. Additional recording periods for these same individuals are
expected in the future. The diaries for all subjects were assessed to calculate the between-person
variance, the within-person variance, and the autocorrelation for various lags in the time spent in
outdoor, residence, indoor (non-residence), and vehicle microenvironments, as well as for time
spent performing high-METS activities. The effectiveness of various day-type definitions (for
example, weekend versus weekday, or workday versus non-workday) for grouping similar diary
days is examined. Seasonal variation in activity patterns is analyzed. These data have the
potential to aid in the development of improved input variance and autocorrelation statistics for
longitudinal diary assembly algorithms in EPA's human exposure models.
INTRODUCTION
Recently, new methods of assembling multi-day diaries in human exposure models from
cross-sectional single-day diaries have been proposed that are based on the variance and
autocorrelation statistics of the simulated population (Glen et al. 2008). Appropriately modeling
intra- and interindividual variability using such algorithms may be essential in producing
appropriate estimates of exposure. In addition, reproducing realistic autocorrelations in key diary
properties may be required for the modeling of episodic exposure patterns.
Previously, longitudinal time activity-location data collected in children in the Southern
California Chronic Ozone Exposure Study (Geyh et al. 2000) have been analyzed to obtain
estimates of appropriate measures of variance and autocorrelation for use in the longitudinal
algorithm. Data from a new study in adults are now presented.
C-2
-------
BACKGROUND
Exposure models require construction of human activity diaries that cover the entire
simulation period of a model run. This period is often several months, a year, or even longer. In
EPA's models, human activity diaries are usually drawn from EPA's CHAD (Consolidated
Human Activity Database; McCurdy et al., 2000; http://www.epa.gov/chadnetl), which typically
includes just one day (24 hours) of activities from each person. A "longitudinal" diary is one
that covers the same person over a long period of time. While the SHEDS modeling period may
be of user-specified duration, it is assumed in this section to be one year, to provide a concrete
example.
Recently, a new longitudinal diary assembly algorithm has been developed (Glen et al.
2007) based on the variance and autocorrelation properties of the modeled simulation. The new
method requires the user to:
1) Select the diary property most relevant to exposure for the current application (such as
outdoor time or time spent in vehicles)
2) Specify the D statistic, which relates the within-person and between-person variances
for this diary property; and
3) Specify the 1-day lag autocorrelation in this diary property.
The new method is currently implemented in EPA's APEX and SHEDS-Air Toxics
models. The new method allows the modeler to apportion the total variance in the key diary
property into the within- and between-person variances ow2 and c\,2 by specifying the D statistic,
defined to be:
D pertains to the population as a whole and is bounded by zero and one. A value of zero implies
all persons have the same average behavior, whereas a value of one implies the greatest possible
difference in mean behavior that is consistent with the total variance.
In addition to targeting the within-person and between-person variances through setting
the D statistic, the new diary assembly method optionally allows targeting of the day-to-day
autocorrelation. This is a measure of the tendency for similar diaries to occur on consecutive
days. The lag-one autocorrelation in a variable y is for a person defined as:
N-1
The population autocorrelation A is the mean of the A values for all individuals.
Autocorrelation could be of interest to the exposure modeler if the concentration time series were
strongly episodic, for example. In the diary assembly, a positive autocorrelation indicates a
tendency for diaries with x-scores near each other to be used on consecutive days, while a
negative autocorrelation indicates a tendency for dissimilar x-scores to be used on consecutive
days. Some preliminary values of A have been derived from the same data that were used to
estimate D (Glen et al., 2008).
C-2
-------
METHODS
Activity Diary Study
Activity-location data were collected from 10 adults. Nine of the adults were working
professionals; one was a stay-at-home parent. Nine of the adults recorded their personal location
and activity for two-week periods in each of four seasons in 2006 and 2007. Additional data
were collected in one of the male subjects in 1999, another male (the 10th adult) in 2002, and in
one of the females in 2008 (collected during maternity leave). The data were recorded 24 hours
a day, in increments as small as one minute. In this preliminary analysis, the time spent
outdoors, indoors, in travel, and performing hard work each day were calculated from the diaries.
"Hard work" was self-reported by each individual, as defined as activities requiring heavy
breathing and/or sweating. Daily high temperatures and precipitation amounts were acquired for
each day of the study.
Variance and Autocorrelation Statistics
Variance and lag-one autocorrelation statistics were calculated for the studied
individuals. Variance statistics were estimated for both the raw measured variables (i.e. time in
minutes) and the scaled ranks of the variable for each person on a given day. The ratio of the
between-person variance to the total variance (the sum of the between- and within-person
variance) was calculated for the population. This ratio, calculated using the raw variables, is the
intraclass correlation coefficient (ICC), while the same ratio, calculated using the ranks, is D, the
diversity statistic. The autocorrelation A was also calculated using both the raw variables and
the scaled ranks of the variables on each day for each person in the study.
Analysis of Time Spent in Locations/Activities
The longitudinal data were assessed to support decisions on optimal diary pools for
exposure modeling. Time spent in each of the examined locations/activities were assessed as a
function of day of the week (weekday versus weekend), day type (workday versus non-
workday), season, temperature, precipitation, and gender. These analyses were undertaken to
assess the utility of different diary pool definitions. Optimal definitions of diary pools can
adequately capture temporal patterns in activities while maximizing the number of activity
diaries available for sampling on a given day for a simulated individual. Differences between
groups were assessed with the Wilcoxon signed rank test (for 2 groups) or the Kruskal-Wallis
test (for more than 2 groups). The Wilcoxon rank sum (two-sample) test was used to test
differences between genders.
RESULTS AND DISCUSSION
Individual Variability
Figure 1 shows an example of the individual variability in time spent in different
locations/activities for a single male subject; a 367-day period from this subject is depicted.
Distributions of time for this subject are also shown. These figures demonstrate the large amount
of intra-individual variability that can be seen in longitudinal activity studies. Distributions of
time spent in locations/activities for the population is shown in Figure 4[sic 2].
C-4
-------
Examples of Day-to-Day Variability For a Single Subject (M)
Time Spent Outdoors
Time Spent in Travel
100200300400500600 0 100200300400500600
Distribution Of Time Spent in travel lor an Individual CumuLriivv Distllbution of Time SpenEinTr.tvrllal .in Indlvidu.t
400
350
. 300
250
109 300 300 «0 500 800
Of Tim* Spent Doing Hard Woik lot M Individual
Time Spent Doing Hard Work
I i i
I
jjLuLLM
.1 l.Hu
300
275
IfTS
: 150
| 125
| 100
1 75
50
25
0
1W 200 200 400 500
mm
400
350
300
I 250
I 20°
I 150
100
50
0
100 200 100 400 500 630
Figure 1. Time series and distributions of time spent in locations/activities for 367 days of
data from a single male subject. Note high degree of interpersonal variability in behavior.
T!nwS[.!i* Doing H*il Weil
Time Spoil Moots
Tim Spein Dolns Hui Work
130
« 1000
3 MO
600
I 400
z 200
0
m
z 20
D
m 'cm tin im
i
Tine Sf «nl Ouldo
Tkn* Spent In Travel
Tkne tyenl Mdgni
Time Spent In Tuvel
200
•y.
ISC
I4t
12C
m
sc
sc
40
20
0
ISO
, 16°
J MO
o 120
= 100
I 90
' 60
I «
20
8
MB w JOB u IBB n
at MO tmyammMvomm
Figure 2. Distributions of time spent in different activities for all days for all subjects.
Variance and Autocorrelation Statistics
D, ICC, and A values for the population for time spent in different locations/activities are
given in Table 1. Values of the ICC are lower than D; while A for the raw variables were higher
than A for the scaled ranks. These trends were also consistent with observed tends in the
Southern California data. Values were also calculated by gender (Table 2), temperature
categories (Table 3), and day types (Table 4) where possible.
C-5
-------
The D and ranked A values were compared to those calculated for children from the
Southern California Chronic Ozone Exposure Study (SCCOES). The diversity (D) for this group
of adults for outdoor time were higher than those calculated for the children (0.38 versus 0.19).
The D values for travel time in the current study were also higher (0.18 in children versus 0.36 in
this study). These differences reflect the increased heterogeneity in these variables in the studied
adults versus the (relatively homogenous) studied children. The A values calculated for outdoor
time in this study were virtually identical to those estimated using data from SCCOES.
In general, differences between D by temperature and day types were notable, even considering
the small number subjects in this study. There were gender differences observed in D; the
mechanism of these differences is unclear, but is likely influenced by the activity patterns of the
female who was not a worker.
There were observed differences in A by temperature, but especially by day type. This is
not unexpected, as it is reasonable that the behavior of working adults is more consistent day-to-
day on workdays. These trends should be confirmed by analysis of other longitudinal data. Note
however, that such differences in are only important when strongly episodic behavior or
exposure is of interest. In general, the values of D are much more relevant to exposure.
Table 1. Variance and Autocorrelation Statistics: All Days/Subjects
Location/ Activity
Indoors
Outdoors
Travel
Hard Work
ICC
0.26
0.16
0.14
0.18
D
0.33
0.38
0.31
0.22
A (Raw)
0.23
0.22
0.12
0.17
A (Ranks)
0.34
0.31
0.19
0.19
Table 2. Variance and Autocorrelation Statistics: By Gender
Location/Activity
Males
Indoors
Outdoors
Travel
Hard Work
Females
Indoors
Outdoors
Travel
Hard Work
ICC
0.36
0.14
0.36
-0.01
0.08
0.07
0.05
0.15
D
0.54
0.22
0.46
0.15
0.09
0.27
0.16
0.24
A (Raw)
0.25
0.24
0.17
0.22
0.37
0.35
0.15
0.16
A (Ranks)
0.16
0.22
0.08
0.20
0.25
0.18
0.11
0.21
C-6
-------
Table 3. Variance and Autocorrelation Statistics: By Temperature
Location/Activity
Days with max temp
less than 50 degrees
Indoors
Outdoors
Travel
Hard Work
Days with max temp
greater or equal to 50 degrees
Indoors
Outdoors
Travel
Hard Work
ICC
0,37
0,20
0,23
0,21
0.12
0,09
0,10
0.01
D
0,37
0,27
0,37
0.31
0,26
0.24
0.24
0.20
A (Raw)
0.23
0.33
0.20
0.14
0.45
0.39
0.34
0.35
A (Ranks)
0.19
0.18
0.09
0.14
0.23
0.20
0.09
0.14
Table 4. Variance and Autocorrelation Statistics: By Daytype
Location/Activity
Workday
Indoors
Outdoors
Travel
Hard Work
NonWorkday
Indoors
Outdoors
Travel
Hard Work
ICC
0.37
0.19
0.45
0.20
0.12
0.11
0.09
0.06
D
0.47
0.31
0.47
0.25
0.21
0.14
0.24
0.07
A (Raw)
0.56
0.78
0.30
0.53
0.59
0.60
0.38
0.43
A (Ranks)
0.05
0.07
0.01
-0.12
0.24
0.19
0.08
0.18
Time Spent in Different Locations/Activities
The time spent in different locations/activities for different day types, seasons,
temperature categories are presented in Figures 3-6. The effects of gender and precipitation
were also studied. There were no significant differences for these categories, and thus plots are
not shown. The plotted data represent all days for all subjects. The medians are represented by
the midline of the boxes, the first and third quartiles by the ends of the boxes, and the means by
the stars. The whiskers extend to cover data that lies beyond the boxed but within the quartiles
plus 1.5 times the interquartile range. Points outside this range are plotted.
Results by day of the week and day type are presented in Figure 3. Day type (workday
versus non-workday) was at least as good as day of the week in categorizing time/activities. This
trend is similar to that seen in a recent analysis of the larger, cross-sectional database of diaries
from The National Human Activity Pattern Survey (NHAPS, data not shown). That analysis
indicated that a workday/non-workday was a better discriminator of time spent outside than a
weekday/weekend split. As such, further comparisons are also presented for both workdays and
non-workdays.
C-7
-------
Time Spent Outdoors By Cay of ine Week
Time Spent in Travel B, Day of the Week
Time Spent Doing Hard Work By Day of the We-
Time Spent Outdoors By Day Type
Time Spent in Travel By Day Type
Time Spent Doing Hard Work By Day Type
Figure 3. Time spent in different locations/activities as a function of day of the week, and
daytype (workday versus non-workdays).
The effect of season on time spent in locations/activities is shown in Figure 4. Seasonal
effects were apparent for time spent outdoors on non-workdays, and for time spent doing hard
work. Travel was also affected by season, likely due to the large number of work-related travel
days in the fall for this particular group of workers.
Tim* Spent Outdoors By Season
Time Spent in Travel By Season
Time Sp«m in Doing Hard Wort By Season
Time Spent Outdoors By Season and Day Type
Time Spent in Tra^l By Season and Day Type
Time Spent Doing Hard Work By Season and Day Type
IIS
|M137 | 17fl | 27* \ 140 I 1O2 | 14B | 202 | 105 j
46 302 105 IN 157 178 275 I 140 102 149 I 202 I 105
-*f°a Ssotrnt Vtmt C* Seneg
Figure 4. Time spent in different locations/activities as a function of season and daytype.
-------
The effect of temperature category is shown in Figure 5. The temperature category was
defined as warmer = maximum temperature greater than or equal to 75 degrees, colder=
maximum temperature less than 75 degrees. Temperature category was better than or as good as
season in discriminating behavior in time spent outdoors, even when day type was considered.
Time Spent Outdoors By Tempeiature Category
Time Spent in Travel By Temperature Categoiy
Time Spent Doing Hard Work By Temperature Category
Time Spent Outdoors By Temperature Category and Day Type
Time Spent in Travel By Temperature Category and Day Type Time Spent Doing Hard Work By Temperature Category and Day Type
Figure 5. Time spent in different locations/activities as a function of temperature category
(colder: max temp< 75 degrees, warmer: max temp> 75 degrees) and day type.
CONCLUSIONS
• The diversity (D) and autocorrelation (A) for this group of adults for outdoor time were
higher than those calculated for children in a previous study. Thus these data provide
some justification for considering age when considering D and A input values for EPA's
exposure models.
• While the current data suggest possible effects of temperature, day type and gender on
diversity (D) and autocorrelation (A), more data from this and other studies are needed to
confirm these findings. Such results could aid in the fine-tuning of the longitudinal diary
algorithm.
• The analysis of the time spent in locations was consistent with recent findings from cross-
sectional diary studies indicating that workdays/non-workdays may be a better grouping
for diary pools than weekdays/weekends.
Temperature category was at least as good as season in discriminating behavior for this
population for time spent outdoors, especially when day type was considered. Such
breakdowns by temperature and day type may eliminate the need for diary pools for
different seasons, providing larger pools for diary sampling on a given day. Further
analysis with other time-activity data can confirm this trend.
C-9
-------
FUTURE WORK
We plan to repeat this type of study periodically. Data will be compared to/combined
with analyses of other available longitudinal time/location/activity studies.
DISCLAIMER
The information in this document has been funded wholly (or in part) by the U. S.
Environmental Protection Agency (EPA contract 68-D-00-206). It has been subjected to review
by the EPA and approved for publication. Approval does not signify that the contents
necessarily reflect the views of the Agency, nor does mention of trade names or commercial
products constitute endorsement or recommendation for use.
REFERENCES
Geyh AS, Xue J, Ozkaynak H, Spengler JD. (2000). The Harvard Southern California Chronic
Ozone Exposure study: assessing ozone exposure of grade school-age children in two southern
California communities. Environ Health Perspect. 108:265-270.
Glen G, Smith L, Isaacs K, McCurdy T, Langstaff J. (2008). A new method of longitudinal
diary assembly for human exposure modeling. J Expo Sci Environ Epidemiol. 18(3):299-311.
McCurdy T, Glen G, Smith L, Lakkadi L. (2000). The National Exposure Research
Laboratory's Consolidated Human Activity Database. J Expo Anal Environ Epidemiol. 10:566-
78.
C-10
-------
Isaacs et al. (2009) in original poster format:
Statistical Properties of Longitudinal Time-Activity Data for Use in EPA Exposure Models
Kristin Isaacs1, Thomas McCurdy2, April Errickson3, Susan Forbes3, Graham Glen1, Stephen Graham4, Lisa McCurdy5, Melissa Nysewander1, Luther Smith1, Nicolle Tulve2, and Daniel Vallero2
1Alion Science and Technology, Research Triangle Park, NC, 2Human Exposure and Atmospheric Sciences Division, National Exposure Research Laboratory, US Environmental Protection Agency, Research Triangle Park, NC, 3School of Information and
Library Science, University of North Carolina at Chapel Hill, Chapel Hill, NC, 4Office of Air Quality Planning and Standards, US Environmental Protection Agency, Research Triangle Park, NC, 5Homemaker, Durham, NC.
Activity Diary Study
category (colder: maxtemp<75 degrees, warmer: max tempi 75 degrees) and
daytype.
C-ll
-------
Appendix D
Microenvironmental Mapping
Figure D-l presents how CHAD codes are mapped to the eight microenvironments used to model exposure
in the CO REA. Table D-l provides the CHAD activity codes used to identify when a simulated individual
was in a work air district.
Figure D-l. Microenvironmental Mapping Input File Showing Mapping of CFIAD Location Codes
to the Eight Microenvironments for Application of APEX4.3 to Carbon Monoxide.
! Mapping of CHAD location codes to nine APEX
! by Option 4 of Memorandum dated 12/8/2009.
CHAD Loc. Description
U
X
30000
30010
30020
30100
30120
30121
30122
30123
30124
30125
30126
30127
30128
30129
30130
30131
30132
30133
30134
30135
30136
30137
30138
30139
30200
30210
30211
30219
30220
30221
30229
Uncertain of correct code
No data
Residence, general
Your residence
Other residence
Residence, indoor
Your residence, indoor
. . . , kitchen
. . . , living room or family room
. . . , dining room
. . . , bathroom
. . . , bedroom
. . . , study or office
. . . , basement
. . . , utility or laundry room
. . . , other indoor
Other residence, indoor
. . . , kitchen
. . . , living room or family room
. . . , dining room
. . . , bathroom
. . . , bedroom
. . . , study or office
. . . , basement
. . . , utility or laundry room
. . . , other indoor
Residence, outdoor
Your residence, outdoor
. . . , pool or spa
. . . , other outdoor
Other residence, outdoor
. . . , pool or spa
. . . , other outdoor
microenvironments defined
APEX
= -1
= -1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
7
7
7
7
7
7
7
U
U
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
D-l
-------
30300
30310
30320
30330
30331
30332
30340
30341
30342
30400
31000
31100
31110
31120
31121
31122
31130
31140
31150
31160
31170
31171
31172
31200
31210
31220
31230
31300
31310
31320
31900
31910
32000
32100
32200
32300
32400
32500
32510
32520
32600
32610
32620
32700
32800
32810
32820
32900
32910
32920
33100
Residential garage or carport =
. . . , indoor =
. . . , outdoor =
Your garage or carport =
. . . , indoor =
. . . , outdoor =
Other residential garage or carport =
. . . , indoor =
. . . , outdoor =
Residence, none of the above =
Travel, general =
Motorized travel =
Car
Truck
Truck (pickup or van) =
Truck (not pickup or van) =
Motorcycle or moped =
Bus
Train or subway =
Airplane =
Boat
Boat, motorized =
Boat, other =
Non-motorized travel =
Walk
Bicycle or inline skates/skateboard =
In stroller or carried by adult =
Waiting for travel =
. . . , bus or train stop =
. . . , indoors =
Travel, other =
. . . , other vehicle =
Non-residence indoor, general =
Office building/ bank/ post office =
Industrial/ factory/ warehouse =
Grocery store/ convenience store =
Shopping mall/ non-grocery store =
Bar/ night club/ bowling alley =
Bar or night club =
Bowling alley =
Repair shop =
Auto repair shop/ gas station =
Other repair shop =
Indoor gym /health club =
Childcare facility =
. . . , house =
. . . , commercial =
Large public building =
Auditorium/ arena/ concert hall =
Library/ courtroom/ museum/ theater =
Laundromat =
1
1
7
1
1
7
1
1
7
1
8
8
8
8
8
8
5
8
8
0
7
7
7
7
7
7
7
7
5
4
8
8
3
3
4
3
3
3
3
3
3
2
3
3
4
1
4
3
3
3
3
H
H
H
H
H
H
H
H
H
H
O
O
O
O
O
O
O
O
O
O
O
0
0
0
0
0
0
0
0
0
0
0
0
0
O
H
O
O
O
O
O
O
O
O
O
O
0
0
0
0
H
D-2
-------
33200
33300
33400
33500
33600
33700
33800
33900
34100
34200
34300
35000
35100
35110
35200
35210
35220
35300
35400
35500
35600
35610
35620
35700
35800
35810
35820
35900
36100
36200
36300
Hospital/ medical care facility =
Barber/ hair dresser/ beauty parlor =
Indoors, moving among locations =
School
Restaurant =
Church
Hotel/ motel
Dry cleaners =
Indoor parking garage =
Laboratory =
Indoor, none of the above =
Non-residence outdoor, general =
Sidewalk, street =
Within 10 yards of street =
Outdoor public parking lot /garage =
. . . , public garage =
. . . , parking lot =
Service station/ gas station =
Construction site =
Amusement park =
Playground =
. . . , school grounds =
. . . , public or park =
Stadium or amphitheater =
Park/ golf course =
Park
Golf course =
Pool/ river/ lake =
Outdoor restaurant/ picnic =
Farm =
Outdoor, none of the above =
4
3
3
4
3
4
3
3
6
3
3
7
5
5
6
6
6
2
7
7
7
7
7
7
7
7
7
7
7
7
7
0
H
0
0
0
H
0
H
O
O
O
O
O
O
O
O
O
O
O
O
H
0
H
0
0
0
0
0
0
0
0
Table D-l. CHAD Work Related Activity Codes Used To identify Work Air Districts.
<10> Work and Other Income Producing Activities
10000: work and other income producing activities, general
10100: work, general
10110: work, general, for organizational activities
10111: work for professional/union organizations
10112: work for special interest identity organizations
10113: work for political party and civic participation
10114: work for volunteer/ helping organizations
10115: work of/ for religious groups
10116: work for fraternal organizations
10117: work for child/ youth/ family organizations
10118: work for other organizations
10120: work, income-related only
10130: work, secondary (income-related)
10200: unemployment
10300: breaks
D-3
-------
Appendix E
Analysis of CHAD Diaries for Time Spent in Vehicles.
The US Census Bureau (2009) provides an on-line facility for accessing the detailed
census data included in their Summary File 3 (SF3). We obtained information on travel time to
work for workers ages 16 years and over specific to Denver County, Colorado and Los Angeles,
CA (US Census Bureau, 2009, Table P31). Staff converted the counts listed in Table P31 for
trips to work places other than home into the percentages listed in Columns 2 and 3 of Table E-l.
Although the P31 statistics apply to people 16 years or older, staff assumed that the statistics
were generally applicable to people 18 years or older.
We next determined the number of 24-hour diaries in EPA's Consolidated Human
Activity Database (CHAD) (US EPA, 2002) that met the following criteria: the subject was >18
years of age and the diary reported at least one minute in a motor vehicle between 6 am and 9
am. The number of these diaries that had in-vehicle times corresponding to the bins listed in
Table E-l are given in Column 4 and were converted to the percentages listed in Column 5.
Table E-l. Comparison of Denver and LA commuting characteristics (US Census, 2009)
to time spent in motor vehicles using CHAD Diaries (US EPA, 2002).
Travel time
(minutes)
(1)
1 to 9
10to19
20 to 29
30 to 39
40 to 59
60 to 89
90+
Total
Percent of commuters
according to SF3
census data for Denver
County
(2)
10.3
32.0
24.2
18.6
9.3
3.8
1.7
100
Percent of commuters
according to SF3
census data for Los
Angeles County
(3)
7.8
25.9
21.0
21.4
13.6
7.0
3.4
100
24-hour diaries meeting
inclusion criteria3
Number in
CHAD
(4)
563
1,676
1,068
1,111
665
407
258
5,748
Percent in
CHAD
(5)
9.79
29.16
18.58
19.33
11.57
7.08
4.49
100
Notes:
a Subjects are 1 8+ years of age. Diaries are those having >one minute in motor vehicle time spent
between 6 AM and 9 AM.
E-l
-------
References
US Census Bureau. (2009). American Fact Finder. Census Summary File 3 (SF3) - custom tables. Available at:
www.factfinder.census.gov.
US EPA. (2002). EPA's Consolidated Human Activities Database. Available at: http://www.epa.gov/chad/.
E-2
-------
Appendix F
Differences in Human Activity Patterns Between Individuals
With and Without Cardiovascular Disease
The following presents a memorandum by Cohen et al. (1999) that was included in
the Johnson et al. (2000) CO exposure assessment (see Appendix J of that report).
It is in its original form, with some minor editing performed by staff for inclusion
into the CO REA.
F-l
-------
MEMORANDUM
TO: Harvey Richmond
FROM: Jonathan Cohen, Sergey Nikiforov, and Arlene Rosenbaum
DATE: January 15, 1999
SUBJECT: EPA 68-DO-0062 Work Assignment 2-24: Task 2: Evaluation of
Differences in Human Activity Patterns Between Individuals With or
Without Cardiovascular Disease
EVALUATION OF DIFFERENCES IN HUMAN ACTIVITY PATTERNS BETWEEN
INDIVIDUALS WITH OR WITHOUT CARDIOVASCULAR DISEASE
SUMMARY
Activity pattern data from the National Human Activity Pattern Survey were used to compare
activity patterns and exertion distributions between subjects with or without angina. The diary
survey provided a 24-hour diary of activities. Exertion rates for each person in the survey were
simulated 100 times. For each person, the body weight was simulated from a log-normal
distribution specific to the age and gender. The resting metabolic rate was simulated using a
regression against body weight, with coefficients depending on age and gender. Finally, the
exertion rate was simulated for each activity and person by multiplying the simulated resting
metabolic rate by a MET exertion ratio with a distribution specific to each type of activity. The
current version of the probabilistic NAAQS Exposure Model for Carbon Monoxide (pNEM/CO),
described in Johnson (1998), begins with the same set of physiological equations and statistical
distributions for probabilistic simulation of exposure. The pNEM/CO model uses the much
broader Consolidated Human Activity Data Base (CHAD) and simulates additional
physiological variables, such as the ventilation rate. The description of the relevant probabilistic
and physiological equations in this memorandum is largely based on Johnson (1998); see that
memorandum for more detailed information.
Differences between angina and non-angina subjects were evaluated for several summary
statistics: average and 95th percentile of the maximum daily 8-hour exertion, percentage of time
spent outdoors or in a vehicle, average percentage of time at light, moderate or heavy exertion
F-2
-------
levels. Age and gender have very significant effects on these summary statistics of activity and
exertion. Since angina patients tend to be much older and tend to include more females than the
general population, it is very important to adjust for age and gender effects when comparing
angina and non-angina groups. Otherwise, one cannot distinguish between the angina effect and
the effects of age and gender. Statistical analyses comparing angina to non-angina subjects were
performed, adjusting for age and gender either by stratification (comparing subjects in a given
age/gender subgroup), or by fitting a general linear model (with separate terms for age, gender,
and angina effects and their interactions). These analyses showed that, overall, angina subjects
tended to have less extreme exertion levels. More specifically, the maximum 8-hour exertion
energies tended to be lower, as did the percentages of time above moderate or high exertion rate
thresholds. The percentages of time spent outdoors or in a vehicle were generally not statistically
significantly different between angina and non-angina subjects.
The large sample of NHAPS subjects produced, in many cases, statistically significant
differences in the exertion rate summaries between angina and non-angina subjects. However,
those differences were generally numerically small compared to the mean values. Therefore we
conclude that the differences in activity and exertion between angina and non-angina subjects,
although statistically significant, are not large enough to severely impact the validity of
pNEM/CO modeling results that do not adjust for an angina/non-angina difference.
METHODOLOGY
For these analyses we used the National Human Activity Pattern Survey (NHAPS) database, a
telephone survey of human activity patterns conducted for the USEPA between October 1992
and September 1994 by the Survey Research Center at the University of Maryland. See Klepeis
et al. (1996, 1998) and Tsang and Klepeis (1996) for more details about the NHAPS study and
various statistical analyses of those data. The NHAPS data (Triplett, 1996) are included in
CHAD. (Other CHAD studies did not include questions about cardiovascular disease and so
could not be used for these analyses comparing angina and non-angina respondents.) A
nationally representative sample of 9,386 respondents completed a detailed diary listing all their
activities and locations over a 24-hour period (either from the previous day or a previous
weekend day). A few respondents did not state their age and/or gender and their data was not
used in our analysis. Our analysis used 9,149 of the surveys. Respondents were also asked
demographic questions, including age and gender, and health questions, including whether or not
they have been told by a doctor that they have angina: 243 respondents (2.6 percent) had angina.
Respondents were asked about employment status (e.g. full-time, part-time, or unemployed) but
not about their occupation. Other follow-up questions (not used in our analyses) related to the
respondent's exposure to either water or air pollution on the diary day. For each household, the
respondent was randomly selected to be either the adult or child (under 18) with the next
birthday; an adult provided proxy responses for a child.
The EPA report (Klepeis, Tsang and Behar, 1996), Section 3, shows that the sample is
reasonably representative of the national population with respect to gender and age distributions.
The NHAPS population slightly underrepresented males (46 % NHAPS compared to 49 % from
the 1990 Census). The fraction of weekend (Saturday or Sunday) respondents was 33 %, close to
F-3
-------
the desired ratio of 2:7, but Thursdays, Fridays and Saturdays were underrepresented. The Fall
season was significantly underrepresented. The database includes weights to adjust for varying
selection probabilities, due to differences in the numbers of adults or children in a selected
household, the numbers of non-business phones in a household, the numbers of non-business
telephones in each census region, and to the survey stratification between weekend or weekdays
and between children and adults. Based on discussions with the EPA WAM, it was decided that
the weights would not be used in these analyses; the raw, unweighted data would be treated as an
approximately simple random sample. Note that the statistical weights: 1) were not used in the
pNEM/CO exposure modeling effort, 2) could not be used to accurately estimate standard errors
of weighted means, and 3) were close to 1 for most respondents.
In pNEM/CO, each activity is assigned a probability distribution of the exertion rate (kilo-
calories per minute). For this analysis, the 24-hour sequence of exertion rates was simulated 100
times for each person in the NHAPS sample; the sequence of activities is fixed but the simulated
exertion rates vary. Following both CHAD and the exposure modeling methodology currently
used in pNEM/CO, a constant simulated exertion rate is assumed throughout the time period of
each listed activity in the 24-hour diary. If the individual repeats the same activity at a later time,
with other activities intervening, the exertion rate is simulated again. SAS statistical software
was used for the simulations and for the statistical analysis.
The assigned exertion rate distribution depends upon the type of activity, and the occupation,
age, gender, and body weight of the respondent. The exertion rate (kilo-calories/minute =
kcal/min), also referred to as average energy expenditure rate, EE, is defined as the product
EE = MET x RMR.
MET is the metabolic equivalent of work, a dimensionless ratio (i.e., exertion compared to the
resting metabolic rate) specific to each activity, and, in some cases, to an age group. RMR is the
resting metabolic rate (kcal/min), approximately equal to the basal metabolic rate. We used the
same set of MET statistical distributions supplied by Tom McCurdy that are currently used in
pNEM/CO (and CHAD). For the work activity "at main job," the MET distribution depends on
the occupation. Since occupation was not recorded in NHAPS, we followed the pNEM
methodology and randomly selected the occupation based on census fractions of persons in each
activity. The same occupation is assumed throughout a simulated person-day (in case the person
repeats the work activity), but is randomly selected again for the next simulated person-day.
Note that this procedure may bias the comparison between angina and non-angina subjects, since
the distribution of occupation is expected to differ between angina subjects and the general
population.
A single RMR value was simulated to represent each person-day. Thus the same person would
have 100 simulated RMRs, one for each of the 100 days simulated. This reflects the assumption
that each person represents the activity pattern for a group of persons with the same age and
gender. As in pNEM/CO, RMR was simulated from a normal distribution where the mean is of
the form a + b (Body Mass), and the standard deviation is the constant _. The values of a, b, and
_ are the values derived by Schofield (1985) for 12 age/gender combinations (this assumes basal
F-4
-------
metabolic rate is equivalent to resting metabolic rate). In turn, the body mass was simulated
using the log-normal distributions estimated by Brainard and Burmaster (1992) and Burmaster
and Crouch (1994). The parameters of the log-normal distributions depend on age and gender.
The statistical analysis used the following summary statistics of the activity and simulated
exertion patterns for each person in the NHAPS study. The selection of these summary statistics
was based on recommendations from the EPA WAM:
Average maximum 8-hour energy expenditure. For each 8-hour period in a simulated person-
day, starting every 10 minutes, integrate the simulated EE to give the energy expenditure in
Meal (millions of calories), i.e. sum the products of activity time and energy expenditure
rate. For each simulated day, compute the maximum 8-hour energy expenditure, treating the
simulated day in circular fashion so that the respondent is assumed to repeat exactly the same
activity and exertion rate patterns on the day after the diary day. For example, the simulated
activities for the period starting at 10 pm are assumed to follow the reported sequence of
activities for the diary day from 10 pm to midnight and then the reported sequence from the
beginning of the diary day until 6 am. To represent a typical value for the selected person,
compute the average maximum 8-hour energy expenditure across the 100 simulations.
95th percentile maximum 8-hour energy expenditure. As in the last bullet, compute the
maximum 8-hour energy expenditure for each simulated day. To represent an extreme value
for the selected person, compute the fifth highest maximum 8-hour energy expenditure
among the 100 simulations.
_ Percentage time spent outdoors. This number is the same for all simulations, since the
activity patterns are held constant.
_ Percentage time spent in a vehicle. This number is the same for all simulations, since the
activity patterns are held constant.
_ Percentage time spent outdoors or in a vehicle. This number is the same for all simulations,
since the activity patterns are held constant.
_ Average percentage time with exertion rate above 2.39 kcal/min. For each simulated person-
day, the percentage of that day with an EE (rate) above the threshold level of 2.39 kcal/min
was computed; then, this percentage was averaged over the 100 simulations for that person.
The statistic estimates the percentage time spent at or above the threshold exertion rate level
over a long period, assuming the daily activity pattern was the same every day. The threshold
of 2.39 kcal/min, which equals 0.010 MJ/min, represents "light" exertion (see below).
_ Average percentage time with exertion rate above 5.97 kcal/min. The threshold of 5.97
kcal/min, which equals 0.025 MJ/min, represents "moderate" exertion (see below).
_ Average percentage time with exertion rate above 9.55 kcal/min. The threshold of 9.55
kcal/min, which equals 0.040 MJ/min, represents "heavy" exertion (see below).
F-5
-------
The exertion rate thresholds used for these analysis were originally defined as 0.010, 0.025, and
0.040 mega-joules per minute, but were converted into the more commonly used calorie units (1
joule equals 0.2388 calories). For purposes of exposure assessment, exertion categories (i.e.,
light, moderate, or heavy exertion) are more usefully defined by the ventilation rate VE (liters air
per minute) rather than the energy expenditure rate EE (kilo-calories per minute). For the EPA's
Ozone Criteria Document, the Environmental Criteria and Assessment Office categorized VE
into ranges of 0-23, 24-43, 44-63, and 64+ liters of air per minute to define light, moderate,
heavy, and very heavy exertion, respectively (based on a reference male adult with body weight
70 kg). To convert from EE to VE, EE is first multiplied by an energy conversion factor, ECF, to
give the oxygen uptake rate VO2 (liters of oxygen per minute). ECF varies across the
population, but is approximately 0.2 liters oxygen per kcal (Esmail, Bhambhani, and Brintnell,
1995). The "ventilatory equivalent rate" (VER) is the dimensionless ratio of VE (liters per
minute) divided by VO2 (liters per minute) and has typical values from about 24 for light
exertion to about 32 for peak exertion. Thus the selected energy expenditure rates are
approximately equivalent to the following ventilation rates:
EE = 0.010 MJ/min = 2.39 kcal/min:
VE = EE _ ECF _ VER = 2.39 _ 0.2 _ 24 = 11.5 liters/min = light exertion
EE = 0.025 MJ/min = 5.97 kcal/min:
VE = EE _ ECF _ VER = 5.97 _ 0.2 _ 28 = 33.4 liters/min = moderate exertion
EE = 0.040 MJ/min = 9.55 kcal/min:
VE = EE _ ECF _ VER = 9.55 _ 0.2 _ 32 = 61.1 liters/min = heavy exertion
The selected summary statistics were computed for each of the 243 angina subjects and 8,906
non-angina subjects in the NHAPS study. A statistical analysis compared the distributions of
these summary statistics for persons with and without angina. For each summary statistic we
compared the mean values between the angina and non-angina groups using standard t tests. The
significance level (p-value) for the difference in means was computed using the Smith-
Satterthwaite procedure, that tests for no difference in population means assuming that the two
populations are normally distributed but may have different variances. P-values at or below 0.05
denote significant differences at the five percent level of significance. By the central limit
theorem, the p-values for the t test comparisons should be reasonably accurate for the large
samples used in the overall analyses, even if the normality assumption does not hold, but the p-
values will be less accurate for the analyses of specific gender and age subgroups. We also
compared variances using a standard F test, that assumes normality of the two populations.
Since the normality assumption may not be a sufficiently good approximation, we also applied
two non-parametric tests that do not require specific parametric distributions. The non-
parametric Wilcoxon test, also known as the Mann-Whitney-Wilcoxon test or the Rank Sum
Test, was used to compare the central tendencies of the two distributions. This test assumes only
that the populations have the same distributional shape, which may or may not be the normal
distribution, but the distribution of values for angina population might be shifted by some
F-6
-------
constant value, and thus might have a different median than the non-angina population. The
Kolmogorov-Smirnov test was used to evaluate any possible differences between the two
distributions, whether due to differences in means, medians, variances, or any other features of
the distribution. This test uses the maximum absolute difference between the two cumulative
distribution functions, assuming only that these distributions are continuous.
The mean, variance, median, and distribution function comparisons were made for all persons
combined, separately for males and females, and then separately for four age groups within the
male and female subgroups. Age groupings were chosen to include approximately 25 percent of
angina subjects in each group. Separate comparisons for males and females are needed to
distinguish whether any overall differences in exertion or activity are explained by the fact that
angina subjects are more likely to be female than in the general population. Since activity
patterns and exertion rates differ between males and females, any overall difference between the
angina and non-angina groups might be explained by the greater propensity for females to get
angina, rather than the direct effect of angina. Similarly, the subsetting by age group evaluates
the effect of the different age distributions for angina subjects compared to the general
population (angina subjects tend to be much older). This statistical analysis does not, and cannot,
address questions as to whether the angina causes the change in exertion or activity patterns, or
vice versa. We only examine whether or not the summary statistics of activity and exertion
patterns are different for the two populations.
A general linear model approach was also used as an alternative method of adjusting for the
effects of age and gender on the angina/non-angina comparison. We focused attention on a
relatively simple statistical model with cubic terms in age (a simple linear function of age fitted
poorly), gender, interactions between age and gender, and a single term for the effect of angina:
Summary Statistic = I(male){_ + _(age) + _(age)2 + _(age)3}
+ I(female){_ + _(a|
+ _I(angina) + error
I(female){_ + _(age) + _(age)2 + _(age)3}
where: I(male) = 1 for males, 0 for females; I(female) = 1 for females, 0 for males; I(angina) = 1
for persons having angina, 0 for persons not having angina. The errors are assumed to be
normally distributed, statistically independent, and have mean zero and some constant variance.
This statistical model assumes that the expected value of the summary statistic is a cubic
function of age, but is a different function for males and females. The selected model has the
same coefficient for the cubic term for males and females, but different coefficients for the
intercept, linear, and quadratic effects. The model also assumes that having angina changes the
mean by a constant amount, which is the same factor for all age groups and both genders. A
more sophisticated model might allow for interactions between angina and the age and gender
variables, to allow for the possibility that the angina effect varies by gender and/or age. Note,
however, that our statistical analysis clearly showed that age and gender were much more
significant predictors of exertion patterns than the angina indicator, explaining most of the
variability in the summary statistics.
F-7
-------
Project resources were insufficient for a detailed exploration of alternative statistical models. We
tried using logarithmic transformations to improve the model fit, but could not reasonably use
such models in view of the large number of cases where the observed summary statistic was zero
(the logarithm is then undefined). The model fit for the selected model (without taking
logarithms) varied with the summary statistic. R squared goodness-of-fit statistics were
extremely low, less than 0.05, for the percentages of time spent outdoors and/or in a vehicle. For
the summary statistics based on the maximum 8-hour exertion and the percentages of time above
exertion rate thresholds, the R squared statistics ranged from a poor fit, 0.25, to a fairly good fit,
0.48. The cases of poor fitting models may be because the selected statistical models poorly
represent the relationship between age, gender, and angina and the activity/exertion summary
statistic and/or because the activity/exertion pattern varies substantially between people of the
same age, gender, and angina status.
RESULTS
Age, Gender, and Angina Disease Distributions
Table 1 shows the number of subjects with or without angina by gender and by age group. The
four age groups were chosen to have approximately the same numbers of angina subjects. The
strong association between angina and age is illustrated by the fact that 52/243 =21 % of angina
subjects are under 55 but 6877/8906 = 77 % of non-angina subjects are under 55. Angina
subjects tend to be significantly older than the general population. The association between
angina and gender is weaker. 103/243 = 42.3 % of angina subjects are male, but 4116/8906 =
46.2 % of non-angina subjects are male.
Overall Comparisons of Activity and Exertion Summary Statistics between Angina and
Non-Angina Subjects
Table 2 compares the means between the angina and non-angina subjects, without stratification
by age or gender. The average and 95th percentile of the maximum eight hour exertion has a
statistically significantly lower mean for angina subjects. Furthermore, for each of the exertion
levels 2.39, 5.97, and 9.55 kcal/min (0.010, 0.025, and 0.040 MJ/min), the mean percentage of
time above each level was statistically significantly lower for the angina subjects. Non-angina
subjects spend an average of 2.8 percent of their time doing activities requiring moderate or
higher levels of exertion, defined by exertion rates above 5.97 kcal/min (0.025 MJ/min); angina
subjects spend an average of 2.2 percent of their time doing such activities. All subjects spend
over 75 percent of time in light or sedentary activities, with extertion rates below 2.39 kcal/min,
including sleeping. All these exertion distribution comparisons show that angina subjects tend to
do activities with less exertion than the general population. However, since the summary
analyses in Table 2 do not take into account the marked differences between the age and gender
distributions of angina and non-angina subjects, the lower exertion rates could be associated with
the tendency for angina subjects to be older (and female) rather than the disease itself. The
average percentages of time spent outdoors are nearly identical, and are not statistically
significantly different between angina and non-angina subjects, but angina subjects spend
statistically significantly less time in vehicles (4.5 % rather than 5.5 %, on average).
F-8
-------
Table 2 compares the standard deviations using a F-test based on the variance ratio for angina vs.
non-angina subjects. In most cases the F tests show statistically significantly different variances
(and, therefore, standard deviations).
Table 2 also uses the non-parametric Wilcoxon test to compare the central tendencies of the two
distributions without the normality assumption required by the T test. Corresponding to the T
test comparisons, the Wilcoxon test finds that the angina and non-angina distributions are
significantly different in almost all cases; the angina subjects have a lower median value for each
of the selected summary statistics. Exceptions are for the average maximum 8-hour exertion, just
significant at the 7 % level, and the percentage of time spent outdoors, which has a non-
significant p-value of 22 %.
Finally, Table 2 compares the distribution functions using the Kolmogorov-Smironov test. The
distributions are statistically significantly different at the five and one percent levels in all cases
except for the percentage of time spent outdoors, which shows no significant difference. For that
variable, the T and Wilcoxon tests showed no statistically significant differences in central
tendency although the F test showed a statistically significant difference in the population
variances. If the population variances are different, so are the two distribution functions. The
discrepancy between the F and Kolmogorov-Smirnov tests is partly explained by the fact that the
F test is very sensitive to the assumption of normal distributions, whereas the Kolmogorov-
Smirnov test only requires the distributions to be continuous. (Both tests assume that the mean
and variances are constant for each population, which is inconsistent with the variation of the
means and variances with age and gender shown in the stratified analyses in Tables 3 and 4.) The
discrepancy is also partly explained by the fact that the Kolmogorov-Smirnov test is less
powerful (less likely to detect a difference) than the other tests, because it makes the fewest
assumptions and considers the widest class of alternative hypotheses.
Stratified Comparisons of Activity and Exertion Summary Statistics between Angina and
Non-Angina Subjects
Tables 3 and 4 provide the same statistical comparisons as Table 2, stratified by gender and age
group. The results show the mean values for the selected summary statistics are not consistently
lower for each age and gender subgroup of angina subjects. For example, Table 2 showed that
the angina subjects had a lower overall mean value of the average maximum 8-hour exertion
than the non-angina subjects. Tables 3 and 4 show the mean is actually higher for angina
subjects 0-54 of either gender and for males 75 or older. The mean average maximum 8-hour
exertions are consistently higher for males of all age groups, with or without angina, compared
to females. Similar patterns are found for the 95th percentile of the maximum 8-hour exertion.
The comparisons of the percentages of time spent outdoors or in a vehicle also vary across age
and gender subgroups. The largest, and most surprising, angina vs. non-angina difference is for
the mean percentage of time spent outdoors by 0-54 year old males: angina subjects have a mean
of 17 % compared to the mean of 9 % for non-angina subjects. However the angina subjects in
the 55-64 and 65-74 age groups of either gender spend less time outdoors, on average, than non-
F-9
-------
angina subjects.
The Table 3 and 4 comparisons of the mean percentages of time above the light, moderate or
high exertion levels show a variety of patterns for different age groups, genders, and exertion
levels.
Comparisons of Activity and Exertion Summary Statistics between Angina and Non-
Angina Subjects Adjusted for Age and Gender Differences
Table 5 gives the results of the fitted general linear model. As explained above, the fitted model
assumes that for each gender, the average value of the summary statistic is a cubic function of
age. Furthermore, having angina changes the expected value by a fixed amount, which is
assumed to be the same value for every age and gender. This angina effect is the coefficient
reported in the table, together with its standard error and p-value. P-values less than or equal to
0.05 indicate summary statistics where the angina effect was statistically significant at the 5
percent significance level. The angina coefficient can be thought of as the effect of angina after
adjusting for age and gender. The effects of age and gender are not reported, but in all cases were
extremely statistically significant compared to the angina effect.
Table 5 also reports the R squared goodness-of-fit statistic, which is the squared correlation
between the observed and predicted values. R squared values vary from 0 (the worst possible fit)
to 1 (a perfect fit), and are often interpreted as the fraction of the variability in the dependent
variable (summary statistic) that is explained by the regression model.
The first two rows of Table 5 show that the angina effect on the average and 95th percentile
maximum 8-hour exertion is a statistically significant reduction (at the 6 and 1 % levels,
respectively) for angina subjects compared to non-angina subjects. However, these reductions of
0.04 Meal and 0.16 Meal are small when compared to the overall mean values of 1.4 and 2.3
Meal (non-angina subjects) reported in Table 2. The next three rows show that angina subjects
tend to spend a little more time (0.7 percentage points) outdoors and a little less time (0.5
percentage points) in a vehicle compared to non-angina subjects; those differences are not
statistically significant. The last four rows show that angina subjects tend to spend less time at
moderate or high levels of exertion, after adjusting for age and gender, although the differences
are at most 1 percentage point and are not statistically significant. For example, the unadjusted
average percentage time above 2.39 kcal/min (0.010 MJ/min) was 23.5 % for non-angina
subjects (Table 2), and the effect of angina is to reduce the expected percentage of time by 0.7.
As shown in Tables 3 and 4, this is due to average reductions of up to 5 percentage points for
ages 55 and older but increases of 6 (males) and 2 (females) percentage points for the 0-54 age
group.
R squared goodness-of-fit statistics were extremely low, 0.05 or less, for the percentages of time
spent outdoors and/or in a vehicle. Thus the regression models for those percentages give very
poor predictions. There are two possible reasons for this. First, the combination of age, gender,
and angina status may be strongly associated with the percentages of time spent outdoors or in a
vehicle but the assumed form of the regression model may poorly represent the functional
F-10
-------
relationship. Second, the combination of age, gender, and angina status may be poorly associated
with the percentages of time spent outdoors or in a vehicle so that those activity percentages vary
mainly with the effects of factors other than age, gender, and angina status. In either case, those
regression models are not recommended for use in predicting the activity percentages.
For the summary statistics based on the maximum 8-hour exertion and the percentages of time
above exertion rate thresholds, the R squared statistics ranged from a poor fit, 0.25, to a
reasonably good fit, 0.48. As above, the cases of poor fitting models may be because the selected
statistical models poorly represent the relationship between age, gender, and angina and the
activity/exertion summary statistic and/or because the activity/exertion pattern varies
substantially between people of the same age, gender, and angina status. Alternative general
linear models, or the more sophisticated generalized linear models, could be developed to
improve the predictive ability of the statistical models.
F-ll
-------
REFERENCES
Brainard, J., and Burmaster, D. E. 1992. "Bivariate Distributions for Height and Weight of Men
and Women in the United States." Risk Analysis 12:2, pp. 267-275.
Burmaster, D. E., and Crouch, E. A. C. 1994. "Lognormal Distributions of Body Weight as a
Function of Age for Males and Females in the United States." Risk Analysis 17:4, pp. 499-
508.
Esmail, Bhambhani, and Brintnell. 1995. "Gender Differences in Work Performance on the
Baltimore Therapeutic Equipment Work Simulator." Amer. J. Occup. Therapy 49, pp. 405-
411.
Johnson, T. R, 1998. Proposed Probabilistic Algorithm for Estimating Ventilation Rate in the
1998 Version ofpNEM/CO. Memorandum submitted to EPA (September 25, 1998).
Klepeis, N. E., Tsang, A. E., and Behar, J. V. 1996. Analysis of the National Human Activity
Pattern Survey (NHAPS) Respondents from a Standpoint of Exposure Assessment.
EPA/600/R-96/074.
Klepeis, N. E., Nelson, W. C., Tsang, A. M., Robinson, J. P., Hern, S. C., Engelmann, W. H.,
and Behar, J. V. 1998. "The National Human Activity Pattern Survey (NHAPS): Data
Collection Methodology and Selected Results." Submitted to Journal of Exposure Analysis
and Environmental Epidemiology.
Schofield, W. N. 1985. "Predicting Basal Metabolic Rate, New Standards, and Review of
Previous Work." HumNutr ClinNutr. 39C (Supp 1), pp. 5-41.
Triplett, T. 1996. National Human Activity Pattern Survey, CD-ROM Version 2.0. University of
Maryland Survey Research Center.
Tsang, A. M. and Klepeis, N. E. 1996. Descriptive Statistics Tables from a Detailed Analysis of
the National Human Activity Pattern Survey (NHAPS) Data. EPA/600/R-96/148.
F-12
-------
Table 1. Distribution of subjects according to their age, gender and disease status
Age Group
0-54
55-64
65-74
75+
Total
Males
Angina (%)
35 (1 .0)
28 (6.5)
23 (7.9)
17(10.7)
103(2.4)
Non-angina
(%)
3307 (98.9)
400 (93.5)
267 (92.1)
142(89.3)
4116(97.6)
All
3342
428
290
159
4219
Females
Angina (%)
17(0.5)
28 (5.4 )
48 (9.6)
47(14.4)
140(2.8)
Non-angina
(%)
3570 (95.5)
491 (94.6)
450 (91 .4)
279 (85.6)
4790 (97.2)
All
3587
519
498
326
4930
All
Angina (%)
52 (0.8)
56 (5.9)
71 (9.0)
64(13.2)
243 (2.6)
Non-angina
(%)
6877 (99.2)
891 (94.1)
717(91.0)
421 (86.8)
8906 (97.4)
All
6929
947
788
485
9149
This table was modified by staff on 2-22-1 0 from the below original version due to issues related to the conversion from Word Perfect to Microsoft
Word.
Gender
Males
Females
All
Age
group
Angina(%)
35(1.0)
3342
28 (6.5)
428
23 (7.9)
290
17(10.7)
159
103(2.4)
4219
Non-angina(%) All
3307 (98.9)
400 (93.5)
267(92.1)
142(89.3)
4116(97.6)
Angina(%)
17(0.5)
3587
28 (5.4 )
519
48 (9.6)
498
47 (14.4)
326
140(2.8)
4930
Non-angina(%) All
3570 (95.5)
491 (94.6)
450(91.4)
279 (85.6)
4790 (97.2)
Angina(%)
52 (0.8)
6929
56 (5.9)
947
71 (9.0)
788
64(13.2)
485
243 (2.6)
9149
Non-angina(%) All
6877 (99.2)
891 (94.1)
717(91.0)
421 (86.8)
8906 (97.4)
F-13
-------
Table 2. Statistical Tests for the Association between Angina and Various Variables Representing Physical
Exertion. All
Variable
Average maximum 8hr
exertion (Meal)
Ninety fifth percentile of
maximum 8hr exertion
(Meal)
Percentage of time spent
outdoors
Percentage of time spent in
vehicle
Percentage of time spent
outdoors or in vehicle
Average percentage of time
with exertion above 2.39
kcal/min
= 0.01 OMJ/min (light)
Average percentage of time
with exertion above 5.97
kcal/min
= 0.025 MJ/ min (moderate)
Average percentage of time
with exertion above 9.55
kcal/min
T Test Comparison of
Means
Mean
Angina
1.28
1.87
6.73
4.55
11.27
19.98
2.17
0.213
Mean P-value
Non-angina
1.40 0.00
2.25 0.00
6.74 0.99
5.55 0.01
12.29 0.27
23.53 0.00
2.78 0.01
0.406 0.00
F Test Comparison of
Standard Deviations
St. Dev.
value
Angina
0.48
0.97
12.87
6.19
14.33
13.56
3.68
0.554
St. Dev. P-
Non-angina
0.49 0.68
1.13 0.00
11.63 0.02
7.13 0.00
13.45 0.15
13.78 0.75
3.56 0.46
0.761 0.00
Wilcoxon Kolmogorov
Test -Smirnov
Test
P-value
P-value
0.00 0.00
0.00 0.00
0.22 0.23
0.00 0.00
0.00 0.00
0.00 0.00
0.00 0.00
0.00 0.00
F-14
-------
= 0.040 MJ/min (heavy)
Table 3. Statistical Tests for the Association between Angina and Various Variables Representing Physical
Exertion. Males
Variable
Average maximum
8hr exertion (Meal)
Ninety fifth
percentile of
maximum 8hr
exertion (Meal)
Percentage of time
spent outdoors
Age
Group
0-54
55-64
65-74
75+
0-54
55-64
65-74
75+
0-54
55-64
65-74
T Test Comparison of F Test Comparison of
Means Standard Deviations
Mean Mean P- St. Dev. St. Dev. P-
value value
Angina
1.85
0.00
1.48
0.00
1.39
0.36
1.27
0.52
2.94
0.17
2.40
0.04
1.91
0.14
1.73
0.71
16.86
0.02
9.28
Non-angina Angina
1.59
1.77
1.49
1.20
2.68
2.90
2.17
1.67
8.85
10.02
0.49
0.34
0.48
0.86
0.47
0.96
0.44
0.65
1.06
0.13
1.21
0.61
0.78
0.31
0.69
0.68
19.16
0.00
14.26
F-15
Non-angina
0.55
0.48
0.48
0.42
1.30
1.14
0.94
0.76
13.75
13.86
Wilcoxon
Test
P-value
0.02
0.01
0.41
0.51
0.22
0.02
0.26
0.55
0.01
0.65
0.12
Kolmogorov
-Smirnov
Test
P-value
0.02
0.02
0.82
0.95
0.06
0.03
0.63
0.88
0.01
1.00
0.13
-------
Table 3. Statistical Tests for the Association between Angina and Various Variables Representing Physical
Exertion. Males
Variable
Percentage of time
spent in vehicle
Percentage of time
spent outdoors or
in vehicle
Average
Age
Group
75+
0-54
55-64
65-74
75+
0-54
55-64
65-74
75+
0-54
T Test Comparison of F Test Comparison of
Means Standard Deviations
Mean Mean P- St. Dev. St. Dev. P-
value value
Angina Non-angina Angina Non-angina
0.79
6.43
0.13
8.23
0.72
5.96
0.92
3.99
0.00
7.20
0.51
2.29
0.14
22.83
0.02
13.27
0.24
13.63
0.45
10.51
0.98
34.76
10.40
7.09
6.09
6.87
5.88
3.34
14.94
16.89
16.29
10.43
27.78
0.78
11.38
0.20
12.63
0.16
7.55
0.63
3.78
0.00
9.04
0.29
2.54
0.05
19.28
0.05
15.30
0.76
15.87
1.00
12.86
0.19
13.33
14.34
10.05
8.10
9.63
7.80
3.94
15.52
16.18
15.91
10.39
15.18
Wilcoxon
Test
P-value
0.58
0.89
0.18
0.82
0.44
0.02
0.17
0.19
0.69
0.02
Kolmogorov
-Smirnov
Test
P-value
0.66
0.52
0.19
0.93
0.54
0.02
0.22
0.29
0.41
0.05
F-16
-------
Table 3. Statistical Tests for the Association between Angina and Various Variables Representing Physical
Exertion. Males
Variable
percentage of time
with exertion
above 2.39
kcal/min
= 0.010MJ/min
(light)
Average
percentage of time
with exertion
above 5.97
kcal/min
= 0.025 MJ/m in
(moderate)
Average
percentage of time
with exertion
above 9.55
kcal/min
= 0.040 MJ/min
(heavy)
Age
Group
55-64
65-74
75+
0-54
55-64
65-74
75+
0-54
55-64
65-74
75+
T Test Comparison of F Test Comparison of
Means Standard Deviations
Mean Mean P- St. Dev. St. Dev. P-
value value
Angina
0.00
24.60
0.06
21.92
0.63
18.24
0.41
6.63
0.05
3.43
0.01
2.27
0.08
2.02
0.53
0.662
0.59
0.565
0.19
0.155
0.01
0.132
Non-angina
30.07
23.32
15.87
4.46
5.44
3.27
1.62
0.735
0.846
0.388
0.157
Angina
0.34
14.51
0.14
13.23
0.64
11.13
0.63
6.37
0.00
3.55
0.17
2.47
0.06
2.42
0.92
0.792
0.11
1.068
0.40
0.361
0.00
0.331
Non-angina
12.03
12.48
10.37
4.31
4.41
3.46
2.41
0.986
1.222
0.716
0.512
Wilcoxon
Test
P-value
0.06
0.67
0.39
0.02
0.01
0.20
0.43
0.55
0.05
0.06
0.55
Kolmogorov
-Smirnov
Test
P-value
0.14
0.84
0.74
0.01
0.01
0.40
0.59
0.15
0.17
0.04
0.96
F-17
-------
Table 3. Statistical Tests for the Association between Angina and Various Variables Representing Physical
Exertion. Males
Variable Age T Test Comparison of F Test Comparison of Wilcoxon Kolmogorov
Group Means Standard Deviations Test -Smirnov
Test
Mean Mean P- St. Dev. St. Dev. P- P-value
value value P-value
Angina Non-angina Angina Non-angina
^^^_^^^_ __ __^^^^^^^_^^^^^^^
Table 4. Statistical Tests for the Association between Angina and Various Variables Representing Physical
Exertion. Females
Variable Age T Test Comparison of F Test Comparison of Wilcoxon Kolmogorov
Group Means Standard Deviations Test -Smirnov
Test
Mean Mean P- St. Dev. St. Dev. P- P-value
value value P-value
Angina Non-angina Angina Non-angina
Average maximum 6-54 1.30 1.27 6.31 6.38 6.34 6.72 6.73
8hr exertion (Meal) 55-64 0.69 0.32 0.33 1.00 0.56 0.22
65-74 1.21 1.27 0.30 0.31 0.94 0.31 0.44
75+ 0.33 0.33 0.30 0.34 0.44 0.66
1.05 1.10
0.29
0.96 0.98
0.63
Ninety fifth 0-54 1.98 2.01 0.82 0.91 0.99 0.86
F-18
-------
Table 4. Statistical Tests for the Association between Angina and Various Variables Representing Physical
Exertion. Females
Variable
percentile of
maximum 8hr
exertion (Meal)
Percentage of time
spent outdoors
Percentage of time
spent in vehicle
Percentage of time
Age
Group
55-64
65-74
75+
0-54
55-64
65-74
75+
0-54
55-64
65-74
75+
0-54
T Test Comparison of F Test Comparison of
Means Standard Deviations
Mean Mean P- St. Dev. St. Dev. P-
value value
Angina Non-angina Angina Non-angina
0.86
1.79
0.41
1.42
0.26
1.27
0.59
3.64
0.42
4.31
0.88
2.84
0.14
3.79
0.40
4.54
0.55
4.21
0.35
5.26
0.23
2.82
0.86
8.18
1.92
1.51
1.31
5.11
4.59
4.14
2.27
5.35
5.60
4.15
2.94
10.46
0.69
0.80
0.68
0.53
0.57
0.56
0.43
7.29
0.20
9.53
0.23
5.51
0.02
12.04
0.00
5.48
0.72
7.53
0.71
5.94
0.77
4.37
0.91
9.93
0.77
0.57
0.52
9.58
8.22
7.34
4.60
5.98
7.23
6.18
4.34
11.27
Wilcoxon
Test
P-value
0.33
0.31
0.43
0.43
0.23
0.49
0.76
0.40
0.06
0.15
0.72
0.18
Kolmogorov
-Smirnov
Test
P-value
0.28
0.62
0.47
0.91
0.63
0.53
1.00
0.35
0.12
0.19
0.96
0.27
F-19
-------
Table 4. Statistical Tests for the Association between Angina and Various Variables Representing Physical
Exertion. Females
Variable
spent outdoors or
in vehicle
Average
percentage of time
with exertion
above 2.39
kcal/min
= 0.010MJ/min
(light)
Average
percentage of time
with exertion
above 5.97
kcal/min
= 0.025 MJ/m in
(moderate)
.Ayjiiac^^
Age
Group
55-64
65-74
75+
0-54
55-64
65-74
75+
0-54
55-64
65-74
75+
0-54
T Test Comparison of F Test Comparison of
Means Standard Deviations
Mean Mean P- St. Dev. St. Dev. P-
value value
Angina Non-angina Angina Non-angina
0.36
8.52
0.48
8.10
0.88
6.61
0.45
23.50
0.49
19.04
0.33
13.97
0.26
11.31
0.73
1.44
0.71
0.79
0.05
0.65
0.87
0.75
0.23
0.101
10.18
8.29
5.21
21.40
21.04
15.46
11.84
1.59
1.31
0.68
0.43
0.184
0.57
12.15
0.22
8.21
0.26
12.36
0.00
12.28
0.86
10.34
1.00
8.45
0.31
9.75
0.24
1.66
0.44
1.27
0.00
1.35
0.22
1.73
0.01
0.170
10.42
9.38
6.25
12.15
10.47
9.53
8.61
1.97
2.07
1.56
1.31
0.324
Wilcoxon
Test
P-value
0.04
0.99
0.77
0.41
0.51
0.43
0.51
0.73
0.04
0.51
0.84
0.63
Kolmogorov
-Smirnov
Test
P-value
0.05
0.86
0.97
0.69
0.54
0.76
0.89
0.74
0.06
0.88
0.67
0.80
F-20
-------
Table 4. Statistical Tests for the Association between Angina and Various Variables Representing Physical
Exertion. Females
Variable
Age T Test Comparison of
Group Means
Mean Mean P-
value
Angina Non-angina
F Test Comparison of
Standard Deviations
St. Dev. St. Dev. P-
value
Angina Non-angina
Wilcoxon
Test
P-value
Kolmogorov
-Smirnov
Test
P-value
percentage of time
with exertion
above 9.55
kcal/min
= 0.040 MJ/min
(heavy)
55-64
65-74
75+
0.06
0.095
0.96
0.028
0.89
0.025
0.44
0.093
0.030
0.016
0.00
0.277
0.01
0.076
0.00
0.075
0.90
0.198
0.110
0.077
0.67
0.85
0.28
1.00
0.99
0.83
F-21
-------
Table 5. General Linear Models for the Association between Angina and Various Variables Representing Physical
Exertion.
Variable
Average maximum 8hr exertion (Meal)
Ninety fifth percentile of maximum 8hr exertion
(Meal)
Percentage of time spent outdoors
Percentage of time spent in vehicle
Percentage of time spent outdoors or in vehicle
Average percentage of time with exertion above
2.39 kcal/min = 0.010 MJ/min (light)
Average percentage of time with exertion above
5.97 kcal/min = 0.025 MJ/min (moderate)
Average percentage of time with exertion above
9.55 kcal/min = 0.040 MJ/min (heavy)
Angina
Coefficient1
-0.0445
-0.1553
+0.6975
-0.4805
+0.2170
-0.7359
-0.1730
-0.0933
Standard
Error
0.0237
0.0581
0.7648
0.4679
0.8777
0.6996
0.1910
0.0439
P-value
0.0608
0.0075
0.3618
0.3045
0.8047
0.2929
0.3650
0.0334
R squared
0.4819
0.4114
0.0388
0.0325
0.0520
0.4239
0.3570
0.2494
1. The angina coefficient is the expected difference (angina minus non-angina) between the summary statistic for angina and
non-angina subjects of the same age and gender.
F-22
-------
OS
o
9
o
PH
W
|
Is
It
o
•
8
PH
a
w
------- |