1. No category

March 17, 2015
Environmental Protection Agency
EPA Docket Center (EPA/DC)
Mailcode 28221T
Attention Docket ID No. OAR–2008–0699, 1200
Pennsylvania Ave. NW.
Washington, DC 20460
By Email.
Re: Docket ID No. EPA–HQ–OAR–2008–0699, National Ambient Air Quality Standards for Ozone.
Dear Sir or Madam,
Air Alliance Houston and the undersigned individuals and organization appreciate this opportunity to
comment on the proposed 2014 National Ambient Air Quality Standards for Ozone. Air Alliance Houston
is a non-profit organization whose mission is to reduce air pollution in the Houston region to protect
human health and environmental integrity through research, education, and advocacy.
The Clean Air Act requires the Environmental Protection Agency (EPA) to establish National Ambient Air
Quality Standards (NAAQS) requisite to protect human health and welfare with an adequate margin of
safety.1 The American Lung Association, in its “State of the Air 2014” report, ranks Houston sixth among
the top ten most ozone-polluted cities in the nation.2 There is an urgent need to strengthen the ozone
standards to protect the people of Houston. Only strong standards will drive the cleanup of ozone in
Houston and across the nation.
I.
1
2
A standard of 60 parts per billion is necessary to protect human health and welfare.
42 U.S.C. § 7409.
See http://www.lung.org/press-room/press-releases/healthy-air/SOTA-2014-National-Press-Release.html.
A. The Clean Air Scientific Advisory Committee recommends an ozone NAAQS between
60 and 70 parts per billion.
According to the Clean Air Scientific Advisory Committee (“CASAC”), the current standard is inadequate
to protect public health. At 70 parts per billion (ppb), there is mounting evidence of adverse effects of
long term exposure on the population. The CASAC suggests that the standard should be set below this
level to meet the statutory requirements of the Clean Air Act to protect public health with an adequate
margin of safety. A standard of 60 ppb is the most protective option and the only option that will
efficiently achieve this goal. “The CASAC concludes that there is adequate scientific evidence to
recommend a range of levels for a revised primary ozone standard from 70 ppb to 60 ppb…The
frequency of lung function decrements and premature mortalityfrom short-term exposure to ozone
decreases is tremendous even further when the alternative standard is lowered to 60 ppb.”3
B. The current standard of 0.075 parts per million is above the level necessary to protect
vulnerable populations.
A review of the Clean Air Science Advisory Committee findings by the Sierra Club reveals that the current
8‐hour average standard of 0.075 parts per million is above the level proven to harm the lungs.4
Populations at risk include children, the elderly, people with asthma, and healthy individuals that work
or exercise outdoors. People with pre‐existing respiratory diseases are at increased risk because they
have less pulmonary reserve and cannot tolerate the reduction in lung function or the increase in
respiratory symptoms. In the United States today, there are more than 25 million people suffering from
asthma, 74 million children, and 40 million senior citizens. There are nearly 17 million outdoor workers.
Based on age criteria alone, more than one‐third of the population is at increased risk of adverse effects
from ozone in the range of 60 to 70 ppb. The CASAC has repeatedly recommended an 8-hour ozone in
the range of 60 to 70 ppb in their most recent letter to the EPA Administrator. Explaining and developing
the Policy Assessment for the Review of the Ozone National Ambient Air Quality Standards, Sierra Club
revealed that the Risk and Exposure Assessment estimates revised standard level of 60 ppb would
tremendously reduce children’s exposures of concern by 95 to 100% compared to the current standard.
3
See EPA, “Heath Risk and Exposure Assessment for Ozone, Final Report,” available at
http://www.epa.gov/ttn/naaqs/standards/ozone/s_o3_2008_rea.html.
4
See Sierra Club, “Ten Reasons Why the Ozone Air Quality Standard Must Be Strengthened,” available at
http://www.law.uh.edu/faculty/thester/courses/Environmental-Practicum2015/Smog%20Rule%20Talking%20Points.pdf.
C. Studies show respiratory effects of both short-term and long-term exposure to ozone.
According to Sierra Club comments on the CASAC’s review, exposure to ozone, in the short-term (acute)
and repeat (chronic) exposure, is the leading cause of an exacerbation of respiratory impacts such as
breathing discomfort (e.g. coughing, wheezing, shortness of breath, pain upon inspiration), decreasing
lung function and capacity, and lung inflammation and injury.5 Research on the relationship between
ozone exposure and respiratory effects is well-documented, and indeed, the EPA’s Integrated Science
Assessments (ISA) made a conclusive determination that ozone is responsible for adverse respiratory
effects. As the ISA revealed, additional controlled human exposure, epidemiologic, and toxicological
studies have strengthened the causal relationship between short-term exposure and respiratory health
effects. New studies have solidified links between short-term increases in ambient ozone concentrations
and health effects. For controlled human exposure studies, new evidence shows significant decreasing
lung function and pulmonary inflammation in healthy adults following exposures at decreasing
concentrations down to even 60 ppb. New epidemiologic studies strengthen evidence on short-term
ozone exposures and respiratory outcomes such as respiratory-related hospital admissions, emergency
department visits, and mortality.
Additionally, evidence of the effects of long-term ozone exposure has grown considerably. New studies
demonstrated the impacts of long-term exposure on respiratory health, such as for pulmonary
inflammation and injury, new onset asthma, and respiratory mortality. The ISA concluded that there is
now a likely causal relationship between long-term exposure and adverse respiratory effects. EPA stated
that scientific studies now offer an “overall strong body of evidence of those adverse health effects.”
D. The Children’s Health Protection Advisory Committee reaffirmed the recommendation
of 60 ppb.
The EPA’s Children’s Health Protection Advisory Committee (CHPAC) strongly reaffirmed the
recommendation of 60ppb based on the expanding scientific evidence. According to CHPAC, children
have increased susceptibility due to increased ventilatory rates and increased outdoor physical activity
compared with adults. The 6.8 million children suffering from asthma in the United States are some of
the most vulnerable to ozone-related respiratory impacts (CDC, 2014). The US EPA 2013 Ozone
Integrated Science Assessment summarized numerous recent epidemiologic studies that found
5
See “Sierra Club Comments regarding EPA’s Clean Air Scientific Advisory Committee’s Review of the Ozone
National Ambient Air Quality Standards,” at 9-15 (March 13, 2014), available at
http://yosemite.epa.gov/sab/sabproduct.nsf/5A248564CC67232585257C9B005B189F/$File/AS+FILED+Sierra+Club
+CASAC+EPA+Ozone+Comments+vfinal.pdf.
relationships between ambient ozone exposure concentrations within and even below the CASAC
previously proposed range, 60-70ppb, and adverse childhood health impacts including: increased
asthma exacerbations, impaired lung development, changes in birth outcomes, and increased upper
respiratory illness (US EPA, 2013). Therefore, the current scientific evidence documenting ozone-related
childhood health impacts is stronger compared to the last review. It warrants a lower recommended
range of standards to adequately protect children’s health and wellbeing. According to the CHPAC
committee, children suffer a disproportionate burden of ozone-related health issues due to critical
developmental periods of lung growth in childhood and adolescence that can result in permanent
disability. Laboratory toxicology studied airways of infant monkeys exposed to ozone air pollution and
noticed structural changes in the respiratory tract. Epidemiological studies provide evidence that ozone
is responsible for increases risk of emergency room visits and hospital admissions for respiratory
problems, and even premature death.
Finally, CHPAC observed that:
One concrete example of how children’s health will be positively impacted by a lower standard
is outlined in the 2014 EPA Second Draft Policy Assessment for the Review of Ozone NAAQS” (US
EPA, 2014). It estimates that 14-19% of children (approximately 952,000–1,292,000 asthmatic
children based on CDC statistics) living in urban centers will have a greater than 10% decrement
in lung function based on a standard of 75ppb, and this percentage decreases to 5-11%
(approximately 340,000–748,000 asthmatic children based on CDC statistics) with a 60ppb
standard. The reduction from 75ppb to 60ppb would lead to approximately 500,000 fewer
children affected by ozone exposure. Therefore, the reduced standard would result in
tremendous quantifiable children’s health protections, and this is only one example of the
numerous childhood health protections afforded.6
E. Research in Houston finds strong evidence of increased asthma attacks and cardiac
arrests due to elevated ozone levels.
Research conducted by Drs. Kathy Ensor and Loren Raun at Rice University in Houston finds strong
evidence of increased asthma attacks and cardiac arrests due to elevated ozone levels.7 For asthmatics
6
See “Children Health Protection Advisory Committee: CASAC Review of the Health Risk and Exposure Assessment
for Ozone and Policy Assessment for the Review of the Ozone NAAQS: Second External Review Draft,” available at
http://www2.epa.gov/sites/production/files/2014-12/documents/2014.05.19_chpac_ozone_naaqs.pdf
7
The following studies are attached to this document and incorporated herein by reference: Ensor, K. B., Raun, L. H.
and Persse, D. (2013). A Case-Crossover Analysis of Out-of-Hospital Cardiac Arrest and Air Pollution. Circulation.
V127, pp 1192-1199.; Raun, L. H., Ensor, K. B. and Persse, D. (2014) Using community level strategies to reduce
in Houston, the risk of an asthma attack increases by 5% when ambient ozone levels increase by 20 ppb
over a three-day period. Considering periods when the ozone levels are between 50 to 70 ppb, the
increased risk of an asthma attack is 13% when ambient ozone levels increase by 20 ppb during a threeday period. The increased risk is 10% for increases of this level in a single day. In the range of 70 to 90
ppb, these risks increase to 45% and 21%, respectively.
For heart patients, the concern is a three-hour window of elevated ozone levels. A 20 ppb increase in
ozone levels resulted in a 4% increase in cardiac arrests. The impact is greatest when the ozone levels
are above 75 ppb.
Negative cardiovascular effects are experienced in response to short-term ozone exposure, including
changes in heart rate variability and blood markers of systemic inflammation and oxidative stress,
supporting certain effects observed in toxicological studies. Ozone exposures are shown to increase risks
of hospitalization for acute myocardial infarction, coronary atherosclerosis, stroke, and heart disease,
even at ambient ozone levels well-below current NAAQS. Evidence on chronic ozone exposure reveals
an increasing number of studies showing a relationship with cardiovascular disease. New studies
associate ozone exposure with increased risks for heart attacks. Exposure to ozone has also been linked
to increased risk for stroke incidents. New research also shows that chronic ozone exposure may put
children at risk for cardiovascular disease later in life. Young adults growing up in areas with higher
ozone concentrations showed a tendency towards early atherosclerotic (hardening of the arteries).
According to studies by Raun and Ensor, out of hospital cardiac arrests (OHCA) are associated with shortterm exposure to air pollutants.8 Unexpected cardiovascular collapses due to an underlying cardiac
cause occur out of the hospital; approximately 300,000 persons in the United States experience an
OHCA each year. More than 90% of those persons who experience OHCA die (Mc Nally et al. 2011). Also,
the study population of interest for the purpose of their analysis is representative of the general
population that experiences an OHCA that is not trauma related. It is not representative of the overall
population which would include sensitive subgroups, because it excludes those with life threatening comorbidities or clinically recognized heart diseases.
II.
EPA Cannot Consider Economic Costs or Difficulty of Implementation when Setting
asthma attacks triggered by outdoor air pollution: a case crossover analysis. Environmental Health, 13:58.; Raun, L.
and Ensor, KB. 2012. “Association of out-of-hospital cardiac arrest with exposure to fine particulate and ozone
ambient air pollution from case-crossover analysis results: are the standards protective?” James A. Baker III
Institute for Public Policy of Rice University.
NAAQS.
Arguments against a NAAQS in the range of 60-70 ppb that rely on implementation costs are irrelevant
and incorrect. The language of section 109 Clean Air Act is absolute and clear that EPA must set a quality
standard requisite to protect public health and welfare “with an adequate margin of safety.”9 The
language of the Clean Air Act unambiguously excluds costs and highlights the need to protect public
health against the impact of pollution. Industry groups have unsuccessfully challenged this assertion in
the past. It is settled that EPA cannot consider economic arguments; public health must not be weighed
against economic costs. Therefore, arguments about cost and feasibility of achieving NAAQS standards
are irrelevant. In Whitman v. American Trucking Assns., 531 U.S. 457, 465 (2001), the United States
Supreme Court unambiguously held that costs could not be considered in setting NAAQS. Likewise, the
Court of Appeal of the District of Columbia held that attainability and technological feasibility are not
relevant considerations in the promulgation of NAAQS. Am. Petroleum Inst. v. Costle., 665 F.2d 1176,
1185 (D.C. Cir. 1981).
III.
The EPA must reject the Texas Commission on Environmental Quality’s contention that no
link exists between ozone and health impacts and the Commission’s argument that human
behavior makes concerns about outdoor air quality irrelevant.
The Texas Commission on Environmental Quality (TCEQ), in a recent article by the Director of the TCEQ
Toxicology Division, Dr. Michael Honeycutt, questions the science linking lowered ozone levels and
health impacts and posits that a lower standard would not result in any measurable benefit to public
health.10 The TCEQ makes a number of claims contending that various health conditions are not
connected to lower ozone levels. The TCEQ also argues that human behavior—the fact that most people
spend 90% of their time indoors—means that outdoor air quality is irrelevant. The EPA must reject these
claims as unsupported by (and indeed contradicted by) the best available science.
First, the TCEQ questions the connection between asthma and ozone levels in order to combat EPA’s
focus on asthmatics in regard to ozone levels. When setting primary ozone NAAQS, the EPA has the
authority to consider effects of the rule on asthmatics and other sensitive groups. Mississippi v. EPA, 744
F.3d 1334 (D.C. Cir. 2013) cert. denied sub nom. Util. Air Regulatory Grp. v. EPA, 135 S. Ct. 53, 190 L. Ed.
2d 30 (2014). And must protect those groups, not just the average population. See American Lung Ass’n
9
24 U.S.C. § 7409.
Michael Honeycutt, “Will EPA’s Proposed New Ozone Standards Provide Measurable Health Benefits?” available
at http://www.tceq.state.tx.us/publications/pd/020/2014/will-epas-proposed-new-ozone-standards-providemeasurable-health-benefits.
10
v. EPA, 134 F. 3d 388, 389 (D.C. Cir. 1998).
Dr. Honeycutt points out that asthma diagnoses are rising nationwide while ozone levels are declining.
This is a generalization that proves nothing about the link, or lack thereof, between asthma and ozone
exposure. The article indicates that some Texas county hospitals’ data show that in the winter when
ozone levels are generally lowest, asthma admissions are highest. This hardly supports the claim that
there is no link between asthma and ozone levels. Honeycutt and Stephanie Shirley’s paper, A
Toxicological Review of the Ozone NAAQS,11 recognizes that there are numerous potential factors of
asthma development and exacerbation, cold weather among them. But this review focuses on
identifying the main cause of asthma and does not rule out ozone as a factor. It is not necessary that
ozone be identified as the main cause or driving factor in asthma. Furthermore, absolute certainty about
the relationship between asthma and ozone is not required. Rather, the EPA has room for some
uncertainty with “an adequate margin of error” that buffers the public from unknown health threats.
See Lead Indus. Ass'n, Inc. v. EPA, 647 F.2d 1130 (D.C. Cir. 1980); 42 USCS § 7409(b)(1). Because the best
available science does show some potential connection between ozone and the development and
exacerbation of asthma, the EPA is required to set a NAAQS level that addresses that connection.
Dr. Honeycutt also posits that ambient ozone concentrations are not actually representative of peoples’
everyday exposure and therefore, the standard should not be lowered. While it may be true that people
spend more than 90% of their time indoors, that behavior is irrelevant. With the implementation of
NAAQS, the Clean Air Act gives people the protection of healthy air regardless of their daily habits. See
42 USCS § 7409(b)(1). “NAAQS must be set at a level that is requisite to public health from adverse
effects of the pollutant in ambient air…”. Whitman v. Am. Trucking Ass'ns, 531 U.S. 457, 121 S. Ct. 903,
149 L. Ed. 2d 1, (2001), emphasis added. It is the ambient air that poses the risk that the NAAQS are
supposed to protect against. Linking the protections afforded by the NAAQS to individual behaviors, or
even the behavior of a majority of people, would defeat the intent of the Clean Air Act.
IV.
EPA Modeling in the Health Risk and Exposure Assessment does not cast doubt on the
science of the harmful effects of ozone.
The TCEQ’s Dr. Honeycutt has exploited a quirk in atmospheric chemistry and in EPA’s modeling to cast
doubt on the science behind and the health benefits of a lower ozone standard:
The EPA’s own modeling in its Health Risk and Exposure Assessment (HREA) indicates that
11
Available at http://www.tceq.com/assets/public/implementation/tox/ozone/superconference.pdf.
lowering ozone concentrations would actually result in more deaths in some cities (Appendix 7,
page 7B-2 of the HREA). Either this indicates that lowering the ozone standard defeats its stated
purpose of protecting human health, or it indicates that something is wrong with the EPA’s
interpretation of the data. Either way, it’s not a good argument for lowering the ozone
standard.12
There is an EPA model that predicts a slight increase initially in premature deaths that could result if
ozone standards are lowered. This is due to the complexities of atmospheric chemistry and the fact that
reducing nitrogen oxides (NOx) in a high-NOx environment such as Houston can temporarily favor the
production of ozone and lead to an increase in ozone levels. Data included in an EPA review of the
health impact of lower ozone standards shows that deaths attributable to ozone could actually increase
in two cities, Houston and New York, were current levels reduced. This could cause a short-term spike in
ozone-related health problems. The effect would be short term and serves to emphasize the point that
exposure to ozone can be deadly. However, the TCEQ is overemphasizing the significance of this
possibility and exploiting it to cast doubt on the very fact of ozone’s health impacts.
A model predicting a possible temporary increase in mortality in Houston is not evidence that ozone is
not a harmful air pollutant, or that the negative health impacts of ozone are in doubt. The best science
of the day, as presented by EPA in this rule, indicates that a lower ozone standard will benefit hundreds
of thousands of people across the country, including in Houston.
This fact was explained to us in a recent email by the EPA’s Dr. Scott Jenkins:
Extensive scientific evidence and analysis shows that reducing high ozone concentrations will
reduce risks broadly across the country – including in Texas. On high ozone days, when ozone is
reduced, the number of deaths goes down.
Complex atmospheric chemistry can also cause ozone to go up in some areas when NOx
emissions go down. This mainly occurs on low ozone days. However, we have much less
confidence in the public health implications of these changes in low ozone concentrations.
In summary, we are much more confident in our estimates that on high ozone days, when ozone
12
Michael Honeycutt, “Will EPA’s Proposed New Ozone Standards Provide Measurable Health Benefits?” available
at http://www.tceq.state.tx.us/publications/pd/020/2014/will-epas-proposed-new-ozone-standards-providemeasurable-health-benefits.
concentrations go down, deaths also go down.13
As we understand it, the model that predicts a temporary increase in mortality in Houston is based on a
scenario in which Houston responds to the new standard with a strategy that reduces only NOx
emissions, as opposed to a strategy that reduces both NOx and VOC emissions. We favor strategies to
reduce all pollutants in Houston, including VOCs, many of which are harmful air pollutants in their own
right.
We understand that implementation of a new ozone NAAQS is left to the states in their state
implementation plans. However, if there is a particular implementation strategy that would be more
effective for Texas and Houston, then it is important that that strategy be used. We would appreciate
guidance from EPA on the best strategy to implement a new standard in Texas and provide the most
health benefits possible.
V.
EPA should provide guidance on and encourage a regulatory approach that decreases both
NOx and VOC emissions in Houston.
Complex atmospheric chemistry can cause ozone to go up in some areas when NOx emissions go down.
This mainly occurs on low ozone days. However, there is much less confidence in the public health
implications of these changes in low ozone concentrations. Furthermore, the HREA found that the air
quality response to a dual reduction approach of reducing both NOX and VOC emissions generally
resulted in larger decreases in mid-range ozone concentrations. By reducing both NOx and VOC
emissions (as opposed to reducing NOx emissions alone), the increases in low ozone concentrations
were smaller in the urban study areas, and this was most apparent in Houston, along with six other
cities. Therefore, the issue identified in the TCEQ’s argument, the reverse effect of NOx only reduction,
could be averted by a dual-reduction approach. These results suggested that by tailoring the reduction
approach to Houston’s atmospheric qualities there could be even larger decreases in ozone-associated
mortality than indicated in the HREA estimates.
The EPA addressed the impact of NOx reduction on ozone in its proposed rule, highlighting how the
reduction cause by this NOx phenomenon is mainly seen during conditions that cause low ambient
ozone concentrations. While there may be short-term reduction of ozone near the emission sources in
urban areas, the NOx will eventually react downwind of the source to form ozone. Photochemical model
simulations have shown that NOx reduction decreases the highest ozone concentrations in outlying
13
Email from Scott Jenkins to Brian Butler (Dec. 5, 2014) (emphasis added).
areas and slightly increases ozone concentrations near the NOx emission site on days with low ozone
concentrations. 79 Fed. Reg. 75270 (2014) In considering both types of risk estimates, in the Proposed
Rule the EPA notes that there is a greater public health concern for adverse ozone effects at higher
ambient ozone concentrations (which drive higher exposure concentrations, section 3.2.2 of the PA (U.S.
EPA, 2014c) as compared to lower concentrations. In summary, the EPA is confident in its estimates that
on high ozone days, when ozone concentrations go down, deaths also go down.
The TCEQ has stated that a lower ozone standard is not justified. According to their studies, ozone has
similar effects on the lungs of asthmatics and non-asthmatics, and children and young adults are equally
sensitive to ozone exposure. This actually supports implementing a lower ozone standard if ozone will
have adverse effects on even healthy lungs. Long-term, there is scientific consensus that reducing ozone
will only lead to public health benefits. Ozone worsens conditions like asthma and lung disease; and the
current standard of 75 parts per billion allows for unacceptable health risks.
The TCEQ argues that since most people spend of their time indoors people are rarely exposed to any
significant levels of ozone. In actuality, this argument further exposes the need for a lower standard
because the current negative effects of ozone are seen even though people spend most of their time
indoors. The Proposed Rule states the lowest ozone exposure concentration for which that respiratory
effects such as decreased lung function and increased airway inflammation have been reported in
healthy adults at 60 ppb after 6.6 hours of exercise, which is why the new ozone standard should be set
at 60 ppb. A decrease in ozone related deaths follows a decrease in ozone ppb.
Thus, setting the standard for ozone at 60 ppb is appropriate and justified by science and will result in
substantial improvements in overall public health. According to the Administrator in the CASAC
conclusion, “there are meaningful reductions in mean premature mortality associated with ozone levels
lower than the current standard” (Frey, 2014a, p. 10). Moreover, the HREA risk estimates for urban
areas have likely understated the ozone-associated mortality and morbidity risk reductions, so the
beneficial consequences of a lower standard will be greater than expected across urban populations.
VI.
The EPA should provide more detailed guidance on communication of health information.
Air Alliance Houston has long been at the forefront of efforts to communicate health information about
air quality to the public. Our “Ozone Theater” youth education program reaches some 5,000 elementary
and middle school students each year and received the EPA’s “Clean Air Excellence Award” in 2007.14
14
See www.ozonetheater.org.
Our Houston Clean Air Network website and smart phone app provides the only neighborhood scale real
time map of ozone pollution available anywhere in the world.15
Air Alliance Houston has long tried to find simple, effective ways to communicate information about
ozone and other types of air pollution to children and adults alike. The Air Quality Index and the Ozone
Action Day systems have proven useful tools, but they have not kept pace with the technology and
methods available to communicate air quality information to the public. The EPA has proposed to revise
the levels association with the Air Quality Index (AQI), but it has not proposed any rules nor offered any
guidance to organizations such as Air Alliance Houston wishing to use the AQI or other comparable
systems to communicate air quality information to the public in new and innovative ways.
The Houston Clean Air Network has proven superior to systems such as the “Ozone Action Day” warning
system. In TCEQ Region 12, which covers the Houston area, eight-hour violations were successfully
predicted by Ozone Action Days 57% of the time in 2014 and only 38% of the time in 2013. The Houston
Clean Air Network map, by contrast, eliminates the need for predictions by reporting ozone values in
real time. But the Houston Clean Air Network’s use of the Air Quality Index to communicate air quality
data on five-minute increments has subjected it to criticism that has limited its appeal to certain
audiences. The EPA’s silence on this and other matters has hampered efforts to produce new tools such
as the Houston Clean Air Network and present them to a wider audience. The EPA should foster
innovation in the field of communication of health and air quality information by providing guidance.
15
See www.houstoncleanairnetwork.com.
VII.
Conclusion
Thank you for the opportunity to provide comments on this proposed rule. If you wish to discuss these
comments further, please contact Adrian Shelley at adrian@airalliancehouston.org, 713-528-3779.
Sincerely,
Adrian Shelley
Executive Director
Air Alliance Houston
/s/ Elaine Wiant
President
League of Women Voters of Texas
/s/ Tom “Smitty” Smith
Director
Public Citizen, Texas Office
/s/ Caroline Dinges
JD Candidate 2016
University of Houston Law Center
/s/ Catherine Irene Mandengue
FLLM
University of Houston Law Center
Health Services and Outcomes Research
A Case-Crossover Analysis of Out-of-Hospital Cardiac
Arrest and Air Pollution
Katherine B. Ensor, PhD; Loren H. Raun, PhD; David Persse, MD
Background—Evidence of an association between the exposure to air pollution and overall cardiovascular morbidity
and mortality is increasingly found in the literature. However, results from studies of the association between acute air
pollution exposure and risk of out-of-hospital cardiac arrest (OHCA) are inconsistent for fine particulate matter, and,
although pathophysiological evidence indicates a plausible link between OHCA and ozone, none has been reported.
Approximately 300 000 persons in the United States experience an OHCA each year, of which >90% die. Understanding
the association provides important information to protect public health.
Methods and Results—The association between OHCA and air pollution concentrations hours and days before onset was
assessed by using a time-stratified case-crossover design using 11 677 emergency medical service–logged OHCA events
between 2004 and 2011 in Houston, Texas. Air pollution concentrations were obtained from an extensive area monitor
network. An average increase of 6 µg/m3 in fine particulate matter 2 days before onset was associated with an increased
risk of OHCA (1.046; 95% confidence interval, 1.012–1.082). A 20-ppb ozone increase for the 8-hour average daily
maximum was associated with an increased risk of OHCA on the day of the event (1.039; 95% confidence interval,
1.005–1.073). Each 20-ppb increase in ozone in the previous 1 to 3 hours was associated with an increased risk of OHCA
(1.044; 95% confidence interval, 1.004–1.085). Relative risk estimates were higher for men, blacks, or those aged >65
years.
Conclusions—The findings confirm the link between OHCA and fine particulate matter and introduce evidence of a similar
link with ozone. (Circulation. 2013;127:1192-1199.)
Key words: sudden death ◼ heart arrest ◼ epidemiology ◼ particulates ◼ pollution ◼ ozone
O
ut-of-hospital cardiac arrest (OHCA) is defined as a condition characterized by an unexpected cardiovascular
collapse due to an underlying cardiac cause occurring outside
the hospital. It is of significant concern given that ≈300 000
persons in the United States experience an OHCA each year,
and >90% of those persons who experience an OHCA die.1
Understanding the role of air pollution in increasing the risk
of OHCA is important to protect public health. Evidence that
short-term exposure to air pollution is associated with cardiovascular morbidity and mortality is increasingly found in
the literature, especially with respect to fine particulate matter with an aerodynamic diameter <2.5 µm (PM2.5), and, to a
lesser extent, ozone.2–6 A handful of case-crossover studies
have specifically examined the association between PM2.5 and
ozone air pollution with a focus on OHCA or out-of-hospital
cardiac death.7–11 However, in these studies, the results of
an association between OHCA and PM2.5 have been inconsistent, and no association has been found between OHCA
and ozone (eg, studies reported a range of −6.0% to 11.0%
increase in risk of OHCA per 10 µg/m3 increase in PM2.5
and −5.5% to 22.8% increase in risk of OHCA per 20-ppb
increase in ozone).
Clinical Perspective on p 1199
In an effort to better understand the association of air
pollution and OHCA, we used an extensive air-monitoring
network and a large emergency medical service (EMS) call
database spanning 8 years. We focused on 2 pollutants, PM2.5
and ozone, both with epidemiological evidence supported by
pathophysiological arguments that link them to cardiac end
points.4,12–18 We also examined the association between nitrogen dioxide, sulfur dioxide, and carbon monoxide with cardiac arrest. Our studies were conducted on both a daily and
hourly time scale.
Methods
Out-of-Hospital Cardiac Arrest Data
The Rice University and Baylor College of Medicine Institutional
Review Board approved all data-collecting procedures for human
subjects. All cases in which EMS performs chest compressions are
considered OHCA cases. The OHCA study data included non–deadon-arrival adults aged ≥18 years from Houston Fire Department
EMS calls over the 8-year period from 2004 to 2011. The database
consisted of 11 677 cases of OHCA events. In addition to recording the
Received August 8, 2012; accepted January 30, 2013.
From Rice University, Department of Statistics (K.B.E., L.H.R.); City of Houston Health and Human Services, Bureau of Pollution Control and
Prevention (L.H.R.); and City of Houston Fire Department, Houston, TX (D.P.).
Correspondence to Loren H. Raun, PhD, Rice University, Department of Statistics, PO Box 1892, MS 138, Houston, TX 77251-1892. E-mail
raun@rice.edu
© 2013 American Heart Association, Inc.
Circulation is available at http://circ.ahajournals.org
DOI: 10.1161/CIRCULATIONAHA.113.000027
Downloaded from http://circ.ahajournals.org/
by guest on February 27, 2015
1192
Ensor et al Out-of-Hospital Cardiac Arrest and Pollution 1193
time and location of the event, other relevant information necessary
for age, sex, race, and preexisting condition stratification were also
available. This additional information was collected by EMS with the
use of Utstein guidelines.19
Ambient Air Quality and Meteorologic Data
Ambient pollution concentration data were obtained from the Texas
Commission of Environmental Quality for the 8-year study period of
2004 through 2011. In this analysis, hourly data from 47 monitors
measuring ozone, 12 measuring PM2.5, 22 measuring nitrogen dioxide, 13 measuring sulfur dioxide, and 12 measuring carbon monoxide
were used. The hourly and daily average values of PM2.5, ozone, nitrogen dioxide, sulfur dioxide, and carbon monoxide were calculated
across monitors. For ozone, we calculated the daily maximum 8-hour
running mean. The number of air monitors measuring a specific pollutant changed through the study years as monitors went on and off
line. However, <1% of the time all monitors were simultaneously
down. All air pollution data were collected by using Environmental
Protection Agency federal reference methods20 and validated by the
Texas Commission of Environmental Quality.
To control for potential confounding meteorologic events,
1-hour ambient meteorologic (temperature, relative humidity, and
wind speed) data were obtained from the Texas Commission of
Environmental Quality for the study years. These data were used to
estimate the average hourly and daily ambient apparent temperature
level during the study period. The apparent temperature was calculated with the method used by O’Neill et al21 originally described by
Steadman22 and Kalkstein and Valimont.23
Statistical Methodology
The OHCA event, pollution, and meteorologic databases were analyzed by using a time-stratified case-crossover design coupled with
conditional logistic regression. The case-crossover design was first
introduced by Maclure24 and is used increasingly in the literature to
assess episodic events following short-term exposure to air pollution.3,4,7–10,25 In the case-crossover design, each individual experiencing a health event serves as his or her own reference; in other words,
individuals act as their own control. Ambient air pollution is used
as a proxy for personal exposure. The ambient air pollution concentrations at times when the study individual is not experiencing the
OHCA health event are the reference concentrations. The reference
concentrations are statistically compared with the concentrations during or around the time the study individual experienced the OHCA
health event. Conditional logistic regression is applied to estimate the
association of pollution and increased relative risk of the health event
while controlling for confounding factors.
In our application of the case-crossover design, we conducted an
exploratory sensitivity analysis with single lag models to examine
the association of air pollution and OHCA on 2 time scales: hour
and day. The hour or day of the individual OHCA event (depending
on the time scale being studied) was the initial exposure period (lag
0) considered for that case. For the hourly time scale analysis, we
examined the association at the hour of onset (lag 0 hour) and 1 to 8
hours before onset (lag 1, 2, 3, 4, 5, 6, 7, 8). For the daily time scale
analysis, we examined the association at the day of onset (lag 0 day)
and the 1 to 5 before onset (lag 1, 2, 3, 4, 5).
We then implemented constrained distributed lag models to estimate the cumulative effect over 2-hour average or 2-day average
increments (lag 0–1, lag 1–2, lag 2–3) for those pollutants where associations were indicated in our exploratory analysis.
Referent exposures, selected by time-stratified sampling, were the
exposures in the day (and hour for the hourly analysis) of the event
on all days falling within the same month and on the same day of the
week as the event. This reference period design has been shown to
limit the bias present due to patterns in air pollution.26 A conditional
logistic regression was used to estimate the relative risk associated
with each pollutant. We included apparent temperature in our conditional logistic regression model by using a nonparametric smoothing
spline of degree 3 with 4 knots optimally chosen.27–31
The EMS data, in which the call time acts as the time of the OHCA,
provided the ability for an analysis on the hourly as opposed to the
daily scale typically assessed. However, because both cardiac arrest
and pollutant data may have diurnal patterns, temporal confounding
must be considered.10 For our analysis of the hourly relationship, we
explored the impact of the cardiac arrest temporal pattern in confounding our understanding of the relationship between OHCA and
hourly air pollution (when an effect was found) by comparing the
OHCA/air pollution relationship when the temporal OHCA pattern
was constant to the finding from the full data set.
When a significant association between individual pollutants and
OHCA was found, we investigated potential confounding between
pollutants. We estimated correlations between pollutants on the daily
and hourly scale and also included pollutants as a covariate in the
model. The main concern was potential confounding between PM2.5
and ozone as indicated by previous researchers.32 When a relationship was found between OHCA and an air pollutant, we stratified
the analyses by age, sex, race, and season to examine the effects by
subgroup. The case-crossover logistic regression was conducted in
SAS version 9.3.33
Results
Figure 1 identifies the location of OHCA events for the 8-year
period (geo-masked for privacy). The characteristics of the
OHCA study group are shown in Table 1. Of the 11 677
qualified cases of OHCA during the study period, the largest
percentage of cases were individuals between the ages of 35
and 64 years, more of the cases were male (59%) than female
(41%), and most of the cases were of black individuals (46%),
followed by white individuals (35%) and Hispanic individuals (16%). The data indicate that 79% of the cases presented
with a preexisting condition, not necessarily cardiac related.
Because of the stressful conditions during the EMS call, the
designation of preexisting conditions by the victim or relatives is considered less reliable by the Houston EMS than the
other data. For this reason, stratification by preexisting condition was not explored in this study. To evaluate the impact of
the season, we broke the year into cold (November to March)
and warm season (April to October). During the study period,
55% of the cases were in the warm season and 45% were in
the cold season.
Statistics of the average hourly and daily pollutant levels
during the study period are listed in Table 2. Pearson correlation coefficients between pollutants and apparent temperature
on both time scales (hourly and daily) and each season (all,
warm, and cold) were calculated (Table 3). Note, correlations
vary between daily and hourly time scales because of different diurnal pollutant patterns. On the daily scale, the strongest
correlation was between carbon monoxide and nitrogen dioxide at 0.75, 0.72, and 0.79 for all year, warm, and cold season.
Ozone is most correlated with PM2.5 on the day scale during
the warm season (0.40, 0.37, and 0.26 for all year, warm, and
cold season). On the hourly scale, there is little to no correlation between ozone and PM2.5 during the warm season (0.01,
0.07, and −0.21 for all year, warm, and cold season).
Conditional logistic regression results for each pollutant on
the hourly and daily time frame are summarized on Table 4 and
Figure 2. The plots and the table offer different information.
The plots graphically show the change in effect estimates with
increasing lags for ozone and PM2.5, whereas the table shows
more limited ozone and PM2.5 lags information and includes
other pollutants.
Downloaded from http://circ.ahajournals.org/ by guest on February 27, 2015
1194 Circulation March 19, 2013
Figure 1. Locations of OHCA events between 2004 and 2011 in Houston, Texas. Subject locations have been randomly shifted to protect
confidentiality. OHCA indicates out-of-hospital cardiac arrest.
PM2.5 Results
The lag model results for PM2.5 on the daily analysis scale
indicate that a daily average increase of 6 µg/m3 in PM2.5 in the
2 days before onset (average of 1 and 2 days) was associated
with an increase of OHCA risk (1.046; 95% confidence interval [CI], 1.012–1.082). This was the strongest effect found.
There was no effect after 3 days (1.021; 95% CI, 0.991–1.051).
significant association between OHCA and ozone on lag 0 day
indicates that this association found on the hourly scale within
the day of onset is not simply reflecting the temporal cardiac
pattern. To further investigate confounding from the cardiac
temporal pattern, we compared the results of the same analysis limited to a time of day when the cardiac temporal pattern
was constant and found no change in the risk.
Ozone Results
Stratification and Sensitivity
The lag model results for ozone on the hourly analysis scale
indicate that each 20 ppb of ozone increase in the average
of the previous 1 to 3 hours was associated with an increase
OHCA risk (1.044; 95% CI, 1.004–1.085). This was the strongest effect found in the distributed lag model. No effect was
found after 3 hours.
Also included in Figure 2 are the results for the single lag
model for lag 0 day. The results indicate that an increase of
20 ppb of ozone for the 8-hour average daily maximum on
the day of the event was associated with an increased risk
of OHCA (1.038; 95% CI, 1.004–1.072). The finding of a
Analysis of stratification of the cases by the demographic
characteristics of the data (age, sex, and race) found that
the risk from exposure to ozone or PM2.5 is highest for men
(1.051; 95% CI, 1.006–1.097), those of black ethnicity
(1.053; 95% CI, 1.003–1.105), and >65 years of age (1.049;
95% CI, 1.000–1.100) (Figure 3). The apparent temperature
is most correlated with ozone on the hourly scale during the
cold season (0.20, 0.03, and 0.39 for all year, warm, and cold
season). The apparent temperature itself was not a significant
predictor for OHCA, nor did the inclusion of apparent
temperature change our conclusions related to the pollutants.
Downloaded from http://circ.ahajournals.org/ by guest on February 27, 2015
Ensor et al Out-of-Hospital Cardiac Arrest and Pollution 1195
Table 1. Study Population Characteristics of OHCA Events in
Houston, Texas, From 2004 to 2011
Total
11 677
Preexisting condition
9196 (79)
Age
Mean
64 SD,16.81
18 to <35
576 (5)
35–64
5153 (44)
65–74
2244 (19)
75+
3704 (32)
Sex
Female
4776 (41)
Male
6901 (59)
Race
White
4065 (35)
Black
5338 (46)
Hispanic
1875 (16)
Other
399 (3)
Season
Warm (April to October)
6411 (55)
Cold (November to March)
5266 (45)
Values are n (%).OHCA indicates out-of-hospital cardiac arrest; and SD,
standard deviation.
Discussion
We find consistent evidence of an association between OHCA
and exposure to ozone in Houston, Texas at short time scales
up to 3 hours in duration and also at the daily level on the day
of the event. For exposure to PM2.5, an association is found for
2 days before the event. Other pollutants were not found to
impact the occurrence of OHCA.
Our findings add to the significant literature relating OHCA
with PM2.5, where findings across studies are inconsistent.
Furthermore, we add to the small but growing scientific conversation relating OHCA and ozone. Finally, we bring the most
comprehensive data set to date to this literature, in terms of
duration of the study, number of pollution monitors included,
and the number of OHCA events studied. The implications
of this work are improved health policy and action with the
objective of reducing the number of annual OHCA currently
at ≈300 000 in the nation and 1460 in Houston.
Association Between PM2.5 and OHCA
The association between PM2.5 and OHCA varies across
studies, which is due in large part to the variation in study
design. A detailed synthesis of recent studies is provided
in Raun and Ensor34 for both PM2.5 and ozone. Some of the
key features that varied across studies included the number
of monitors used, the area covered, sample size of cases, the
designation of health end point, the comorbidities studied,
the method of pollution measurement, the composition of
particulates, and the level of ambient concentration. The
early studies, which did not find an association, had fewer
OHCA events, lower PM2.5 concentrations, and different
PM2.5 composition than the later studies that did find an
association.3,7–11
Association Between Ozone and OHCA
Although a few studies have examined the link between ozone
and OHCA, there is growing evidence of a pathophysiological
link. In the effects seen in animal toxicology studies after
human ozone exposure, as well, researchers have found a
reduction in serum tocopherol (free radical scavenger),35 an
increase in the gradient of alveolar-to-arterial Po2 potentially
due to alveolar-arterial oxygen impairment,36 and, most
recently, changes in several proinflammatory cytokines in
blood.12 The lack of investigation of the association between
ozone and OHCA may stem from practical considerations
such as data limitations. In some locations, ozone is only
monitored periodically. When the association is investigated,
the lack of significant findings may be a product of the
additional complexity of controlling accurately for the impact
from temperature. Ozone is clearly found more often at higher
Table 2. Description of data
Variable
Percentile
No. of
Monitors
% of Missing
Data
Mean (SD)
5%
25%
50%
75%
95%
IQR
PM2.5, μg/m hourly
12
0
11.42 (5.89)
3.87
7.34
10.3
14.37
22.8
7.03
O3, ppb hourly
47
0
25.52 (16.14)
4.3
13.23
22.92
34.61
57.25
21.38
6.87
3
NO2, ppb hourly
22
1
9.16 (5.76)
2.84
4.96
7.52
11.84
21.18
SO2, ppb hourly
13
1
1.97 (3.23)
0.28
0.75
1.45
2.48
5.18
1.73
CO, ppb hourly
12
1
281.91 (202.45)
121.23
171.94
225.09
315.75
632.76
143.81
Apparent temperature daily, °F
15
0
73.37 (17.39)
42.58
59.70
75.89
88.89
95.53
29.19
PM2.5, μg/m daily
12
0
11.42 (4.73)
5.50
8.18
10.45
13.71
20.96
5.52
NO2, ppb daily
22
0
9.11 (4.17)
3.51
6.01
8.41
11.66
16.87
5.65
SO2, ppb daily
13
0
1.96 (2.38)
0.44
0.97
1.66
2.55
4.27
1.57
CO, ppb daily
12
0
279.90 (130.90)
139.89
194.69
249.89
332.36
526.16
137.67
3
CO indicates carbon monoxide; IQR, interquartile range; PM2.5, fine particulate matter with an aerodynamic diameter <2.5 µm; NO2, nitrogen dioxide; O3, ozone; SD,
standard deviation; and SO2, sulfur dioxide.
Downloaded from http://circ.ahajournals.org/ by guest on February 27, 2015
1196 Circulation March 19, 2013
Table 3. Pearson correlation coefficients between pollutants and apparent temperature.
PM2.5, µg/m3
O3, ppb
NO2, ppb
SO2, ppb
CO, ppb
AT, °F
0.24
−0.33
0.05
0.34
0.22
0.11
−0.32
0.20
All pollution data
PM2.5, µg/m3
O3, ppb
0.01
0.40
NO2, ppb
0.08
0.24
SO2, ppb
0.07
0.08
0.11
0.23
CO, ppb
0.21
0.23
0.75
0.22
AT, °F
0.31
0.14
−0.57
−0.14
0.71
−0.49
0.08
−0.08
Hourly
−0.24
−0.25
Daily
Warm/cold season (hour)
PM2.5, µg/m3
O3, ppb
−0.21
0.07
0.44
0.11
0.46
0.07
−0.49
0.12
−0.46
0.39
0.72
−0.32
NO2, ppb
0.27
−0.21
SO2, ppb
0.06
0.12
0.05
CO, ppb
0.34
−0.23
0.72
0.05
AT, °F
0.11
0.03
−0.50
−0.04
−0.28
0.29
0.20
0.33
0.19
0.09
0.18
0.15
0.47
0.79
−0.32
0.21
0.13
Cold (Nov to Mar)
0.00
−0.20
Warm (April to Oct)
Warm/cold season (day)
PM2.5, µg/m3
0.26
O3, ppb
0.37
NO2, ppb
0.20
0.61
SO2, ppb
0.09
0.13
0.13
CO, ppb
0.27
0.42
0.72
0.15
AT, °F
0.15
−0.33
−0.53
−0.10
0.43
0.19
Cold (Nov to Mar)
−0.09
−0.13
−0.26
Warm (April to Oct)
Apr indicates April; AT, apparent temperature; CO, carbon monoxide; Mar, March; NO2, nitrogen dioxide; O3, ozone; Oct, October; Nov, November; PM2.5, fine
particulate matter with an aerodynamic diameter <2.5 µm; and SO2, sulfur dioxide.
temperatures, and an increased risk of OHCA is closely tied
to the combined effect. Finally, our results indicate that the
association may be more readily found at the hourly level over
the daily, with the daily level the more frequently studied time
frame.
Examining 3 recent large studies in comparison with our
findings, we find differences in 2 of the studies3,7 regarding the
number of cases, the number of monitors, the specific health
end point considered, and the magnitude and variation in pollution levels studied.34 In the third study, Silverman et al10 of
New York City (n=8216) found an increased risk (1.045; 95%
CI, 0.991–1.1) for a daily average increase of 20 ppb. Our
study design is most similar to Silverman et al10; both studies
have a large number of cases extracted from an EMS 911 database, limited exposure concentration uncertainty, and similar
ozone interquartile range. The results found in New York City
and Houston are consistent with findings from an important
case-crossover study with a more encompassing health end
point. Stafoggia et al30 examined susceptibility factors to
ozone mortality. Of interest to our objective is their examination of ozone-related mortality in those with preexisting cardiovascular conditions. The researchers estimated an increase
risk (1.093; 95% CI, 1.044–1.145) in mortality for a 20-ppb
increase in the daily 8-hour ozone running maximum average.
Given the comparability between the study of Houston and
New York City and the corroborating study by Stafoggia et
al,30 the current results of the comparable studies support the
likelihood that there is an increased risk of OHCA with exposure to ozone.
Limitations
A potential limitation of this study is selection bias from
the exclusion of cases in which chest compressions were
not initiated because the adults were considered dead on
arrival. Resuscitation was withheld if the individual was dead
on arrival as defined by decapitation, rigor mortis, dependent lividity, decomposition, incineration or obvious mortal
wounds, absence of any signs of life (pulse, respirations, or
any spontaneous movement) on EMS arrival associated with
a penetrating head injury (gunshot wound, stab, etc), or penetrating extremity injury with obvious exsanguination, absence
of any signs of life (pulse, respirations, or any spontaneous
movement) for >5 minutes associated with a penetrating
injury to the chest or abdomen and a >10-minute transport
time to a trauma center, or the absence of any signs of life
(pulse, respirations, or any spontaneous movement) associated with blunt trauma. However, the large size of this study
minimizes risks from selection bias.
Downloaded from http://circ.ahajournals.org/ by guest on February 27, 2015
Ensor et al Out-of-Hospital Cardiac Arrest and Pollution 1197
Table 4. Percentage Change in Risk of OHCA for an Interquartile Increase in Air Pollutants
PM2.5
IQR 6 µg/m3
% (95% CI)
O3
IQR 20 ppb
% (95% CI)
0
2.7 (−0.3 to 5.8)
3.8 (0.4 to 7.2)
0.9 (−3.0 to 5.0)
−0.2 (−2.1 to 1.7)
1.8 (−0.9 to 4.6)
1
3.5 (0.5 to 6.6)
1.8 (−1.4 to 5.2)
−0.7 (−4.4 to 3.0)
−1.2 (−3.2 to 0.8)
−0.2 (−2.8 to 2.4)
2
3.7 (0.7 to 6.8)
2.7 (−0.6 to 6.1)
−0.4 (−4.1 to 3.4)
−0.7 (−2.9 to 1.5)
0.9 (−1.7 to 3.6)
3
2.1 (−0.9 to 5.1)
−0.6 (−3.8 to 2.7)
0.9 (−2.8 to 4.7)
−1.3 (−3.3 to 0.7)
0.4 (−2.2 to 3.1)
4
0.2 (−2.7 to 3.2)
-1.2 (−4.3 to 2.1)
0.3 (−3.4 to 4.1)
−0.9 (−2.6 to 0.8)
0.3 (−2.4 to 3.0)
0–1
3.9 (0.5 to 7.4)
3.6 (0.0 to 7.4)
−0.1 (−4.3 to 4.3)
−0.9 (−3.0 to 1.3)
0.9 (−2.1 to 4.0)
1–2
4.6 (1.2 to 8.2)
3.0 (−0.6 to 6.8)
−0.8 (−4.9 to 3.5)
−1.3 (−3.7 to 1.1)
0.4 (−2.5 to 3.4)
0
0.9 (−1.4 to 3.4)
3.7 (−0.1 to 7.7)
−0.1 (−0.6 to 0.4)
0.4 (−0.2 to 1.0)
0.0 (0.0 to 0.0)
1
1.1 (−1.3 to 3.5)
4.2 (0.4 to 8.2)
0.0 (−0.5 to 0.5)
0.0 (−0.7 to 0.8)
0.0 (0.0 to 0.0)
2
1.1 (−1.2 to 3.5)
4.6 (0.8 to 8.7)
0.0 (−0.5 to 0.5)
0.2 (−0.5 to 0.9)
0.0 (0.0 to 0.0)
3
0.3 (−2.0 to 2.7)
4.0 (0.2 to 8.0)
0.1 (−0.3 to 0.6)
0.2 (−0.6 to 0.9)
0.0 (0.0 to 0.0)
4
0.9 (−1.5 to 3.3)
3.4 (−0.5 to 7.4)
0.2 (−0.3 to 0.7)
0.0 (−0.8 to 0.7)
0.0 (0.0 to 0.0)
Lag
NO2
IQR 6 ppb
% (95% CI)
SO2
IQR 2 ppb
% (95% CI)
CO
IQR 141 ppb
% (95% CI)
Daily lag
Hourly lag
O3 was based on an 8-hour maximum. Statistics reflect the adjustment for apparent temperature. CI indicates confidence interval; CO, carbon monoxide; IQR,
interquartile range; NO2, nitrogen dioxide; O3, ozone; OHCA, out-of-hospital cardiac arrest; PM2.5, fine particulate matter with an aerodynamic diameter <2.5 µm; and
SO2, sulfur dioxide.
Another limitation of the study is the absence of stratification by preexisting conditions and personal risk factors owing
to the lack of this information. Finally, the exposure concentrations in the study are limited to the use of the average
pollutant concentration across the city over the use of more
local pollutant concentrations. This is especially true when
the study area is large and the pollutant varies spatially. We
chose to use the average concentration rather than potentially
misclassifying the associated reference concentrations if the
individual experienced the OHCA in a location not representative of his usual exposure. This limitation is inherent in the
case-crossover study design.
Future Research
Although this study identifies an association between PM2.5 and
ozone air pollution and OHCA, future research to better define
the exposure time period associated with triggering an OHCA
is needed. Epidemiological studies have found the time to trigger a cardiac event from exposure to PM2.5 or ozone ranges from
the day or previous day of onset to hours before onset.4,7,9,10 Part
of this inconsistent range of time to trigger is due to exposure
time misclassification. This could be better handled by addressing the uncertainty in combining the disparate data sets such as
OHCA recorded at the minute and continuous across space and
air pollution data recorded hourly at fixed locations.37
Figure 2. Forest plot of Houston relative risk of OHCA associated with 20-ppb
increase in ozone or 6 µg/m3 increase in
PM2.5. CI indicates confidence interval;
EMS, emergency medical service; OHCA
indicates out-of-hospital cardiac arrest;
PM2.5, fine particulate matter with an aerodynamic diameter <2.5 µm; and RR, relative risk..
Downloaded from http://circ.ahajournals.org/ by guest on February 27, 2015
1198 Circulation March 19, 2013
Figure 3. Forest plot of relative risk of
OHCA associated per an interquartile range
increase in the average of 1- to 3-hour
lagged ozone and 1- to 2-day lagged PM2.5
by age, ethnicity, sex, and season. CI
indicates confidence interval; OHCA, outof-hospital cardiac arrest; PM2.5, fine particulate matter with an aerodynamic diameter
<2.5 µm; and RR, relative risk.
Acknowledgments
The authors thank the anonymous reviewers and associate editor for
comments leading to a greatly improved version of this manuscript.
Furthermore, the authors thank Laura Campos and Jiao Li for their
assistance.
Sources of Funding
This work was funded by Houston Endowment and the City of
Houston.
Disclosures
None.
References
1. McNally B, Robb R, Mehta M, Vellano K, Valderrama AL, Yoon PW, Sasson C, Crouch A, Perez AB, Merritt R, Kellermann A; Centers for Disease
Control and Prevention. Out-of-hospital cardiac arrest surveillance—Cardiac Arrest Registry to Enhance Survival (CARES), United States, October 1, 2005–December 31, 2010. MMWR Surveill Summ. 2011;60:1–19.
2. Bell ML, McDermott A, Zeger SL, Samet JM, Dominici F. Ozone and
short-term mortality in 95 US urban communities, 1987-2000. JAMA.
2004;292:2372–2378.
3. Forastiere F, Stafoggia M, Picciotto S, Bellander T, D’Ippoliti D, Lanki T.
A case-crossover analysis of out-of hospital coronary deaths and air pollution in Rome, Italy. Am. J Respir Crit Care Med. 2005;172:1549–1555.
4. Peters A, Dockery DW, Muller JE, Mittleman MA. Increased particulate
air pollution and the triggering of myocardial infarction. Circulation.
2001;103:2810–2815.
5. Pope CA III, Dockery DW. Health effects of fine particulate air pollution:
lines that connect. J Air Waste Manag Assoc. 2006;56:709–742.
6. Brook RD, Rajagopalan S, Pope III A, Brook JR, Bhatnagar A, Diez-Roux
AV, Holguin F, Hong Y, Luepker RV, Mittleman MA, Peters A, Siscovick
D, Smith SC, Whitsel L, Kaufman JD. Particulate matter air pollution
and cardiovascular disease: an update to the scientific statement from the
American Heart Association. Circulation. 2010;121:2331–2378.
7. Dennekamp M, Akram M, Abramson MJ, Tonkin A, Sim MR, Fridman
M, Erbas B. Outdoor air pollution as a trigger for out-of-hospital cardiac
arrests. Epidemiology. 2010;21:494–500.
8. Levy D, Sheppard L, Checkoway H, Kaufman J, Lumley T, Koenig J, Siscovick D. A case-crossover analysis of particulate matter air pollution and
out-of-hospital primary cardiac arrest. Epidemiology. 2001;12:193–199.
9. Rosenthal F, Carney JP, Olinger ML. Out-of-hospital cardiac arrest and
airborne fine particulate matter: a case-crossover analysis of emergency
medical services data. Environ Health Perspect. 2008;116:631–636.
10. Silverman RA, Ito K, Freese J, Kaufman BJ, De Claro D, Braun J, Prezant
DJ. Association of ambient fine particles with out-of-hospital cardiac arrests in New York City. Am J Epidemiol. 2010;172:917–923.
11. Sullivan J, Ishikawa N, Sheppard L, Siscovick D, Checkoway H, Kaufman
J. Exposure to ambient fine particulate matter and primary cardiac arrest
among persons with and without clinically recognized heart disease. Am J
Epidemiol. 2003;157:501–509.
12. Devlin RB, Duncan KE, Jardim M, Schmitt MT, Rappold AG, Diaz-Sanchez D. Controlled exposure of healthy young volunteers to ozone causes
cardiovascular effects. Circulation. 2012;126:104–111.
13. Dockery DW, Luttmann-Gibson H, Rich DQ, Link MS, Mittleman MA, Gold
DR, Koutrakis P, Schwartz JD, Verrier RL. Association of air pollution with
increased incidence of ventricular tachyarrhythmias recorded by implanted
cardioverter defibrillators. Environ Health Perspect. 2005;113:670–674.
14. Gold DR, Litonjua A, Schwartz J, Lovett E, Larson A, Nearing B, Allen
G, Verrier M, Cherry R, Verrier R. Ambient pollution and heart rate variability. Circulation. 2000;101:1267–1273.
15.Peters A, Döring A, Wichmann HE, Koenig W. Increased plasma viscosity during an air pollution episode: a link to mortality? Lancet.
1997;349:1582–1587.
16. Peters A, Liu E, Verrier RL, Schwartz J, Gold DR, Mittleman M, Baliff J,
Oh JA, Allen G, Monahan K, Dockery DW. Air pollution and incidence of
cardiac arrhythmia. Epidemiology. 2000;11:11–17.
17. Riediker M, Cascio WE, Griggs TR, Herbst MC, Bromberg PA, Neas L,
Williams RW, Devlin RB. Particulate matter exposure in cars is associated
with cardiovascular effects in healthy young men. Am J Respir Crit Care
Med. 2004;169:934–940.
18. Srebot V, Gianicolo EA, Rainaldi G, Trivella MG, Sicari R. Ozone and
cardiovascular injury. Cardiovasc Ultrasound. 2009;7:30.
19. Jacobs I, Nadkarni V, Bahr J, Berg RA, Billi JE, Bossaert L, Cassan P,
Coovadia A, D’Este K, Finn J, Halperin H, Handley A, Herlitz J, Hickey
R, Idris A, Kloweck W, Larkin GL, Mancini ME, Mason P, Mears G, Monsieur K, Montgomery W, Morley P, Nichol G, Nolan J, Okada K, Perlman
J, Shuster M, Steen PA, Sterz F, Tibballs J, Timerman S, Truitt T, Zideman
D. ILCOR Task Force on Cardiac Arrest and Cardiopulmonary resuscitation Outcomes, Cardiac arrest and cardiopulmonary resuscitation outcome
reports: update and simplification of the Utstein templates for resuscitation registries. Resuscitation. 2004;63:233–249.
20. Environmental Protection Agency. List of Designated Reference and
Equivalent Methods. MD-D205-03. Research Triangle Park, NC: U.S. Environmental Protection Agency; 2012. http://www.epa.gov/ttnamti1/files/ambient/criteria/reference-equivalent-methods-list.pdf Accessed July 1, 2012.
Downloaded from http://circ.ahajournals.org/ by guest on February 27, 2015
Ensor et al Out-of-Hospital Cardiac Arrest and Pollution 1199
21.O’Neill MS, Zanobetti A, Schwartz J. Modifiers of the tempera
ture and mortality association in seven US cities. Am J Epidemiol.
2003;157:1074–1082.
22. Steadman RG. The assessment of sultriness. Part I: a temperature-humidity index based on human physiology and clothing science. J Appl Meteorol. 1979;18:861–873.
23. Kalkstein LS, Valimont KM. An evaluation of summer discomfort in the
United States using a relative climatological index. Bull Am Meteorol Soc.
1986;67:842–848.
24. Maclure M. The case-crossover design: a method for studying transient
effects on the risk of acute events. Am J Epidemiol. 1991;133:144–153.
25. Wellenius GA, Burger MR, Coull BA, Schwartz J, Suh HH, Koutrakis P,
Schlaug G, Gold DR, Mittleman MA. Ambient air pollution and the risk
of acute ischemic stroke. Arch Intern Med. 2012;172:229–234.
26. Janes H, Sheppard L, Lumley T. Case-crossover analyses of air pollution
exposure data: referent selection strategies and their implications for bias.
Epidemiology. 2005;16:717–726.
27.Basu R, Samet JM. Relation between elevated ambient temperature
and mortality: a review of the epidemiologic evidence. Epidemiol Rev.
2002;24:190–202.
28. Braga AL, Zanobetti A, Schwartz J. The time course of weather-related
deaths. Epidemiology. 2001;12:662–667.
29. Curriero FC, Heiner KS, Samet JM, Zeger SL, Strug L, Patz JA. Temperature and mortality in eleven cities of the Eastern United States. Am J
Epidemiol. 2002;155:80–87.
30. Stafoggia M, Forastiere F, Agostini D, Biggeri A, Bisanti L, Cadum E,
Caranci N, de’ Donato F, De Lisio S, De Maria M, Michelozzi P, Miglio
R, Pandolfi P, Picciotto S, Rognoni M, Russo A, Scarnato C, Perucci CA.
Vulnerability to heat-related mortality: a multicity, population-based,
case-crossover analysis. Epidemiology. 2006;17:315–323.
31. Eilers PH, Marx BD. Flexible smoothing with B-splines and penalties.
Statistical Science 1996;11:89–121.
32. Bell ML, Kim JY, Dominici F. Potential confounding of particulate matter
on the short-term association between ozone and mortality in multisite
time-series studies. Environ Health Perspect. 2007;115:1591–1595.
33. SAS/STAT software. Version [9.3] of the SAS System for [Unix]. Cary,
NC: SAS Institute Inc; 2010. http://www.sas.com/presscenter/guidelines.
html Accessed July 16, 2012.
34. Raun L, Ensor KB. Final Report. Association of out-of-hospital cardiac
arrest with exposure to fine particulate and ozone ambient air pollution from case-crossover analysis results: are the standards protective?
Houston, TX: James A. Baker III Institute for Public Policy; 2012. http://
bakerinstitute.org/publications/HPF-pub-RaunEnsorParticulateExposure-101212.pdf Accessed August 2012.
35. Foster WM, Stetkiewicz PT. Regional clearance of solute from the respiratory epithelia: 18-20 h postexposure to ozone. J Appl Physiol.
1996;81:1143–1149.
36. Gong H Jr, Wong R, Sarma RJ, Linn WS, Sullivan ED, Shamoo DA, Anderson KR, Prasad SB. Cardiovascular effects of ozone exposure in human
volunteers. Am J Respir Crit Care Med. 1998;158:538–546.
37. Young LJ, Gotway CA, Yang J, Kearney G, DuClos C. Linking health
and environmental data in geographical analysis; It’s so much more
than centroids. Spatial and Spatio-temporal. Epidemiology. 2009;1:
73–84.
Clinical Perspective
The implications of this work are improved health policy and action with the objective of reducing the number of annual
out-of-hospital cardiac arrest currently at ≈300 000 in the nation and 1460 in Houston, Texas. We find consistent evidence
of an association between out-of-hospital cardiac arrest and exposure to ozone in Houston at short time scales up to 3 hours
in duration and also at the daily level on the day of the event. For exposure to fine particulate matter an association is found
for 2 days before the event. Other pollutants were not found to impact the occurrence of out-of-hospital cardiac arrest. Our
findings add to the significant literature relating out-of-hospital cardiac arrest and fine particulates. Furthermore, we add to
the small but growing scientific conversation relating out-of-hospital cardiac arrest and ozone. Finally, we bring the most
comprehensive data set to date to this literature, in terms of the duration of the study, the number of pollution monitors
included, and the number of out-of-hospital cardiac arrest events studied.
Downloaded from http://circ.ahajournals.org/ by guest on February 27, 2015
A Case-Crossover Analysis of Out-of-Hospital Cardiac Arrest and Air Pollution
Katherine B. Ensor, Loren H. Raun and David Persse
Circulation. 2013;127:1192-1199; originally published online February 13, 2013;
doi: 10.1161/CIRCULATIONAHA.113.000027
Circulation is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
Copyright © 2013 American Heart Association, Inc. All rights reserved.
Print ISSN: 0009-7322. Online ISSN: 1524-4539
The online version of this article, along with updated information and services, is located on the
World Wide Web at:
http://circ.ahajournals.org/content/127/11/1192
Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published
in Circulation can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial
Office. Once the online version of the published article for which permission is being requested is located,
click Request Permissions in the middle column of the Web page under Services. Further information about
this process is available in the Permissions and Rights Question and Answer document.
Reprints: Information about reprints can be found online at:
http://www.lww.com/reprints
Subscriptions: Information about subscribing to Circulation is online at:
http://circ.ahajournals.org//subscriptions/
Downloaded from http://circ.ahajournals.org/ by guest on February 27, 2015
2/6/2015
Environmental Health | Full text | Using community level strategies to reduce asthma attacks triggered by outdoor air pollution: a case crossover analysis
2.71
Research
Using community level strategies to reduce asthma attacks triggered by outdoor air pollution: a
case crossover analysis
Loren H Raun1 2 * , Katherine B Ensor 1 and David Persse 3 4
* C orresponding author: Loren H Raun raun@rice.edu
1
Department of Statistics, Rice University, 6100 Main Street, Houston, TX 77005, USA
2
C ity of Houston Health and Human Services Bureau of Pollution C ontrol and Prevention, 7411 Park Place Blvd, Houston, TX 77087, USA
3
C ity of Houston Emergency Medical Services, 600 Jefferson Suite 800, Houston, TX 77002, USA
4
Department of Medicine, Baylor C ollege of Medicine, One Baylor Plaza Houston, Houston, TX 77030, USA
For all author emails, please log on.
Environmental Health 2014, 13:58
doi:10.1186/1476-069X-13-58
The electronic version of this article is the complete one and can be found online at: http://www.ehjournal.net/content/13/1/58
Received:
Accepted:
Published:
10 February 2014
2 July 2014
11 July 2014
© 2014 Raun et al.; licensee BioMed C entral Ltd.
This is an Open Access article distributed under the terms of the C reative C ommons Attribution License (http://creativecommons.org/licenses/by/4.0), which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The C reative C ommons Public Domain
Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Abstract
Background
Evidence indicates that asthma attacks can be triggered by exposure to ambient air pollutants, however, detailed pollution information is missing from asthma action
plans. Asthma is commonly associated with four criteria pollutants with standards derived by the United States Environmental Protection Agency. Since multiple
pollutants trigger attacks and risks depend upon city-specific mixtures of pollutants, there is lack of specific guidance to reduce exposure. Until multi-pollutant
statistical modeling fully addresses this gap, some guidance on pollutant attack risk is required. This study examines the risks from exposure to the asthma-related
pollutants in a large metropolitan city and defines the city-specific association between attacks and pollutant mixtures. Our goal is that city-specific pollution risks be
incorporated into individual asthma action plans as additional guidance to prevent attacks.
Methods
C ase-crossover analysis and conditional logistic regression were used to measure the association between ozone, fine particulate matter, nitrogen dioxide, sulfur
dioxide and carbon monoxide pollution and 11,754 emergency medical service ambulance treated asthma attacks in Houston, Texas from 2004-2011. Both single
and multi-pollutant models are presented.
Results
In Houston, ozone and nitrogen dioxide are important triggers (RR = 1.05; 95% C I: 1.00, 1.09), (RR = 1.10; 95% C I: 1.05, 1.15) with 20 and 8 ppb increase in
ozone and nitrogen dioxide, respectively, in a multi-pollutant model. Both pollutants are simultaneously high at certain times of the year. The risk attributed to these
pollutants differs when they are considered together, especially as concentrations increase. C umulative exposure for ozone (0-2 day lag) is of concern, whereas for
nitrogen dioxide the concern is with single day exposure. Persons at highest risk are aged 46-66, African Americans, and males.
Conclusions
Accounting for cumulative and concomitant outdoor pollutant exposure is important to effectively attribute risk for triggering of an asthma attack, especially as
concentrations increase. Improved asthma action plans for Houston individuals should warn of these pollutants, their trends, correlation and cumulative effects. Our
Houston based study identifies nitrogen dioxide levels and the three-day exposure to ozone to be of concern whereas current single pollutant based national
standards do not.
Keywords: Asthma; Air pollution; Risk; Ozone; Nitrogen dioxide; Action plans
Background
Asthma is a serious and sometimes life-threatening chronic respiratory disease that affects almost 25 million Americans and costs the nation $56 billion per year
[1]. In 2009, 3.3 deaths per 100,000 people were attributed to asthma and there were 1.9 million asthma related emergency department visits [2,3]. Asthma
prevalence increased from 7.3% in 2001 to 8.4% in 2010, when 25.7 million persons were diagnosed with asthma [4].
Although the association between air pollutants and asthma attacks is well documented [5,6], the lack of specific guidance in asthma intervention programs to
reduce exposure beyond broad nationally set air quality alerts may severely limit effectiveness of the air quality alert approach. In a recent review of the literature
http://www.ehjournal.net/content/13/1/58
1/10
2/6/2015
Environmental Health | Full text | Using community level strategies to reduce asthma attacks triggered by outdoor air pollution: a case crossover analysis
two effective intervention methods were identified, namely self-management education and more general comprehensive home-based multi-trigger reduction
interventions [7]. However, in their umbrella review the authors were unable to identify any reviews related to the effectiveness of alerts for air quality.
C urrently air quality alerts in the United States address pollutants in isolation from each other, but individuals are exposed to a mixture of pollutants. A barrier to
pollution specific asthma education is that detailed pollutant-asthma guidance is contingent upon the development of methods to address multi-pollutant mixtures
[4]. Asthma attacks are triggered by multiple Environmental Protection Agency (EPA) criteria pollutants, namely ozone (O 3 ), nitrogen dioxide (NO 2 ), particulate
matter (PM), sulfur dioxide (SO 2 ), and carbon monoxide (C O) [8-12]. An additional complication is that cities have different mixtures of these pollutants. The need
to address multiple pollutants in the criteria pollutant review and standard setting process was identified in 2004 [13]. Researchers focused on developing national
criteria pollutant standards in the multi-pollutant exposure context have made limited progress over the last decade [14-19]. Until multi-pollutant statistical modeling
fully addresses this gap, more specific guidance to mitigate risk from exposure to pollutants is required. One solution is to develop guidance on a city-specific basis,
especially in highly polluted cities.
To demonstrate the city-specific guidance, asthma risks were evaluated using an environmental public health tracking model framework [20] for Houston, Texas.
With this model, our focus is on tracking the “association” within a city. The concern is the pollutant mixture, not the individual pollutants. Moreover, regular tracking
of the association within a city can be used to evaluate trends reflecting the effectiveness of regulatory measures, interventions, or identify changing potency of air
pollutants [20].
Houston is the fourth largest city in the United States and has recognized air pollution problems [21]. The Houston-Galveston region has an extensive air-monitoring
network. Furthermore, Houston emergency medical service (EMS) responded to 11,754 emergency calls for asthma from 2004 through 2011. Each EMS response
cost approximately $1,400 for a total estimated cost of $17 million [22]. C urrently, asthma action plans vary in Houston. Some plans address only medications and
lung function and others extend to a check box of asthma symptom triggers. For example, the Houston Independent School District asthma action plan includes air
pollution as an asthma symptom trigger but more detailed information is needed regarding which pollutants are of concern, their trends, correlation, and cumulative
effects.
A case-crossover analysis was used to measure the association between asthma associated criteria pollutant levels and EMS calls for asthma attacks from 2004 to
2011 in Houston. The pollutants examined were daily ozone, PM with an aerodynamic diameter less than 2.5 microns (PM2 .5 ), NO 2 , SO 2 and C O. In our analysis we
first developed single, and then multi-pollutant models of the association. We also segmented the overall model by time to examine trends and by demographics to
examine effect modification. We then analyzed the association based on ranges of concentration to formulate concentration-risk curves. This analysis is followed by
a discussion of recommendations for asthma action plans for asthmatics in Houston. This research uncovers some critical new data that may be helpful in
developing guidance on a city-specific basis.
Methods
Study design and setting
The data used in this study were obtained from the Houston Fire Department EMS call database segmenting by two fields, working assessment and treatment. The
selection for working assessment was asthma and for treatment administered was nebulized albuterol (n = 11,754). The working assessment input is determined by
EMS personnel and identifies the primary reason for treatment. The data were obtained during the eight-year period (2004-2011). Rice University and Baylor
C ollege of Medicine Institutional Review Boards approved all data-collecting procedures for human subjects.
Participant data
Included in the study were all patients older than two years of age. Patients two years and younger were excluded from the study because the diagnosis is less
reliable. If EMS responded to the same person multiple times within two weeks, the first call was retained and subsequent calls were removed from the database for
analysis [20,23]. There were no other exclusion criteria. The EMS database consists of data collected according to National EMS Information Systems guidelines
[24]. In addition to recording the working assessment and the administration of albuterol, the database also includes the following relevant information: time of call,
location, age, sex, and race of patient.
Ambient air quality, meteorological, and other data
Ambient pollution concentration data were obtained from the Texas C ommission of Environmental Quality (TC EQ). In this analysis, hourly data from 35 ozone, 13
NO 2 , nine C O, nine PM2 .5 , and eight SO 2 monitors in the Houston Metropolitan Area were used. The daily average values of ozone, NO 2 , C O, PM2 .5 , and SO 2 were
calculated across monitors. Researchers commonly use the average concentrations across monitors to obtain one average pollution level in case cross-over
analysis [25-28]. The use of the average, over other spatial exposure estimation methods (e.g, inverse distance or kriging), is preferred when the activity patterns
of the subject are not known or cannot be reasonably assumed to be similar on case and control periods.
The daily maximum 8 hour running mean was also calculated for ozone. The number of air monitors measuring a specific pollutant changed through the study years
as monitors went on and off line. However, more than 99% of the time at least one monitor was operating for each pollutant. All air pollution data were collected
using EPA federal reference methods [29] and validated by the TC EQ.
Ambient apparent temperature was used to control for meteorological conditions. The apparent temperature was calculated with the method used by O’Neill et al.
[30] originally described by Steadman and Kalkstein and Valimont [30-32]. Aeroallergen data available for the study area are in the form of daily pollen and mold
spore counts collected by the Houston Department of Health and Human Services at a single location using a Burkard Spore Trap sampling at 10 liters/minute [33].
During the study period, these data were largely incomplete. The percent of complete daily tree, grass and weed pollen data was 42.1%, 53.5% and 61.0%,
respectively, and the percent of complete daily ascomycetes and basidiomycetes spore data was 61.0% and 62.6%. Therefore, these data could not be included in
the analysis. Weeks with reported influenza and major U.S. holidays were flagged with an indicator and incorporated in the model [31].
Statistical analysis
The data were analyzed using a time-stratified case-crossover design coupled with conditional logistic regression [34]. All tests are conducted at a significance level
of 0.05. The case-crossover design was first introduced by Maclure (1991) and is increasingly used in studies to assess episodic events following short-term
exposure to air pollution [25-27,35-39]. In the case-crossover design each individual experiencing a health event serves as his or her own reference, in other
words, individuals act as their own control. Ambient air pollution was used as a proxy for personal exposure. The ambient air pollution concentrations at times when
the study individual is not experiencing the asthma attack are the reference concentrations. Referent exposures, selected by time stratified sampling, were the
exposures on all days falling within the same month and on the same day of the week as the event [40]. This reference period design has been shown to limit bias
caused by patterns in air pollution [40]. C onditional logistic regression was applied to estimate the association of pollution and increased relative risk of the health
event while controlling for confounding factors.
Following exploratory data analysis, the association of EMS calls for asthma attacks and the potential confounding variables (apparent temperature, holidays and
http://www.ehjournal.net/content/13/1/58
2/10
2/6/2015
Environmental Health | Full text | Using community level strategies to reduce asthma attacks triggered by outdoor air pollution: a case crossover analysis
influenza season) was examined. The form and lags of these variables showing the strongest association with EMS calls for asthma attacks according to the lowest
Akaike Information C riterion (AIC ) score were included universally in the pollution models. All further modeling included the confounding variables.
Sensitivity analysis with pollution lag models was conducted to examine the association of single air pollutants and asthma attacks. The association at the day of
onset (lag 0 day), one to three days prior to onset (lag 1, 2, 3) and constrained distributed lag models (0-1 day, 1-2 day, and 0-2 day) were examined.
Significant associations found in the exploratory single pollutant analysis were combined in a multi-pollutant model. Interaction terms were also explored.
Regression diagnostics were used to define the final multi-pollutant model. The final multi-pollutant model was used to examine stratification, time segments, and
concentration-risk curves. The case-crossover logistic regression was conducted in SAS version 9.3 [41].
Results
Exploratory data analysis
A breakdown of the EMS calls for asthma attacks occurring during the study period are presented in Table 1 by age group, sex, race, season and year. The age
group of patients less than 24 years old comprises of 31% of the calls. Of the remaining 69% of patients 27% fall in age group 25 to 45, 29% in age group 46 to
66 years, and those 67 and above comprise 13% of the sample. There are approximately 9% more female patients compared to male and the predominant race of
patients is African American. The statistics of those requiring EMS ambulance treatment for asthma attacks are consistent with C enter for Disease C ontrol statistics
on asthma prevalence based on data from 2008 to 2010 [4]. Generally, there were fewer calls in the cold season than the warm.
Table 1. Number of EMS-treated asthma attacks by age group, sex, race, season and year
Daily levels of the pollution and meteorological data are presented in Table 2 by all year and season. Less than 1% of pollution and meteorological data were
missing during the study period. In general, NO 2 and C O appear higher in the winter than summer while the opposite is true for PM2 .5 and ozone. The locations of
the air monitors used in the study in relation to the EMS-treated asthma attacks are shown in Figure 1. The median concentrations by month over all years and all
monitors are plotted with the monthly counts of EMS-treated asthma in Figure 2. Ozone and NO 2 concentrations dip in June and July as do the number of EMS cases.
In addition to the median, July has the lowest frequency of days when the maximum eight hour average concentration of ozone met or exceeded 76 parts per billion
at a monitor (not shown) [42]. These lower ozone concentrations in June and July coincide with high daily rain frequency in these months [42]. Pearson correlation
coefficients between daily measures of air pollutant concentrations and apparent temperature indicate the strongest correlations between daily pollutants were
between NO 2 and C O (r = 0.74) followed by NO 2 and SO 2 (r = 0.57), by C O and SO 2 (r = 0.56), daily PM2 .5 and ozone (r = 0.42). The correlation between ozone and
NO 2 was (r = 0.23). The strongest correlation between a pollutant and apparent temperature was for NO 2 (r = -0.54). As discussed below, the interaction terms
between model variables were not significant.
Table 2. Daily pollution and meteorological levels 2004 to 2011
Figure 1. EMS-treated asthma attacks and pollution monitors in Houston, Texas (2004-2011).
Figure 2. Number of EMS-treated asthma calls in Houston by month and median pollution concentration (2004-2011).
Case-crossover and conditional logistic regression analysis
Analyzing the association between asthma and average apparent temperature using the conditional logistic regression model showed that the previous day was the
relevant exposure period. The minimum AIC was used to select the best model. A similar study found the same result [20]. The logistic regression assumption of
linearity in relative risk is appropriate in this case. In addition, controlling for holidays slightly increased the relative risk of an asthma attack while influenza season
had no effect. Again, these results were found by other researchers, with confounding from asthma attacks around the holidays given in [43] and the non-effect of
influenza reported in multiple studies [43-46].
Results of the single pollutant exploratory analysis lag models are shown in Table 3 where we list the adjusted relative risk of EMS calls for asthma attacks from
exposure to an increase in interquartile range (IQR) of the respective pollutants. The lag for the statistically significant relative risk model with the minimum AIC is
indicated in Table 3 with an asterisk (*).
Table 3. Analysis of relative risk for EMS-treated asthma attacks per IQR increase in single pollutant
An IQR increase in single pollutants on the day of the attack was associated with a relative risk of 1.12 (RR = 1.12; 95% C I: 1.07, 1.17) for exposure to NO 2 , 1.05
(RR = 1.05; 95% C I: 1.02, 1.08) for exposure to C O, and 1.02 (RR = 1.02; 95% C I: 0.99, 1.05) for PM2 .5 . The variable with the best model fit in the exploratory
analysis for ozone was an average of lag 0, 1 and 2 days. The relative risk for the cumulative variable for an IQR increase in ozone was found to be 1.07 (RR = 1.07; 95% C I: 1.03, 1.11). No effect was found for SO 2 .
The multi-pollutant pollutant models showing the highest risk were for levels of NO 2 and ozone. The acute asthma attack risk for these pollutants by year of the
study is shown in Figure 3. The adjusted relative risks shown in Figure 3 were obtained from a multi-pollutant model in relation to an increase in the IQR. C ontrols
for apparent temperature and holidays were included in the models. Although the number of cases differ between years, (i.e., there were more cases in 2007 and
2008 than the other years and fewer in 2004 (Table 1)), the pattern indicates that the risk associated with ozone is somewhat inverse that of NO 2 . While at first this
may seem logical given that NO 2 is a component in the formation of ozone, the relationship is not as simply defined. Taken together between the two pollutants,
http://www.ehjournal.net/content/13/1/58
3/10
2/6/2015
Environmental Health | Full text | Using community level strategies to reduce asthma attacks triggered by outdoor air pollution: a case crossover analysis
there is no apparent downward trend in the risk of an asthma attack. It is worth noting that ozone and NO 2 correlations on the daily level were not strong, (r = 0.23), and the monthly patterns differ (see Figure 2).
Figure 3. Change in association between EMS-treated asthma attacks and NO2 and ozone (2004-2011).
Regression modeling including the significant associations (i.e., those lags marked with *) discussed above (i.e., all pollutants but SO 2 ), and controlling for
confounding variables, was used to identify the final multi-pollutant model for the pollutants ozone and NO 2 . Again, based on an IQR level increase in pollution, the
relative risk due to ozone is 1.05 (RR = 1.05; 95% C I: 1.00, 1.09) and for NO 2 is 1.10 (RR = 1.10; 95% C I: 1.05, 1.15). The interaction terms between ozone and
NO 2 or these pollutants with apparent temperature were not significant. The single pollutant risks were dampened slightly compared with the multi-pollutant model.
The risk from both pollutants decreased 2% in multi compared with single pollutant models.
Multi-pollutant model analysis
Results of the multi-pollutant model stratified by age, sex and race groups are shown in Figure 4. Adjusted relative risks shown in Figure 4 were obtained from a
multi-pollutant model in relation to an increase in the IQR controlling for apparent temperature and holidays.
Figure 4. Change in association between EMS-treated asthma attacks and NO2 and ozone by demographics.
The overall multi-pollutant model association was generally stable across age groups (relative risk for an increase in IQR): 24 years and less for ozone and NO 2 ,
(RR = 1.06; 95% C I: 0.98, 1.14), (RR = 1.09; 95% C I: 1.01, 1.19); 25 to 45 years for ozone and NO 2 , (RR = 1.05; 95% C I: 0.96, 1.14), (RR = 1.11; 95% C I: 1.01,
1.22); 46 to 66 years old for ozone and NO 2 , (RR = 1.03; 95% C I: 0.95, 1.11), (RR = 1.16; 95% C I: 1.06, 1.26) and for 67 years and older for ozone and NO 2 , (RR = 1.05; 95% C I: 0.95, 1.17), (RR = 1.00; 95% C I: 0.89, 1.13). Stratification by sex indicated ozone and NO 2 , (RR = 1.08; 95% C I: 1.02, 1.14), (RR = 1.07; 95% C I:
1.00, 1.14), relative risk for an increase in IQR, respectively, had a similar effect on females. However, NO 2 had a stronger effect on males than ozone, (RR = 1.13;
95% C I: 1.06, 1.21) and (RR = 1.01; 95% C I: 0.95, 1.07), relative risk for an increase in IQR respectively. Stratification by race indicated, per increase in IQR, NO 2
dominated the risk for African Americans (RR = 1.13; 95% C I: 1.07, 1.19) for NO 2 , (RR = 1.03; 95% C I: 0.98, 1.08) for ozone, while ozone dominated the risk for
C aucasian (RR = 1.12; 95% C I: 1.00, 1.24) for ozone, (RR = 0.98; 95% C I: 0.87, 1.11) for NO 2 . The risk per IQR for Hispanics was similar between NO 2 and ozone
but slightly shifted toward ozone, (RR = 1.11; 95% C I: 0.98, 1.25) for ozone, (RR = 1.08; 95% C I: 0.95, 1.24) for NO 2 . The lower percentage of EMS calls for all
races except African Americans likely impedes useful comparisons between races (see Table 1).
Analysis by levels of ozone and NO2
In a separate analysis, the dataset was divided into bins by ozone and NO 2 levels over the study period. The segmentation based on level of each pollutant was
used to examine the difference in risk with respect to two important factors to consider when constructing guidance for asthma action plans.
The first important factor is the relevant exposure period. We examine the difference in risk when exposure prevention guidelines are focused on concentrations for
the day of the asthma event, to those that include the day of the event and the two previous days. The latter time period is the relevant exposure time for Houston,
however current warnings focus only on daily levels of pollutants. Figure 5 shows the concentration-risk plot for the ozone single pollutant model for lag 0 compared
with the cumulative effect of lag 0 to 2 days. The results shown in Figure 5 are adjusted for apparent temperature and holidays. Scales differ in the figure for each
pollution level. Modeling results indicate that as the ozone concentrations increase, accounting for the cumulative effect of lag 0 to 2 days becomes more important.
At the highest bin level of 70 to 90 ppb for the maximum daily 8 hour average concentration, the point estimate risk from the cumulative effect of lag 0 to 2 days is
twice as high as the risk from lag 0 day. The risk at this level is also more variable for the cumulative effect of lag 0 to 2 days than the risk from lag 0 day. The
finding that cumulative ozone, of up to three days, has a stronger impact than single day past levels has been found by other researchers [46].
Figure 5. Ozone single pollutant model concentration-risk plot comparing lag 0 day with cumulative lag 0-2 day.
The second issue to consider when constructing asthma action plans regards the city-specific multi-pollutant mixture. We examine the difference in risk when
pollutants are considered in isolation compared with a multi-pollutant context. For Houston, the significant multi-pollutant model includes ozone and NO 2 . We found
that compared with the single pollutant models of these constituents, the risk attributed to NO 2 is slightly dampened when ozone is considered, and the risk for
ozone is greatly reduced when NO 2 is considered. Forrest plots of the relative risk are shown in Figure 6 where the upper half of the plot reflects the difference in
the risk of single and multi-pollutant models as NO 2 concentrations increase and the bottom half of the plot reflects these differences when ozone increases.
Figure 6. Concentration-risk plot comparing single pollutant and multi-pollutant models for NO2 and ozone.
http://www.ehjournal.net/content/13/1/58
4/10
2/6/2015
Environmental Health | Full text | Using community level strategies to reduce asthma attacks triggered by outdoor air pollution: a case crossover analysis
The results of a multi-pollutant model evaluated for combinations of bins by quartiles of concentrations are shown in Table 4. Quartiles are based on the NO 2 daily
concentrations and the eight-hour highest daily average for ozone. For example the lower right cell of the table contains the risk from exposure to ozone and NO 2
from the multi-pollutant model run on the segmented data containing only those asthma cases that occurred when both NO 2 and ozone were in the fourth quartile.
This cell also contains the number of days where both ozone and NO 2 were high (i.e., in the fourth quartile) and the number of asthma cases occurring in that
segment. These results indicate that in general the exposure to NO 2 is associated with a greater risk than ozone, in Houston, and that during the 217 days during
the study period when both of the pollutants were simultaneously high (both in quartile 4), there were 952 calls to EMS for asthma attacks requiring albuterol. The
relative risk attributed to NO 2 during that time was 1.44 (RR = 1.44; 95% C I: 1.38, 1.50) and 1.06 (RR = 1.06; 95% C I: 1.04, 1.07) for ozone. The full model
indicated that there was no statistically significant interaction between ozone and NO 2 . However when we look at the bin results of Table 4 when NO 2 is in the fourth
quartile, the risk from NO 2 increases as ozone increases (i.e. the last column in Table 4). Figure 7 is a plot of the risk as the concentrations of ozone and NO 2
increase.
Table 4. Relative risk for multi-pollutants by quartile bins of concentrations during study period
Figure 7. Forest plots of rolling concentration bins NO2 and ozone multi-pollutant model concentration-risk.
Discussion
The results indicate that when the pollutants in Houston were considered together in a multi-pollutant model, two pollutants stood out as triggers for attacks: ozone
and NO 2 . C linical studies indicate that these two pollutants appear to act similarly in triggering an attack because they are both oxidant gases that cause
inflammation of the deep lung and respiratory tract [8,9]. Exposure to them may prime eosinophils to subsequent activation by inhaled allergens in atopic patients
[47]. The effect of exposure to both pollutants in a mixture has been explored to some degree using bolus-response studies in humans. These studies found that
previous continuous exposure to ozone decreases the absorption of a bolus of ozone. This decrease is likely due to depletion of compounds able to absorb ozone.
However, the absorption of the ozone bolus increased when there was simultaneous exposure to NO 2 [48]. EPA’s review of studies that examine ozone and NO 2
binary mixtures concluded that, “very generally, additivity occurred after acute exposure and synergism occurred with prolonged exposure.” While laboratory
exposure patterns can’t accurately simulate real-world exposure, findings from the laboratory appear to be consistent with those seen in the population: there is an
increase in risk when both pollutants are present, especially at higher concentration. Regardless of the degree of interaction, it is reasonable to expect that
exposure to high levels of both pollutants simultaneously increases the risk.
This study found that relative risk in the multi-pollutant model due to an 8 ppb increase in NO 2 is 1.05 (RR = 1.05; 95% C I: 1.00, 1.09), whereas with a 20 ppb
increase in ozone the relative risk is 1.10 (RR = 1.10; 95% C I: 1.05, 1.15). For ozone, the cumulative effect of exposure on the day of the attack and the two days
prior pose the greatest risk (0-2 day lag), while for NO 2 the greatest risk occurs from exposure on the day of the attack. Failing to account for risk, that is attributed
to pollutants differently when considered together, especially as concentrations increase, can lead to faulty assumptions regarding which pollutants to attribute the
risk.
All age groups below 67 years are at risk from increased levels of the pollution mixture. The risk from NO 2 exposure appears to increase with increasing age. The
risk from NO 2 is higher for males than females, although more females required EMS treatment for asthma in this study. Ozone and NO 2 concentrations dip in June
and July similar to case numbers. However, in the fall and spring both pollutants can be simultaneously high and case numbers also trend up in this period. The
linear dose-response assumption for a plot of the risk as the concentrations of ozone and NO 2 increase (Figure 7) is a good fit for ozone and a reasonable fit for
NO 2 until very high concentrations. Days with both high ozone and high NO 2 in Houston can be partially explained by a component of the conceptual model for
ozone formation in the Houston-Galveston Area [49]. Land/sea breeze flow reversal occurs when high pressure dominates the area, resulting in light synoptic scale
forcing. The light winds and subsidence allow high concentrations of pollutants to accumulate during the night and morning hours, and the land breeze carries the
pollutants out over Galveston Bay and into the Gulf of Mexico. During the afternoon, the sea breeze flow reversal carries the ozone back into the city and potentially
over freshly emitted NO 2 .
Comparison with other studies
The association between ambient air pollution and asthma related health effects have been explored by several researchers in single city analyses e.g., [27,28,5052]. However, reviews and meta-analysis of studies have not found a consistent message [53]. For example, in a review of 19 studies focused on children,
exposure to 10 μg/m3 of NO 2 , nitrous oxide, and C O were associated with an increased prevalence of asthma ((meta-OR: 1.05, 95% C I: 1.00, 1.11; meta-OR:
1.02, 95% C I: 1.00, 1.04; and meta-OR: 1.06, 95% C I: 1.01, 1.12), SO 2 was associated with an increased prevalence of wheeze (meta-OR: 1.04, 95% C I: 1.01,
1.07), NO 2 was associated with an increased incidence of asthma (meta-OR: 1.14, 95% C I: 1.06, 1.24) and particulate matter was associated with an increased
incidence of wheeze (meta-OR: 1.05, 95% C I: 1.04, 1.07) but no common thread was found for exposure to ozone [54].
One reason for the inconsistencies may be a result of using different indicators to measure air pollution [55]. Studies in some locations focus on a subset of
pollutants because pollutant concentration information is not consistently available. For example, in a study in Detroit ozone was excluded because it was only
collected in the warm season and daily PM2 .5 was imputed from data collected every third day [28]. The number and spatial coverage of monitors measuring
pollution is also highly variable. Where the Detroit study used data from four monitors to derive the average pollutant concentration, two were used in a study in
Spain [52], 24 for PM2 .5 and 13 for ozone were used in a study in New York [27], and our study used 9 for PM2 .5 and 35 for ozone. As discussed previously,
differences in results may also be a function of differences between cities (e.g., pollutant mixtures, geography, ethnicity, socioeconomic status, climate, time activity
patterns, study cohort including age group and other reasons [28]).
Finally, a direct comparison between the results from the Houston study and other studies is not possible because to our knowledge, this is the first study to
examine the association between air pollution and ambulance-treated asthma attacks. The difference in either attack onset or severity a patient experiences
requiring the use of an ambulance over traditional emergency department visits is not known.
Still, a comparison of the Houston study results with a meta-analysis and a multi-city study [54,56] was conducted. In the meta-analysis of nineteen studies [54],
http://www.ehjournal.net/content/13/1/58
5/10
2/6/2015
Environmental Health | Full text | Using community level strategies to reduce asthma attacks triggered by outdoor air pollution: a case crossover analysis
the relative risk for incidence of asthma associated with NO 2 exposure, odds ratios converted to the same scale as the Houston study, was 1.11 (RR = 1.11; 95%
C I: 1.05, 1.19). This relative risk for NO 2 is similar to the relative risk of 1.10 found from NO 2 in the Houston study multi-pollutant model. Recall that for Houston,
ozone and NO 2 are important triggers (RR = 1.05; 95% C I: 1.00, 1.09), (RR = 1.10; 95% C I: 1.05, 1.15) with 20 and 8 ppb increase in ozone and NO 2 , respectively,
in a multi-pollutant model. However, the meta-analysis [54], found no risk from ozone exposure whereas the Houston study did.
While in the multi-city study of 14 hospitals in seven cities [56], the relative risk for respiratory related emergency department visits from exposure to ozone was
1.03 (RR converted to Houston scale: 1.03, 95% C I: 1.00, 1.07% per 20 ppb increase). This relative risk from ozone exposure is similar to the ambulance treated
relative risk of 1.05 from ozone exposure in Houston. This study found no risk from NO 2 exposure [56] while the Houston study did.
When the association was found, NO 2 in the meta-analysis and ozone in the multi-city analysis, the relative risks were of a similar magnitude. Yet, where the
Houston study found both ozone and NO 2 to be of importance, neither the meta-analysis nor the multi-city study found both pollutants to be significant. Studies
which examined the association between and asthma and ozone and NO 2 as co-pollutants found inconsistent results with respect to statistical significance and
relevant exposure period e.g., [12,46,50,52,57,58].
Conclusions
This work was conducted to inform guidance to reduce exposure to air pollution triggered asthma attacks and avert asthma related public health emergencies in a
major city with significant air pollution. As discussed previously, the current approach to reduce exposure to air pollution triggered asthma through air pollution
alerts based on national single pollutant warning levels may be severely limited and too vague. In the ten years since the National Research C ouncil identified the
need to address the effect of mixtures of multi-pollutant that trigger asthma, no new or specific guidance has been established for asthmatics.
The method used here, case cross-over analysis with conditional logistic regression applied to a specific city, was chosen until multi-pollutant statistical modeling
methods evolve and are able to dictate specific national guidance for public health intervention. C ase cross-over analysis has drawbacks when applied to more than
one pollutant at a time (e.g., diminished statistical power as pollutants are added, difficulties identifying higher order interaction, confounding from correlation)
[14,16-18]. To preserve power, we focus only on five pollutants, all previously linked to asthma attack triggers. Of those five, the two pollutants, ozone and NO 2
that stood out most as triggers in the single pollutant model remained in the multi-pollutant model. The contaminant with the third highest relative risk in the single
pollutant model, C O, was correlated with NO 2 (r = 0.74). Fully understanding the confounding between these pollutants is not a concern for this application because
an asthma plan tracking NO 2 would be protective for C O.
These results seen in both the single and multi-pollutant model provide confidence in the conclusion that the asthma related pollutants of concern in Houston can be
tracked with ozone and NO 2 . While Houston health care workers have likely been concerned with ozone impacts on asthmatics because Houston ozone levels are
above the EPA criteria pollutant standard for ozone, this study provides local quantitative evidence of the link. Since the area NO 2 levels are below the EPA criteria
pollutant standard, the link with this pollutant is new and important information to Houstonians.
Beyond identification of two pollutants of concern that increase the risk of an adverse health effect, important information related to the relevant exposure period
prior to triggering an adverse health effect was found (e.g., hours, a day, or extended days). This study concluded that in Houston, the relevant exposure period for
NO 2 is on the order of one day, but the cumulative effect of ozone over a three-day period posed a significantly different and higher risk as concentrations
increased compared with the single day risk estimates. This concept of a cumulative effect from ozone is also new and important information for a community
member.
On a city-specific level, this analysis provides detailed results that could help prevent attacks by identifying: those individuals in Houston, Texas that may be most at
risk of an acute asthma attack requiring EMS treatment triggered from air pollutants; which pollutants trigger the attack, the relevant time period of exposure; and
the magnitude of increased risk as concentrations increase.
Asthma action plans in Houston may identify these pollutants as important asthma attack triggers, especially when they are simultaneously warn of the cumulative
effect of ozone, and recommend tracking personal sensitivity as pollutants increase, especially for the most at-risk demographics.
Abbreviations
EPA: Environmental Protection Agency; O 3 : Ozone; NO 2 : Nitrogen dioxide; PM: Particulate matter; SO 2 : Sulfur dioxide; C O: C arbon monoxide; EMS: Emergency
medical services; PM2.5: Particulate matter with diameter less than 2.5 microns; TC EQ: Texas C ommission on Environmental Quality; AIC : Akaike’s information
criterion; IQR: Interquartile range; RR: Relative risk; C I: C onfidence intervals; ppb: Parts-per-billion; meta-OR: Meta-analysis odds ratio.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
LR conducted the exploratory data analysis, the single and multiple pollutant case-crossover analysis, provided interpretation and was the main author of the
methodology and results section of the manuscript. KBE conducted the bin analysis, peer reviewed all of the results, provided interpretation, edited the manuscript
and authored the remaining sections of the paper. DP conceived the project, developed the health effects database, provided interpretation of the results and edited
the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors gratefully acknowledge the aid from our Rice University student assistants Jiao Li and Elizabeth Ramirez Ritchie, and Bobbie Harris, Jerry C orpening and
Arturo Blanco from the C ity of Houston Department of Health and Human Services. We further would like to thank Laura C ampos for her expert assistance with all
aspects of the project including producing the GIS maps and aiding with data management and statistical analysis. Funding Source: Houston Endowment. There are
no competing financial interests.
http://www.ehjournal.net/content/13/1/58
6/10
2/6/2015
Environmental Health | Full text | Using community level strategies to reduce asthma attacks triggered by outdoor air pollution: a case crossover analysis
References
1. Barnett SB, Nurmagambetov TA: Costs of asthma in the United States: 2002-2007.
J Allergy Clin Immunol 2011, 127:145-152.
2. Health data interactive
Center for Disease Control and Prevention and National Center for Health Statistic 2009. [Online Server. Public Reports. Mortality and life expectancy. Mortality
by underlying cause, ages 18+ 2009]
http://205.207.175.93/hdi/ReportFolders/ReportFolders.aspx webcite
3. National C enter for Environmental Health: Asthma’s impact on the nation.
Center for Disease Control and Prevention 2012.
http://www.cdc.gov/asthma/impacts_nation/AsthmaFactSheet.pdf webcite
4. Akinbami LJ, Moorman JE, Bailey C , Zahran HS, King M, Johnson C A, Liu X: Trends in asthma prevalence, health care use, and mortality in the
United States, 2001-2010.
NCHS Data Brief 2012, 94:1-8.
5. Ozone and your health
United States Environmental Protection Agency 2009.
http://www.epa.gov/airnow/ozone-c.pdf webcite
6. Indoor Environments Division: Asthma facts.
United States Environmental Protection Agency 2013.
EPA-402-F-04-019. [http://www.epa.gov/asthma/pdfs/asthma_fact_sheet_en.pdf webcite ]
7. Labre MP, Herman EJ, Dumitru GG, Valenzuela KA, C echman C L: Public health interventions for asthma: an umbrella review, 1990-2010.
Am J Prev Med 2012, 42:403-410.
8. Air quality criteria for ozone and related photochemical oxidants (2006 Final)
United States Environmental Protection Agency 2006.
EPA 600/R-05/004aF 1:1-821. [http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=149923 webcite ]
9. Integrated review plan for the primary national ambient air quality standard for nitrogen dioxide
United States Environmental Protection Agency 2007.
http://www.epa.gov/ttn/naaqs/standards/nox/data/20070823_nox_review_plan_final.pdf webcite
10. U.S. Federal Register: Secondary national ambient air quality standards for oxides of nitrogen and sulfur.
United States Environmental Protection Agency 2012.
http://www.gpo.gov/fdsys/pkg/FR-2012-04-03/pdf/2012-7679.pdf webcite
11. U.S. Federal Register: National ambient air quality standards for particulate matter.
United States Environmental Protection Agency 2013.
http://www.gpo.gov/fdsys/pkg/FR-2013-01-15/pdf/2012-30946.pdf webcite
12. Schildcrout JS, Sheppard L, Lumley T, Slaughter JC , Koenig JQ, Shapiro GG: Ambient air pollution and asthma exacerbations in children: an eightcity analysis.
Am J Epidemiol 2006, 64:505-517.
13. National Research C ouncil of the National Academies: Air quality management in the United States. Washington, D.C : The National Academies Press; 2004.
14. Dominici F, Peng RD, Barr C D, Bell ML: Protecting human health from air pollution: shifting from a single-pollutant to a multi-pollutant
approach.
Epidemiology 2010, 21:187-194.
15. Johns D: Overview of EPA multi-pollutant science and risk analysis workshop: February 24, 2011, Chapel Hill, NC.
Presentation. United States Environmental Protection Agency 2011.
http://www.epa.gov/ncer/publications/workshop/04_07_2011/djohns_cac_110408.pdf webcite
16. Sun Z, Tao Y, Li S, Ferguson KK, Meeker JD, Park SK, Batterman SA, Mukherjee B: Statistical strategies for constructing health risk models with
multiple pollutants and their interactions: possible choices and comparisons.
Environ Health 2013, 12:1-85.
17. Bateson TF, C oull BA, Hubbell B, Ito K, Jerrett M, Lumley T, Thomas D, Vedal S, Ross M: Panel discussion review: session three: issues involved in
interpretation of epidemiologic analyses: statistical modeling.
J Expo Sci Environ Epidemiol 2007, 17:S90-S96.
http://www.ehjournal.net/content/13/1/58
7/10
2/6/2015
Environmental Health | Full text | Using community level strategies to reduce asthma attacks triggered by outdoor air pollution: a case crossover analysis
18. Johns DO, Stanek LW, Walker K, Benromdhane S, Hubbell B, Ross M, Devlin RB, C osta DL, Greenbaum DS: Practical advancement of multipollutant
scientific and risk assessment approaches for ambient air pollution.
Environ Health Perspect 2012, 120:1238-1242.
19. Brown JS, Graham JA, C hen LC , Postlethwait EM, Ghio AJ, Foster WM, Gordon T: Panel discussion review: session four-assessing biological
plausibility of epidemiological findings in air pollution research.
J Expo Sci Environ Epidemiol 2007, 17:S97-S105.
20. Paulu C , Smith AE: Tracking associations between ambient ozone and asthma-related emergency department visits using case-crossover
analysis.
J Publ Health Manag Pract 2008, 14:581-591.
21. Bethel HL, Sexton K, Linder S, Delclos G, Stock T, Abramson S, Bondy M, Fraser M, Ward J: A closer look at air pollution in Houston: identifying
priority health risks.
The Mayor’s Task Force 2007, 1-15.
http://www.epa.gov/ttnchie1/conference/ei16/session6/bethel.pdf webcite
22. Dr. David E, Persse MD: Medical Director Department of Health & Human Services City of Houston.
832-394-6819. 600 Jefferson, Ste. 800 Houston, Texas 77002 United States. David.persse@houstontx.gov
23. Pereira G, C ook A, De Vos AJ, Holman C D: A case-crossover analysis of traffic-related air pollution and emergency department presentation
for asthma in Perth, Western Australia.
Med J Aust 2010, 193:511-514.
24. National EMS Information Systems: Report: EMS data standard.
National Highway Traffic Safety Administration. Office of Emergency Medical Services 2013.
[http://www.nemsis.org/media/XSD_v3/_nemsis_v3.3.4/3.3.4.140328/DEMEMS/DataDictionary/NEMSISDataDictionary_v3.3.4.140328.pdf webcite ]
25. Ensor KB, Raun LH, Persse D: A case-crossover analysis of out-of-hospital cardiac arrest and air pollution.
Circulation 2013, 127:1192-1199.
26. Levy D, Sheppard L, C heckoway H, Kaufman J, Lumley T, Koenig J, Siscovick D: A case-crossover analysis of particulate matter air pollution and
out-of-hospital primary cardiac arrest.
Epidemiology 2001, 12:193-199.
27. Silverman RA, Ito K, Freese J, Kaufman BJ, De C laro D, Braun J, Prezant DJ: Association of ambient fine particles with out-of-hospital cardiac
arrests in New York City.
Am J Epidemiol 2010, 172:917-923.
28. Li S, Batterman S, Wasilevich E, Wahl R, Wirth J, Su F, Mukherjee B: Association of daily asthma emergency visits and hospital admissions with
ambient air pollutants among the pediatric Medicaid population in Detroit: Time-series and time-stratified case-crossover analyses with
threshold effects.
Environ Res 2011, 111:1137-1147.
29. List of designated reference and equivalent methods: List of designated reference and equivalent methods.
United States Environmental Protection Agency 2012. [MD-D205-03]
http://www.epa.gov/ttnamti1/files/ambient/criteria/reference-equivalent-methods-list.pdf webcite
30. O'Neill MS, Zanobetti A, Schwartz J: Modifiers of the temperature and mortality association in seven US cities.
Am J Epidemiol 2003, 157:1074-1082.
31. Kalkstein LS, Valimont KM: An evaluation of summer discomfort in the United States using a relative climatological index.
Bull Am Meteorol Soc 1986, 67:842-848.
32. Steadman RG: The assessment of sultriness. part I: a temperature-humidity index based on human physiology and clothing science.
J Appl Meteorol 1979, 18:861-873.
33. Influenza surveillance (2004-2011)
Texas Department of State Health Services
http://www.dshs.state.tx.us/idcu/disease/influenza/surveillance/ webcite
34. Maclure M: The case-crossover design: a method for studying transient effects on the risk of acute events.
Am J Epidemiol 1991, 133:144-153.
35. Forastiere F, Stafoggia M, Picciotto S, Bellander T, D’Ippoliti D, Lanki T: A case-crossover analysis of out-of hospital coronary deaths and air
pollution in Rome, Italy.
Am J Respir Crit Care Med 2005, 172:1549-1555.
36. Peters A, Dockery DW, Muller JE, Mittleman MA: Increased particulate air pollution and the triggering of myocardial infarction.
Circulation 2001, 103:2810-2815.
37. Dennekamp M, Akram M, Abramson MJ, Tonkin A, Sim MR, Fridman M, Erbas B: Outdoor air pollution as a trigger for out-of-hospital cardiac arrests.
http://www.ehjournal.net/content/13/1/58
8/10
2/6/2015
Environmental Health | Full text | Using community level strategies to reduce asthma attacks triggered by outdoor air pollution: a case crossover analysis
Epidemiology 2010, 21:494-500.
38. Rosenthal F, C arney JP, Olinger ML: Out-of-hospital cardiac arrest and airborne fine particulate matter: a case-crossover analysis of
emergency medical services data.
Environ Health Perspect 2008, 116:631-636.
39. Wellenius GA, Burger MR, C oull BA, Schwartz J, Suh HH, Koutrakis P, Schlaug G, Gold DR, Mittleman MA: Ambient air pollution and the risk of acute
ischemic stroke.
Arch Intern Med 2012, 172:229-234.
40. Janes H, Sheppard L, Lumley T: Case-crossover analyses of air pollution exposure data: referent selection strategies and their implications for
bias.
Epidemiology 2005, 16:717-726.
41. SAS/STAT software, version [9.3] of the SAS system for [Unix]
Copyright © [2010] SAS Institute Inc. SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute
Inc., Cary, NC, USA
http://www.sas.com/presscenter/guidelines.html webcite
42. Texas C ommission on Environmental Quality: One-hour ozone high value days. http://www.tceq.state.tx.us/cgibin/compliance/monops/ozone_exceedance.pl webcite
43. Lee SL, Wong WH, Lau YL: Association between air pollution and asthma admission among children in Hong Kong.
Clin Exp Allergy 2006, 36:1138-1146.
44. Villeneuve PJ, C hen L, Rowe BH, C oates F: Outdoor air pollution and emergency department visits for asthma children and adults: a casecrossover study in northern Alberta, Canada.
Environ Health 2007, 6:1-40.
45. Wilson AM, Wake C P, Kelly T, Salloway JC : Air pollution, weather, and respiratory emergency room visits in two northern New England cities: an
ecological time-series study.
Environ Res 2005, 97:312-321.
46. Anderson HR, Ponce de Leon A, Bland JM, Bower JS, Emberlin J, Strachan DP: Air pollution, pollens, and daily admissions for asthma in London 198792.
Thorax 1998, 53:842-848.
47. D’Amato G: Urban air pollution and plant-derived respiratory allergy.
Clin Exp Allergy 2000, 30:528-536.
48. Office of Research and Development: Air quality criteria for ozone and related photochemical oxidants. Volume I of III.
Unit States Environ Protection Agency 2006, 1:4-14.
49. Technical Support Section. Technical Analysis Division: Conceptual model for ozone formation in the Houston-Galveston Area.
Texas Commission on Environmental Quality 2002, 1-40. [Appendix A to Phase I of the mid course review modeling protocol and technical support document]
http://www.tceq.texas.gov/assets/public/implementation/air/am/docs/hgb/protocol/HGMC R_Protocol_Appendix_A.pdf webcite
50. Galan I, Tobias A, Banegas JR, Aranguez E: Short-term effects of air pollution on daily asthma emergency room admissions.
Eur Respir J 2003, 22:802-808.
51. Lin M, C hen Y, Burnett RT, Villeneuve PJ, Krewski D: Effect of short-term exposure to pollution on asthma hospitalization in children: Bidirectional case-crossover analysis.
J Epidemiol Community Health 2002, 57:50-55.
52. Tenias JM, Ballester F, Rivera ML: Association between hospital emergency visits for asthma and air pollution in Valencia, Spain.
Occup Environ Med 1998, 55:541-547.
53. Nordling E, Berglind N, Melén E, Emenius G, Hallberg J, Nyberg F, Pershagen G, Svartengren M, Wickman M, Bellaner T: Traffic-related air pollution and
childhood respiratory symptoms, function and allergies.
Epidemiology 2008, 19:401-408.
54. Gasana J, Dillikar D, Mendy A, Forno E, Vieira ER: Motor vehicle air pollution and asthma in children: a meta-analysis.
Environ Res 2012, 117:36-45.
55. Penard-Morand C , Raherison C , C harpin D, Kopferschmitt C , Lavaud F, C aillaud D, Annesi-Maesano I: Long-term exposure to close-proximity air
pollution and asthma and allergies in urban children.
Eur Respir J 2010, 36:33-40.
56. Stieb DM, Szyszkowicz M, Rowe BH, Leech JA: Air pollution and emergency department visits for cardiac and respiratory conditions: a multi-city
time-series analysis.
Environ Health 2009, 8:1-25.
57. Strickland MJ, Darrow LA, Klein M, Flanders WD, Sarnat JA, Waller LA, Sarnat SE, Mulholland JA, Tolbert PE: Short-term associations between ambient
air pollutants and pediatric asthma emergency department visits.
http://www.ehjournal.net/content/13/1/58
9/10
2/6/2015
Environmental Health | Full text | Using community level strategies to reduce asthma attacks triggered by outdoor air pollution: a case crossover analysis
Am J Respir Crit Care Med 2010, 182:307-316.
58. Farhat SC L, Paulo RLP, Shimoda TM, C onceicao GMS, Lin C A, Braga ALF, Warth MPN, Saldiva PHN: Effect of air pollution on pediatric respiratory
emergency room visits and hospital admissions.
Braz J Med Biol Res 2005, 38:227-235.
Sign up to receive new article alerts from Environmental Health
http://www.ehjournal.net/content/13/1/58
Sign up
10/10
JAMES A. BAKER III INSTITUTE FOR PUBLIC POLICY
RICE UNIVERSITY
ASSOCIATION OF OUT–OF–HOSPITAL CARDIAC ARREST
WITH EXPOSURE TO FINE PARTICULATE AND OZONE
AMBIENT AIR POLLUTION FROM CASE–CROSSOVER
ANALYSIS RESULTS: ARE THE STANDARDS PROTECTIVE?
BY
LOREN RAUN, PH.D.
FACULTY FELLOW, DEPARTMENT OF STATISTICS
RICE UNIVERSITY
AND
KATHERINE B. ENSOR, PH.D.
CHAIR, DEPARTMENT OF STATISTICS
RICE UNIVERSITY
OCTOBER 12, 2012
Exposure to Fine Particulate and Ozone Ambient Air Pollution
THESE PAPERS WERE WRITTEN BY A RESEARCHER (OR RESEARCHERS) WHO PARTICIPATED IN A
BAKER INSTITUTE RESEARCH PROJECT. WHEREVER FEASIBLE, THESE PAPERS ARE REVIEWED BY
OUTSIDE EXPERTS BEFORE THEY ARE RELEASED. HOWEVER, THE RESEARCH AND VIEWS
EXPRESSED IN THESE PAPERS ARE THOSE OF THE INDIVIDUAL RESEARCHER(S), AND DO NOT
NECESSARILY REPRESENT THE VIEWS OF THE JAMES A. BAKER III INSTITUTE FOR PUBLIC POLICY.
© 2012 BY THE JAMES A. BAKER III INSTITUTE FOR PUBLIC POLICY OF RICE UNIVERSITY
THIS MATERIAL MAY BE QUOTED OR REPRODUCED WITHOUT PRIOR PERMISSION,
PROVIDED APPROPRIATE CREDIT IS GIVEN TO THE AUTHOR AND
THE JAMES A. BAKER III INSTITUTE FOR PUBLIC POLICY.
2
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Abstract
About 300,000 cardiac arrests occur outside of hospitals in the United States each year; most are
fatal. Studies have found that a small but significant percentage of the cardiac arrests appear to
be triggered by exposure to increased levels one of two air pollutants: fine particulate matter and
ozone.
We analyzed seven key studies to determine if Environmental Protection Agency (EPA)
standards protect the public from out-of-hospital cardiac arrests (OHCA) triggered by exposure
to fine particulate matter and ozone. Using Houston, Texas, data, we found evidence of an
increased risk of cardiac arrest on the order of 2% to 9% due to an increase of fine particulate
levels (a daily average increase of 10 µg/m3) on the day of, or day before, the heart attack. The
EPA fine particulate standard of 35 µg/m3 (35 micrograms per cubic meter of air) therefore does
not effectively protect the public from OHCA triggered by exposure to fine particulates.
However, the EPA’s ozone standard does appear to adequately protect public health from OHCA
triggered by exposure to ozone.
Introduction
The first decisive regulatory move toward protecting public health from impacts of air pollution
occurred in 1971 through passage of the Clean Air Act (CAA). Section 109 directs the
Environmental Protection Agency (EPA 1971) to promulgate standards for certain pollutants
found in ambient air. These pollutants—ozone, carbon monoxide, nitrogen dioxide, sulfur
dioxide, particulate matter and lead—were believed to represent a present or future threat to
public health.
The CAA further requires that the standards be set at a level sufficient to protect health with an
adequate margin of safety. The phrase “adequate margin of safety” has been defined to be the
maximum permissible ambient air level that will protect the health of any sensitive group, while
accounting for uncertainties with risk assessment and toxicology studies and still protecting
against hazards not yet identified.
3
Exposure to Fine Particulate and Ozone Ambient Air Pollution
While the congressional action was decisive 40 years ago, even today it is unclear whether the
public health has been adequately protected with regard to at least two of the criteria pollutants:
fine particulate matter and ozone. Fine particulate matter (PM2.5) is made up of particles 2.5
microns in size or smaller (EPA 2005). Ozone, on the tropospheric level, is a photochemical
oxidant formed when volatile organic compounds and nitrogen oxides combine under certain
atmospheric conditions (EPA 2006). The standards for these two pollutants have existed in a
state of flux, changing as the health effects, magnitude of association, and latency of effect are
better understood with emerging evidence.
Out-of-hospital cardiac arrest (OHCA) is an important new example of a health effect with an
association with short-term exposure to air pollutants. Defined as a condition characterized by an
unexpected cardiovascular collapse due to an underlying cardiac cause occurring out of the
hospital, approximately 300,000 persons in the United States experience an OHCA each year;
more than 90% of those persons who experience an OHCA die (McNally et al. 2011).
It appears that one trigger for OHCA is exposure to PM2.5 air pollution (Ensor et al. 2012;
Dennekamp et al. 2010; Silverman et al. 2010); recently researchers have found that another
trigger of OHCA may be ozone (Ensor et al. 2012). This epidemiological evidence is supported
by pathophysiologic arguments that link PM2.5 and ozone air pollutants to cardiac endpoints
(see Dockery et al. 2005; Gold et al. 2000; Peters et al. 1997; Peters et al. 2000; Riediker et al.
2004; Srebot et al. 2009).
Given these recent findings, we ask, “Are the current air pollution standards for PM2.5 and
ozone protective of health in terms of OHCA for the public and sensitive subpopulations?”
We seek to answer the question by first, presenting the case-crossover analysis studies that have
examined the risk of OHCA from exposure to PM2.5 and ozone, and then assessing the findings
in relationship to existing PM2.5 and ozone standards.
4
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Overview of Case-Crossover Method
The method increasingly used to study the association between air pollution and OHCA is a
case-crossover design analyzed with conditional logistic regression. In addition to OHCA, this
methodology has been used to assess asthma attacks (Lin 2002), congestive heart failure (Kwon
et al. 2001), strokes (Tsai et al. 2003; Wellenius et al. 2012), and other episodic health events
that have followed short-term exposure to air pollution.
Case-crossover design was first introduced 20 years ago (Maclure 1991). In the case-crossover
design, each individual experiencing a health event serves as his or her own reference; in other
words, individuals serve as their own control. Ambient air pollution is used as a proxy for
personal exposure. The ambient air pollution concentrations at times when the study individual is
not experiencing the OHCA health event are the reference concentrations. The reference
concentrations are statistically compared with the concentrations during or around the time the
study individual experienced the OHCA health event. Conditional logistic regression is applied
to estimate the association of pollution and increased relative risk of the health event while
controlling for confounding factors.
The number of events used in case-crossover analysis for OHCA ranges from a little more than
350 (Levy et al. 2001) to more than 11,000 (Ensor et al. 2012). More cases are needed when the
exposure concentration range is narrower (i.e., smaller interquartile range). The number of cases
included in these studies has been increasing over time, resulting in stronger statistical power of
the analysis.
The reference concentration periods are chosen to minimize multiple competing biases present
from the absence of stationarity in air pollution time series. Researchers have found that
reference periods are best taken the same day of the week, hour of the day, and month as the
event (Bateson and Schwartz 1999; Greenland 1996; Lumley and Levy 2000; Levy et al. 2001;
Navidi 1998 ).
5
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Temperature is included in the case-crossover analysis to control for effects of heat or cold on
OHCA. Apparent temperature—the body’s perceived temperature—is calculated from
temperature and dew point (Steadman 1979) and is often used over temperature alone. The effect
of temperature or apparent temperature on OHCA may be nonlinear depending upon the
temperature range in the study area (Baus and Samet 2002; Braga et al. 2001; Curriero et al.
2002; Stafoggia et al. 2006). In a case-crossover analysis of air pollution and OHCA in New
York City (Silverman et al. 2010), immediate and delayed nonlinear temperature effects were
found and adjusted using natural cubic splines of the same-day and the average of the past three
days’ apparent temperature. Alternatively, in a similar study conducted in Melbourne
(Dennekamp et al. 2010), temperature effects were found to be linear.
Conditional logistic regression is used and a linear exposure-effect model is assumed. The
relative risk and 95% confidence intervals from lags of daily or hourly concentrations are
estimated usually for a concentration change equivalent to the interquartile range where lag 0 day
refers to the day of the event, while a lag 1 day is the day before the event, etc. To check the
validity of the linear exposure-effect assumption, lag estimates of effect are found by quartile of
pollution for a given lag and by using regression spline smoothers of one and three knots (Levy
et al. 2001). If the effect is linear it should be constant across quartiles.
Cardiac hospital admissions and daily mortality statistics are often used in these case-crossover
studies. These health statistics are available as daily counts. The weakness of daily counts is that
more transient effects (hours to minutes) cannot be assessed. In addition, there is unknown error
in results for lag 0, defined as the day of the event, because the researcher cannot rule out that the
time of the cardiac arrest may have preceded most of the exposure (Levy et al. 2001). Some
studies use emergency medical service (EMS) health data. The EMS data, in which the 911 call
time acts as the time of the OHCA, provides the ability for a more refined analysis on the hourly,
as opposed to daily, scale. The hourly scale of both the health event and the pollution
concentration enable an analysis of ambient concentration standards of less than a day. However,
since both cardiac arrest and pollutant data may have diurnal patterns, temporal confounding
must be considered (Silverman et al. 2010).
6
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Some researchers extend their full analysis into a subset by season: warm or cold. This is
conducted to better understand if effects are more extreme in a given season as a consequence of
changes in the pollution profile by season. For example, ozone is higher in the warmer months,
and PM2.5 concentrations and chemical composition may vary by season. In addition, exposure
patterns (e.g., time spent outdoors) could vary by season. Researchers also generally explore the
pollution and OHCA association by gender and age, and some have examined presenting heart
rhythm or pre-existing co-morbidities.
Pollution data are obtained from ambient air monitors used to measure hourly concentrations.
The most accurate measurements will be those analyzed using an EPA analytical reference
method. If multiple monitors are available, the majority of researchers average the concentrations
across the monitors to obtain one pollution level (Ensor et al. 2012; Levy et al. 2001; Silverman
et al. 2010). Because there can be spatial variability in both PM2.5 and ozone concentrations, the
average concentration is more representative of the general pollution in the area than data from a
single monitor.
Case-Crossover Method vs. the Alternative
The case-crossover method is an alternative to Poisson or Extended Cox traditional time-series
regression models used to assess the short-term effects of air pollution. The methods have
produced almost identical results (Tsai et al. 2003; Peters et al. 2006). For example, in a
reanalysis that compared the two methods, Neas et al. (1999) confirmed the association between
total suspended particulate pollution and daily mortality in Philadelphia, Pennsylvania, using a
Poisson regression analysis with case-crossover analysis.
The weaknesses of the regression analyses are that they are more sophisticated and require more
investigator decisions than the case-crossover approach. For example, the regression requires the
adequate controlling for confounding from trends of pollution by time and season. Typically,
non-parametric smoothing functions of time are used to model and control seasonality. The
smoothing functions, used to fit each model term, are sensitive to the degrees of freedom
determined by the investigator. Generally Poisson regression requires knowledge of the size of
7
Exposure to Fine Particulate and Ozone Ambient Air Pollution
the population at risk. This parameter enters the regression analysis through the offset term. The
assumption of a constant, and thereby unnecessary, offset is reasonable if the population at risk is
very large relative to the daily number of events and the size and makeup of the population at
risk does not vary with the exposure of interest. If, for example, the susceptible portion of the
total population at risk increases over time from multiple exposures or decreases over time from
harvesting, the assumption that the risk does not vary with the exposure would not be met (Neas
et al. 1999). In addition, the researcher must be aware of and incorporate in the model, or remove
from consideration, periods of anomalous events (e.g., sickness, natural disaster, power outage).
For studies with a large number of cases, anomalous events should have little to no impact on the
results, but the potential impact of the outliers should be considered.
The strength of the case-crossover method is that, in contrast with traditional time series models,
confounding is controlled by design rather than by modeling, thereby obviating the need for
sophisticated modeling. Time-invariant and subject-specific variables are not confounders.
Because the subjects serve as their own control, the size of the population at risk is not an issue.
The pollution reference periods are chosen so that times of day, day of week, seasonality, or
longer term pollution trends are not possible confounding factors.
One weakness of the case-crossover design compared to Poisson regression is that the casecrossover has lower statistical power (Neas et al. 1999). In a comparison of methods conducted
by Neas et al. (1999), larger standard errors were found using case-crossover compared with
Poisson regression. In their application in New York City, Silverman et al. (2010) explain that
the risk estimates from case-crossover were less significant than those found from time series
analysis because the case-crossover method effectively used 12 degrees of freedom/year while
the times series used 7 degrees of freedom/year. Another weakness of the case-crossover method
compared to the Poisson regression model is that times series accounts for over dispersion of the
Poisson variance while the case-crossover analyses typically do not (Lu and Zeger 2007).
Finally, Peters et al. (2006) feels that the Poisson regression may be preferred simply because it
is less time consuming.
8
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Researchers have moved from viewing time series and case-crossover models as competing
methods to application of the models in tandem to verify and validate findings. Many researchers
avoid discussion of weakness in their results from using one model over another by applying
both (e.g., Dennenkamp et al. 2010; Silverman et al. 2010).
Key Components of Qualifying Studies
For use in the present work, the strongest evidence would come from studies that incorporate key
components of the state-of-the-science in the design and analysis. The key components drawn
from the literature described above are summarized into eight items in Table 1.
Table 1. Key Components of Qualifying Studies
1. Number of cases
2. Referent selection
3. Temperature control
4. Health data
5. Pollution data
analysis
6. Pollution data spatial
coverage
7. Validation of results
8. Study populations
Studies contain a large number of cases to increase the statistical
power of the analysis.
Referents are selected on same day of the week (and hour of the
day, if appropriate) and month as the event to minimize bias.
Temperature is controlled and the temperature effect relationship
investigated to allow nonlinear modeling, if appropriate.
The health endpoint is out-of-hospital-cardiac arrest.
The analytical method to determine the pollution concentration for
PM2.5 and ozone is an EPA federal reference method to ensure
high-quality analytical results.
Pollution data is available on an hourly scale from multiple
monitors to ensure the ambient exposure concentration is
representative of the area.
The case-crossover result is verified using time series modeling.
The study population is representative of the population of
individuals that have experienced an OHCA in that area.
Studies that Examine the Link
Case-crossover studies that looked specifically at the association between PM2.5 and ozone and
OHCA were found by searching PubMed and Google Scholar for the following key words in
the title and/or abstract: (1) ozone, O3, air pollution, fine particulate, or PM2.5, (2) case cross
over, case cross-over, case-crossover (3) out-of-hospital cardiac arrest. While there are numerous
studies that use a different statistical analysis method or examine PM2.5 and/or ozone
9
Exposure to Fine Particulate and Ozone Ambient Air Pollution
association with hospital admissions, overall mortality, or cardiac mortality (Bell et al. 2004;
Guo et al. 2010; Ito et al. 2005; Ji et al. 2011; Lee et al. 1999; Moore et al. 2010; Neas et al.
1999; Xu et al. 2008), we focused only on those that best fit the criteria identified for this study
and were conducted in the last 15 years. This search resulted in seven studies that were most
applicable according to the key components listed in Table 1. These studies and their qualifying
components are listed in Table 2. There are two case-crossover studies, Peter et al. (2001) and
Stafoggia et al. (2010), that while not explicitly included in Table 2, provide valuable supporting
information and warrant mention in the discussion.
Table 2. Qualifying Case-Crossover Studies and Key Components
Study Title, Author, Location, Pollutant
1
A case-crossover analysis of particulate matter air
pollution and out-of-hospital primary cardiac arrest.
Levy, D., L. Sheppard, et al.; Seattle and King County,
WA (PM2.5)
Exposure to ambient fine particulate matter and primary
cardiac arrest among persons with and without clinically
recognized heart disease. Sullivan, J., N. Ishikawa, et al.;
Seattle, WA (PM2.5)
A case-crossover analysis of out-of-hospital coronary
deaths and air pollution in Rome, Italy. Forastiere, F., M.
Stafoggia, S. Picciotto, et al. (ozone)
Out-of-hospital cardiac arrest and airborne fine
particulate matter: a case-crossover analysis of
emergency medical services data in Indianapolis, IN.
Rosenthal, F.S., J.P. Carney, M.L. Olinger (PM2.5)
Association of ambient fine particles with out-of-hospital
cardiac arrests in New York City. Silverman, R.A., Ito,
K., Freese, J., et al. (PM2.5 and ozone)
Outdoor air pollution as a trigger for out-of-hospital
cardiac arrests. Dennekamp, M., Akram, M., Abramson,
M.J., et al.; Melbourne, Australia (PM2.5 and ozone)
A case-crossover analysis of out-of-hospital arrest and air
pollution in Ensor, K., Raun, L., Persse, D.; Houston, TX
(PM2.5 and ozone)
*refers to PM2.5 method
10
Presence of Key Components
2 3 4 5 6 7 8 Total
√
√
√
√
3
√
√
5
√
√
√
√
√
√
√
√
√
√
√
√
√
√
√
√
√
√
√*
√
√
√
√
√
√
√
√
√
4
√
6
√
√
8
√
√
7
√
√
8
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Study Descriptions
Table 3 lists a summary of the study details. The number of cases of OHCA examined in the
studies ranges from 362 to 11,677. Overall, there has been an increasing trend in the number of
cases included in studies over time.
All studies included the preferred referent selection. In addition, all studies controlled for
ambient temperature or apparent temperature and all studies except, possibly, Levy et al. (2001)
and Forastiere et al. (2005) allowed for nonlinear modeling of temperature, if appropriate. The
methods used to incorporate temperature in the model in the Levy et al. (2001) and Forastiere et
al. (2005) studies are unknown.
The health endpoint of interest for this analysis, OHCA, is the endpoint identified by Rosenthal
et al. (2008), Silverman et al. (2010), Dennekamp et al. (2010), and Ensor et al. (2012). Levy et
al. (2001), and Sullivan et al. (2003) report an endpoint of primary cardiac arrest, which is
interpreted here as OHCA. The word “primary” is usually used to indicate there is no other
suspected cause (e.g., trauma, drugs, or environmental factors). Finally, Forastiere et al. (2005)
reports an endpoint of out-of-hospital cardiac death. While this endpoint does not completely
encompass all OHCA events, it can be argued that it is likely representative because the majority
of those experiencing an OHCA do not survive.
Studies will be most comparable if the PM2.5 and ozone pollution samples are analyzed using an
EPA federal reference method or equivalent. An EPA federal reference method is a method
explicitly specified using a combination of design-and performance-based criteria. Approval as
an equivalent method is based on the degree of similarity to the reference method and reference
method specification. All studies that analyzed PM2.5 except two, Levy et al. (2001) and
Sullivan et al. (2003), used an EPA reference method. These two studies used a different method,
nephelometry, to estimate the PM2.5 fraction. The quality of the measurements obtained from
nephelometry as a proxy for estimation of PM2.5 does not meet the standards of measurement
set by EPA reference method. Of the studies that analyzed ozone, the methods used in the study
11
Exposure to Fine Particulate and Ozone Ambient Air Pollution
by Forastiere et al. (2005) and Dennekamp et al. (2010) are not reported; the other studies used
EPA reference methods.
The number of monitor locations used in the studies ranges from one to 33 for PM2.5 and one to
47 for ozone. The larger the study location or the more variable the concentration in the area, the
more monitors are needed; however, most studies that used only one location consider the small
number a possible limitation of their research regardless of spatial coverage needs or variability
(e.g., Dennekamp et al. 2010). Rosenthal et al. (2008), Forastiere et al. (2005), and Dennekamp
et al. (2010) relied on one location. Rosenthal et al. (2008) reported that the surrounding
monitors were correlated to the extent that one monitor was representative. Levy et al. (2001)
and Sullivan et al. (2003) relied on three locations. Silverman et al. (2010) relied on 33 ozone
and 16 PM2.5 monitors while Ensor et al. (2012) relied on 47 ozone and 12 PM2.5 monitors.
The validation of case-crossover results with the Poisson time series method was reported by
Silverman et al. (2010), Dennekamp et al. (2010), and Ensor et al. (2012).
The study population of interest for the purposes of the present analysis is the general population
that experiences an OHCA that is not trauma related. The study population for Levy et al. (2001)
is more limited and possibly not as representative of the overall population, which would include
sensitive subgroups, because it excludes those with a life threatening co-morbidities or clinically
recognized heart disease. Those who belonged to a health maintenance organization made up the
study population for Sullivan et al. (2003). Because this choice limited the population to those
who were insured, this population may not have been representative of the overall population.
Rosenthal et al. (2008), Silverman et al. (2010), and Ensor et al. (2012) used Emergency Medical
Service 911 call databases, and therefore the population was limited to the portion of cases that
used this service. The other studies relied on medical records.
All studies stratified by some age grouping and all but Levy et al. (2001) stratified by gender.
The next most common stratification parameters include pre-existing condition (Levy et al.
2001; Sullivan et al. 2003; and Forastiere et al. 2005) and race (Levy et al. 2001; Sullivan et al.
2003; and Ensor et al. 2012), followed by warm versus cold season (Silverman et al. 2010 and
12
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Ensor et al. 2012). In order to conduct an analysis on an association with pollution on the hourly
level, Rosenthal et al. (2008) subset the OHCA event dataset into only those events that were
witnessed.
Table 3. Study Descriptions
Time Frame
Authors
No. of Cases
Seattle &
King
County,
WA
Oct. 1988
to July
1994
Levy et
al.
362
Referent
Selection
Seattle &
King County,
WA
Rome, Italy
Indianapolis,
IN
New York, NY
Melbourne,
AUS
Houston,
TX
1985 to 1994
1998 to 2000
July 2002 to
July 2006
2002-2006
2003 to 2006
2004 to
2011
Sullivan et al.
Forastiere et al.
Rosenthal et al.
Silverman et al.
Dennekamp et
al.
Ensor et
al.
1,206
5,144
1,374
8,216
8,434
11,677
All studies used preferred referent selection same day of week (and or hour) within a month of the event
Method to
Control for
Temperature
Ambient
temp.;
unknown
Ambient
temp.; linear
and
quadratic
Apparent temp.
of day of event
and day 1-3
before;
unknown
Ambient temp.;
two-segment
linear model
Apparent temp.;
natural cubic
Ambient temp;
linear functional
form (GAM)
Apparent
temp.;
natural
cubic
Endpoint
Primary
Cardiac
Arrest
Primary
Cardiac
Arrest
OHCA Death
OHCA
OHCA
OHCA
OHCA
Monitors
3
3
1
1
33 PM2.5, 16
Ozone
1
12 PM2.5,
47 ozone
HMO
members
No pre-existing
cardiac
condition, age
35 and older
911 EMS data
Assumed
primary cardiac
arrest, 911 EMS
data
Age 35 and
older
Age 18
and older,
911 EMS
data
Gender;
race; preexisting
condition;
smoking
Age; gender;
pre-existing
condition
Age; gender;
race; heart
rhythm;
witnessed
Age; gender;
season
Age; gender
Age;
gender;
race;
season
Study
Population
Stratification
No preexisting
cardiac
condition or
life
threatening
disease
Age;
season; preexisting
condition;
smoking;
physical
activity;
alcohol;
aspirin use;
time of day
13
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Study Findings and Comments
The seven studies either implemented single lag models or, in a few cases, a constrained
distributed lag model consisting of an average over two days. The single lag models provide an
estimate of a relative risk for increase in pollution levels equivalent to the interquartile range
(IQR) of the pollutant during the stated lag period. The constrained distributed lag models
provide an estimate of the cumulative effect over more than one lag. Table 4 and 5 list the
increase in risk from the main result for each study from single and/or constrained distributed
lags by IQR of concentration used in the study location for PM2.5 and ozone, respectively.
Because different locations used different IQRs, the odds ratios in the table are not directly
comparable.
Levy et al. (2000) conducted a case-crossover analysis of particulate matter and OHCA and
found a null result (e.g., no effect). However, there were several aspects to the study that do not
meet the criteria of strong evidence for use in the present analysis. The study by Levy et al.
(2001), used nephelometry to measure fine particulate; nephelometry is not an EPA reference
method for measuring fine particulates. In addition, it was unclear how temperature was included
in the logistic regression and there was not an alternative statistical validation. Finally, the small
number of cases (n=362) may have resulted in low statistical power. Ozone was not investigated.
Following Levy et al. (2000), the next case-crossover analysis investigating the OHCA and
particulate association was conducted by Sullivan et al. (2002). This investigation also found null
results, except for a subgroup of smokers with preexisting heart disease. While this study had
more cases (n=1206) than the previous study and controlled for temperature with
accommodation for nonlinearity, it also did not use an EPA reference method (nephelometry) to
measure PM2.5 or include an alternative statistical validation. In addition, the study subjects
consisted of members of a health maintenance organization (HMO) and may not be generalizable
to the population at risk for OHCA. Known risk factors for heart disease are low income, low
education populations and African American ethnicity (Roger et al. 2011). Only 6% of the cases
were of African American ethnicity and low income groups may not have insurance and may not
be represented in an HMO. Again, ozone was not investigated.
14
Exposure to Fine Particulate and Ozone Ambient Air Pollution
In 2005, Forastiere et al. (2005) published a case-crossover analysis (n=5144) showing a
statistically significant association between OHCA and ultrafine and coarse particulate air
pollution in Italy. This study did not assess fine particulate pollution. However, the results
provide evidence of the likelihood there is an effect at the PM2.5 range since a significant effect
was found at the lower and higher range than PM2.5. This study did investigate ozone but only
during the warm season; they did not find a significant association.
Rosenthal et al. (2008) conducted a case-crossover analysis to explore the association between
OHCA and fine particulate matter and if risk was dependent on subject characteristics or preexisting heart rhythm. This study, based on EMS data (n=1374), was the first to look at hourly
association of pollution and OHCA. The study found a null result overall, except for a subgroup
of OHCA witnessed by bystanders (n=511). For this witness group, OHCA risk significantly
increased with a 10 µg/m3 increase in PM2.5 exposure during the hour of the event.
As summarized in Table 2, the three most recent studies, Silverman et al. (2010), Dennekamp et
al. (2010), and Ensor et al. (2012), had the strongest evidence because they incorporate the most
key components of the state-of-the-science in the design and analysis.
Silverman et al. (2010) conducted a case-crossover analysis of air pollution and OHCA (n=8216)
and found an increase of 10.0 µg/m3 in PM2.5 over two days (average of lag 0 and 1) was
associated with an increase in OHCA risk of 4% (95% CI -1 to 8). During the warm season, the
case-crossover analysis yielded a result of stronger effect; an increase of 10.0 µg/m3 in PM2.5
over two days (average of lag 0 and 1) in the warm season was associated with an increase in
OHCA risk of 8% (95% CI 2 to 15). They did not find a difference in risk between men and
women and between age groups. Ozone was not found to be significant using case-crossover
analysis and was not explicitly reported in the paper; however, an estimate of the Poisson timeseries design results indicates that there was an increase in OHCA risk of 5% (95% CI -1 to 11)
per 22 ppb ozone using the eight-hour average daily maximum for the average of 0-1 day lagged.
Dennekamp et al. (2010) conducted a case-crossover analysis of air pollution and OHCA
(n=8434) and found an increase of 4.26 µg/m3 in PM2.5 over two days (average of lag 0 and 1)
15
Exposure to Fine Particulate and Ozone Ambient Air Pollution
was associated with an increase of OHCA risk of 3.6% (95% CI 1.3 to 6.0). While the two-day
distributed lag model was the strongest effect, significant effects were also found at the single lag
models: day of the event (2.44%, 95% CI 0.54 to 4.37) and previous day (2.46%, 95% CI 0.33 to
4.65). The study did not find an effect for longer lags of PM2.5 or for any lag of ozone. Men were
found to be more susceptible than women, and the largest effect was seen in age group 65-74.
Ensor et al. (2012) conducted a case-crossover analysis of air pollution and OHCA (n=11,677)
and found an increase of 6 µg/m3 in PM2.5 over two days (average of lag 1 and 2) was
associated with an increase of OHCA risk of 4.6% (95% CI 1.2 to 8.2). While the two day
average lag was the strongest effect, significant effects were also found for the previous day
3.5% (95% CI 0.5 to 6.6) and two days prior 3.7% (95% CI 0.7 to 6.8). The study did not find an
effect on OHCA rates for PM2.5 levels more than two days out. The study also found an increase
of 20 ppb ozone for the eight-hour average daily maximum was associated with an increased risk
of OHCA on the day of the event, and the 20 ppb ozone increase was associated with an increase
of OHCA on the previous one, two, and three hours before the event: day of event 3.8% (95% CI
0.4 to 7.2); one hour prior 4.2% (95% CI 0.4 to 8.2); two hours before 4.6% (95% CI 0.8 to 8.7);
three hours before 4.0% (95% CI 0.2 to 8.0). This was the first study to find a significant effect
with respect to ozone. The study acknowledges the potential confounding from OHCA time of
day with ozone time of day since OHCAs are known to have an hourly pattern (e.g., peaking in
the morning). However, the significance seen on the day of the event (lag 0) is indicative that the
hour result is not simply picking up on the OHCA hourly pattern.
16
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Table 4. PM2.5 Study Findings
Time Frame
Authors
No. of Cases
PM2.5 % Change
(95% CI) (µg/m3)
Seattle and
King County,
WA
October 1988
to July 1994
Seattle and
King County,
WA
Indianapolis, IN
New York,
NY
Melbourne, AUS
Houston,
TX
1985 to 1994
July 2002 to July
2006
2002 to 2006
2003 to 2006
2004 to 2011
Dennekamp et al.
Ensor et al.
8,434
11,677
3.61 (1.29 to 5.99)
3.5 (.5 to
6.6)
Levy et al.
Sullivan et al.
Rosenthal et al.
Silverman et
al.
362
1,206
1,374
8,216
-10.7 (-22.1 to
2.4)
-6 (-12 to 2)
2 (-6 to 11)
12 (1 to 25)
4 (-1 to 8)
2.44 (.54 to 4.37)
Daily average
(24-hr avg.)
Daily average
(24-hr avg.)
Daily average
Hour
IQR at Location
Nephelometry
(.51 X 10-1 km1
bsp)
13.8 µg/m3
based on
Nephelometry
(.54 X 10-1 km1
bsp)
10 µg/m3
10 µg/m3
4.26 µg/m3
6 µg/m3
Lag
1 (day)
1 (day)
0 (day)
0 to 1 (day)
0 to 1 (day)
0 (day)
1 (day)
PM2.5 Metric
Daily average
Daily average
Daily
average
Table 5. Ozone Study Findings
Rome, Italy
New York, NY
Melbourne, AUS
Houston, TX
Time Frame
1998 to 2000
2002 to 2006
2003 to 2006
2004 to 2011
Authors
Forastiere et al.
Silverman et al.
Dennekamp et al.
Ensor et al.
No. of Cases
5,144
8,216
8,434
11,677
O3 % Change
(95% CI) (ppb)
5 (-1 to 11)
(estimated from time-series
design) not statistically
significant
2.94 (-2.42 to 8.59)
3.8 (.4 to 7.2)
-1 (-12.5 to 12)
3.78 (-1.32 to 9.14)
4.6 (.8 to 8.7)
Ozone Metric
Daily average
(Apr-Sep)
Daily average
Daily average
8 hour daily max
Hour
IQR at Location
71.4 µg/m3 (Apr-Sep)
35.7 ppb
22 ppb
8.02 ppb
20 ppb
Lag
0 (day)
0 to 1
0 to 1 (day)
0 (day)
0 (day)
2 (hr)
The Link is Established
The odds ratio of increased risk of OHCA associated with an increase of 10 µg/m3 daily average
PM2.5 on the day, day before, or the average of the day of OHCA onset and the day before onset
are presented in Figure 1 for four of the six studies examining the relationship. The results shown
in the figure have been scaled from the IQR of the pollution in the study location to an equivalent
17
Exposure to Fine Particulate and Ozone Ambient Air Pollution
IQR, 10 µg/m3 of PM2.5, for comparison between studies. The IQR of the PM2.5 in the study
location is noted in the figure.
The two studies not included in the plot, Levy et al. (2001) and Sullivan et al. (2003), are
different from the other four because they did not use an EPA reference method to measure
PM2.5 and are therefore not directly comparable. These two studies both found a null result (i.e.,
no association). The authors speculate that low statistical power may be an issue given that the
number of cases was small (n= 362 and 1206). Equally important, the null results these studies
found may be due to the composition of the particulate in the study area, both in Seattle. Seattle
particulate is relatively sparse in transition metals and sulfites and is dominated by wood smoke.
There is growing evidence that the composition of particulates is an important consideration
when studying the health impact (Franklin et al. 2008; De Hartog et al. 2009).
As depicted in Figure 1, statistically significant effects are found with an increasing number of
cases. While the study of the association in New York City (Silverman et al. 2010) is not quite
significant at an increased risk of 4.5% (95% CI -.9 to 10) for a daily average increase of 10
µg/m3, the point estimate is in line with those in Melbourne (Dennekmap et al. 2010) and
Houston (Ensor et al. 2012).
Taken as a whole, results from studies that had more than 8,000 cases support the likelihood that
there is an increased risk of OHCA of perhaps 2% to 9% associated with 10 µg/m3 daily average
increase of PM2.5 on the day before, the day of, or the average of the day before and the day of
the OHCA onset.
The study of Indianapolis (Rosenthal et al. 2008) also found an association of OHCA and hourly
PM2.5 for OHCA that were witnessed of 12% (CI 1 to 25). This is shown as the last study on
Figure 1. While this is the first case-crossover study on OHCA to find an association on the
hourly scale, the results are supported by a case-crossover on a different cardiac endpoint. Peters
et al. (2001) examined the association of increased particulate air pollution and the triggering of
myocardial infarction. This study with a rather small number of cases (n=772) by Peters et al.
(2001) found a significant increase in relative risk with an increase in PM2.5 concentration and
18
Exposure to Fine Particulate and Ozone Ambient Air Pollution
found no effect for ozone. The unique finding of the study by Peters et al. (2001) was that the
association with PM2.5 was found at both the 24-hour average lag exposure time scale and the
hourly exposure time scale (two hours). The vast majority of cardiac and PM2.5 association
research has been focused on the daily scale only. The Peters et al. (2001) study, while not a
perfect match with respect to health endpoint, is similar enough to be considered as supporting
evidence of possibly more transient (hourly) time scale effects of PM2.5 air pollution on OHCA
as seen by Rosenthal et al. (2008).
Figure 1. Forest plot of city-specific odds ratios of OHCA associated with a 10 µg/m3 daily
average increase in PM2.5
*hourly scale, witnessed OHCA
Figure 2 presents the odds ratio of four studies examining the increased risk of OHCA associated
with an increase of 20 ppb in the daily average of ozone for various day lags (single lags 0,1,2
19
Exposure to Fine Particulate and Ozone Ambient Air Pollution
and/or average of 0-1 lag) and hour lags (0,1,2). Similar to the scaling of PM2.5 results, the
results shown in Figure 2 have been scaled from the IQR of the study location to an equivalent
IQR, 20 ppb of ozone, for comparison across studies. As before, the IQR of the ozone in the
study location is noted in the figure.
While the evidence in the figure regarding the association between ozone air pollution and
OHCA events is inconsistent, a closer examination of the comparability of the studies is
warranted.
The first study shown in Figure 2 of Rome (Forastiere et al. 2005) indicates a null result, or no
association between OHCA and ozone. This study may not be as representative as some of the
other studies for several reasons. First, the event size is smaller than the other studies; the study
of Rome was based on 5144 events, while there were 8216, 8434, and 11677 events used in the
studies in New York, Melbourne, and Houston, respectively. In addition, the ozone exposure
concentration in the Rome study was based on one monitor and the analytical method used to
quantify the ozone concentration was not reported (i.e., EPA reference method); the studies in
New York and Houston were based on data analyzed using an EPA reference method and the
exposure concentration metric represented data from a network of monitors—16 and 47
respectively—instead of a single location. Finally, the health endpoint is not specifically OHCA
but out-of-hospital coronary death.
The Melbourne study focused on OHCA as an endpoint and also was based on a large number of
cases. However, as in the Rome study, the Melbourne study used only one monitor location and
no ozone analytical method was reported. The potential for a larger uncertainty with the
exposure concentration defined by one location, coupled with the fact that the ozone pollution in
Melbourne is both the lowest in magnitude and variation (has the smallest IQR), also renders the
study of Melbourne less directly comparable to the study of New York City and Houston. Box
plots of ozone concentrations, reconstructed or estimated from the publications, and monthly
average temperature of the study location are shown in Figure 3.
20
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Overall, the studies of New York City and Houston are most comparable. They both have a large
number of cases extracted from an EMS 911 database, limited exposure concentration
uncertainty, and similar ozone IQR. The only obvious difference, besides the larger number of
cases in the study of Houston, is the difference in climate. Except that the results of New York
City are not significant at the 95% confidence level—4.5 (-0.9 to 10) for a daily average increase
of 20 ppb—the point estimate of the association between OHCA and ozone in New York is
approximately the same as that found to be statistically significant in Houston (Ensor et al.
2012): as much as a 4% increased risk of OHCA with a daily or eight-hour running maximum
average daily increase of 20 ppb ozone on the day or average of day and day before the onset of
the OHCA.
The results found in New York City and Houston are consistent with findings from an important
case-crossover study with a more encompassing health endpoint. Stafoggia et al. (2010)
examined susceptibility factors to ozone mortality. Of interest to our objective is their
examination of ozone-related mortality in those with preexisting cardiovascular conditions. The
researchers estimated a 5.1% (95% CI 0.65 to 19.45) increase in mortality for a 20 ppb increase
in the daily eight-hour ozone running maximum average.
Given the comparability between the studies of Houston and New York City and the
corroborating study by Stagoggia et al. (2010), current results of the comparable studies support
the likelihood that there is an increased risk of OHCA of a range of 1% to 8% associated with a
daily eight-hour maximum increase of 20 ppb on the day, or the average of the day of OHCA
onset and the day before onset.
As noted previously, the study of Houston (Ensor et al. 2012) also found an hourly association of
OHCA and ozone (e.g., 4.6% increase, 95% CI 0.8 to 8.7) at hour lag two. This is the first casecrossover study of OHCA to find an association with ozone on the hourly scale.
21
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Figure 2. Forest plot of city-specific odds ratios of OHCA associated with a 20 ppb daily
average increase in ozone
22
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Figure 3. Box plots of city-specific ozone concentrations and monthly average temperature
*The analytical method used to measure ozone concentrations in Rome and Melbourne was not
reported. Note: Concentrations are reconstructed or estimated from those reported in the
publications.
Are the Standards Protective?
The case-crossover analysis research discussed above indicates that there is evidence of an
increased risk of OHCA of approximately
•
2% to 9% associated with a 10 µg/m3 increase of the daily average PM2.5 on the day
before, the day of, or the average of the day before and day of the OHCA onset as well as
some evidence of risk on an hourly scale; and
•
1% to 8% associated with a daily or daily eight-hour running average maximum increase
of 20 ppb ozone on the day, or average of day and day before the onset of the OHCA, and
also some evidence of risk on an hourly scale.
23
Exposure to Fine Particulate and Ozone Ambient Air Pollution
While there are many possible methods to evaluate if the EPA standards of these pollutants are
protective (i.e., dose or concentration response curves), a logical way to evaluate the standards in
this study is to use the same data and approach that initially identified the risk.
Using the Houston study data, we conduct a case-crossover analysis identical to the initial
research examining the association of OHCA and exposure to PM2.5 and exposure to ozone;
however, we remove the OHCA events from the analysis where the concentration on the day of
the event was above the respective standards. Then we compare the increased risk from the
initial research and the risk from the analysis where days above the standard are eliminated and
see if the risk changes and by how much. If the hypothetical dataset where only concentrations
below the standard occur results in no increased risk, we deem the standard protective for
OHCA.
The current (as of June 2012) National Ambient Air Quality Standard (NAAQS) for PM2.5 is
two-pronged; to protect against short-term effects, the 24-hour average must not exceed 35
µg/m3 and to protect against long-term effects, the annual average must be less than 15 µg/m3.
The current (as of June 2012) NAAQS for ozone is an eight-hour maximum of 75 ppb of the 24
possible running eight-hour average concentrations for each day.
The focus of this research is on short-term health effects; therefore, only the PM2.5 24-hour
average standard of 35 µg/m3 and the ozone daily eight-hour average maximum of 75 ppb are of
interest. After removing all events (and referents) from the data set where the PM2.5
concentration was 35 µg/m3 or above, the increased risk associated with OHCA and PM2.5
remained 3.4% (95 CI 0.4 to 6.5) (see Figure 4). In fact, there were only two events that were
eliminated from the original dataset because they exceeded the standard. Based on this analysis,
the 35 µg/m3 standard is not effective at protecting the public with respect to OHCA triggered
from exposure to PM2.5. The increased risk of OHCA from exposure to PM2.5 is occurring at
levels lower than the standard.
24
Exposure to Fine Particulate and Ozone Ambient Air Pollution
On the other hand, after removing all events (and referents) from the data set where the ozone
concentrations were above the daily eight-hour maximum of 75 ppb, the results drastically
changed, and there was no statistically significant risk of OHCA associated with ozone at any
short-term metric: daily eight-hour maximum running average and one- or two-hour average
(Figure 4). In this case, there were 139 events that were eliminated from the original dataset
because they exceeded the standard. This fraction of events eliminated would not substantially
reduce the statistical power of the analysis. The ozone standard appears to be effective at
protecting the public with respect to OHCA triggered from exposure to ozone.
Figure 4. Forest plot of increased risk of OHCA associated concentrations of PM2.5 and ozone
comparing cases of those observed in Houston from 2004 to 2011, and a hypothetical database of
those during this time frame with only events where the concentration was below the standard.
25
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Conclusion
The seven studies identified for review occurred within the last 15 years, addressed out-ofhospital cardiac arrest as the health endpoint, and considered the ambient pollutants of PM2.5
and ozone. Further, the chosen method of analysis discussed was the case-crossover design
coupled with a conditional logistic regression of the resulting case-control events. The casecrossover design allows subjects to serve as their own control and thereby mitigates the impact
of confounding subject factors. Several of the seven studies also performed a comparative and/or
confirmatory analysis using Poisson time series regression.
The studies vary in the number of cases considered as well as the quality and certainty of the
measure of pollution. These factors are discussed extensively in the paper. One of the most
critical issues to consider is the quality of the pollution measurements. Another issue with PM2.5
is the composition of the particulate matter limits our comparability, as differential composition
has shown differential impact on health outcomes. Studies such as those in New York City and
Houston, which rely on a large number of monitors throughout the study region, provide a better
representation of the pollution for the region over studies that rely on a single monitor.
In general, there is evidence of a 2% to 9% range of increased risk of OHCA due to increases of
10 µg/m3 daily average PM2.5 levels on a given day, as well as the day before. Further, there is
nascent evidence that the impact may be more local in time, namely within an hour of the
OHCA. When examining the ozone, we observe an approximate 1% to 10% range of increased
risk of OHCA when the maximum daily eight-hour average levels increase 20 ppb ozone on the
day and day before the onset of the OHCA. Again, there is evidence of increased risk due to
ozone changes at an hour time scale.
A further examination of the data from Houston demonstrates the effectiveness of the ozone
standard in protecting the population from an OHCA event triggered from exposure to ozone but
a failure of the standard in achieving this goal for PM2.5.
This study was completed for the Health Policy Forum at Rice University’s Baker Institute.
26
Exposure to Fine Particulate and Ozone Ambient Air Pollution
References
Basu, R., and J.M. Samet. 2002. “Relation between elevated ambient temperature and mortality:
a review of the epidemiologic evidence.” Epidemiologic Reviews 24:190-202.
Bateson, T.F., and J. Schwartz. 1999. “Control for seasonal variation and time trend in casecrossover studies of acute effects of environmental exposures.” Epidemiology 10:539–
544.
Bell, M.L., A. McDermott, S.L. Zeger, et al. 2004. “Ozone and short-term mortality in 95 U.S.
urban communities, 1987–2000.” 2004. The Journal of the American Medical
Association 292:2372–2378.
Braga, A.L.F., A. Zanobetti, and J. Schwartz. 2001. “The time course of weather-related deaths.”
Epidemiology 12:662-667.
Curriero, F.C., K.S. Heiner, J.M. Samet, et al. 2002. “Temperature and mortality in eleven cities
of the Eastern United States.” American Journal of Epidemiology 155:80-87.
Dennekamp, M., Akram M., Abramson M.J., et al. 2010. “Outdoor air pollution as a trigger for
out-of-hospital cardiac arrests.” Epidemiology 21:494-500.
De Hartog, J.J., and Lanki T., et al. 2009. “Associations between PM2.5 and heart rate variability
are modified by particle composition and beta-blocker use in patients with coronary heart
disease.” Environmental Health Perspectives 117:105-111.
Dockery, D.W., H. Luttmann-Gibson, D.Q. Rich, et al. 2005. “Association of air pollution with
increased incidence of ventricular tachyarrhythmias recorded by implanted cardioverter
defibrillators.” 2005. Environmental Health Perspectives 113(6):670–674.
Ensor, K.B., L. Raun, and D. Persse. Submitted June 2012. “A case-crossover analysis of out-ofhospital arrest and air pollution,” submitted to Environmental Health Perspectives.
Environmental Protection Agency (EPA). 1971. Clean Air Act, Section 108 and 109. Available
at http://www.epa.gov/oar/caa/title1.htm.
EPA. 2005. “Air Quality Criteria Particulate Matter.” EPA/600/P-99/002bf. National Center for
Environmental Assessment, Office of Research and Development, Research Triangle
Park, NC.
27
Exposure to Fine Particulate and Ozone Ambient Air Pollution
EPA. 2006. “Air Quality Criteria for Ozone and Related Photochemical Oxidants.” EPA/600/R05/004aF-cF. National Center for Environmental Assessment, Office of Research and
Development, Research Triangle Park, NC.
EPA. Reference method. http://www.epa.gov/ttnamti1/files/ambient/criteria/referenceequivalent-methods-list.pdf.
Forastiere F., M. Stafoggia, and S. Picciotto, et al. 2005. “A case-crossover analysis of out-of
hospital coronary deaths and air pollution in Rome, Italy.” American Journal of
Respiratory Critical Care Medicine 172(12):1549–1555.
Franklin M., and P. Koutrakis, et al. 2008. “The role of particle composition on the association
between PM2.5 and mortality.” Epidemiology 19:680-689.
Gold D.R., A. Litonjua, and J. Schwartz, et al. 2000. “Ambient pollution and heart rate
variability.” Circulation 101(11):1267–1273.
Greenland, S. 1996. “Confounding and exposure trends in case-crossover and casetime-control
designs.” Epidemiology 7:231–239.
Guo, Y., A.G. Barnett, Y. Zhang, et al. 2010. “The short-term effect of air pollution on
cardiovascular mortality in Tianjin, China: Comparison of time series and case–crossover
analyses.” Science of the Total Environment 409:300-306.
Hanley, J.A. 2001. “A heuristic approach to the formulas for population attributable fraction.”
Journal of Epidemiology and Community Health 55:508-514.
Ito, K., F.D.L. Samantha, and L. Morton. 2005. “Association between ozone and daily mortality:
Analysis and meta-analysis.” Epidemiology 16(4): 446-457.
Ji, M., D.S. Cohan, and M.L. Bell. 2001. “Meta-analysis of the association between short-term
exposure to ambient ozone and respiratory hospital admissions.” Environmental Research
Letters 6:11.
Kwon, H.J., S.H. Cho, F. Nyberg, et al. 2001. “Effects of ambient air pollution on daily mortality
in a cohort of patients with congestive heart failure.” Epidemiology 12: 413-419.
Lee, J.T. and J. Schwartz. 1999. “Reanalysis of the Effects of Air Pollution on Daily Mortality in
Seoul, Korea: A Case-Crossover Design. Environmental Health Perspectives 107: 633636.
Levy, D., T. Lumley, L. Sheppard, et al. 2001. “Referent selection in case-crossover analyses of
acute health effects of air pollution.” Epidemiology 12:186–92.
28
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Levy, D., L. Sheppard, et al. 2001. “A case-crossover analysis of particulate matter air pollution
and out-of-hospital primary cardiac arrest.” Epidemiology 12:193-199.
Lin, M., Y. Chen, R.T. Burnett, et al. 2002. “The influence of ambient coarse particulate matter
on asthma hospitalization in children: case-crossover and time-series analyses.”
Environmental Health Perspectives 110: 575-581.
Lu, Y., and S.L. Zeger. 2007. “On the equivalence of case-crossover and time series methods in
environmental epidemiology.” Biostatistics 8:337–344.
Lumley, T., and D. Levy. 2000. “Bias in the case-crossover design: implications for studies of air
pollution.” Environmetrics 11:689 -704.
Maclure, M. “The case-crossover design: a method for studying transient effects on the risk of
acute events.” 1991. American Journal of Epidemiology 133(2): 144-153.
McNally, B., R. Robb, M. Mehta, et al. 2011. “Out-of-Hospital Cardiac Arrest Surveillance—
Cardiac Arrest Registry to Enhance Survival (CARES), United States, October 1, 2005–
December 31, 2010.” National Center for Chronic Disease Prevention and Health
Promotion, CDC 60:1-19.
Moore, K., R. Neugebauer, F. Lurmann, et al. 2010. “Ambient Ozone Concentrations and
Cardiac Mortality in Southern California 1983–2000: Application of a New Marginal
Structural Model Approach.” American Journal of Epidemiology 171: 1233-1243.
Navidi, W. 1998. “Bidirectional case-crossover designs for exposures with time trends.”
Biometrics 54: 596–605.
Neas, L.N., J. Schwartz, and D. Dockery. 1999. “A case-crossover analysis of air pollution and
mortality in Philadelphia.” Environmental Health Perspectives 107: 629-631.
Peters, A., A. Doring, H.E. Wichmann, et al. 1997. “Increased plasma viscosity during an air
pollution episode: a link to mortality?” Lancet 349 (9065): 1582–1587.
Peters, A., D.W. Dockery, et al. 2001. “Increased particulate air pollution and the triggering of
myocardial infarction.” Circulation 103: 2810-2815.
Peters, A., E. Liu, R.L. Verrier, et al. 2000. “Air pollution and incidence of cardiac arrhythmia.”
Epidemiology 11(1): 11–17.
Peters, A., S. von Klot, N. Berglind, et al. 2006. “Comparison of different methods in analyzing
short-term air pollution effects in a cohort study of susceptible individuals.”
Epidemiologic Perspectives & Innovations 3:10.
29
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Pope, C.A. III, and D.W. Dockery. 2006. “Health effects of fine particulate air pollution: lines
that connect.” Journal of the Air & Waste Management Association 56: 709–742.
Riediker, M., W.E. Cascio, T.R. Griggs, et al. 2004. “Particulate matter exposure in cars is
associated with cardiovascular effects in healthy young men.” American Journal of
Respiratory and Critical Care Medicine 169(8): 934–940.
Roger, V.L., A.S. Go, D.M. Lloyd-Jones, et al. 2011. “Heart Disease and Stroke Statistics–2011
Update: A Report From the American Heart Association.” Circulation 123: e18-e209.
Rosenthal, F., J.P. Carney, and M.L. Olinger. 2008. “Out-of-hospital cardiac arrest and airborne
fine particulate matter: a case-crossover analysis of emergency medical services data.”
Environmental Health Perspectives 116 (5): 631-636.
Silverman, R.A., K. Ito, J. Freese, et al. 2010. “Association of ambient fine particles with out-ofhospital cardiac arrests in New York City.” American Journal of Epidemiology 172: 917923.
Srebot,V., E.A. Gianicolo, G. Rainaldi, et al. 2009. “Ozone and cardiovascular injury.”
Cardiovascular Ultrasound 7:30.
Stafoggia, M., F. Forastiere, D. Agostini, et al. 2006. “Vulnerability to heat-related mortality: a
multicity, population-based, case-crossover analysis.” Epidemiology 17:315-323.
Steadman, R.G. 1979. “The assessment of sultriness. Part I: a temperature-humidity index based
on human physiology and clothing science.” Journal of Applied Meteorology 18: 861–
873.
Sullivan, J., N. Ishikawa, L. Sheppard, et al. 2003. “Exposure to ambient fine particulate matter
and primary cardiac arrest among persons with and without clinically recognized heart
disease.” American Journal of Epidemiology 157(6): 501-509.
Sunyer, J., J. Schwartz, A. Tobias, et al. 2000. “Patients with chronic obstructive pulmonary
disease are at increased risk of death associated with urban particle air pollution: a casecrossover analysis.” American Journal of Epidemiology 151: 50-56.
Tsai, S., W. Goggins, H. Chiu, et al. 2003. “Evidence of an association between air pollution and
daily stroke admissions in Kaohsiung, Taiwan.” Stroke. 34: 2612-2616.
Wellenius, G.A., M.R. Burger MR, B.A. Coull, et al. 2012. “Ambient Air Pollution and the Risk
of Acute Ischemic Stroke.” Archives of Internal Medicine 172(3): 229-234.
30
Exposure to Fine Particulate and Ozone Ambient Air Pollution
Xu, X., J.V. Zborowski, V.C. Arena, et al. 2008. “Case-Crossover Analysis of Air Pollution and
Cardiorespiratory Hospitalizations, Using Routinely Collected Health and Environmental
Data for Tracking, Science and Data.” Journal of Public Health Management Practice
14(6): 569–576.
31