Triggering of Transmural Infarctions, but Not Nontransmural Infarctions, by Ambient Fine Particles

Background Previous studies have reported increased risk of myocardial infarction (MI) after increases in ambient particulate matter (PM) air pollution concentrations in the hours and days before MI onset. Objectives We hypothesized that acute increases in fine PM with aerodynamic diameter ≤ 2.5 μm (PM2.5) may be associated with increased risk of MI and that chronic obstructive pulmonary disease (COPD) and diabetes may increase susceptibility to PM2.5. We also explored whether both transmural and nontransmural infarctions were acutely associated with ambient PM2.5 concentrations. Methods We studied all hospital admissions from 2004 through 2006 for first acute MI of adult residents of New Jersey who lived within 10 km of a PM2.5 monitoring site (n = 5,864), as well as ambient measurements of PM2.5, nitrogen dioxide, sulfur dioxide, carbon monoxide, and ozone. Results Using a time-stratified case-crossover design and conditional logistic regression showed that each interquartile-range increase in PM2.5 concentration (10.8 μg/m3) in the 24 hr before arriving at the emergency department for MI was not associated with an increased risk of MI overall but was associated with an increased risk of a transmural infarction. We found no association between the same increase in PM2.5 and risk of a nontransmural infarction. Further, subjects with COPD appeared to be particularly susceptible, but those with diabetes were not. Conclusions This PM–transmural infarction association is consistent with earlier studies of PM and MI. The lack of association with nontransmural infarction suggests that future studies that investigate the triggering of MI by ambient PM2.5 concentrations should be stratified by infarction type.


Research
Most previous studies (D'Ippoliti et al. 2003;Peters et al. 2001Peters et al. , 2005Pope et al. 2006;Zanobetti and Schwartz 2005), but not all , that have investigated the triggering of myocardial infarction (MI) by particulate matter (PM) air pollution con centrations in the hours and days before MI onset have reported an association. Other studies have reported increased mortality due to MI or increased mortality or cardiovascular admissions among MI survivors associated with increases in PM over the previous few days (Braga et al. 2001;von Klot et al. 2005;Zanobetti and Schwartz 2007). Several studies have investigated whether certain subgroups are particularly susceptible and have reported increased cardiovascular effects among those with diabetes (Dubowsky et al. 2006;Goldberg et al. 2001;Liu et al. 2007;O'Neill et al. 2005;Zanobetti and Schwartz 2002) and among patients with chronic obstruc tive pulmonary disease (COPD) (Naess et al. 2007;Zanobetti and Schwartz 2005;Zanobetti et al. 2000). However, Zanobetti and Schwartz (2005) reported increased sus ceptibility among patients with COPD but not among persons with diabetes.
Numerous researchers have reported that the percentage of infarctions that are nontrans mural has been increasing (Goff et al. 2000;Hellermann et al. 2003; Kostis et al. 2007; Myerson et al. 2009;Roger et al. 2006Roger et al. , 2010Rogers et al. 2008). More recently, Shao (2008) noted a similar secular trend in clinical presenta tion of MI to emergency departments (EDs) in New Jersey. In short, 72% of persons admitted to hospitals in New Jersey from 1990 through 1992 for MI had had trans mural infarctions, and only 28% had had nontransmural infarc tions. Since then, however, this pattern has reversed. From 2002 through 2004, most of the admissions for MI were for nontransmural infarctions (63%), with only 37% for trans mural infarctions (Shao 2008). These changes may be due in part to improvements in preven tive pharmacotherapies (statins, beta blockers, aspirin), interventional procedures [angioplasty, coronary artery bypass graft (CABG)], more sensitive diagnostic tests (troponins), and treat ment of the MI upon ED arrival (reperfusion therapy, increased use of antiplatelet agents). It has not been reported whether PM with aero dynamic diameter ≤ 2.5 µm (PM 2.5 ) triggers all infarctions, whether PM-MI associations differ in magnitude, or whether these asso ciations are restricted to PM-transmural or PM-nontransmural infarctions alone. Such investigations may provide insight into the mechanisms by which PM may trigger cardio vascular events.
Using the same data as Shao (2008), we attempted to replicate, in New Jersey, previous MI-PM 2.5 studies that were con ducted in other U.S. and European cities. We hypothesized that increases in mean PM 2.5 concentration on the same day and a few days before ED arrival for MI may be associ ated with increased risk of MI and that these effect estimates would be greater for persons with diabetes and those with COPD than for individuals without these conditions. Further, we explored whether there were differences in risks associated with increases in ambient PM 2.5 concentration between transmural and nontransmural infarctions.

Study population and outcome definition.
Using the Myocardial Infarction Data Acquisition System (MIDAS), a statewide sur veillance system in New Jersey that combines hospital discharge data and death certificate registration data (Kostis et al. 1994(Kostis et al. , 2001, we extracted all records with a primary diag nosis of acute MI [International Classification of Diseases, version 9 (ICD9), codes 410. 01, 410.11, 410.21, 410.31, 410.41, 410.51, 410.61, 410.71, 410.81, and 410.91] for patients who were admitted between 1 January 2004 and 31 December 2006, were ≥ 18 years of age, were residents of New Jersey at the time of their MI, and who were without a previous diagnosis of MI (ICD9 code 412). Because the "1" in the fifth digit of the ICD9 code (e.g., 410.01) indicates a first MI for the subject, we also used this designation to exclude those persons who had had a previ ous MI. Those who were not admitted into the hospital (e.g., patient declined admission, Background: Previous studies have reported increased risk of myocardial infarction (MI) after increases in ambient particulate matter (PM) air pollution concentrations in the hours and days before MI onset. oBjectives: We hypothesized that acute increases in fine PM with aerodynamic diameter ≤ 2.5 µm (PM 2.5 ) may be associated with increased risk of MI and that chronic obstructive pulmonary disease (COPD) and diabetes may increase susceptibility to PM 2.5 . We also explored whether both transmural and nontransmural infarctions were acutely associated with ambient PM 2.5 concentrations. Methods: We studied all hospital admissions from 2004 through 2006 for first acute MI of adult residents of New Jersey who lived within 10 km of a PM 2.5 monitoring site (n = 5,864), as well as ambient measurements of PM 2.5 , nitrogen dioxide, sulfur dioxide, carbon monoxide, and ozone. results: Using a time-stratified case-crossover design and conditional logistic regression showed that each interquartile-range increase in PM 2.5 concentration (10.8 µg/m 3 ) in the 24 hr before arriving at the emergency department for MI was not associated with an increased risk of MI overall but was associated with an increased risk of a transmural infarction. We found no association between the same increase in PM 2.5 and risk of a nontransmural infarction. Further, subjects with COPD appeared to be particularly susceptible, but those with diabetes were not. conclusions: This PM-transmural infarction association is consistent with earlier studies of PM and MI. The lack of association with nontransmural infarction suggests that future studies that investigate the triggering of MI by ambient PM 2.5 concentrations should be stratified by infarction type. volume 118 | number 9 | September 2010 • Environmental Health Perspectives patient died before admission) were not included in this data set. We classified patients with an MI coded as 410.7 (subendocardial infarction) as having a nontransmural infarc tion, and those with codes 410.0 (infarction of anterolateral wall), 410.1 (infarction of other anterior wall), 410.2 (infarction of infero lateral wall), 410.3 (infarction of infero posterior wall), 410.4 (infarction of other inferior wall), 410.5 (infarction of other lateral wall), and 410.6 (true posterior wall infarction) as having transmural infarctions.
This study and the original MIDAS study were approved by the University of Medicine and Dentistry of New Jersey-New Brunswick Institutional Review Board. MIDAS was also approved by the New Jersey Department of Health and Senior Services Institutional Review Board.
Air pollution. Using ambient pollut ant measurements from the New Jersey Department of Environmental Protection and from U.S. Environmental Protection Agency (EPA) Web sites (U.S. EPA 2008), we used hourly concentrations of PM 2.5 (7 monitoring stations), nitrogen dioxide (NO 2 ; 9 stations), sulfur dioxide (SO 2 ; 14 stations), carbon mon oxide (CO; 13 stations), and ozone (O 3 ; 15 stations) for the study period 1 January 2004 to 31 December 2006. For each patient, we calculated the distance between each PM 2.5 monitor (in operation at the time of the MI) and the patient's residence and assigned PM 2.5 measurements from the closest monitor to their residence. Those patients who lived > 10 km from a PM 2.5 monitoring station were excluded from PM 2.5 analyses. We calculated mean PM 2.5 concentrations for each succes sive 24hr period before ED arrival for the MI (e.g., mean of hours 0-23 before ED arrival; 24-47 mean of hours before ED arrival, etc.). If > 6 hr (> 25%) of a 24hr period of PM 2.5 concentrations were missing, we set the mean for this 24hr period of PM 2.5 concentration to missing. We used these mean concentra tions in all analyses. We repeated this monitor matching and mean concentration calculation process for each of the other pollutants.
Weather. Hourly temperature and dew point measurements were made at the Newark, Caldwell, Somerset, and Trenton, New Jersey, airports during the study period. We used the airport monitor closest to each patient's residence to provide the weather observations for that patient during the study period. After calculating 24hr mean tem perature and dew point, in the same man ner as for the pollutant concentrations, we calculated 24hr mean apparent temperature (Steadman 1979;Zanobetti and Schwartz 2005) as a measure of each patient's perceived air temperature given the humidity and used these values in all analyses.
Study design. We used a timestratified casecrossover design (Levy et al. 2001;Maclure 1991) that has previously been used in studies of ambient air pollution and MI (D'Ippoliti et al. 2003;Pope et al. 2006;Sullivan J et al. 2005;  Schwartz 2005), as well as other cardiovascular outcomes (Rich et al. 2005(Rich et al. , 2006a(Rich et al. , 2006bWellenius et al. 2005). In this design, each patient contributed information as a case dur ing the period immediately before the MI and as a matched control during times when an MI did not occur. The casecrossover design is analogous to a matched case-control study, but instead of estimating the relative risk of MI comparing exposure between persons (i.e., cases vs. controls), we estimated the relative risk of MI comparing exposure during differ ent time periods within the followup time of each case of MI. Because case periods and their matched control periods are derived from the same person and a conditional analysis is con ducted, nontime-varying confounders such as age, comorbidities, and history of longterm smoking were controlled by the study design. However, variables that may be related to both air pollution and the incidence of MI that vary over short time periods, such as weather conditions, which vary day to day, were pos sible confounders and were included in our analytic models. Case periods were defined as the 24hr period before ED admission for MI, whereas control periods (three or four per case depending on the number of days in the cal endar month) were matched to the case period by day of the week, time of the day, year, and month. For example, if a person arrived at the ED with an MI at exactly 0000 hours on 20 March 2009, then the 24hr case period was the prior 24 hr (0000 hours to 2359 hours on Thursday 19 March), and the control peri ods were 0000 hours to 2359 hours on the three previous Thursdays (5, 12, and 26 March 2009). Pollutant concentrations corresponding to these case and control periods were con trasted in all analyses. Main analysis. Using a conditional logistic regression model stratified on each MI (one case and three or four control periods), we regressed case-control status (i.e., case period = 1, control period = 0) against the mean PM 2.5 concen tration in the 24 hr before ED arrival. Using Akaike's information criterion to select the opti mal lag time and number of degrees of freedom for apparent temperature, we also included a natural spline (3 degrees of freedom) of the mean apparent temperature in the 48 hr before ED arrival in the model. We then reran this same model, replacing the 24hr mean PM 2.5 concentration (for both case and control peri ods) with one of six lagged mean PM 2.5 con centrations (i.e., mean of hours  to estimate the risk of MI associated with each lagged PM 2.5 concentration. From each model, we present the odds ratio (OR) and its 95% confidence interval (CI) scaled to the interquartile range (IQR) of PM 2.5 observed during the study period. A pvalue < 0.05 was used to indicate statistical significance.
Next, we examined whether the risk of MI associated with an IQR increase in mean PM 2.5 concentration in the 24 hr before ED arrival was different for patients with transmu ral versus nontransmural infarctions. We reran the same model described above for patients with transmural infarctions only, and then for patients with nontransmural infarctions only. To examine potential effect modification by other factors, including COPD, diabetes, age (< 65 years, ≥ 65 years), sex, race (white, black, other), and season (winter = December, January, February; summer = June, July, August), we used an interaction term (e.g., COPD × PM 2.5 ) in the main model. Sensitivity analyses. To assess the stability of our PM 2.5 relative risk estimates after adjust ing for gaseous pollutant concentrations, we created twopollutant models (PM 2.5 + NO 2 , PM 2.5 + CO, PM 2.5 + SO 2 , PM 2.5 + O 3 ) using mean pollutant concentrations from the 24 hr before ED arrival. We also evaluated whether any association between ED arrival for MI and the mean PM 2.5 concentration in the previous 24 hr was independent of concentrations from previous 24hr periods. We reran the same conditional logistic regression model described above, including all seven lagged PM 2.5 con centrations (i.e., the mean PM 2.5 concentra tions from the 24 hr before ED arrival for MI, and the mean concentrations from the previous six lagged 24hr periods).
All data sets included in the analyses were constructed using SAS software (version 9.1.3; SAS Institute Inc., Cary, NC), and all analy ses were conducted using R (version 2.6.1; R Foundation for Statistical Computing, Vienna, Austria). The authors had full access to the data and take responsibility for its integrity.

Results
During the study period, a total of 37,791 patients were admitted to nonfederal New Jersey hospitals for a first acute MI; of these, 5,864 lived within 10 km of a PM 2.5 monitor ing site and had PM 2.5 data available for analy sis (i.e., PM 2.5 data available for at least 18 of the 24 hr before ED arrival for the index MI). Study patients were predominantly older (59% > 65 years of age), white (69%), and male (56%). Sixtytwo percent had hyper tension, 61% had a history of ischemic heart dis ease, and 30% had diabetes (Table 1). Those who were excluded from our analysis were similar to persons who were included in age (61% > 65 years of age), sex (57% male), and comorbidities (e.g., 59% with hyper tension), but they were slightly more likely to be white (82%). Subjects included in the analy sis who experienced a nontransmural infarction were generally older and were more likely to have had cardiorespiratory comorbidities than were those who experienced a transmural infarc tion. Compared with the subjects who had a nontransmural infarction, those who had a transmural infarction were more likely to have had an angioplasty and/or catheteriza tion postinfarction but less likely to have had CABG (Table 1). Of note, there were seven monitors that made PM 2.5 measure ments in 2006 but only three in 2004 and 2005, which resulted in a larger number of patients living within 10 km of a PM 2.5 monitoring site and thus available for analysis in 2006 (Table 2).
Of the seven monitors that were con tinuously measuring PM 2.5 during the study period, patients were most often assigned to the Elizabeth Lab (32%), Camden Lab (26%), and New Brunswick (18%) monitors. The dis tribution of mean daily PM 2.5 concentrations at each of these seven monitors is shown in Table 2. When combining all the monitors, PM 2.5 concentrations had a median of 11.7 µg/m 3 ; 5th and 95th percentiles of 3.6 and 31.5 µg/m 3 , respectively; and 25th and 75th percentiles of 7.4 and 18.3 µg/m 3 , respectively (IQR = 10.8 µg/m 3 ). We scaled all our effect estimates presented below by this IQR. Table 3 shows Pearson correlation coefficients for indi vidual pollutants and apparent temperature.
Next, we separately estimated the risk of ED admission for MI associated with each IQR increase in the mean PM 2.5 concentra tion in the previous 24 hr, and lagged 24hr periods (hours 24-47, 48-71, 72-95, 96-119, 120-143, and 144-167), adjusting for appar ent temperature during that same lag period. Each 10.8µg/m 3 increase in the mean PM 2.5 concentration in the 24 hr before ED arrival was not associated with an increased risk of MI (Table 4). Similarly, we found no associations between ED admission for overall MI and any of the lagged PM 2.5 concentrations. However, when we then restricted our analysis to trans mural infarctions only we found a significantly increased risk associated with each 10.8µg/m 3 increase in PM 2.5 concentration in the previous . When examining lagged PM 2.5 concentrations, the relative risk estimates were mostly > 1.0, but none was statistically significant. In contrast, when we restricted our analysis to nontrans mural infarctions, we found no increased risks associated with any lagged PM 2.5 concentra tions (Table 4). Because we found an association with transmural infarctions only, we restricted all further analyses to this infarction type and evaluated whether several factors modified this association. Patients with preexisting COPD and those < 65 years of age had substantially larger risks of transmural infarction associ ated with each 10.8 µg/m 3 increase in PM 2.5 concentration in the previous 24 hr than did patients without preexisting COPD and those ≥ 65 years of age (Table 5). We found no dif ference in PM 2.5 relative risk estimates by race, sex, season, or whether subjects had diabetes (Table 5).
We then included mean PM 2.5 and NO 2 concentrations from the 24 hr before ED arrival simultaneously in a model (n = 1,262 patients with transmural infarctions with both PM 2.5 and NO 2 mean concentrations). The PM 2.5 relative risk estimate in this twopollutant model was not substantially different from the singlepollutant model on the same n = 1,262 patients (Table 6). This was also true when adjusting for CO, SO 2 , and O 3 . IQR increases in CO, SO 2 , and O 3 were not associated with significantly increased risks of transmu ral infarctions in any single or twopollutant model (Table 6). Next, when including all seven lagged PM 2.5 concentrations in the same model, the risk associated with the 24hr mov ingaverage PM 2.5 concentration was larger and still statistically significant (OR = 1.15; 95% CI, 1.04-1.28). All the PM 2.5 concen tration relative risk estimates for the other lag periods were smaller than the 24hr moving average relative risk estimate, and none was statistically significant (data not shown).

Discussion
Using a large multiyear (2004)(2005)(2006) statewide data set of hospital admissions for first MI, we found no association between admission for MI overall and PM 2.5 concentrations in the previous week. However, when we restricted the analysis to patients with transmural infarc tions, we found a significant 10% increase in the risk of a transmural infarction associated with the mean PM 2.5 concentration in the 24 hr before ED arrival. This association per sisted with adjustment for gaseous pollutant concentrations in the prior 24 hr and with adjustment for PM 2.5 concentrations during each of the previous six 24hr periods. Further, patients with COPD, but not diabetes, were particularly susceptible to acute increases in ambient PM 2.5 concentrations. We found no association with this same PM 2.5 concentra tion and nontransmural infarctions.
Our findings are consistent with previous studies (D'Ippoliti et al. 2003;Peters et al. 2001;Peters et al. 2005;Pope et al. 2006;Zanobetti and Schwartz 2005) in indicating an acute association between PM and MI onset, although the method of estimating MI onset time (day of hospital admission for MI, time of symptom onset estimated by patient, etc.), lags of pollutant concentrations examined (lag days 0-6, 0-6 lagged 24hr periods before ED arrival for MI), and specific particle sizes exam ined (PM 10 , PM 2.5 ) are not uniform. Previous studies reported PM associations with all MIs, not just transmural infarctions (D'Ippoliti et al. 2003;Peters et al. 2001;Peters et al. 2005;Pope et al. 2006;Zanobetti and Schwartz 2005). However, these studies drew data from earlier time periods [e.g., Peters et al. (2001), Boston, Massachusetts, 1995-1996Peters et al. (2005), Augsburg, Germany, 1999D'Ippoliti et al. (2003), Rome, Italy, 1995-1997Zanobetti and Schwartz (2005), 21 U.S. cities, 1985-1999] when most infarc tions were likely transmural. Therefore, their reports of increased relative risk of MI are consistent with our finding of increased rela tive risk of transmural infarctions only. Thus, our study adds to the body of knowledge that relates increased risk of MI with increases in PM in the previous hours and days, but ours is the first to report that associa tions are restricted to transmural infarctions.
Pathophysiologic pathways proposed as mechanisms underlying previously reported PM/MI associations include systemic inflam mation, endothelial dysfunction, disturbance of autonomic tone, and enhanced coagulation/ thrombosis (Brook et al. 2004). Our finding of an acute (within 24 hr) association between PM and transmural infarctions, but not nontransmural infarctions, suggests another mechanism or pathway, perhaps related to those listed above, by which particles may trigger an MI. In a recent study, Bartoli et al. (2009) reported decreased myocardial blood flow in canines associated with concentrated air particles exposure, but not with filtered air exposure, after experimental occlusion of the coronary artery. Cardiac work was not  increased by PM in this dog model, and data suggested that blood flow reductions were due to increased coronary vascular resistance, perhaps related to more limited recruitment of collateral vessels. Thus, particle inhalation, by reducing compensatory mechanisms, may transform limited (nontransmural) injury into more extensive (transmural) infarction. This finding is consistent with other research in which PM has been associated with increased rates of STsegment depression in humans Pekkanen et al. 2002) or in canine model markers of ischemia (Wellenius et al. 2003). There is also increasing support in the literature for the prothrombotic effects of air pollutants that indicates the possibil ity that limitations in revascularization after a primary plaquerelated thrombotic event could be due to enhanced thrombosis in gen eral (Delfino et al. 2009;Jacobs et al. 2010;Lucking et al. 2008). A recent study that examined the association between PM and acute coronary events demonstrated greater relative risk in those individuals with angio graphically documented previous coronary disease (Pope et al. 2006). Future analyses using data from MIDAS will include exam ining differences in susceptibility to PM by prior disease events, including MI. Ambient PM has previously been asso ciated with increased hospital admissions for COPD and asthma (Chen et al. 2004;Dominici et al. 2006;MedinaRamon et al. 2006;Peel et al. 2005;Zanobetti and Schwartz 2003). COPD is associated with a greater propensity to hypoxia, reduced pul monary reserve, and a generally heightened inflammatory state, which all may predispose to transmural infarcts.
Although several researchers have reported acute associations between PM and cardio vascular outcomes in panels of patients with diabetes or have reported greater changes in cardiorespiratory biomarkers among those with diabetes than those without diabetes (Dubowsky et al. 2006;Goldberg et al. 2001;Liu et al. 2007;O'Neill et al. 2005;Zanobetti and Schwartz 2002), we did not find greater susceptibility to PM among patients with dia betes. Similar to the findings of Zanobetti and Schwartz (2005), we found larger relative risks for patients with COPD than for those with out this condition, but no difference in relative risk estimates for patients with diabetes versus those without diabetes.
Although our study has several strengths, including a large sample size, with subcatego ries of infarct and availability of continuous PM 2.5 concentrations from seven monitor ing stations during the study period, several limitations should be considered when inter preting our results. First, this study relies on administrative data only, and therefore some of the outcomes coded as an MI may not have been an MI, or those classified as transmural may have been miscoded as nontransmural, or vice versa. Specifically, we were limited to only ICD9 codes to classify an MI as a transmural or nontransmural. However, a data audit was previously conducted that verified the accu racy of MI diagnoses as well as the accuracy of information in the MIDAS data set (Kostis et al. 2001). Thus, any outcome misclassifi cation (MI vs. nonMI) was likely minimal. Further, any misclassification of transmural versus nontransmural was likely unrelated to ambient PM 2.5 concentrations, resulting in nondifferential exposure misclassification and underestimates of relative risk.
Second, our data set included only those subjects who experienced an MI and were admitted to a hospital, and excluded MI patients who died before hospital admission. If PM does trigger transmural infarctions, which might have a higher outofhospital mortality rate, our data set would include a lower number of PMtriggered transmural infarctions than actually occurred during the study period. Therefore, our estimate of a 10% increase in the risk of a transmural infarction associated with each 10.8 µg/m 3 increase in PM 2.5 may be conservative.
Third, linking MIs to hourly and daily air pollution fluctuations requires minimizing mis classification of the estimate of MI onset time, so as to minimize bias. Although four previous studies (Peters et al. 2001;Peters et al. 2005;Pope et al. 2006;Sullivan J. et al. 2005) have used patients' selfreported time of pain and symptom onset, this was not practical for our study, which relied on an administrative data set without personal interviews. Previous stud ies have reported median and mean delay times from symptom onset to ED arrival of 2.3-4.7 hr (Goldberg et al. 2002), with 44% of MI patients in Massachusetts arriving within 2 hr, and 78% arriving within < 6 hr (Goldberg et al. 2000). Thus, because there is often a delay of several hours from symptom onset to ED arrival, use of ED arrival hour instead of hour of symptom onset results in greater exposure mis classification and bias toward the null. Further, in our analysis, the mean PM 2.5 concentra tions in the 6, 12, and 24 hr before ED arrival were highly correlated (rvalues = 0.83-0.95). Therefore, we did not examine associations with ambient PM 2.5 concentrations < 24 hr before ED arrival, but instead focused on asso ciations between infarctions and the ambient air pollution concentration in the 24 hr before ED arrival for that MI. Our finding of an asso ciation between increased 24hr mean PM 2.5 concentrations and transmural infarction needs to be replicated in a study that better estimates MI symptom onset time. This study would also allow a more proper investigation of the risk of a transmural infarction associated with PM 2.5 concentrations < 24 hr before MI symptom onset than is possible in our analysis.
Fourth, because all of the 24 hr used to calculate the mean PM 2.5 concentration in the 24 hr before ED arrival may not have been before the onset of MI symptoms, this again was a source of nondifferential expo sure misclassification that may have resulted under estimates of relative risk. As reported by Lokken et al. (2009), this error in estima tion of symptom onset time and the result ing relative risk-rate underestimation may be substantial in studies of acute cardiovascular and cerebro vascular events and shortterm increases in PM. When comparing relative risk estimates based on symptom onset of stroke versus those based on hospital presen tation for stroke, they observed an approxi mately 40% under estimation of the relative risk of stroke when using time of hospital presentation (Lokken et al. 2009).
Last, we assigned PM 2.5 concentrations to all patients who lived < 10 km from a PM 2.5 monitor, regardless of how close they lived to the monitor or how much time they spent at locations other than their residence. Because this error is not likely differential with respect volume 118 | number 9 | September 2010 • Environmental Health Perspectives to when a patient had the MI, it likely resulted in nondifferential exposure misclassification and underestimates of relative risk. In the two pollutant analyses adjusting for gaseous pol lutant concentrations, gaseous pollutants may have greater degrees of spatial variability than does PM 2.5 within this 10km radius, resulting in residual confounding.

Conclusions
We found increased risk of transmural infarc tions, but not nontransmural infarctions, asso ciated with each 10.8 µg/m 3 increase in mean PM 2.5 concentration in the 24 hr before ED arrival. Further, patients with COPD, but not those with diabetes, appeared particularly sus ceptible to effects of ambient particles. If our findings are confirmed, future investigations of PM and MI triggering should stratify by type of MI, with particular emphasis on transmural infarctions. Future work should also investigate mechanistic explanations for these findings.