Changes in triggering of ST-elevation myocardial infarction by particulate air pollution in Monroe County, New York over time: a case-crossover study

Background Previous studies have reported that fine particle (PM2.5) concentrations triggered ST elevation myocardial infarctions (STEMI). In Rochester, NY, multiple air quality policies and economic changes/influences from 2008 to 2013 led to decreased concentrations of PM2.5 and its major constituents (SO42−, NO3−, elemental and primary organic carbon). This study examined whether the rate of STEMI associated with increased ambient gaseous and PM component concentrations was different AFTER these air quality policies and economic changes (2014–2016), compared to DURING (2008–2013) and BEFORE these polices and changes (2005–2007). Methods Using 921 STEMIs treated at the University of Rochester Medical Center (2005–2016) and a case-crossover design, we examined whether the rate of STEMI associated with increased PM2.5, ultrafine particles (UFP, < 100 nm), accumulation mode particles (AMP, 100-500 nm), black carbon, SO2, CO, and O3 concentrations in the previous 1–72 h was modified by the time period related to these pollutant source changes (BEFORE, DURING, AFTER). Results Each interquartile range (3702 particles/cm3) increase in UFP concentration in the previous 1 h was associated with a 12% (95% CI = 3%, 22%) increase in the rate of STEMI. The effect size was larger in the AFTER period (26%) than the DURING (5%) or BEFORE periods (9%). There were similar patterns for black carbon and SO2. Conclusions An increased rate of STEMI associated with UFP and other pollutant concentrations was higher in the AFTER period compared to the BEFORE and DURING periods. This may be due to changes in PM composition (e.g. higher secondary organic carbon and particle bound reactive oxygen species) following these air quality policies and economic changes. Electronic supplementary material The online version of this article (10.1186/s12940-019-0521-3) contains supplementary material, which is available to authorized users.


Introduction
Over the past decade, policy initiatives to improve air quality have been implemented nationwide and in New York State [1]. These measures included the lowering of the sulfur content of diesel and home heating fuels, use of particle regenerative traps, and nitrogen oxide controls on new heavy-duty diesel vehicles. Additionally, several actions to reduce sulfur dioxide (SO 2) and nitrogen oxide (NOx) emissions from power plants in upwind source areas have also been enacted (i.e. Cross-State Air Pollution Rule). Furthermore, economic changes have also led to changes in emissions [1,2]. The economic drivers include the recession of 2008 that resulted in generally decreased economic activity and thus lower emissions, and a dramatic decline in the price of natural gas that displaced coal as a fuel for electricity generation.
Short-term increases in PM 2.5 concentrations in the previous hours and days have been repeatedly associated with the triggering of myocardial infarction [8,[11][12][13][14][15][16][17], including prior Rochester studies reporting triggering of ST segment elevation myocardial infarction (STEMI) by short-term increases in ambient PM 2.5 concentrations in the previous hour [11,12]. Furthermore, Rich et al. reported a higher rate of MI associated with increased PM 2.5 concentrations when the PM 2.5 mixtures had high mass fractions of secondary PM species (sulfate, nitrate, and/or organics), compared to when the mixtures had low mass fractions of secondary PM species [18]. This result suggests that PM 2.5 high in secondary species, formed through active, oxidative atmospheric photochemistry, may be more likely to trigger myocardial infarctions than primary fine particles [18]. Given these findings, reduced concentrations of secondary PM species sulfate and nitrate following the policy initiatives described above may result in a lower rate of STEMI per unit mass of PM, compared to periods when PM contained a higher proportion of secondary PM.
Using data on patients with a STEMI treated at the University of Rochester Medical Center (URMC) and air pollutant concentrations measured in Rochester, we examined whether the rate of STEMI associated with increased PM (PM 2.5 , UFP, and AMP) concentrations in the previous few hours and days was modified by periods defined based on the time of pollutant source changes following these environmental policies. We hypothesized that increases in the rate of STEMI associated with the same incremental increase in PM concentrations would be smaller AFTER these pollutant changes occurred (2014-2016), compared to BEFORE (2005BEFORE ( -2007, and DUR-ING this period (2008)(2009)(2010)(2011)(2012)(2013). We also explored effect modifications by periods (BEFORE, DURING, and AFTER) of the rate of STEMI associated with concentrations of BC and gaseous pollutants SO 2 , O 3 , and CO.

Study population
We included all patients treated at the Cardiac Catheterization Laboratory (Cath Lab) at URMC in Rochester, NY for STEMI between January 1, 2005 and December 31, 2016, who resided within 15 miles of the pollution monitoring station in Rochester. If a patient experienced multiple STEMIs during the study period, we only included STEMI that occurred at least 72 h after a previous STEMI, resulting in 921 STEMI events in 912 patients available for analyses. Patients' demographic and clinical characteristics were also obtained. This study was approved by the University of Rochester Medical Center Research Subjects Review Board. Informed consent was not required since data were past events and the research could not affect treatment, there was no contact with study subjects, and all results are presented in aggregate only.

Outcome assessments
American College of Cardiology (ACC)/American Heart Association (AHA) guidelines were used at the time of Cath Lab admission to diagnose STEMI [19]. Specifically, STEMI was defined as ST segment elevation on the presenting electrocardiogram (ECG) of > 1 mm in ≥2 contiguous precordial leads, or in ≥2 adjacent limb leads, or new or presumed new left bundle branch block in the presence of angina or angina equivalent. ECG criteria are both necessary and sufficient for the diagnosis of STEMI in a patient deemed by the treating physician to have symptoms consistent with cardiac chest pain. Symptom onset time (date and hour) was self-reported by each patient (or kin if the patient was unable to communicate) upon arrival to the URMC Cath Lab.

Air pollution and meteorology measurements
The New York State Department of Environmental Conservation operates a routine air quality monitoring site in Rochester, New York, where PM 2.5 mass concentrations, SO 2 , O 3 , and CO were measured continuously throughout the study period (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016). However, the CO and SO 2 monitors were replaced near the end of 2010 with higher sensitivity models. The measured CO concentrations from 2007 to the installation of the new monitor were generally close to the instrument's detection limit (0.2 ppm). This site operated a single-wavelength (880 nm) aethalometer (Magee Scientific, Berkeley, CA, USA) to measure black carbon (BC) until mid-2008, and a two-wavelength unit (370 and 880 nm) from July 2008 to the present. Particle number size distributions were measured using a Scanning Mobility Particle Sizer (SMPS, TSI, Inc., Shoreview, MN). The concentrations were aggregated into several size groups including UFP (≤ 100 nm) and AMP (100-500 nm). The UFP were further divided into UFP11-50 nm (marker for nucleation and spark ignition vehicle emissions) [20] and UFP50-100 nm (marker of diesel vehicle emissions and residential wood burning; Aitken mode) [21][22][23]. These data have been presented and their trends analyzed in several recent studies [3,24]. NOx data were not available for the entire study period and thus were not included in our study. Temperature and relative humidity data measured at the Rochester International Airport were obtained from the National Climate Data Center. Hourly means of each pollutant and weather variables were used in all statistical analyses described below.

Study design
We used a time-stratified case-crossover design to estimate the rate of STEMI associated with air pollutant concentrations [25,26]. This design contrasts pollutant concentrations immediately before the acute STEMI (case-period) to other time periods when the patient did not have an acute STEMI (control period). Case periods of this study were defined as the 1, 3, 12, 24, 48, and 72-h periods prior to the time of STEMI symptom onset, with control periods (3-4 per case, depending on the number of days in calendar month) matched by calendar year, month, weekday, and hour of the day. Since these case and control periods were separated by 1 week, there was no overlap in the air pollutant concentrations included in the case and control periods. Further, since the case and control periods were from the same patient, time-invariant confounders such as age, gender, and chronic comorbidities were controlled by design. Factors that varied between the case and control periods (e.g., temperature and relative humidity) were potential confounders included in our models.

Statistical analysis
First, we computed the average concentrations of each pollutant (PM 2.5 , AMP, UFP, UFP 11-50 nm, UFP 50-100 nm, BC, SO 2 , O 3 , and CO) in the 1 (lag hour 0), 3 (lag hours 0-2), 12 (lag hours 0-11), 24 (lag hours 0-23), 48 (lag hours 0-47), and 72 (lag hours 0-71) hours prior to each symptom onset (i.e. case) time and its matched control time. If the time of symptom onset of a STEMI event was estimated to be in the first 29 min of the hour (e.g. 12:28), then lag hour 0 was defined as the previous hour (i.e. 11:00-11:59). If the STEMI symptom onset time was estimated to be in the 30th minute or after (e.g. 12:49), then lag hour 0 was defined as that same hour (i.e. 12:00-12:59). We then regressed the case-control status (i.e., case = 1, control = 0) against the mean PM 2.5 concentration at lag hour 0 of the case and control periods, using conditional logistic regression models stratified by each case-control set, adjusting for mean temperature and relative humidity during the same lag hours. We used Akaike's information criterion to select the optimal functional form (natural spline with 2, 3, 4, or 5 degrees of freedom versus 1 degree of freedom/linear) for these two covariates. One degree of freedom (linear) was selected for both variables and used in the model. We also estimated the rate of STEMI associated with each interquartile (IQR) increase in the average PM 2.5 concentration at longer lag hours (lag hours 0-2, 0-11, 0-23, 0-47, and 0-71) in separate models in the same manner. We then repeated these analyses for the other pollutants (AMP, UFP, UFP 11-50 nm, UFP 50-100 nm, BC, SO 2 , O 3 , CO). Three sensitivity analyses were performed. First, we examined whether the exposureresponse functions (i.e. relative rates in the model described above) were linear or non-linear, by comparing AIC from the model described above, to models where the rate of STEMI associated with increased pollutant concentration was modeled using a natural spline with 2, 3, 4, or 5 degrees of freedom. Second, we examined whether the pollutant/ STEMI association changed after further adjustment for holiday in the model. Third, we examined whether the pollutant/STEMI association was modified by age (< 65 vs. ≥65 years), sex (male vs. female), and history of smoking, diabetes, dyslipidemia, and heart failure (yes vs. no) in separate models by adding an interaction term (e.g. UFP *male) to the model described above. The IQRs used to scale effect estimates were the IQRs of the pollutant specific lag times, from the control periods during the entire study years (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016). Since the correlation among data resulting from multiple STEMI events contributed by the same patient is likely negligible (921 events contributed by 912 patients), we did not account for within-subject correlation in our analyses.
To examine whether the rate of STEMI associated with increased PM 2.5 concentration in the previous hour (lag hour 0) was modified by the period of pollutant source changes (BEFORE [2005][2006][2007], DURING [2008][2009][2010][2011][2012][2013], and AFTER [2014-2016]), we added two interaction terms of 'period' (a categorical variable with three levels) and PM 2.5 concentration to the model of main effect analysis described above. We then repeated this analysis for other lag hours (lag hours 0-2, 0-11, 0-23, 0-47, and 0-71) and the other pollutants (AMP, UFP, UFP 11-50 nm, UFP 50-100 nm, BC, SO 2 , O 3 , CO). To account for the unbalanced distribution of patients' age across the three periods, in sensitivity analysis, we added an interaction term between pollutant concentration and age (< 65 years vs. ≥65 years) to the model of effect modification by period described above. We then repeated this analysis for all patient characteristics with unbalanced distributions across the three periods (sex, smoking, diabetes, dyslipidemia, heart failure). Analyses were performed using SAS (version 9.4; SAS institute Inc., Cary, NC) and R (version 3.0.1, R foundation for statistical computing, Vienna, Austria, splines package). A p < 0.05 was used to define statistical significance.

Results
Most STEMI patients were male (71%) and non-Hispanic white (83%), and were on average 63 years of age (standard deviation: ± 14 years) ( Table 1). Thirty-one percent of patients were smokers. Patients in the BE-FORE period were generally older, less often male, and less often smokers compared to patients in the DURING and AFTER periods. The prevalence of a prior history of coronary artery bypass grafting (CABG), diabetes, and Denominator is all STEMIs (N = 921 for "All years", N = 249 for "Before", N = 372 for "During", and N = 300 for "After") heart failure was higher in patients in the BEFORE period, than other periods. However, the prevalence of dyslipidemia was lower in patients in the BEFORE period (Table 1). There were substantial decreases in the concentrations of almost all pollutants during the study period except for O 3 ( Table 2). From the BEFORE to the AFTER period, the median concentration of PM 2.5 decreased by~30%, AMP by4 0%, UFP by~45%, BC and CO by~50%, and SO 2 by9 0%. However, the median O 3 concentration increased by3 0% from the BEFORE period to the DURING and AFTER periods. Further description of these changes in Rochester and across New York State has been provided previously [1][2][3]. Pearson correlations between hourly pollutants concentrations are shown in Additional file 1: Table S1. PM 2.5 was moderately correlated with AMP (r = 0.62-0.69), but poorly correlated with UFP (r = 0.12-0.22) in all years. AMP was moderately correlated with UFP in the BEFORE (r = 0.53) and DURING (r = 0.57) period, but the correlation was weaker in the AFTER period (r = 0.29).
Over all periods, estimates of the relative rate of STEMI associated with increased PM 2.5 and AMP concentrations were all close to 1.0 and not statistically significant (Table 3). However, increased rates of STEMI were associated with increased UFP and UFP 11-50 nm concentrations in the previous 1 (lag hour 0) and 3 h (lag hours 0-2). IQR increases in UFP concentrations in the previous 1 and 3 h were associated with 12% (95%CI: 3%, 22%) and 11% (95%CI: 2%, 22%) increases in the rate of STEMI, respectively. Similar to UFP, IQR increases in the particle number concentrations of UFP 11-50 nm (nucleation mode and spark ignition vehicle emissions [20][21][22]) at lag hour 0 and lag hours 0-2 were associated with 13% (95%CI: 4%, 22%) and 12% (95%CI: 3%, 21%) increases in the rate of STEMI, respectively. However, increases in UFP 50-100 nm (diesel vehicle emissions and residential wood burning; Aitken mode [20][21][22]) concentrations were not significantly associated with increases in the rate of STEMI, with effect estimates for all lag hours slightly greater than 1.0.
Increases in the rate of STEMI associated with increased concentrations of UFP, UFP 11-50 nm, BC, and SO 2 were larger in the AFTER period compared to those in the BEFORE and DURING periods ( Fig. 1; Table 4). In the previous hour (lag hour with the largest UFP relative rate in Table 3), each IQR (3702 particles/cm 3 ) increase in UFP concentration was associated with 9% (95%CI: − 4%, 23%), 5% (95%CI: − 12%, 24%), and 26% (95%CI: 7%, 48%) increases in the rate of STEMI, in the BEFORE, DURING, and AFTER periods, respectively. This pattern remained even when separately including interaction terms between air pollution and age, sex, smoking, diabetes, dyslipidemia, and heart failure in the model (Additional file 1: Table S4). There were similar patterns for UFP in the previous 3 h, UFP 11-50 nm in the previous hour and previous 3 h, BC in the previous hour, and SO 2 in the previous 1 to 24 h. However, the pattern for CO relative rates was different. Each IQR (0.23 ppm) increase in CO concentration at lag hour 0 was associated with 9% (− 6 to 26%), 31% (2 to 68%), and 29% (− 15 to 96%) increases in the rate of STEMI, in the BEFORE, DURING, and AFTER periods, respectively, showing similarly larger relative rates in the DUR-ING and AFTER periods. We did not observe a pattern of effect modification by period for the associations between STEMI and PM 2.5 , UFP 50-100 nm, AMP, or O 3 .

Discussion
In Rochester, NY from 2005 to 2016, increased rates of STEMI were associated with increased concentrations of UFP, UFP 11-50 nm, BC, and CO in the previous 1 to 3 h, as well as increased SO 2 concentration in the previous 3 to 72 h. Further, we found larger rates of STEMI associated with UFP, UFP11-50 nm, BC, and SO 2 in the period after air quality policy implementation and economic changes (AFTER period: 2014-2016) compared to the periods BEFORE (2005)(2006)(2007) or DURING (2008-2013) these changes. Given concomitant changes in pollutant composition over the three periods, this difference in relative rate by period may reflect differential triggering of STEMI by specific PM constituents.
In our previous studies, Evans et al. reported a 17% increase in the rate of STEMI associated with each 7.1 μ g/ m 3 increase in PM 2.5 concentrations in the previous    [12]. When using data from 2005 to 2016, but the same analysis methods for subjects living within 15 miles of the monitoring station, we found only a 3% (− 5 to 12%) increased rate of STEMI associated with each 7.1 μg/m 3 increase in PM 2.5 in the previous hour. Differences in PM composition between the study periods and differences in study populations may explain, in part, this discrepancy. Most, but not all other casecrossover studies also reported increased risks/odds/rates of STEMI associated with short-term increases in PM 2.5 concentrations [13][14][15][16][17]27]. However, most of these studies did not have symptom onset time and thus were not able to examine triggering of MI by increased pollutant concentrations in the previous few hours, perhaps missing an etiologically important time window. Most of these studies observed the largest increased rate of STEMI associated with pollutant concentrations in the concurrent or previous day, with excess relative rate estimates (associated with each 10 μg/m 3 increase in PM 2.5 concentration) ranging from 5 to 15% [13,14,16,17].
In contrast with the lack of findings for PM 2.5 , we found a 12% increase in the rate of STEMI associated with each 3702 particles/cm 3 increase in UFP concentration in the previous hour. Analyses of different UFP size fractions (11-50 nm and 50-100 nm) suggested that this observed increase in STEMI rate associated with UFPs was primarily driven by UFP 11-50 nm, which are mostly gasoline emissions and nucleation mode particles [20,22]. UFPs comprise a large particle number concentration, but contribute little mass to PM 2.5 [28,29]. In the present study, PM 2.5 mass concentration and UFP particle number concentration were poorly correlated (r = 0.21). Both UFPs and PM 2.5 have been associated with adverse cardiovascular responses including systematic inflammation, oxidative stress, endothelial dysfunction, thrombosis and coagulation, and autonomic dysfunction [7,10,30,31]. However, compared to larger particle fractions (e.g. AMP), UFPs deposit more deeply into the lung and may even escape clearance mechanisms and translocate into the circulatory system [32,33]. Therefore, UFPs may exert adverse health effects via pathophysiological pathways different from PM 2.5 and AMP [30].
Consistent with a meta-analysis of air pollutants and MI by Mustafic et al. [8], we observed increased rates of STEMI associated with increased CO and SO 2 concentrations, but not O 3 . Our findings of an increased rate of STEMI associated with increased concentrations of BC in the previous hour (Excess Rate = 9%, 95% CI: 1%, 18%; per 0.44 μg/m 3 ) and previous 3 h (Excess Rate = 9%, 95% CI: 0%, 18%; per 0.43 μg/m 3 ) are generally consistent with a study by Zanobetti and Schwartz that reported an 8.3% (95% CI: 0.2%, 15.8%) increase in the risk of hospitalization for MI associated with each 1.7 μg/m 3 increase in BC concentrations in the previous 2 days in Boston, MA from 1995 to 1999 [34].
An interesting finding of our study is the pattern of larger rates of STEMI associated with increased concentrations of UFP, UFP 11-50 nm, BC, and SO 2 in the period AFTER pollutant source and economic changes compared to the periods BEFORE or DURING these changes, which is contradictory to our a priori hypothesis. However, this is consistent with our other work conducted in New York   Odds ratios were estimated from conditional logistic regression models adjusting for mean temperature and relative humidity during the same lag hour(s) STEMI ST-elevation myocardial infarction Fig. 1 Rate of STEMI associated with each interquartile range increase in pollutant concentration, by time period. For each pollutant, the lag hour with the largest main effect that was statistically significant was selected for presentation. Error bars are 95% confidence intervals  [35]. Our finding of a greater excess rate of STEMI associated with increased UFP concentrations in the previous hour in the AFTER period (26%) compared to the DURING (5%) and BEFORE (9%) periods, is similar. Together, these findings may reflect the same change in PM composition and differential triggering of acute cardiovascular events by one or more PM components and/or PM mixtures. We suspect that secondary organic compounds and particle-bound reactive oxygen species (ROS) may play a role in the stronger adverse effect of these pollutants (i.e., UFP, UFP 11-50 nm, BC, SO 2 ,) after economic changes and policy initiatives that reduced SO 2 and NOx emissions. It is also possible that this lower relative rate estimated in the AFTER period was attributable to non-linearity of the exposure-response function. However, this seems unlikely, since our sensitivity analyses of the rate of STEMI associated with UFP in the previous hour suggested a linear (1 df) exposure-response function fit the data best. Concentrations of sulfate and nitrate have decreased from 2005 to 2016 across Rochester and New York State following air quality policy initiatives and economic changes [1,36]. However, Squizzato et al. and Zhang et al. reported that over this same period, secondary organic carbon increased across New York State [35,37]. These decreases in SO 2 and NOx emissions may have led to an increased oxidation rate of volatile organic compounds by hydroxyl radicals that would have otherwise reacted with the pollutant gases [37,38]. The oxidation of SO 2 and NO 2 forms strong acid, but oxidation of organic compounds forms secondary organic aerosol and produces ROS such as peroxy radicals and peroxide compounds [39]. ROS along with the secondary organics can deposit on particles (particle-bound ROS), which can then be inhaled leading to oxidative stress [40]. Oxidative stress and associated inflammation are hypothesized to be key mechanisms underlying the adverse cardiovascular health effect associated with air pollutants [7]. Although it is well known that ROS can be formed in situ after particles are deposited Odds ratios were estimated from conditional logistic regression models adjusting for mean temperature and relative humidity during the same lag hour(s) IQR interquartile range, STEMI ST-elevation myocardial infarction, OR odds ratio, 95%CI 95% confidence interval in the respiratory tract, little attention has been focused on particle-bound ROS [39,40]. We suspect that PM high in secondary organics and ROS may be more toxic than PM high in secondary sulfate and nitrate in triggering MI. Furthermore, recent source apportionment analyses across New York State including Rochester suggest that although PM 2.5 concentrations and contributions from most major source types were decreasing, spark-ignition vehicle emissions were increasing [4,37]. Rich et al. (2019) found increased rates of cardiovascular hospitalizations associated with increased concentrations of spark ignition vehicles and diesel emissions in the previous few days [41]. The correlation coefficient between secondary organic carbon and the spark-ignition vehicle contribution to PM 2.5 mass in Rochester was 0.73, suggesting that the increased formation of secondary organic carbon in the AFTER period may be associated with spark-ignition automotive emissions. This result is consistent with our findings that the UFP 11-50 nm fraction (marker for nucleation and spark ignition vehicle emissions [23]) was the primary contributor for the increased rate of STEMI associated with UFP. Diesel emissions also contribute to nucleation mode particles [23]. UFP 50-100 nm fraction (a larger mode indicator of diesel vehicle emissions) was not associated with increased rates of STEMI. Taken together, the observed stronger effect of UFP and BC on the rate of STEMI in the AFTER period may be explained by increased formation of secondary organics and ROS in the atmosphere, in part from spark-ignition vehicle emissions, and the subsequent increase in particle-bound ROS. These findings suggest that although the concentrations of PM have decreased following policy initiatives and economic changes, changing composition of PM may make the same dose of PM (e.g. each 3702 particles/cm 3 of UFP) more toxic in triggering STEMI. SO 2 also had a stronger effect in the AFTER period, which may represent the influence of the remaining emissions from diesel traffic near the monitoring site including heavy-duty diesel traffic on the major roads and the adjacent diesel trains. Although on-road diesel fuel went to ultralow sulfur in 2006 and non-road diesel sold in NYS went to ultralow in 2012, out-of-state non-road diesel such as that used in diesel railroad engines was not ultralow sulfur until 2014. There would also be an influence from upwind coal-fired power plants in Dunkirk, NY (shutdown December 31, 2015) and Tonawanda, NY (Huntley Generating Station closed March 1, 2016). Thus, SO 2 may be a surrogate for fresh secondary organic aerosol being produced in the upwind domain and transported with the SO 2 . CO had a stronger influence in the DURING and AFTER periods. CO could be a surrogate for spark-ignitions vehicles and residential wood combustion. Squizzato et al. [37] reported that car registrations increased in Monroe County after 2010. However, this also may be a result of a more sensitive CO monitor starting in 2011, resulting in reduced exposure misclassification and less downward bias, and larger relative rates in the DURING and AFTER periods. Health effect studies based on apportioned source contributions are needed to further assess these associations.
This study had several strengths, including a large sample size and multiple years of exposure data, resulting in increased statistical power. Additionally, we used symptom onset time, estimated by the patient and treating physician, to define the start time of each STEMI event, likely resulting in less exposure error than previous studies that used only hospital arrival date/time [8,[13][14][15][16][17]27]. Further, it allowed us to examine whether the rate of STEMI increased a few hours after increases in ambient pollutant concentrations, rather than just on or after one or more days.
However, our study also had several limitations. First, we used air pollution measurements from a central monitoring station to assign exposure to all subjects living within a 15mile radius of the site, regardless of the specific distance from the monitoring station to subject's residence and/or where they worked and spent time. However, this exposure error is likely to be non-differential with regard to case and control periods and thus, should result only in underestimates of the relative rate of STEMI. The magnitude of this measurement error and underestimation may vary for different pollutants. UFP is thought to be more spatially variable than PM 2.5 and AMP, and our previous work observed high spatial heterogeneity in BC across the Rochester area [42]. Although UFP and BC may be subject to greater exposure error, we observed increased rates of STEMI associated with these two pollutants. The true effect estimates in the absence of exposure measurement error might be even larger.
Second, the three time periods BEFORE (2005-2007), DURING (2008)(2009)(2010)(2011)(2012)(2013), and AFTER (2014-2016) were selected based on times of emission changes. However, for many of the policies and emission changes, there are no well-defined time windows for the change because the interventions were implemented over time. For example, emission standards are only for new vehicles, but we do not know how quickly people buy new cars and trucks and take old vehicles out of service. We determined the best approach was to select groups of years where there were moderately constant or specifically changing emissions. As is shown in Additional file 1: Figure S1 from Squizzato et al. [2], SO 2 emissions from coal-fired power plants in western New York were relatively high before 2008, and since 2008 it dropped substantially and remained at a relatively low level after 2013. The differences in air pollution concentrations and compositions among the 3 periods were analyzed and reported in Masiol et al. [4], Zhang et al. [35], and Croft et al. [43]. In the present study, the BEFORE, DURING, and AFTER period is only a proxy for the time of air quality policy implementation/intervention or economic change.
Third, patients in the three periods were perhaps different regarding several clinical characteristics associated with an increased rate of MI [44]. STEMI patients included in the BEFORE period were older and had a higher prevalence of diabetes and heart failure than patients in the AFTER period. By contrast, STEMI patients in the AFTER period were more often male, smokers, and more often had dyslipidemia. Although these characteristics may be related to susceptibility to acute effects of short-term air pollution exposure, our sensitivity analyses show that the pattern of larger increased rate of STEMI associated with increased UFP concentrations at lag hour 0 in the AFTER period compared to the BEFORE and DURING periods remained when also adjusting for interactions between air pollutant and age, sex, smoking, diabetes, dyslipidemia, and heart failure in separate models (Additional file 1: Table S4), suggesting these differences in period-specific effect estimates were not likely due to differences in the characteristics of the STEMI patients by period and interactions between those characteristics and air pollutant concentrations. Furthermore, as discussed above, the pattern of a larger increased rate of STEMI associated with increased UFP in the AFTER period in the present study is similar to the larger rate of ischemic heart disease hospital admissions associated with increased PM 2.5 in the AFTER period reported by Zhang et al. across New York State [35]. This correspondence suggests that difference in study population by period may not be the sole reason for our observed period-specific effects. It is also of note that since the proportion of missing data on clinical characteristics was large in the BEFORE period (18% missing), we may not have accurately compared STEMI patient characteristics between periods.
Fourth, we did not have data on influenza in our study subjects or influenza epidemics occurring during the study period, and thus did not adjust for this variable in our analyses. However, influenza may have been a mediator of the association between air pollution and STEMI, rather than a confounder, as several studies have reported triggering of respiratory infection by short term increases in air pollutant concentrations [43,[45][46][47][48], and others have suggested that influenza may trigger acute myocardial infarction [49][50][51]. Finally, although our inference was primarily made by considering the pattern of STEMI response to pollutants across multiple lag times and not just whether each was statistically significant, type I error in the assessment of the main effect of air pollution on STEMI might be inflated due to multiple comparisons.

Conclusions
In summary, increased rates of STEMI associated with increased UFP, UFP-11-50 nm, BC, and SO 2 concentrations were higher in the AFTER period (2014-2016) compared to the BEFORE (2005-2007) and DURING (2008)(2009)(2010)(2011)(2012)(2013) periods. This may be due to changes in PM composition (i.e. higher secondary organic carbon and particle-bound ROS) following multiple air quality policies and economic changes. The sources and the role of secondary organics and particle-bound ROS in triggering MI need to be further studied.