Acute Adverse Effects of Fine Particulate Air Pollution on Ventricular Repolarization

Background The mechanisms for the relationship between particulate pollution and cardiac disease are not fully understood. Objective We examined the effects and time course of exposure to fine particulate matter ≤ 2.5 μm in aerodynamic diameter (PM2.5) on ventricular repolarization of 106 nonsmoking adults who were living in communities in central Pennsylvania. Methods The 24-hr beat-to-beat electrocardiogram (ECG) data were obtained using a high-resolution 12-lead Holter system. After visually identifying and removing artifacts and arrhythmic beats, we summarized normal beat-to-beat QTs from each 30-min segment as heart rate (HR)-corrected QT measures: QT prolongation index (QTI), Bazett’s HR-corrected QT (QTcB), and Fridericia’s HR-corrected QT (QTcF). A personal PM2.5 monitor was used to measure individual-level real-time PM2.5 exposures for 24 hr. We averaged these data and used 30-min time-specific average PM2.5 exposures. Results The mean age of the participants was 56 ± 8 years, with 41% male and 74% white. The means ± SDs for QTI, QTcB, and QTcF were 111 ± 6.6, 438 ± 23 msec, and 422 ± 22 msec, respectively; and for PM2.5, the mean ± SD was 14 ± 22 μg/m3. We used distributed lag models under a framework of linear mixed-effects models to assess the autocorrelation-corrected regression coefficients (β) between 30-min PM2.5 and the HR-corrected QT measures. Most of the adverse ventricular repolarization effects from PM2.5 exposure occurred within 3–4 hr. The multivariable adjusted β (SE, p-value) due to a 10-μg/m3 increase in lag 7 PM2.5 on QTI, QTcB, and QTcF were 0.08 (0.04, p < 0.05), 0.22 (0.08, p < 0.01), and 0.09 (0.05, p < 0.05), respectively. Conclusions Our results suggest a significant adverse effect of PM2.5 on ventricular repolarization. The time course of the effect is within 3–4 hr of elevated PM2.5.

volume 118 | number 7 | July 2010 • Environmental Health Perspectives Research Numerous studies have consistently found a significant association between fine particu late matter ≤ 2.5 µm in aerodynamic diameter (PM 2.5 ) air pollution and the shortterm risk of clinical cardiovascular mortality (Franklin et al. 2007(Franklin et al. , 2008Ostro et al. 2006;). The mechanisms responsible for such an association have been the focus of recent environmental health studies. In popula tionbased studies of healthy individuals (Dekker et al. 1994;Goldberg et al. 1991;Peters et al. 1990;Rautaharju et al. 2006aRautaharju et al. , 2006bSchouten et al. 1991), longer repolarization within normal range was significantly associated with cardiac events, especially sudden cardiac death. Similar findings from clinical populations (Siscovick et al. 1996;Whitsel et al. 2000Whitsel et al. , 2001 have also been reported. Recent studies have suggested that one of the underlying mechanisms linking air pollution and increased risk of cardiovascular disease (CVD) is the effect of PM on ventricular repolarization (Campen et al. 2006;Ghelfi et al. 2008;Henneberger et al. 2005;Lux and Pope 2009;Samet et al. 2009;Yue et al. 2007). For the time course of PM effects on cardiac elec trophysiology, several published studies have suggested shorter time effects, such as within the same day or 1-2 days prior to electrocardiogram (ECG) measurements (Elder et al. 2007;Liao et al. 1999Liao et al. , 2004Liao et al. , 2009Lux and Pope 2009;Park et al. 2005;Yue et al. 2007;Zhang et al. 2009). For patients who wore implanted cardioverter defibrillators, Dockery et al. (2005) reported significantly increased incidence of arrhythmias associated with the 2day average of various pollutants, which also suggested acute effects of pollution on clinically relevant arrhythmias. In one study specifically designed to investigate the time course of PM on heart rate variability (HRV), Cavallari et al. (2008) reported an early and a laterphase response, with the early effects at 2 hr and delayed effects at 9-13 hr after exposure.
We therefore designed this study to investigate the effects and time course of individuallevel exposures to PM 2.5 on the ventricular repolarization in a sample of non smoking adults who lived in communities in central Pennsylvania.

Materials and Methods
Population. For this report, we used the data collected for the Air Pollution and Cardiac Risk and its Time Course (APACR) study, which we designed to investigate the mecha nisms and the time course of the adverse effects of PM 2.5 on cardiac electro physiology, blood coagulation, and systemic inflam mation. The APACR study has maintained approval by Penn State University College of Medicine institutional review board. All par ticipants gave written informed consent prior to their participation in the study. All study participants were recruited from communi ties in central Pennsylvania, mostly from the Harrisburg metropolitan area. The inclusion criteria for the study included nonsmoking adults > 45 years old who had not been diag nosed with severe cardiac problems (defined as diagnosed valvular heart disease, congenital heart disease, acute myocardial infarction or stroke within 6 months, or congestive heart failure). Community recruitment specialists from the General Clinical Research Center (GCRC), which is funded by the National Institutes of Health, at the Penn State College of Medicine, and the GCRCorganized com munity outreach activities, supported the recruitment of the participants. The GCRC maintains a list of individuals who live in cen tral Pennsylvania communities for various healthrelated studies. The APACR study par ticipants were numerated from the GCRC's list of potential participants; approximately 75% of the individuals who were contacted and who met our inclusion criteria were enrolled in the study. Our targeted sample size was 100 individuals, and we enrolled and examined 106 individuals. The examination of two participants per week was conducted from November 2007 to June 2009 for the entire examination period except for major holidays.
Study participants were examined in the GCRC in the morning between 0800 and 1000 hours. All participants fasted for at least 8 hr before the clinical examination. After completing a health history questionnaire, a trained research nurse measured seated blood pressure (BP) three times, height, and weight, and drew 50 mL blood for biomarker assays according to the blood sample preparation protocols. A trained investigator connected the PM 2.5 and Holter ECG recorders. Participants were given an hourly activity log to record special events that occurred in the next 24 hr, including outdoor activities, exposure to traf fic on the street, travel in an automobile, and any physical activities. The entire examination session lasted for about 1 hr. Participants were then released to proceed with their usual daily routines. The next morning, they returned to the GCRC to remove the PM and Holter monitors, to deliver the completed activity log, and to have their seated BP measured three times and another 50 mL of blood drawn. An exercise echocardiogram was then performed to measure the ventricular function and struc ture for each participant. The entire second day session lasted for about 1 hr and 45 min. A description of the participants' characteris tics are presented in Table 1.
The study protocol was approved by Penn State University College of Medicine institu tional review board. Each participant received $50 and two certificates for breakfast in the hospital cafeteria, and they were reimbursed for their transportation costs. PM 2.5 concentration. The APACR study used personal PM 2.5 DataRam (pDR; Thromo Scientific, Boston, MA) for realtime 24hr personal PM 2.5 exposure assessment. The pDR used lightscattering physics of the fine particles to detect the realtime con centrations of particles of various sizes. The size selection was achieved by using an active pump with a validated sizeselection cyclone inlet (KTL SCC1.062; BGI Inc., Waltham, MA) at a flow rate of 1.5 L/min. The stan dardized operation procedures (SOP) for the use of the pDR, including the calibration, application, data transfer, and data validation, as well as chemical analysis of filters for major PM 2.5 species, were developed by the APACR investigators (Penn State University 2008a). The standardized procedures were rigorously followed in the PM 2.5 data collection. The realtime PM 2.5 concentrations were initially recorded at 1min intervals. For each par ticipant, we calculated the 30min segment specific averages, based on the top and the bottom of the clock time, as our PM 2.5 expo sure variable in the APACR study. Therefore, the PM 2.5 exposure variables were treated as repeated measures, and each individual con tributed 48 exposure data points.
Continuous ambulatory ECG. A high fidelity (sampling frequency 1,000 Hz) 12lead HScribe Holter System (Mortara Instrument, Inc., Milwaukee, WI) was used to collect the 24hr Holter beattobeat ECG data. The highfidelity ECG significantly increases the resolution and enhances the accuracy of vari ous wave form measurements. The Holter ECG data were scanned to a designated com puter for offline processing by an experienced investigator using specialized SuperECG soft ware (Mortara Instrument, Inc.). The SOPs for the APACR study were developed by the study investigators (Penn State University 2008b, 2008c and were rigorously fol lowed in the data collection and interpreta tion processes. Briefly, the Holter ECG Data Collection and Analysis Procedures (Penn State University 2008b) were followed to pre pare, hook up, calibrate, and start the Holter digital recorder. After 24 hr of recording, a trained investigator followed the SOP to retrieve and archive the beattobeat ECG data for offline processing. The main objec tive of the offline processing was to verify the Holteridentified ECG waves and to iden tify and label additional electronic artifacts and arrhythmic beats in the ECG recording. Finally, a single research investigator followed the SuperECG Manual (Penn State University 2008c) to perform beattobeat ECG analysis to calculate ECG parameters.
QT interval and QRS duration variables. We used the above described 24hr beatto beat Holter ECG, after removing artifacts with standardized visional inspection and sta tistical filters, to calculate beattobeat QT intervals using the SuperECG software, which defined QT interval as the start of Q wave to the end of T wave. None of the normal sinus QRS durations was > 120 msec. QT is heart rate (HR) dependent, and only after HR correction does the QT interval have the electrophysiological property of ventricu lar repolarization. We then calculated the following three HRcorrected QT duration indices as the measures of ventricular repolari zation on a 30min basis.
• Prolongation index (QTI) = 100 × (QT max ÷ QT predicted ), where QT max is the maximum QT duration across all normal cardiac cycles within the segment and QT predicted = 656 ÷ (1 + 0.01 × HR) (Rautaharju et al. 1991). • Bazett's HRcorrected QT interval (QTcB) (Bazett 1920 These HRcorrected indices were chosen because the QTI has been reported to be (Rautaharju et al. 1991) less rate sensitive and because it has a higher rate of repeatability than does QT c B, which is the first HRcorrection QT index; QTcF has similar properties but better HR correction compared with QTcB. These ECG measures were treated as repeated outcome measures, and each individual con tributed 48 outcome data points on each of the three ventricular repolarization variables.
The QRS duration (QRS) was calculated as the duration from the start of Q wave to the end of S wave. QRS duration was used as a measure of ventricular depolariation. HRV variables. We performed time and frequency domain HRV analysis on the ECG recording after removing artifacts with stan dardized visional inspection and statistical fil ters. We calculated HRV indices from overall 24hr recording, 30min, and 5min segment specific recordings using the SuperECG pack age (Mortara Instrument, Inc.) according to the current recommendations (Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology 1996). The following HRV indices were used as indices of cardiac auto nomic modulation: SD of all RR intervals Weather variables. We obtained real time temperature and relative humidity using the HOBO H8 logger (Onset Computer Corporation, Bourne, MA). The realtime temperature and relative humidity were recorded at 1min intervals initially. For each participant, we calculated 30min segment specific averages, corresponding to the PM 2.5 and Holter measures. Therefore, these weather covariables were treated as repeated measures, and each individual contributed 48 data points for each variable.
Other participant-level covariables. A standardized questionnaire administered on day 1 of the study was used to collect the fol lowing individuallevel information: a) demo graphic variables, including age, race, sex, and highest education level; b) medication uses, including antianginal medication, antihyper tensive medication, and antidiabetic medi cation; and c) physiciandiagnosed chronic disease history, including CVD (including revascularization procedures and myocardial infarction), hypertension, and diabetes. The averages of the second and third measures of seated systolic and diastolic BPs on day 1 were used to represent BP levels of a par ticipant. Day 1 fasting glucose was measured by Penn State GCRC central laboratory. CVD was defined by antianginal medication use or a history of CVD. Hypertension was defined by antihypertensive medication use, physiciandiagnosed hypertension, systolic BP ≥ 140 mmHg, or diastolic BP ≥ 90 mmHg. Diabetes was defined by antidiabetic medica tion use, a diagnosis of diabetes by a physi cian, or fasting glucose > 126 mg/dL. Body mass index (BMI) was defined as the ratio of weight (kilograms) to height squared (meters).
Statistical analysis. We used distributed lag models (Almon 1965;Pope and Schwartz 1996;Schwartz 2000) under a framework of linear mixedeffects models (Laird and Ware 1982) with a firstorder autoregres sive covariance structure to assess the auto correlation corrected regression coefficients between 30min PM 2.5 and the HRcorrected QT measures. Residual diagnostics were used to assess the appropriateness of mod eling assumptions, and no sizeable depar tures were detected. In these models, one lag indicated a 30min separation between the exposure and outcome. Thus, lag 0 indi cated the spontaneous relationships between PM 2.5 and the HRcorrected QT measures, and lag 1 indicated 30 min between the PM 2.5 and HRcorrected QT measures, and so on. Because QT interval includes QRS duration, with the latter mostly represent ing the ventricular depolarization process, we analyzed QRS interval in models identi cal to those for the QT variables. We chose a constrained distributed lag model, the polynomial distributed lag model, to reduce the potential collinearity of PM 2.5 between individual lags using a seconddegree poly nomial. Another advantage of the distributed lag model is its ability to provide interpre tation of the cumulative effects of the lags included in the model, as well as individual lag effects. Because the PM 2.5 and ECG vari ables were assessed in parallel over 48 lags (24 hr), we predetermined to model no more than 10 lags, which allowed us to fit the dis tributed lag models using at least 80% of the data. We started from the largest number of lags (lag 0-lag 10), and identified significant (p < 0.05) cumulative effects of PM 2.5 on the ECG variables. In this report, the cumulative effect on QTI was significant in the 10lag model. From this 10lag significant cumula tiveeffect model, we reduced the total num ber of individual lags by backeliminating the nonsignificant longer lags (e.g., lag 10), one lag at a time until a significant individual lag was identified (lag 7 in this report). We then identified this model as our final model for all ECG outcomes. All results were expressed per 10µg/m 3 increase in PM 2.5 . The distributed lag models are summarized in Table 2, where model 1 was adjusted for basic demographic variables and model 2 included an additional adjustment for diabetes, hypertension, and CVD. We repeated model 2 by adjusting for each of the HRV variables to examine the impact of cardiac autonomic modula tion on PM 2.5 and QT associations. These results are summarized in Table 3. In these models, all timedependent covariables, such as weather and HRV variables, were entered in the model using the same distributed lag structure as the PM 2.5 variable.

Results
The demographic and CVD risk profiles of the study population are presented in Table 1. The mean age of the participants was 56 years old, 74% were nonHispanic white, 26% were minorities (including blacks, Hispanics, and Chinese), 59% were female, and 43% had chronic diseases (mostly hypertension). Nine individuals (8.5%) reported being diag nosed with coronary heart disease by a physi cian > 2 years ago. At the population level, the distributions of both the PM 2.5 exposure and ventricular repolarization outcome variables are approximately normal. The time of the dayspecific distributions of the PM 2.5 and QTI, as an example of ventricular repolari zation measures, are presented in Figures 1  and 2, respectively. Both the PM 2.5 and QTI showed sufficient variations, both between time points and between individuals, within the 24hr time frame. The cumulative effects and individual lag effects of PM 2.5 on each of the HRcorrected QT measures and QRS duration are sum marized in Table 2 as multivariable adjusted regression coefficients (± SE) associated with a 10µg/m 3 increment of PM 2.5 exposure. In summary, the cumulative effect based on lag 0-lag 7 (3.5 hr) on QTI was signifi cant (p < 0.01). Examining the individual lag effects, most of the adverse ventricular repolarization effects from direct PM 2.5 expo sure occurred around 3-3.5 hr (lag 6-lag 7) after the elevation of PM 2.5 . The multivari able adjusted regression coefficients for the association between a 10µg/m 3 increase in PM 2.5 and QTI were also statistically signifi cant at lag 0 and lag 1, suggesting both very early and laterphase responses. Additional adjustment for chronic disease did not change the pattern of association from the models adjusted only for major demographic and weatherrelated variables. PM 2.5 was not asso ciated with QRS duration in this population ( Table 2).
The final multivariable adjusted PM 2.5 and QT measures (model 2) in Table 2 were rerun with additional adjustment for HRV variables, using one HRV variable at a time. The HRV adjusted PM 2.5 and QT associations are pre sented in Table 3. In summary, the overall pattern of associations between PM 2.5 and HRcorrected QT interval measures did not change substantially with additional adjust ment for HRV variables as measures of car diac autonomic modulation. However, the estimated lag 7 PM 2.5 effects on QTI were attenuated, and the pvalues from this lag were marginally significant (corresponding pvalues were > 0.05 but < 0.10). To examine the potential confounding by the circadian variations of both QT variables and the PM 2.5 exposures, we also stratified our final mod els (model 2) as daytime versus nighttime for ECG measures (using 2100 hours as a cut off). These stratified models indicated similar patterns of association as those from the 24hr overall models (data not shown).
We tested the interaction terms between PM 2.5 and chronic conditions and found no statistical significance at p < 0.05 level (data not shown). Therefore, the estimated effects of PM 2.5 on ventricular repolarization meas ures did not differ significantly depending on whether a person had previous health prob lems. We also performed stratified analysis according to chronic disease status, using the models in Table 2, model 1. We found simi lar associations by chronic disease status (data not shown). It should be noted that the sam ple size of this study is small, and individuals with chronic conditions consisted mostly of wellcontrolled hypertensives. The statistical power was limited to detect significant effect modification by chronic disease status.

Discussion
A large number of epidemiologic studies have found an association between shortterm exposure to increased particulate air pol lution and CVD morbidity and mortality (Franklin et al. 2007(Franklin et al. , 2008Ostro et al. 2006;). However, the mechanisms responsible for such an associa tion have not been fully identified. Previous studies have suggested several promising underlying mechanisms, including cardiac autonomic impairment as measured by lower HRV (Creason et al. 2001;Gold et al. 2000;Liao et al. 1999Liao et al. , 2004Pope et al. 1999), and ventricular repolarization (Campen et al. 2006;Ghelfi et al. 2008;Henneberger et al. 2005;Lux and Pope 2009;Samet et al. 2009;Yue et al. 2007). Various studies, including patientbased population, panel study, large populationbased cohort, controlled expo sures, or ambient fixedlocation air pollution measures, have indirectly suggested short term PM effects on cardiac electrophysiology and clinically relevant arrhythmias (Dockery et al. 2005;Elder et al. 2007;Liao et al. 1999Liao et al. , 2004Liao et al. , 2009Lux and Pope 2009;Park et al. 2005;Yue et al. 2007;Zhang et al. 2009); these cardiac parameters included HRV, ventricular repolarization, Twave alternans, myocardium ischemia, and arrhythmias. The actual time course from PM exposure to effects on cardiac repolarization measures has not been investigated system atically in a communitydwelling sample, nor has it been determined whether the PM and ventricular repolarization association would be mediated through its impact on cardiac auto nomic modulation. Cavallari et al. reported an early and a laterphase HRV response, with the early effects at 2 hr and delayed effects at 9-13 hr after exposure (Cavallari et al. 2008).
For the healthy individuals in this com munitybased study, PM 2.5 had a significant adverse association with ventricular repolariza tion, regardless of which HRcorrected QT intervals were used. Specifically, the estimated cumulative effects of PM 2.5 on QTI were sta tistically significant, and the direction of point estimates from the other two HRcorrected QTs (QtcB and QTcF) indicated the same anticipated direction as QTI, although the estimated cumulative effects for these latter two QTs were not significant at the p < 0.05 level. For the individual lag effects, all three QT meas ures were consistently associated with PM 2.5 exposures at lag 6 and lag 7 (approxi mately 3-3.5 hr after elevated PM 2.5 ), with the estimated effect on QTI also showing evi dence of an earlierphase effect (lag 0 and lag 1, approximately within 1 hr of PM exposure elevation). We did not find any association between PM 2.5 and QRS duration. Therefore, these data support that elevated PM 2.5 levels can lead to longer ventricular repolarization but have no immediate impact on ventricular depolarization (QRS duration). The appar ent effect courses within 3-3.5 hr of elevated PM 2.5 exposure. The results presented in Table 3 indicate that the PM and QT and PM-QRS association remains unchanged, even after adjusting for HRV variables as potential intermediating factors. These data further sug gest that the PM 2.5 and ventricular repolari zation measures were not mediated through adverse effects on cardiac autonomic modula tion, at least not effects with the same lag from exposure. Furthermore, the time from exposure to apparent effects was approximately 3-3.5 hr. Considered with other studies that have indi cated shorterterm acute PM effects on cardiac electrophysiology and arrhythmias, our current timecourse study results are consistent with previous studies and are suggestive of an acute PM 2.5 mediated disturbance of the ventricular repolarization process, which may contribute to acute cardiac events, particularly arrhyth mias and sudden cardiac death. To our knowl edge, this is the first study to demonstrate such findings in a communitybased sample.
A number of approaches have been devel oped to estimate ventricular repolarization including the HRcorrected QT intervals that we used in this study-QTI, QTcB, and QTcF. We used these three HRcorrected QT measures because QTcB is the first HRcorrected QT index, and QTcF has simi lar properties but better HR correction com pared with QTcB. Both QTcB and QTcF are widely used in the literature. We also used the less frequently seen QTI because it was reported to be less sensitive to changes in HR and more repeatable than QTcB (Rautaharju et al. 1991). On the other hand, it was suggested that these HRcorrected QT measures should not be used to deter mine the treatment effects in patients with QTprolongation syndrome (Rautaharju et al. 2009). However, this study sample repre sents communitydwelling individuals, and none have long QT syndrome. In effect, we designed this study to investigate the impact of PM on the variation of HRcorrected QT intervals within normal range in relatively healthy individuals. Moreover, these simple HRcorrected QT variations among normal healthy individuals have been associated with significant prediction of future risk of cardiac events (Rautaharju et al. 2006a(Rautaharju et al. , 2006b, sup portive of our use of these indices.
It should be noted that the effect sizes we estimated in this study are relatively small. For example, for every 10µg /m 3 increase in PM 2.5 , the associated cumulative increase in QTI from previous 3.5 hr is only 0.32, cor responding to less than a 1% increase in this variable, which has a mean of 110 and SD of 6.6. For another instance, for every 10µg /m 3 increase in lag 7 PM 2.5 , the associated increase in QTcB is only 0.22 msec, corresponding to less than a 1% increase in this variable, which has a mean of 438 msec and SD of 23 msec. Although such small changes in QTI may not be clinically meaningful, it can be argued that the entire population is exposed to PM 2.5 in the ambient air, and from personal and indoor sources, on a continuous daily basis. Thus, elevated PM 2.5 levels have a potentially large public health impact. Moreover, the minor effect on QTI estimated in this study was measured in generally healthy individuals. It is possible that PM 2.5 effects on ventricular repolarization might be greater in individu als with underlying structural heart disease, ischemic heart disease, channelopathies, or druginduced effects on repolarization. Future studies should target these clinical subgroups likely to be more susceptible to the effects of PM 2.5 , especially those made more vulnerable by residing near sources of PM 2.5 , for example, near highways.
There are several limitations. First, the APACR study excluded smokers and per sons with acute cardiac events within the past 6 months. Thus, our findings may not apply to smokers or persons with a recent acute cardiac event. Second, the majority of participants reported that they stayed indoors most of the time during the 24hr study period, except when they had to travel by automobile. This behavior pattern is reflected in the relatively low levels of exposure to PM 2.5 . In general, our participants had limited indoor expo sures, such as secondhand smoking. Thus, we were unable to assess whether exposures at much higher levels would exhibit similar asso ciations. However, we purposely used the per sonal monitors and realtime Holter system to collect the true individual level exposure and routine ECG data, respectively. We argue that the associations we observed in these individuals are more reflective of their routine exposure and outcome associations. Third, the ECG data from Holter were not collected under a controlled, supineposition setting. Thus, the shortterm variation of other factors that may impact the ventricular repolarization cannot be fully accounted for. However, it is not feasible to keep a healthy participant in a supine indoor position for 24 hr. Even if this were achieved, the results from such a study design would likely have limited variation in PM 2.5 exposure levels, and the data from such a study would not be generalizable to a realworld situation. In contrast, our study captures the range of activities occurring in real life, including time spent outdoors, time spent commuting in an automobile, and vari ous other activities associated with a disease free, communitydwelling individual. Finally, PM 2.5 was the only pollutant on which data were collected. The observed associations could be due to other unmeasured pollutants highly correlated with PM 2.5 .
In summary, acute exposure to PM 2.5 at the individual level is associated with longer HRcorrected QT interval measures, and the time to the apparent effect is about 3-3.5 hr. The estimated effect of PM on ventricu lar repolarization was independent of major confounding factors and cannot be attributed solely to effects of PM on cardiac autonomic modulation (HRV). There was no association between PM 2.5 and QRS duration, suggest ing no effects on ventricular depolarization. Overall, these findings support that PM may affect ventricular repolarization, and partly through such a mechanism, PM increases car diovascular risk, such as sudden cardiac death.