Epidemiologic studies suggest that variation in ambient particulate matter air pollution is a risk factor for acute cardiovascular events.1–6 Time-series analyses have demonstrated that daily average levels of particulate matter are associated with increased cardiovascular mortality, hospital admissions for myocardial infarction (MI), and automatic implantable cardioverter-defibrillator discharges.3,7–11 A case-crossover study of hospital admissions for MI demonstrated a 0.8% increase in MI admissions per 10-μg/m3 increase in fine-particulate matter (PM).12
Moreover, short-term peaks in fine-particulate matter exposure may trigger cardiac ischemia. A recent case-crossover analysis in 772 individuals with acute MI participating in the Boston Determinants of Myocardial Infarction Onset Study demonstrated that an increase of 25 μg/m3 PM2.5 during a 2-hour period before the onset of MI elevates risk of MI (OR = 1.48; 95% CI = 1.09–2.02) after controlling for PM2.5 on the previous day.13 In addition, there was a delayed response associated with a 24-hour average increase of 20 μg/m3 PM2.5 1 day before the onset of symptoms (OR = 1.62; 1.13–2.34) when the PM2.5 exposure before the event was held constant. This short-term effect suggests that increases in PM2.5 air pollution contribute to the triggering of MI in susceptible individuals.
In contrast to the results from the Boston study and other studies with similar results, prior case-crossover analyses of air pollution effect on primary cardiac arrest in Seattle have failed to detect an association between daily averages of fine-particulate matter and primary cardiac arrest.14,15 This difference may reflect the longer averaging times for exposure used in the primary cardiac arrest (PCA) studies, differences in particulate matter composition, or differences in statistical analytic methods between studies.
The daily average levels of particulate matter and range of 24-hour averaged particulate matter to the Boston Onset Study and other time-series studies that found adverse cardiovascular effect from particulate matter.8,11–15 However, the Boston Onset Study had very short-term average particulate matter levels with resolution of 1-hour compared with daily averages in the Seattle studies of air pollution effect on PCA. The differences in effect are potentially captured by higher short-term peak excursions of particulate matter in the studies performed in the Northeastern United States.
To verify the association found in the Boston study, we studied the relation between particulate matter levels and onset of MI in a much larger, community-based study of MI. We hypothesized that an increase in fine-particulate matter would be associated with an elevated risk for MI onset within 2 hours, especially among individuals with preexisting cardiac disease. To replicate the results of the Boston study most directly, we also repeated our analyses using the identical time periods, referent selection strategy, and model adjustments reported by Peters et al.13
We obtained data from a population-based community-wide database (the Myocardial Infarction Triage and Intervention [MITI] Project registry) containing precise information on symptom onset time in acute MIs occurring between 1988 and 1994 (n = 11,983) in King County, Washington. We linked this information with central site air pollution monitoring data on fine-particulate matter (particles <2.5 μm in diameter; PM2.5), PM10 (particles <10 μm in diameter), carbon monoxide (CO), and sulfur dioxide (SO2). These data were further enhanced by meteorologic variables: daily averages of relative humidity and temperature from the Seattle Tacoma Airport (National Oceanic and Atmospheric Administration). We used the case-crossover study design.16–19 To control for confounding and ensure there was no overlap bias, we used a time-stratified referent selection design that compares each event hour to a set of referent exposure hours occurring on the same day of the week during the same month of the case event.19–21 The study protocol was approved by the University of Washington Institutional Review Board.
Our cases were restricted to those MITI participants with a discharge diagnosis of acute MI confirmed by cardiac enzyme and electrocardiography criteria.22–24 This strategy excluded individuals with unstable angina, cardiac arrhythmias, and noncardiac causes of chest pain. Cases were also excluded if they lived outside of King County or South Snohomish County because the air pollution data may not have represented actual exposures for individuals. Our final analysis included 5793 case days and 20,134 referent exposure days from these same case individuals.
Briefly, the MITI project registry is a community-wide database linking emergency medical service (EMS) response with hospital outcome. The registry was established in 1988 to study care for all Seattle-area patients with suspected acute MI. Case identification is based on activation of the EMS system. The MITI project was a collaborative effort that includes time of EMS activation, symptom duration before activation, and a registry of all patients admitted to 19 hospitals for suspected MI in the Seattle metropolitan area. The registry contains detailed data about all patients who had an acute MI at discharge or death, as confirmed by EMS activation time and medical records. For patients transferred to a different institution during the index hospital stay, charts were abstracted at the subsequent receiving facility such that each patient had a continuous care record. We ascertained the onset time of confirmed acute MI by subtracting the duration of pain before the EMS call or emergency room (ER) admission from the EMS call time or ER admit time, respectively.
The primary exposure metric was short-term average (1-hour, 2-hour, 4-hour, and 24-hour averaged) fine-particulate matter measured by nephelometry from 3 King County monitoring sites (Lake Forest Park, Duwamish, and Kent). Nephelometry data correlate well with gravimetric particle measurements in the 0.1–1.4 aerodynamic range. King County nephelometry measures are also highly correlated with PM2.5 levels (Pearson's r2 = 0.85).25 We also considered 1-hour, 2-hour, 4-hour, and 24-hour average measures of SO2 from the Duwamish site and 1-hour, 2-hour, 4-hour, and 24-hour average measures of CO averaged over 4 sites in King County.
The data were analyzed using the statistical packages SAS (version 8.0; SAS Institute, Cary, NC) and SPSS (version 10; SPSS, Inc., Chicago, IL). We performed simple descriptive analyses including summary statistics, and gra-phic plots of the exposure and demographic data. Levels of correlation between covariates were assessed using Pearson's correlation coefficient.
Air pollution exposures occurring on index hours were compared with exposures that occurred on all referent days in the same time stratum as the index date. These were defined as all observations over 1-hour, 2-hour, 4-hour, and 24-hour average time periods that preceded MI onset time on a single day of the week during 1 month and year. This time-stratified referent selection scheme is not subject to overlap bias, and it also minimizes bias resulting from nonstationarity of air pollution time-series data.19–21
Conditional logistic regression provided estimates of odds ratios (ORs) and 95% confidence intervals (CIs). The primary outcome was onset of incident MI. The primary exposure variable was short-term increase in fine-particulate matter defined as a 10-μg/m3 increase in 1-hour, 2-hour, and 4-hour average PM2.5. We also analyzed 24-hour averaged fine PM on day of event with lags of 1 and 2 days, as well as assessing other pollutant variables (PM10, CO, and SO2). Onset time was rounded to the previous hour if the minute of onset occurred at less than 30 minutes past the hour and was rounded forward for onset after 30 minutes. Lagged analyses used time strata defined by the lagged index date and hour. Relative humidity and temperature were included as confounding variables and entered as both linear and quadratic terms in the conditional logistic regression models. Stratified analyses were performed to assess effect modification of the association between MI onset and particulate matter. We considered age (<50, 50–69, ≥70 years), sex, race (white or nonwhite), and smoking status. Smoking status was categorized as current smoker versus nonsmoker at time of event (never smoker or exsmoker). To assess for potentiation of effect by known risk modifiers for MI, we performed further subanalyses to assess the effects of hypertension and diabetes on cardiac susceptibility to particulate matter. To determine whether type of heart disease influenced the association of MI and particulate matter level, separate models considered all forms of heart disease as a single variable and as subtypes of heart disease (ischemic heart disease and congestive heart failure). To make the study populations more comparable to the Boston study participants, we performed subanalyses of the acute MI population that stratified cases by survival to hospital discharge. We also looked for a greater effect in the younger age group (<50 years of age) because the Boston study population was represented by a slightly larger percent of individuals in this younger age group (21% compared with 13% of MITI participants).
To replicate the Boston study, we repeated our analysis using a unidirectional referent sampling approach. We selected referent exposures on 3, 4, and 5 days before the index exposure. For the 2-hour average particulate matter exposure, the index time was the 2 hours before the onset hour of the event. The 24-hour average particulate matter index time was lagged 1 day (ie, 24–48 hours before MI onset). Thus, the control periods were 24-hour averages lagged by 4, 5, and 6 days. The onset hour was truncated to the previous hour for both the 2-hour and 24-hour averages. Using these exposure times, we performed a conditional logistic regression analysis. We replicated the analyses for PM2.5 given in Tables 3 and 4 of the paper by Peters et al.13 We repeated their crude analysis by quintile of exposure for both the 2-hour (1-hour before onset) and the 24-hour average (on the previous day) exposures (Peters’ Table 3). The final model (Peters’ Table 4) jointly estimates the effect of hourly (2-hour averages) and daily (24-hour averages, lagged 1 day) exposure as linear terms. Temperature and relative humidity are included as linear and quadratic terms. The model controls for season using the single term:
where t = day of year and P = 1-, 1/2-, 1/3-, 1/4-, 1/5-, 1/6-year period. The adjusted model also used indicator variables for day of the week.
Over the 7 years of the study, 5793 cases of MI met the case definition, an average of 856 events per year (range, 387–1330 events/year.). The lower end of the range reflected a reduced number of months of recruitment during the initial year of study. The case population is predominantly composed of older, white men (Table 1). Thirty-one percent of those experiencing MI had a previously documented MI. Twenty-three percent were smokers at the time of MI. Our case number was reduced to 5533 after eliminating cases with missing air pollution exposure or meteorologic covariates.
Table 2 summarizes the distribution of 1-hour averaged means of air pollution and temperature variables in our study. Analyses of correlation between pollutants reveal that the 1-hour averaged light scattering measure of fine-particulate matter is highly correlated with gravimetric PM10 (Pearson's r = 0.78). The 1-hour averaged nephelometry was less well correlated with CO (r = 0.47) and SO2 (r = 0.16).
In single-pollutant analyses, increases in the individual pollutants (PM, CO, SO2) were not associated with MI after adjusting for relative humidity and temperature (Table 3). Stratification of cases by race, sex, age, and smoking status did not modify the association between 1-hour, 2-hour, 4-hour, and 24-hour averaged fine PM and MI (Table 4). Further analyses using 24-hour averaged nephelometry and copollutants lagged by 24 and 48 hours before MI onset found no association between ambient pollutant levels and MI onset (data not shown).
Analyses stratified by preexisting cardiac disease or cardiovascular risk factors are displayed in Table 5. These analyses did not show an association between an increase in particulate matter 1 hour through 24 hours before MI onset in individuals with any of these previously diagnosed conditions. Further analyses using 24-hour averaged nephelometry lagged by 24 and 48 hours before MI onset found no association between ambient fine-particulate matter levels and MI onset in those with a history of cardiac disease or cardiovascular risk factors (data not shown).
The sources of fine-particulate matter differ by season in the Seattle airshed, with wood smoke contributing a substantial fraction in the heating season (November–February) and motor vehicle sources predominating in the nonheating season (March–October). Therefore, we performed a stratified analysis by season of event. These analyses did not find an association between an increase in fine-particulate matter 1 hour before and MI onset with an OR of 1.01 (95% CI = 0.98–1.05) for events occurring in the heating season and an OR of 0.99 (0.91–1.09) for events occurring in the nonheating season. We found similar results using 2-hour, 4-hour, and 24-hour averaged PM exposures. To assess differential effect of wood smoke and traffic-generated PM on MI onset, we performed time-stratified analyses with MI onset between 7 am and 7 pm reflecting traffic-related PM effect and 7 pm and 7 am reflecting wood smoke contribution in the heating season. These analyses using 1-hour to 4-hour averaged PM exposures did not demonstrate differences in effect by time of day in heating and nonheating seasons (data not shown).
We repeated the detailed single-pollutant analyses with CO as the exposure. These models did not find an association between increased CO levels and MI onset (data not shown).
In addition to our application of the time-stratified case-crossover referent selection approach, we performed a replication of the Boston study analysis13 to determine whether differences in referent selection, case onset-time definition, or model adjustment significantly influenced results. Figure 1 graphically depicts the MITI and Boston study results shown in Table 3 of the paper by Peters et al. To facilitate between-study comparisons, we plotted quintile-specific estimates by the midrange of each quintile, except for the highest quintile where we plotted the midrange of the 80th–95th percentiles. The exposure distributions are similar, with Seattle having more variability and slightly higher exposures. This unadjusted analysis did not suggest an increase in the risk of MI onset with higher quintiles of fine-particulate matter exposure in MITI, in contrast to the associations that were demonstrated in the Boston study analysis. In a second analysis of both the 2-hour and 24-hour average PM exposures adjusted for meteorologic variables and season, we also found no associations. The odds ratio of MI onset for the 24-hour averaged PM in the jointly estimated model was 1.05 (0.94–1.17), whereas the odds ratio for the 2-hour averaged fine-particulate matter was 0.99 (0.91–1.09). Results were similar for models that separately estimated association of 2-hour and 24-hour average fine-particulate matter to MI onset. In contrast, the Boston study demonstrated adjusted odds ratios in the same jointly estimated models of 1.22 (1.04–1.42) for 2-hour average PM2.5 and OR 1.27 (1.06–1.53) for 24-hour average PM2.5.
Despite prior evidence that elevated levels of particulate matter are associated with an increased risk of MI onset, this study did not demonstrate any association between an increase in fine-particulate matter and MI onset in individuals either with or without preexisting cardiac disease. Our study, with its much larger sample size than the Boston study, provides evidence against elevated levels of fine-particulate matter exerting an effect on MI onset in Seattle larger than a relative risk of 1.07 per 10 μg/m3 of PM2.5.
We explored possible differences between these studies that may explain their discrepant findings. First, there are differences in fine-particulate matter composition between the 2 cities. Seattle has fine-particulate matter that is relatively low in sulfate and transition metal content. In contrast, Boston has fine-particulate matter that is rich in sulfates and transition metals from long-range transport of emissions from power plants and industrial sources.26 Prior epidemiologic, in vitro, and animal studies demonstrate that particulate matter composition can influence cardiovascular effect. The National Morbidity and Mortality from Air Pollution Study (NMMAPS) found a stronger association of particulate matter effect with cardiovascular and respiratory mortality in northeastern cities relative to West Coast cities.27–31 Furthermore, compositional analyses of ambient air in Quebec found that particulate matter with high sulfate fractions were more strongly associated with increased hospitalizations for cardiac disease.2
The transition metal and sulfate content of the particles may determine the level of oxidative stress induced by PM2.5.32,33 Physiological and histologic studies in susceptible rodents show cardiac injury after episodic, long-term inhalation of ambient-like combustion PM2.5 with high concentrations of transition metals34,35 Tracheal instillation studies of single metal components document that vanadium exposure induces cardiac arrhythmias in pulmonary hypertensive rats and nickel induces a late onset bradycardia.36
We had concerns that PM2.5 generated from wood burning in the heating season may have less effect on MI onset than traffic-related particulate matter. However, our analyses stratified by season showed no blunting of effects in the winter. Moreover, a further analysis stratified by time of day of MI onset, with particulate matter between 7 am and 7 pm more likely attributable to a traffic-related source, did not find an association with fine-particulate matter. These analyses argue against the hypothesis that fine-particulate matter from wood smoke sources diluted the cardiovascular effect from traffic-related fine-particulate matter.
The associations may differ as a result of differences in susceptibility to fine-particulate matter between the study populations. Prior epidemiologic studies have found elevated risk associated with increased PM2.5 air pollution in those with prior heart disease and the elderly.7,37 Despite a higher percentage of individuals with preexisting angina in the MITI study (40% compared with 23% of participants in the Boston study), we were unable to find an association between particulate matter levels and MI onset. Furthermore, in subanalyses of the MITI data, we found no difference in effect of fine-particulate matter in individuals with preexisting MI or congestive heart failure compared with those with incident MI. From the available Boston study data, the cases appear similar in frequency of diabetes and hypertension. Moreover, the MITI study cases are represented by a slightly higher percentage of individuals greater than age 70 (44% compared with 32% in the Boston study). We have insufficient information from the Boston study to comment on differences in use of cardiac medications that might modify risk of PM2.5 on MI onset.
It is noteworthy that the Boston study used a fundamentally different referent selection strategy. Our prior work suggests that for certain referent selection strategies, it is possible for the estimated effect of exposure to be the result of bias.19–21 This bias can overwhelm the small effects of air pollution on cardiac morbidity and mortality. We used an unbiased time-stratified referent sampling strategy in MITI and found no effect of air pollution. However, we were unable to replicate the Boston Onset Study results for the PM2.5 effect on MI onset even after using the same unidirectional referent selection strategy, onset-time definition, adjustment variables, and statistical analysis. In our replication analysis, we did observe a larger effect estimate (OR = 1.07; CI = 0.96–1.19) for the 24-hour average exposure using the unidirectional referent strategy than with any of our time-stratified analyses. This larger effect estimate is likely the result of overlap bias.20 Overlap bias is a bias in the conditional logistic estimating equations (ie, estimating equations do not have mean zero) as a result of misalignment of the analysis with the referent selection approach.20 The unidirectional referent selection strategy is an example of a nonlocalizable referent selection approach.21 It defines the referent window as a 1:1 function of index time and is consequently susceptible to overlap bias. The localizable and ignorable time-stratified referent selection approach chooses referent windows in a disjoint, a priori manner. Consequently, conditional logistic regression is the appropriate analysis approach.20,21
Because the MITI case-crossover study differs by population size, outcome, case ascertainment, and analytic method from time-series studies of cardiac outcomes, it is difficult to make objective comparisons of results. However, we will highlight 2 additional practical differences between case-crossover and time-series studies: power and analysis approach. The large time-series studies (eg, NMMAPS, APHEA) had a 100- to 500-fold greater number of cardiac events. Because of their size, they have much greater power to detect the small effects of particulate matter on cardiac morbidity and mortality. They found that an interquartile range increase of daily-averaged PM10 was associated with approximately a 1% increase in cardiac mortality.27,38 However, the case-crossover studies can use outcomes with a more precise case definition, potentially offsetting some of the power advantage from the time-series studies.
In contrast to prior time-series studies that found a small effect of PM on hospital admission for MI,39–42 the Boston study had a much larger effect estimate (OR = 1.62; 95% CI = 1.13–2.34) for induction of MI onset for an increase of 20 μg/m3 PM2.5. The larger effect estimate in the Boston study relative to the time-series studies may be explained by different outcome measures and approaches to case ascertainment. To date, time-series studies have relied on hospital admission data. Those data lack specificity and may lead to misclassification of cases. In contrast, the Boston study used observational data with stringent event adjudication to determine case status. Differences in classification could severely attenuate estimates from time-series studies relative to those from well-defined case series. Furthermore, it is possible that MI onset is more sensitive to elevated PM2.5 than is hospitalization for MI. However, the MITI data found no fine-particulate matter effect, although it also represents a well-defined and adjudicated case series of MI onset. This suggests that outcome measures and approaches to case ascertainment are not primary explanations of the differences between time-series and case-crossover study results.
With an unrestricted referent window and adjustment in the statistical model for time-varying confounders, the case-crossover analysis is equivalent to the time-series regression analysis.21 Case-crossover studies control for time-varying confounders by restriction of the referent window (ie, matching) rather than modeling; therefore, their analysis differs from time-series study analysis. Because of their reliance on matching rather than modeling, it is likely that case-crossover studies can better control for seasonal confounding than time-series studies. Furthermore, because the data and analysis focus on individual cases, case-crossover studies can more easily assess effect modification as a result of time-invariant personal factors.
Our study had some important strengths. Our sample size is large and drawn from a population-based registry that included individuals with and without previous known cardiac disease, the case ascertainment period is long and continuous, and the greater Seattle airshed has been well characterized. Moreover, our referent selection strategy is based on a sound theoretical foundation supported by good performance in simulation studies.21
However, there are limitations to our study. We were unable to control for use of cardioprotective medications before MI onset, including angiotensin-converting enzyme inhibitors, beta-blockers, statins, and aspirin, because the information were not available. Moreover, data on blood pressure control, decompensation of cardiac disease before MI, and heart rate before MI onset were not available for analysis.
Our study did not detect an association between very short-term elevations in PM and onset of MI. We propose that the next steps to clarifying differences in results between studies include: reanalysis of prior case-crossover studies, particularly those using nonlocalizable referent selection strategies; performance of a prospective multicity study of MI onset that includes more intensive exposure monitoring with greater compositional information on PM; and further in vitro and animal studies of cardiovascular effects that use fine-particulate matter representative of both the long-range transport particles and locally generated particles to which humans are exposed.
1.Dockery DW, Pope CA III, Xu X, et al. An association between air pollution and mortality in six US cities. N Engl J Med
2.Goldberg MS, Burnett RT, Bailar JC III, et al. The association between daily mortality and ambient air particle pollution in Montreal, Quebec. 2. Cause-specific mortality. Environ Res
3.Hoek G, Brunekreef B, Fischer P, et al. The association between air pollution and heart failure, arrhythmia, embolism, thrombosis, and other cardiovascular causes of death in a time series study. Epidemiology
4.Samet JM, Dominici F, Curriero FC, et al. Fine particulate air pollution and mortality in 20 US cities, 1987–1994. N Engl J Med
5.Schwartz J. Air pollution and hospital admissions for heart disease in eight US counties. Epidemiology
6.Schwartz J, Morris R. Air pollution and hospital admissions for cardiovascular disease in Detroit, Michigan. Am J Epidemiol
7.Goldberg MS, Bailar JC III, Burnett RT, et al. Identifying subgroups of the general population that may be susceptible to short-term increases in particulate air pollution: a time-series study in Montreal, Quebec. Res Rep Health Eff Inst
; discussion 115-120.
8.Peters A, Skorkovsky J, Kotesovec F, et al. Associations between mortality and air pollution in central Europe. Environ Health Perspect
9.Prescott GJ, Cohen GR, Elton RA, et al. Urban air pollution and cardiopulmonary ill health: a 14.5 year time series study. Occup Environ Med
10.Izzotti A, Parodi S, Quaglia A, et al. The relationship between urban airborne pollution and short-term mortality: quantitative and qualitative aspects. Eur J Epidemiol
11.Zanobetti A, Schwartz J, Dockery DW. Airborne particles are a risk factor for hospital admissions for heart and lung disease. Environ Health Perspect
12.D'Ippoliti D, Forastiere F, Ancona C, et al. Air pollution and myocardial infarction in Rome: a case-crossover analysis. Epidemiology
13.Peters A, Dockery DW, Muller JE, et al. Increased particulate air pollution and the triggering of myocardial infarction. Circulation
14.Levy D, Sheppard L, Checkoway H, et al. A case-crossover analysis of particulate matter air pollution and out-of-hospital primary cardiac arrest. Epidemiology
15.Sullivan J, Ishikawa N, Sheppard L, et al. Exposure to ambient fine particulate matter and primary cardiac arrest among persons with and without clinically recognized heart disease. Am J Epidemiol
16.Mittleman MA, Maclure M, Robins JM. Control sampling strategies for case-crossover studies: an assessment of relative efficiency. Am J Epidemiol
17.Levy D, Lumley T, Sheppard L, et al. Referent selection in case-crossover analyses of acute health effects of air pollution. Epidemiology
18.Maclure M. The case-crossover design: a method for studying transient effects on the risk of acute events. Am J Epidemiol
19.Janes H, Sheppard L, Lumley T. Assessing the health effects of air pollution using the case-crossover design. (Unpublished paper)
20.Janes H, Sheppard L, Lumley T. Overlap bias in the case-crossover design, with application to air pollution exposures. Stat Med
. 2003. In press.
21.Lumley T, Levy D. Bias in the case-crossover design: implications for studies of air pollution. Environmetrics
22.Every NR, Larson EB, Litwin PE, et al. The association between on-site cardiac catheterization facilities and the use of coronary angiography after acute myocardial infarction. Myocardial Infarction Triage and Intervention Project Investigators. N Engl J Med
23.Weaver WD, Cerqueira M, Hallstrom AP, et al. Prehospital-initiated vs hospital-initiated thrombolytic therapy. The Myocardial Infarction Triage and Intervention Trial. JAMA
24.Kudenchuk PJ, Maynard C, Martin JS, et al. Comparison of presentation, treatment, and outcome of acute myocardial infarction in men versus women (the Myocardial Infarction Triage and Intervention Registry). Am J Cardiol
25.Liu L-JS, Slaughter JC, Larson T. Comparison of light scattering devices and impactors for particulate measurements in indoor, outdoor, and personal environments. Environ Sci Technol
26.Sarnat JA, Long CM, Koutrakis P, et al. Using sulfur as a tracer of outdoor fine particulate matter. Environ Sci Technol
27.Samet JM, Zeger SL, Dominici F, et al. The National Morbidity, Mortality, and Air Pollution Study. Part II: morbidity and mortality from air pollution in the United States. Res Rep Health Eff Inst
. 2000;94:5–70 ; discussion 71-79.
28.Samet JM, Dominici F, Zeger SL, et al. The National Morbidity, Mortality, and Air Pollution Study. Part I: methods and methodologic issues. Res Rep Health Eff Inst
; discussion 75–84.
29.Sarnat JA, Schwartz J, Suh HH. Fine particulate air pollution and mortality in 20 US cities. N Engl J Med
30.Samoli E, Schwartz J, Wojtyniak B, et al. Investigating regional differences in short-term effects of air pollution on daily mortality in the APHEA project: a sensitivity analysis for controlling long-term trends and seasonality. Environ Health Perspect
31.Dominici F, McDermott A, Zeger SL, et al. National maps of the effects of particulate matter on mortality: exploring geographical variation. Environ Health Perspect
32.Burnett RT, Cakmak S, Brook JR, et al. The role of particulate size and chemistry in the association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases. Environ Health Perspect
33.Molinelli AR, Madden MC, McGee JK, et al. Effect of metal removal on the toxicity of airborne particulate matter from the Utah Valley. Inhal Toxicol
34.Kodavanti UP, Schladweiler MC, Ledbetter AD, et al. The spontaneously hypertensive rat as a model of human cardiovascular disease: evidence of exacerbated cardiopulmonary injury and oxidative stress from inhaled emission particulate matter. Toxicol Appl Pharmacol
35.Moyer CF, Kodavanti UP, Haseman JK, et al. Systemic vascular disease in male B6C3F1 mice exposed to particulate matter by inhalation: studies conducted by the National Toxicology Program. Toxicol Pathol
36.Campen MJ, Nolan JP, Schladweiler MC, et al. Cardiovascular and thermoregulatory effects of inhaled PM-associated transition metals: a potential interaction between nickel and vanadium sulfate. Toxicol Sci
37.Anderson HR, Bremner SA, Atkinson RW, et al. Particulate matter and daily mortality and hospital admissions in the west midlands conurbation of the United Kingdom: associations with fine and coarse particles, black smoke and sulphate. Occup Environ Med
38.Le Tertre A, Medina S, Samoli E, et al. Short-term effects of particulate air pollution on cardiovascular diseases in eight European cities. J Epidemiol Community Health
39.Schwartz J. Air pollution and hospital admissions for heart disease in eight US counties. Epidemiology
40.Ballester F, Tenias JM, Perez-Hoyos S. Air pollution and emergency hospital admissions for cardiovascular diseases in Valencia, Spain. J Epidemiol Community Health
41.Dockery DW. Epidemiologic evidence of cardiovascular effects of particulate air pollution. Environ Health Perspect
. 2001;109(suppl 4):483–486.
42.Prescott GJ, Cohen GR, Elton RA, et al. Urban air pollution and cardiopulmonary ill health: a 14.5 year time series study. Occup Environ Med