There is substantial epidemiologic evidence linking cardiovascular mortality and morbidity with short- and long-term exposures to ambient air pollution.1–4 The main candidate mechanisms are release of prothrombotic and inflammatory cytokines from the lung, and effects on the electrical activity and autonomic function of the heart.5 The latter hypothesis is supported by associations between air pollution and heart rate variability and ectopy, observed in experimental studies and in some, though not all, ambient studies of human subjects.6–14 Time-series studies of hospital admissions have not shown consistent positive associations between air pollution and arrhythmic disorders of the heart,15–24 but in general the size of any effect tends to be similar to those observed with other diagnoses such as acute myocardial infarction or congestive cardiac failure.3 Studies of air pollution and sudden death or acute cardiac arrest, in which arrhythmia is likely to be an important mechanism, have not shown convincing associations.25,26
A promising epidemiologic approach has been to study the associations between ambient air pollution and the activity of implantable cardioverter defibrillators. These devices are implanted in persons who have experienced, or are at high risk of experiencing, life-threatening ventricular arrhythmias, principally ventricular tachycardia and ventricular fibrillation. These defibrillators continuously monitor the heart rate and, when certain threshold conditions are exceeded, intervene either by pacing the heart or by delivering an electric shock to reverse the abnormal rhythm. At clinic review, the dates and times of activations, together with the ventricular rate and electrocardiogram, are obtained telemetrically.
To date, studies of air pollution and implantable-defibrillator records have been confined to 9 reports from 4 North American cities.27–35 The results have been inconsistent as to whether associations can be detected and (where these have been observed), which pollutant or pollution source is responsible. Some of this heterogeneity is probably explained by the low statistical power of some studies and by differences in methodology. The largest of these studies (that from Atlanta) failed, with one exception (the coarse fraction of particulate matter) to find convincing evidence of associations between defibrillator-recorded tachyarrhythmias and a comprehensive range of primary and secondary pollutants.
We report an investigation into the association between a range of particulate and gaseous pollutants and the activation of implantable cardioverter defibrillators among patients attending clinics in London. It comprises the largest series of defibrillator activations yet studied and is the first from a European city.
All 9 implantable-cardioverter-defibrillator clinics in London National Health Service hospitals participated in the study. Clinic records were examined to identify those patients with an implanted defibrillator that recorded the date of discharge (“Generation 2” or later devices) and who had a record of an activation event in 1995 through 2003. Some investigators have referred to tachycardias that did not result in treatment as “events.” We instead focused on tachycardias for which the defibrillator delivered treatment (pacing or shock). To avoid the ambiguous use of the term “event” we shall refer to these as activations.
From records in the clinic, data were extracted for each patient on the presenting clinical problem, underlying cardiac diagnosis, ejection fraction (% value), a qualitative indication of ventricular function (normal or mild, moderate, severe impairment), device implanted (name and model, chambers sensed, date of first implantation) and cardiac-related drugs (beta-blockers, antiarrhythmics, digoxin, diuretics, antihypertensives, anticoagulants) at implantation.
Data extracted for each activation included date of the event, time of day, type of rhythm as described in the downloaded log of events or in the notes of the clinic review, and any mention of cardiac-related drugs likely to have been taken at the time of the event. Where more than one activation occurred in one 24-hour period, only the first of these was described in detail, although the total number for that day was recorded. Implantable defibrillators are programmed to respond to EKG R-R intervals above a set rate. This is primarily to detect and treat ventricular tachycardia and ventricular fibrillation. Tachycardias may also occur as a result of sinus tachycardia and supraventricular arrhythmias; in such cases, the device will record and treat the arrhythmia as a ventricular one.36 The most common supraventricular arrhythmias are atrial fibrillation, atrial flutter, and supraventricular tachycardia. The electrocardiogram was found for two-thirds of the activations and this was assessed by a trained study cardiologist. The cardiologist diagnosis was based on information about the R-R interval (time between 2 beats), rate, morphology of P wave, duration of P wave (if any), and the morphology and duration of the R-R intervals according to their regularity.
All monitors measuring background concentrations of particulate and gaseous pollutants in London and the South East were identified. With certain exceptions (such as particle number concentrations), the criterion for selection was the availability of ratified data for 75% or more of days between the beginning of 1995 and the end of 2003. The monitoring stations, summary statistics for each monitor, and methods of measurement are described in eTables 1 and 2 (http://links.lww.com/EDE/A375). Data on the following pollutants were obtained: particulate matter less than 10 and 2.5 μm aerodynamic diameter (PM10 and PM2.5, respectively), black smoke, particle sulfate (SO4=), particle number concentrations, nitrogen dioxide (NO2), nitric oxide (NO), nitrogen oxides (NOx), ozone (O3), sulfur dioxide (SO2), and carbon monoxide (CO). With the exception of O3 (8 hours maximum), the averaging time for all pollutants was 24 hours. The units of measurement are μg/m3 except for carbon monoxide (mg/m3) and particle counts (number per cm3).
A geographic information system was used to assign to each patient's home address, in the month of any defibrillator event, the daily average ambient concentrations of pollutants recorded by the nearest background monitor (eFigure 1, http://links.lww.com/EDE/A375). Based on expert judgment as to the distance up to which monitoring data would be relevant, the following maximum distances were adopted: 10 km for particle number concentrations (1 monitor); 20 km for PM10 (12 monitors), black smoke (11 monitors), SO2 (16 monitors), NOx (21 monitors), and CO (15 monitors); 40 km for PM2.5 (1 monitor) and O3 (20 monitors) and the whole study area for SO4=(1 monitor). A person-month was included in the analysis if the above distance criteria were met and a measure of pollutant was available for more than 90% of days of that month. Because all monitors used in this analysis were in urban areas (with the exception of SO4=), we excluded all subjects in rural areas (defined as the bottom quintile of 5 km smoothed population density at ward level). Table 1 summarizes these as the distribution of all daily concentrations from the linked site used in the analysis of activations. Means of minimum and maximum daily temperatures were obtained from Heathrow Airport.
Our statistical approach was equivalent to the fixed-stratum case-crossover method,37 which can also be seen as a particular type of case-series method.38 To allow for variations in activation episode rates between patients, and over medium-long time periods in the same patient, we stratified analysis into risk sets comprising person-lunar months of 28 days. The cases were the day or days of activation of the defibrillator. The control days were all the other days of the lunar month. Data for the first month after implantation were excluded. Conditional logistic regression was used to assess the association of the occurrence of activation days with pollution within person-lunar months.
It is possible that an event occurring on any given day may make events on a subsequent day more likely (autocorrelation). This was allowed for (to avoid distortion of standard errors) by introducing, as explanatory variables, indicators representing the occurrence of events on previous days.39 For events occurring at the beginning of each lunar month, information from previous days was not missing, as the occurrence or nonoccurrence of events for each person was known even during noncontiguous periods. The indicators were included until there was little evidence for further positive dependence (P < 0.1), which resulted in inclusion at lags 1 and 2.
Preliminary analysis found that temperature was likely to be an important confounder over several lags (eFigure 2, http://links.lww.com/EDE/A375). We controlled for temperature by using natural cubic splines (3 knots) of daily mean temperature; separate spline functions were used for the following lag structures of the temperature measure: lags 0, 1, and 2 days, average lags 3–4 and 5–7days. Humidity was not included in the model because it was not related to activations. Indicator variables for day of the week and holidays were also included. Our approach to the control of temperature was a stringent one and, in fact, if the degree of control for temperature was unnecessary, this would reduce the precision of the estimates though should not bias them.40 We therefore also investigated the sensitivity of the results to (1) no temperature control, (2) reduced control (mean lag 0–1, lag 2–5), and (3) increased control (+ mean lags 8–12). Sensitivity to inclusion of relative humidity (using a 3-knot cubic spline) was also examined.
The association between ambient pollution concentrations and activations was modeled using unconstrained distributed lags from 0 to 5 days. That is, we included 6 pollution variables simultaneously in the model, one for each lag 0–5, so that immediate or delayed effects could be detected. These results are reported as the sum of individual lag terms at lag 0, 0–1, 0–2, 0–3, 0–4, and 0–5. A priori, the emphasis was on lags 0–1 and 0–5 because these are lags frequently reported by other daily time-series studies. This choice also reduces the number of associations tested. When an event occurred in the first days of the lunar month, data were obtained from the continuous series. The primary analytic data set is that of individually linked data for all patients living in London and the surrounding counties in the South-East. In addition, as a sensitivity analysis, we analyzed only London patients to compare the results obtained by individual linking of pollution exposure with those obtained using the daily time-series of pollution data constructed for the whole of London using all eligible monitors within London with at least 75% days of data.
For those pollutants showing positive associations, we examined the interactions with age, sex, underlying ischemic heart disease, ejection fraction, frequency of events, history of storm days, prior event within 3 days, concurrent use of beta-blockers and diagnosis of supraventricular tachycardia. Ethics approval was first obtained from the South Thames Multiple Research Ethics Committee and this was then endorsed by the Local Ethics Committee for each of the participating hospitals.
Records of 2188 patients were examined and of these 726 (33%) had data relating to an activation day. The majority (60%) of the defibrillators sensed a single heart chamber. Of 5654 activations, 98 were excluded because they occurred outside the defined study time range, a further 94 were excluded because of data anomalies. The final sample was 705 patients for whom there were data on 5462 activation days. Forty-four percent lived in Greater London. The demographic and clinical features of the patients are described in Table 2. The mean follow-up was 1204 days. The mean activation rate was similar for London and out-of-London patients. For days on which any activation occurred, 68% had 1 activation, 13% had 2, 6% had 3, 3% had 4, and 10% had 5 or more (up to 122).
For the 5462 days with one or more activations, the therapies recorded at the clinic review were antitachycardia pacing (73%), shock (25%) or “diverted” due to spontaneous reversion (3%). Of the 5462 tachyarrhythmias, 3272 had electrocardiograms assessed by the study cardiologist. A diagnosis of ventricular arrhythmia was made in 99% of the clinic assessments and 87% of study cardiologist assessments. Of the electrocardiograms assessed by the cardiologist, those not diagnosed as ventricular arrhythmia were mostly supraventricular arrhythmia (5%) or sinus tachycardia (5%). The main supraventricular arrhythmias were atrial fibrillation (n = 141) and supraventricular tachycardia (n = 76).
Of all 5462 activation days, linkage was achieved in 100% for SO4=, 75% for O3 and about 50% for PM10, CO, NO2, and SO2. PM10 did not correlate highly with any other pollutant (apart from PM2.5 [r = 0.88], which is expected, being its major component). PM2.5 correlated best with SO4= (r = 0.55) and vice versa. Black smoke correlated best with NOx (r = 0.60). Particle number concentrations correlated poorly with the other pollutants, the strongest correlations being with CO (r = 0.39) and NO2 (r = 0.38). O3 was negatively associated with all of the other pollutants, most strongly with NO2 (r = −0.65) and least with SO4= (r = −0.05) (eTable 3, http://links.lww.com/EDE/A375).
Table 3 shows the associations between each of the 11 pollutants and defibrillator activations. Odds ratios (ORs) are presented for the cumulative effect of lags 0–1 days and 0–5 days. Results are presented for a 10 unit change in each pollutant, except CO and SO4= (1 unit change) and particle number concentrations (1000 unit change). The estimates for cumulative effects to each lag (0, 0–1, 0–2, 0–3, 0–4, and 0–5) are shown for the full linked analysis in Figures 1 and 2. Each of these estimates represents the sum of the individual-day lags up to that day. The proportion of activations linked with specific pollutants varied considerably, and thus there are variations in statistical power.
Among the 22 estimates in Table 3, approximately half were positive. Statistically, the strongest association was for SO4=, lag 0–1 days (OR for 1 μg/m3 = 1.025 [95% CI = 1.003 to 1.047]). The lag 0–5 estimates for SO4= were also in a positive direction. Estimates for PM2.5, O3, and SO2 were also consistently positive in direction while that for PM10 was effectively null for lag 0–1 and somewhat below 1 for lags 0–5. The estimates for NO2 and CO were negative. The estimates for black smoke and particle number concentrations were around unity.
Patterns were similar in the London-only conventional case-crossover ecologic analysis, though precision was less than in the full analysis, as only about half the activation-days occurred in patients residing in London. For London residents only, the results for the linked and unlinked analyses were similar (eTable 4, http://links.lww.com/EDE/A375).
To examine the potential sensitivity of the results to our method of temperature control, we compared 3 approaches for all 11 pollutants. These were (1) no control for temperature, (2) the control used in the current analysis (spline functions for lags 0, 1, 2, average 3–4, and average 5–7), and (3) the second extended to include a spline function for the average of lag 8–12 days temperature. The only analysis that showed any noticeable sensitivity to temperature control was for SO4=, in which temperature control increased the size and precision of the estimate (for no temperature control, OR = 1.011 (95% CI = 0.992 to 1.032); current analysis, 1.025 (1.003–1.047); and extended to 8–12 days, 1.025 (1.003 to 1.047). None of the remaining 11 pollutants showed important evidence of sensitivity to temperature control. Inclusion of relative humidity in the models made no difference. The results were not sensitive to the level of seasonal control when varied by using 14-day rather than 28-day stratification.
We looked for evidence of effect modification by a series of potentially important factors: age over 65 years, sex, ejection fraction, underlying ischemic heart disease, frequent activations, use of beta-blockers, supraventricular arrhythmia, storm days, and previous activation within 3 days (eTable 5, http://links.lww.com/EDE/A375). SO4=, O3, PM10 and PM2.5 were chosen because they showed effects in a positive direction. Of 70 interactions examined, only 2 were statistically significant (less than expected by chance). Among users of beta-blocker drugs, there was a negative association with PM2.5 at lag 0–1 (OR = 0.6 [95% CI = 0.37 to 0.98]) compared with a weak positive association (1.05 [0.91 to 1.23]) in those for whom there was no record of the drug in the follow-up notes (test for interaction, P = 0.03). Results were similar for PM2.5 and beta-blockers for lag 0–5. The other interaction was a greater risk associated with SO4= at lag 0–1 in subjects 65 years or over (1.18 [1.03 to 1.36]) compared with those under 65 (0.96 [0.83 to 1.11]) (test for interaction, P = 0.04).
We investigated the associations between 11 ambient air pollutants and implantable cardioverter defibrillator interventions in 705 patients who had experienced a total of 5462 activation days. Using a fixed-stratum case-crossover approach, we considered lags up to 5 days, controlling for temperature. Associations tended to be positive for SO4=, O3, PM10, PM2.5, and SO2, the strongest association being with SO4=. Associations with pollutants more directly related to traffic emissions (particle number concentrations, black smoke, CO, NO, NO2, and NOx) were generally negative in direction. There were few interactions with clinical factors, and those observed were not consistent in direction.
There is little consistency between previous studies of arrhythmias detected by implantable defibrillators in respect of the pollutants identified, the significance of associations, and the way the defibrillator record was used to define the type of arrhythmia (Table 4). 27–29,31–34 Possible reasons for the heterogeneity in results include variations in statistical power, demographic and clinical characteristics of the patients, indications for implantation, definition of outcome, pollution exposure assessment, statistical approach and the underlying toxicity of the ambient pollution mixture. The results of the largest North American study (Atlanta) were essentially null for all events (irrespective of activation), apart from an unexpected association with coarse particles (which we did not study in London).35 This could be interpreted as a chance finding, especially considering the lack of spatial homogeneity of coarse particles in such a large city. Our study focused only on defibrillator-activation days and in this respect was larger than the Atlanta study (5462 vs. 2539 activation days, respectively). Another difference between our study and some others is that we did not require independent cardiologic confirmation that an activation was caused by a ventricular arrhythmia. However, cross-validation with a cardiologist diagnosis in two-thirds of the activations indicated that our results were unlikely to be substantially affected by the inclusion of an estimated 10% of supraventricular arrhythmias (mainly atrial fibrillation and sinus tachycardia) among the activation days.
Exposure measurement error is inevitable in studies that rely on community monitors; this will tend to mask real underlying effects. However, we do not think that our study was more prone to measurement error than those carried out in North America; our patients were, if anything, living closer to the relevant air-pollution-monitoring stations. For example, Greater London is covered by one airshed and, while large (1580 km2), is considerably smaller in area than Atlanta, which has an area of 21,694 km2 Further, where there was more than one monitor, linkage to the nearest will provide improved precision. Some studies have used rolling average 24-hour concentrations.33 While this approach has some advantages, we do not think that we are likely to have missed any effects of importance by using fixed 24-hour averages. One study has reported associations with hourly concentrations,34 but we consider that this approach to be problematic because it introduces confounding due to diurnal factors that affect both pollution and risk of activations.
Time-series studies in London have observed positive associations between a range of pollutants and hospital admissions for cardiac arrhythmias,16 and similar results have been obtained in most hospital admissions studies in other cities.15,17–24 A meta-analysis of 7 studies of daily admissions for dysrhythmias obtained a random-effects summary estimate of 0.4% (95% CI = 0.00 to 0.8) for a 10 μg/m3 increase in PM10.3 In the current study, the central estimate for PM10 and ventricular arrhythmias was similar (0.3% increase per 10 μg/m3 increase in PM10 (95% CI = −0.5 to 6.5). The latter result does not exclude the increased risk of about 0.4% observed in daily time-series studies.
We had not predicted the results for SO4=, and this should be interpreted in the context of being one of 11 pollutants considered. However, the association has some plausibility; SO4= has been shown to be associated with health effects in both short-term and long-term exposure studies.41 Since SO4= is unlikely to be toxic per se, particulate matter with a higher concentration of SO4= is likely to be acting as a marker for pollution derived from the combustion of sulfur-containing fossil fuels. Such pollution contains a variety of potentially toxic components including metals, and there is increasing epidemiologic and toxicologic evidence suggesting that these may be responsible for health effects.42 Thus, it would not be surprising if the toxicity that SO4= represents varies geographically, depending on the source. This might explain why other studies such as the study in Atlanta did not find an association.
Notwithstanding the above discussion, when this and previous epidemiologic studies are considered as a whole, the evidence that ambient air pollution triggers the activation of implantable cardioverter defibrillators appears to be weak. This conclusion is based on 2 main arguments. The first is that most results for the pollutants studied have been null, including in those studies with higher statistical power such as Atlanta and the present London study. The second is that no study, including the present one, has obtained evidence relating to a particular pollutant as a result of an a priori hypothesis; the usual approach has, apparently, been to analyze all the available pollutants and emphasize those, if any, with positive associations. The earliest report27 implicated a primary mobile source pollutant (NO2), and it is probable that most investigators have tended toward this hypothesis, if only implicitly, from a belief that exhaust emissions are likely to be relatively more toxic. The Atlanta study found no support for an association with mobile-source pollutants, and the results of the present study were similar in this respect. We note that SO4= and O3 (the 2 strongest positive associations observed) together with 2 other pollutants with weak positive associations (PM2.5 and PM10) all tend to have a strong regional character. This might indicate that the associations we have observed, if not due to chance, are due to confounding by uncontrolled weather factors, despite our rigorous control for temperature and humidity.
We thank the following consultants for facilitating access to the records of their clinics: Michael Cooklin, Kings College Hospital; Wyn Davies, St. Mary's Hospital; Jas Gill, St. Thomas' Hospital; David Lefroy, Hammersmith Hospital; Mark Mason, Harefields Hospital; Anthony Nathan, St. Bartholomew's Hospital; Edward Rowland, St. George's Hospital; Simon Sporton, The Heart Hospital; and Richard Sutton, The Royal Brompton Hospital. We also thank these colleagues: Cathy McKay, Research Nurse and Xiao Hua Guo, cardiologist, for their painstaking field work; Dr. Douglas Dockery for the initial idea and advice on study design; Sue Jones, Manager of the St. George's Pacing Clinic, for practical guidance; Mariangela Caputi for assembling the air pollution data; and Mary Field-Smith for office support.
1. Brook RD, Franklin B, Cascio W, et al. Air pollution and cardiovascular disease: a statement for healthcare professionals from the Expert Panel on Population and Prevention Science of the American Heart Association. Circulation
2. Pope CA III, Burnett RT, Thurston GD, et al. Cardiovascular mortality and long-term exposure to particulate air pollution: epidemiological evidence of general pathophysiological pathways of disease. Circulation
3. Department of Health Committee on Medical Effects of Air Pollution. Cardiovascular Disease and Air Pollution.
London: Department of Health; 2006.
4. Pope CA III, Dockery DW. Health effects of fine particulate air pollution: lines that connect. J Air Waste Manag Assoc
5. Health Effects Institute. Understanding the Health Effects of Components of the Particulate Matter Mix: Progress and Next Steps.
HEI Perspectives. Cambridge, MA: Health Effects Institute; 2002.
6. Liao D, Creason J, Shy C, Williams R, Watts R, Zweidinger R. Daily variation of particulate air pollution and poor cardiac autonomic control in the elderly. Environ Health Perspect
7. Godleski JJ, Verrier RL, Koutrakis P, et al. Mechanisms of Morbidity and Mortality From Exposure to Ambient Air Particles.
Research Report 91. Cambridge MA: Health Effects Institute; 2000.
8. Nadziejko C, Fang K, Narciso S, et al. Effect of particulate and gaseous pollutants on spontaneous arrhythmias in aged rats. Inhal Toxicol
9. Berger A, Zareba W, Schneider A, et al. Runs of ventricular and supraventricular tachycardia triggered by air pollution in patients with coronary heart disease. J Occup Environ Med
10. Sarnat SE, Suh HH, Coull BA, Schwartz J, Stone PH, Gold DR. Ambient particulate air pollution and cardiac arrhythmia in a panel of older adults in Steubenville, Ohio. Occup Environ Med
11. Routledge HC, Manney S, Harrison RM, Ayres JG, Townend JN. Effect of inhaled sulphur dioxide and carbon particles on heart rate variability and markers of inflammation and coagulation in human subjects. Heart
12. Sullivan JH, Schreuder AB, Trenga CA, et al. Association between short term exposure to fine particulate matter and heart rate variability in older subjects with and without heart disease. Thorax
13. Corey LM, Baker C, Luchtel DL. Heart-rate variability in the apolipoprotein E knockout transgenic mouse following exposure to Seattle particulate matter. J Toxicol Environ Health A
14. Liao D, Whitsel EA, Duan Y, et al. Ambient particulate air pollution and ectopy–the environmental epidemiology of arrhythmogenesis in Women's Health Initiative Study, 1999–2004. J Toxicol Environ Health A
15. Schwartz J, Morris R. Air pollution and hospital admissions for cardiovascular disease in Detroit, Michigan. Am J Epidemiol
16. Poloniecki JD, Atkinson RW, de Leon AP, Anderson HR. Daily time series for cardiovascular hospital admissions and previous day's air pollution in London, UK. Occup Environ Med
17. Wong TW, Lau TS, Yu TS, et al. Air pollution and hospital admissions for respiratory and cardiovascular diseases in Hong Kong. Occup Environ Med
18. Burnett RT, Smith-Doiron M, Stieb D, Cakmak S, Brook JR. Effects of particulate and gaseous air pollution on cardiorespiratory hospitalizations. Arch Environ Health
19. Tolbert PE, Klein M, Metzger KB, et al. Interim results of the study of particulates and health in Atlanta (SOPHIA). J Expo Anal Environ Epidemiol
20. Lippmann M, Ito K, Nadas A, Burnett RT. Association of Particulate Matter Components With Daily Mortality and Morbidity in Urban Populations.
Research Report 95. Cambridge, MA: Health Effects Institute; 2000.
21. Linn WS, Szlachcic Y, Gong H, Kinney PL, Berhane KT. Air pollution and daily hospital admissions in metropolitan Los Angeles. Environ Health Perspect
22. Stieb DM, Beveridge RC, Brook JR, et al. Air pollution, aeroallergens and cardiorespiratory emergency department visits in Saint John, Canada. J Expo Anal Environ Epidemiol
23. McGowan JA, Hider PN, Chacko E, Town GI. Particulate air pollution and hospital admissions in Christchurch, New Zealand. Aust NZ J Public Health
24. Santos UP, Terra-Filho M, Lin CA, et al. Cardiac arrhythmia emergency room visits and environmental air pollution in Sao Paulo, Brazil. J Epidemiol Community Health
25. Checkoway H, Levy D, Sheppard D, Kaufman J, Koenig JQ, Siscovick D. A Case-Crossover Analysis of Fine Particulate Matter Air Pollution and Out-Of-Hospital Sudden Cardiac Arrest.
HEI Research Report 99. Boston: Health Effects Institute; 2000.
26. 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
27. Peters A, Liu E, Verrier RL, et al. Air pollution and incidence of cardiac arrhythmia. Epidemiology
28. Vedal S, Rich K, Brauer M, White R, Petkau J. Air pollution and cardiac arrhythmias in patients with implantable cardioverter defibrillators. Inhal Toxicol
29. Rich KE, Petkau J, Vedal S, Brauer M. A case-crossover analysis of particulate air pollution and cardiac arrhythmia in patients with implantable cardioverter defibrillators. Inhal Toxicol
30. Dockery DW, Luttmann-Gibson H, Rich DQ, et al. Association of air pollution with increased incidence of ventricular tachyarrhythmias recorded by implanted cardioverter defibrillators. Environ Health Perspect
31. Dockery DW, Luttmann-Gibson H, Rich DQ, et al. Particulate Air Pollution and Nonfatal Cardiac Events. Part II. Association of Air Pollution With Confirmed Arrhythmias Recorded by Implanted Defibrillators. Research Report 124. Boston: Health Effects Institute; 2005.
32. Rich DQ, Schwartz J, Mittleman MA, et al. Association of short-term ambient air pollution concentrations and ventricular arrhythmias. Am J Epidemiol
33. Rich DQ, Kim MH, Turner JR, et al. Association of ventricular arrhythmias detected by implantable cardioverter defibrillator and ambient air pollutants in the St Louis, Missouri metropolitan area. Occup Environ Med
34. Rich DQ, Mittleman MA, Link MS, et al. Increased risk of paroxysmal atrial fibrillation episodes associated with acute increases in ambient air pollution. Environ Health Perspect
35. Metzger KB, Klein M, Flanders WD, et al. Ambient air pollution and cardiac arrhythmias in patients with implantable defibrillators. Epidemiology
36. Grimm W, Flores BF, Marchlinski FE. Electrocardiographically documented unnecessary, spontaneous shocks in 241 patients with implantable cardioverter defibrillators. Pacing Clin Electrophysiol
37. Lumley T, Levy D. Bias in the case-crossover design: implications for studies of air pollution. Environmetrics
38. Farrington CP, Whitaker HJ. Semiparametric analysis of case series data. Appl Stat
39. Bonney GE. Logistic regression for dependent binary observations. Biometrics
40. Schisterman EF, Cole SR, Platt RW. Overadjustment Bias and Unnecessary Adjustment in Epidemiologic Studies. Epidemiology
41. Committee on the Medical Effects of Air Pollutants (COMEAP). Long-Term Exposure to Air Pollution: Effect on Mortality.
London: Department of Health; 2009.
42. Lippmann M, Ito K, Hwang JS, Maciejczyk P, Chen LC. Cardiovascular effects of nickel in ambient air. Environ Health Perspect