OUP user menu

Rapid effects of air pollution on ventricular arrhythmias

Petter L.S. Ljungman, Niklas Berglind, Christina Holmgren, Fredrik Gadler, Nils Edvardsson, Göran Pershagen, Mårten Rosenqvist, Bengt Sjögren, Tom Bellander
DOI: http://dx.doi.org/10.1093/eurheartj/ehn463 2894-2901 First published online: 12 November 2008

Abstract

Aims Air pollution has been associated with ventricular arrhythmias in patients with implantable cardioverter defibrillators (ICDs) for exposure periods of 24–48 h. Only two studies have investigated exposure periods <24 h. We aimed to explore such effects during the 2 and 24 preceding hours as well as in relation to distance from the place of the event to the air pollution monitor.

Methods and results We used a case-crossover design to investigate the effects of particulate matter <10 µm in diameter (PM10) and nitrogen dioxide (NO2) in 211 patients with ICD devices in Gothenburg and Stockholm, Sweden. Events interpreted as ventricular arrhythmias were downloaded from the ICDs, and air pollution data were collected from urban background monitors. We found an association between 2 h moving averages of PM10 and ventricular arrhythmia [odds ratio (OR) 1.31, 95% confidence interval (CI) 1.00–1.72], whereas the OR for 24 h moving averages was 1.24 (95% CI 0.87–1.76). Corresponding ORs for events occurring closest to the air pollution monitor were 1.76 (95% CI 1.18–2.61) and 1.74 (95% CI 1.07–2.84), respectively. Events occurring in Gothenburg showed stronger associations than in Stockholm.

Conclusion Moderate increases in air pollution appear to be associated with ventricular arrhythmias in ICD patients already after 2 h, although future studies including larger numbers of events are required to confirm these findings. Representative geographical exposure classification seems important in studies of these effects.

Keywords
  • Implantable cardioverter defibrillator
  • Air pollution
  • Tachycardia
  • Ventricular fibrillation
  • Heart arrest
  • Particulate matter

Introduction

Air pollution has been associated with short-term cardiovascular health effects in a number of studies.15 The autonomic regulation of the heart appears to be affected by air pollution causing decreased heart rate variability,610 a known risk factor for cardiovascular morbidity, mortality, and arrhythmic complications.11,12 Studies investigating air pollution effects on endpoints related to ventricular arrhythmias have shown mixed results.1315

Patients with a high risk for life-threatening ventricular arrhythmias receive implantable cardioverter defibrillators (ICDs), which can sense ventricular arrhythmias and administer anti-tachycardia pacing (ATP) and/or shock therapy. Studies conducted in North America have used such patients to study the relationship between ventricular arrhythmias and changes in air pollution.1622 Various air pollutants have been studied, and different designs have been used. Studies using case-crossover designs1618 control more efficiently for co-morbidities and time trends. Studies using moving averages of air pollution data16,17 as opposed to calendar day averages have consistently shown associations, emphasizing the importance of correct allocation of exposure prior to the studied event and that the important exposure period may be <24 h. Two of the previous studies have examined the effect of exposure periods shorter than 24 h, namely 3 and 6 h in Boston17 and 6 and 12 h in St Louis.16 None of the previous ICD studies have investigated time windows shorter than 3 h, although short-term exposure (1–4 h) to traffic, ambient air pollution, or experimentally controlled air pollution has been associated with ischaemic changes,2325 vasoconstriction,26 and changes in autonomic control of the heart.6,10,27 Importantly, none of the earlier studies have assessed the difference in effect stratified by distance from the air pollution monitor.

We conducted a case-crossover study in a European setting using highly representative temporal and geographical allocation of air pollution exposure, hypothesizing a possible rapid effect on arrhythmias.

Methods

Study sample and protocol

Patients—at Sahlgrenska University Hospital, Gothenburg, Sweden; Karolinska University Hospital, Stockholm, Sweden; and Stockholm South General Hospital, Stockholm, Sweden—who previously received ICDs or were implanted with ICDs during the course of the study period were recruited, provided that they were deemed mentally fit to cooperate. All included patients except one had implants due to earlier episodes of ventricular tachyarrhythmias. After the patients had been informed of the study and had signed a consent form, data were extracted from the records of each patient concerning pre-implantation diagnosis, type of arrhythmia leading to implantation of the device, ejection fraction, type of ICD, medication, and co-morbidities. The study was approved by the Ethics Committee of Karolinska Institutet.

We focused on symptomatic discharges to be able to collect specific information concerning the context of the event. The patients were therefore asked to contact the clinic within 3 days after sensing arrhythmia being treated by their ICD. Information from the ICD concerning time and date of arrhythmia, type of arrhythmia, therapy administered, and any change in programming was downloaded and documented. We included all ventricular tachyarrhythmias leading to shock therapy or symptomatic ATP therapy, excluding those occurring in rural areas (outside built-up areas more than 60 km from the monitoring station). In addition, ventricular tachyarrhythmias occurring in clusters after an initial event were excluded because of practical difficulties of distinguishing one clinical event from another with overlapping symptomatic and activity data for these events. We hypothesized that these events would not be independent of each other and perhaps reflect a different association with air pollution as a trigger. Consequently, a 7-day blanking period after each ventricular event was chosen as an adequate trade-off between practical considerations and securing independence of events. All intracardiac electrograms were reviewed by an electrophysiologist, and only ventricular tachycardias and ventricular fibrillations were accepted. Patients were interviewed concerning symptoms and possible triggering activities for both the 2 and 24 h periods preceding the ICD discharge as well as the location indoors or outdoors and address at the time of event in order to calculate the distance from the monitor.

Data were collected retrospectively for patients with verifiable time, date, and diagnosis for the specific documented ventricular arrhythmia qualifying the patient for ICD implantation, from here on referred to as the index arrhythmia. Documentation was collected through electrocardiograms, external defibrillator printouts, and patient charts. Given the possible imprecision of starting time for these events compared with those occurring after ICD implantation, only index arrhythmias with complete documentation were included. A total of 211 patients with ICDs were enrolled in the study: 99 from Gothenburg (April 2003 to December 2006) and 112 from Stockholm (August 2001 to December 2006) with an average observational time of 33 months (range 2–69). Prior to adding air pollution data to the data set, from 266 logged events, a total of 126 events were not included as they were non-ventricular, occurred in rural areas, were self-terminating events, or were events occurring in the same individual within 7 days from previous ventricular arrhythmia (Figure 1). In total, 88 patients had 140 symptomatic ventricular events fulfilling inclusion criteria, of which 114 were ventricular arrhythmias downloaded from the ICDs and 26 were index arrhythmias retrospectively collected from hospital records (Figure 1). Events downloaded from the ICDs and index arrhythmias were analysed separately.

Figure 1

Tree diagram describing inclusion of events; 266 events were logged from 211 individuals recruited from a total of four cardiology clinics in Gothenburg and Stockholm.

Air pollution measurements

Air pollution and meteorological data were obtained from one fixed centrally located roof-top monitor in each city reflecting urban background levels (see Supplementary material online, Figures S1 and 2 maps of centres). In Stockholm, NO2 values were available from two monitoring sites, and the mean was used. In the construction of moving averages, complete air pollution data were required for the 2 h exposure analysis, and a maximum of 25% of air pollution data was deemed an acceptable level of missing data for 24 h analysis. Missing values in calculating the mean of NO2 were imputed, according to the method described by Berglind et al.28 The median distance from the indicated positions at the time of event to the monitors was 15 km in Gothenburg and 10 km in Stockholm [for 16 events (14%) lacking information of location, we used their residential addresses as proxies]. During the entire study period, hourly means were provided for PM10 and NO2 for both Gothenburg and Stockholm. Data for particle matter with <2.5 µm in diameter (PM2.5) were available only for Stockholm. In addition, weather data including temperature, relative humidity, and barometric pressure were obtained. Pearson's correlation coefficients were calculated between the different air pollution parameters, showing an overall correlation between PM10 and NO2 of 0.36 for the 2 h average and of 0.29 for the 24 h average.

Statistical analyses

The association between ventricular arrhythmias and air pollution exposure was analysed using a case-crossover design.29 In this design, the time period immediately preceding the time of the event is considered the case period, and other time periods when the event did not occur are considered control periods for that subject. In this manner, each case serves as its own control, and covariates that are constant within a subject are adjusted for by design. The association is then analysed with a conditional logistic regression model.

The end of the case period was defined as the starting time of a confirmed ventricular arrhythmia rounded back to the nearest preceding hour. Control periods were matched to the case period on time of day, day of week, calendar month, and year within each subject.30 Air pollution levels and meteorological data were averaged in 2 and 24 h windows preceding the time of arrhythmia for the case period and the corresponding time for the control periods. The conditional logistic regression model included a linear term for air pollution and penalized splines for temperature, relative humidity, and barometric pressure for the same averaging times as the air pollution parameters. The associations are expressed as odds ratios (ORs) with 95 percent confidence intervals (CIs) for an interquartile range (IQR) increase in mean concentration for each pollutant and averaging time.

In addition to the main analyses, we used the event- and subject-specific data for a series of interaction analyses, exploring possible effect modification. Covariates that were assessed for effect modification were the following predictors of cardiovascular events: ischaemic heart disease, ejection fraction, diabetes, use of beta-blockers, age, and body mass index. We also included the number of arrhythmias as a covariate, which has been previously reported to show associations,20 and factors affecting exposure or misclassification such as indoor or outdoor location at the time of arrhythmia and distance from event to air pollution monitor. For time-variant factors such as location indoors/outdoors and distance from monitor, we had access to observations only for the case period. Interaction analyses were performed by including an interaction term between air pollution and the potential effect modifier in the conditional logistic regression model.

All data management and summary statistics were performed using Stata version 9. The conditional regression analysis was performed using S-plus version 7.

Results

Seventy-three of 211 patients included had 114 ventricular arrhythmias. Primary shock therapy was administered for a total of 58 events, whereas primary ATP was administered for 17 events. Both shock and ATP were administered for 39 events. The most common arrhythmia in our study was ventricular tachycardia (76 percent). Fifty-seven of the ventricular events occurred at home (50 percent) and 5 at work (4 percent). Pre-existing ischaemic heart disease was the most common cardiovascular condition of the study patients (Table 1). Most patients had dual-chamber ICDs, and approximately half had experienced discharges prior to the study period. Patients had a mix of anti-arrhythmic medication shown in Table 1. Gothenburg generally showed higher air pollution levels than Stockholm, particularly for NO2 and 2 h consecutive means of PM10 (Table 2).

View this table:
Table 1

Descriptive statistics of study patients (n = 88) in Stockholm and Gothenburg with ventricular tachyarrhythmic events

Patient characteristicsMean (SD)Range
Age at recruitment (years)62 (13)28–85
Body mass index26 (4)19–42
Categorical variablesnPercentage
Caucasian8698
Males7990
Ischaemic heart disease5664
Other (no ischaemic heart disease)2731
Lung disease89
Diabetes mellitus1214
Arrhythmia leading to ICDa implantation
 Ventricular tachycardia6978
 Ventricular fibrillation1922
Earlier atrial fibrillation78
Ejection fraction (n = 79)
 <30%2731
 30–50%3437
 >50%1820
Medication
 Beta-blockers6574
 Sotalol1416
 Amiodarone1719
 Digoxin1011
 Other anti-arrhythmic agents22
Number of patients with
 ICD events only6270
 ICD and index events1113
 Only index events1517
  • aImplantable cardioverter defibrillator.

View this table:
Table 2

Air pollution and meteorology in the two cities during the study period (Gothenburg: 2003–2006 and Stockholm: 2001–2006)

CityParameterCoverageMedianMinMaxIQR*
GothenburgPM10 (μg/m3)
2 h100%18.950.00203.7514.16
24 h100%19.922.1378.0111.49
NO2 (μg/m3)
2 h100%19.951.45337.0019.98
24 h100%22.203.87130.4714.98
Temperature (°C)
2 h100%8.90−15.4028.8012.51
24 h100%9.09−13.8024.4912.24
Relative humidity (%)
2 h100%80.719.6100.421.9
24 h100%78.532.299.816.4
Pressure (h Pa)
2 h100%1011966104414
24 h100%1011973104214
StockholmPM10 (μg/m3)
2 h94.0%14.620.33159.7911.59
24 h93.1%15.233.9690.509.59
PM2.5 (μg/m3)
2 h90.8%9.170.1599.256.69
24 h89.7%9.492.9747.075.27
NO2 (μg/m3)
2 h98.5%13.670.54102.3213.36
24 h97.7%15.662.4271.129.94
Temperature (°C)
2 h91.8%8.61−19.1730.2412.90
24 h91.6%8.59−16.1725.2013.08
Relative humidity (%)
2 h91.6%82.022.1100.020.7
24 h91.3%80.138.497.516.0
Pressure (h Pa)
2 h96.1%1008958104615
24 h95.9%1008963104515
  • Distribution of consecutive 2 and 24 h means.

  • * IQR, interquartile range.

Associations for 2 h moving averages

The risk of ventricular arrhythmias was associated with increased levels of air pollution in the preceding 2 h exposure period. The strongest association was seen for PM10. NO2 showed a positive association, although weaker in effect. PM2.5, available only to Stockholm, also showed associations similar in magnitude to PM10 arrhythmias, although the 95% CI included unity (Table 3). Sensitivity analyses using 24 h temperature rather than 2 h temperature in the model as well as excluding barometric pressure gave similar results (data not shown).

View this table:
Table 3

Odds ratios of ventricular tachyarrhythmia for an interquartile range increase in air pollutants in patients with implantable cardioverter defibrillators

Pollutant (µg/m3)Moving average (h)IQRaNo. of subjectsNo. of eventsOR95% CI
PM10213.2651011.31(1.00–1.72)
2410.3671061.24(0.87–1.76)
PM2.527.533491.23(0.84–1.80)
245.235531.28(0.90–1.84)
NO2215.3691091.09(0.84–1.42)
2411.0701101.07(0.81–1.42)
  • OR, odds ratio; CI, confidence interval.

  • aInterquartile range for joint distribution of pollutants in Stockholm and Gothenburg used in case-crossover analyses except for PM2.5, where only data from Stockholm are presented.

Ventricular arrhythmias showed a stronger association with PM10 in Gothenburg than in Stockholm (Figure 2A). Ventricular arrhythmias occurring within the median distance from the air pollution monitor [15 km in Gothenburg (range 2–172) and 10 km in Stockholm (range 1–65)] demonstrated an OR of 1.76 (95% CI 1.18–2.61) for an IQR increase in PM10, whereas events further away showed no association (Figure 2A). No such patterns were seen for NO2 exposure. PM10 tended to be stronger associated with arrhythmias in patients spending the preceding 2 h outdoors. During the study period, ventricular arrhythmias in patients with more than two interventions tended to show a stronger association with PM10 than in those with only one or two interventions.

Figure 2

Odds ratios of ventricular arrhythmia for an interquartile range increase in 2 h (A) and 24 h (B) moving averages of air pollutants in different subgroups (P-values for interaction). For variables distance from monitor and location at arrhythmia observation data available only for case periods. Number of observations for each subgroup with complete data on analysed covariate and air pollution indicated by n.

NO2 demonstrated associations in the same direction as for PM10 but generally with weaker estimates and less clear patterns across groups, with the exception of the difference in effect of being outdoors or indoors at the time of event that showed a P-value of 0.04 for interaction.

Associations for 24 h moving averages

Positive associations were seen between increases in PM10, NO2, and PM2.5 for 24 h means and the onset of ventricular arrhythmias but all included unity (Table 3). The point estimates for PM10 and PM2.5 were similar to those for the 2 h mean.

Compared with the 2 h moving average analysis, the 24 h moving average tended to show less difference in effects between study centres (Figure 2B). Events closest to the air pollution monitor were strongly associated with increased PM10, and events farther away showed no associations. For the 24 h moving averages, no consistent pattern was seen for patients with more frequent arrhythmias during the study.

Increased PM10 and NO2 were strongly associated with ventricular index arrhythmias, which occurred prior to the implantation of the device. For NO2, there was a difference in effect between the events collected from the ICDs and the index events.

Age, ejection fraction, pre-existing ischaemic heart disease, and anti-arrhythmic medication did not show any association. Results for sex interaction analyses are not presented, as the subgroups had less than 10 events.

Discussion

We found an association between the triggering of ventricular arrhythmias in patients with ICDs and preceding 2 h increases of air pollution. This is the shortest time frame of air pollution exposure studied in relation to ventricular arrhythmias. The associations appeared weaker for the preceding 24 h moving averages. In contrast, the associations were stronger for arrhythmias occurring closer to the air pollution monitor in each city. The findings for the index arrhythmias occurring before implantation of the ICD further support the main results. Only one association (2 h exposure to PM10) exceeded unity, but the study sample was small and the results were consistent in direction. Taken together, these results seem to confirm our hypothesized rapid effect of air pollution on the onset of ventricular arrhythmias.

This is the first European study investigating the association between air pollution and ventricular arrhythmias in patients with ICDs. Results from some earlier ICD studies reflect a similar pattern,16,17,20,22 whereas others found no association18,19,21 (see Supplementary material online, Table S1). The most important difference between our study and earlier studies is air pollution exposure classification. The ideal exposure measurement would be individual breathing zone samples taken at short intervals. For practical purposes, urban background monitors are used as a proxy for individual exposure and reflect an average exposure in a limited area. Changes in pollution levels need some time to spread across the air shed, and therefore instant changes in local air pollution will not be reflected immediately in the monitor readings. In addition, time is required for the exposure to lead to ventricular arrhythmia. We based exposure assessment on hourly means and rounded back to the nearest preceding full hour. Therefore, the end of the exposure window was on average 30 min earlier than the actual starting time of the event, but no exposure data were collected after the event. In contrast, calendar day averages would, however, imply on average collection of 12 h of exposure data before and after the event. Temporal exposure classification in Boston using 24 h moving averages compared with calendar day averages (midnight-to-midnight) gave substantially higher ORs.17 Furthermore, in our study, distance to the air pollution monitor was a factor influencing association, suggesting that geographically representative exposure classification is also of importance. For PM10 and NO2, local sources are important in Sweden,31 so that vicinity to air pollution monitor increases both the accuracy and precision when interpreting fixed monitor values as proxies for personal exposure. It is possible that the lack of positive association reported in a large study21 was due to the use of 24 h calendar day averages or daily maximum concentrations of pollutants in addition to a large catchment area of 20 counties in Atlanta, not accounting fully for local sources.

In Boston, patients experiencing multiple arrhythmias were particularly sensitive to increased air pollution levels, whereas in Vancouver and Atlanta, this was not seen. Despite exclusion of events occurring within 7 days in the same individual in our study, we found a tendency of increased susceptibility in individuals characterized by repeated events, supporting the results from Boston. However, our sample size is too small to address these questions appropriately.

The association between higher levels of PM10 and ventricular arrhythmias appeared stronger in patients being outdoors at the time of event. Air pollution collected from urban background monitors should better reflect the true individual exposure outdoors than indoors. Outdoor exposure may be related to a higher physical activity, which could act as an independent trigger32 or increase the air pollution exposure by increased respiratory rate.

Mechanisms

The mechanisms of sudden cardiac death, precipitated by ventricular arrhythmias, have been described as a consequence of processes involving ischaemia, previously damaged myocardium, and fluctuating myocardial vulnerability possibly due to changes in ion channel function, inflammation, and autonomic reflexes.33 Although the pathophysiological mechanism by which air pollution exerts its effects on cardiovascular disease is still unclear, it has been proposed to act within several hours to days through increased inflammatory response and oxidative stress and within a few hours through changes in autonomic tone.1 We saw associations between air pollution and ventricular arrhythmias already within 2 h. Rapid effects of air pollution have also been reported on ischaemic heart disease in an epidemiological study23 and in humans experimentally exposed to diesel exhaust.25 Spending time in traffic during 1 h increased the risk of myocardial infarction,24 and other studies have demonstrated associations between short-term air pollution increases and decreased heart rate variability,6,10,27 thus indicating a direct effect on cardiac autonomic control. ICD patients with previous myocardial infarction have a substrate for development of re-entry ventricular arrhythmias. The effects of air pollution may raise the vulnerability to re-entry by altering autonomic tone or increasing ischaemia. Using heart rate variability as a measure of autonomic tone to predict ventricular arrhythmias in ICD patients irrespective of air pollution has yielded mixed results, perhaps depending on the method of analysis.34,35 Hence, although the rapid effects seen in our study suggest an effect mediated through disturbances in the autonomic regulation of the heart, an ischaemic response remains a possibility25 or possibly a combination of pathways.

Strengths and limitations

This is the first study of its kind that has attempted to assess how distance from the air pollution monitor can influence the observed associations and, despite small sample size, has demonstrated associations. Earlier studies allocating air pollution exposure to individuals in a large geographical area may suffer from substantial imprecision in the exposure assessment and thus have less power than the number of events suggests. In addition, we used moving hourly averages to improve temporal exposure classification. Individual factors and co-morbidities were efficiently controlled for by using the case-crossover approach, and we verified our classification of outcome prospectively and independently of exposure. Our study had the additional advantage of collecting specific data concerning activity and location within a few days after the events occurred.

Some potential limitations should also be recognized. The total number of analysed events was rather small and contributes to imprecision of our main results, in particular, in the explorative interaction analyses. The study was based on local routines in four different clinical centres involving follow-up visits 3–6 months periodically, and although desirable, it was not possible to retrieve detailed information on events occurring in clusters. Therefore, by design, we included a 7-day blanking period. The exclusion of arrhythmias occurring within 7 days in the same individual, however, may have unnecessarily decreased the number of observations, considering the evidence that these events may be especially prone to the effects of air pollution.17,22 The intracardiac electrograms and markers from the ICD memory may lead to some misdiagnosis, but this would be independent of exposure classification and therefore lead to underestimated associations. Furthermore, using urban background monitors for air pollution always presents the possibility of non-differential misclassification of individual exposure also attenuating associations.

We considered confounding by meteorological factors such as temperature, air pressure, and relative humidity and adjusted for them in regression analyses. Time-varying factors such as traffic congestion and stress may conceivably be associated with both air pollution and ventricular arrhythmias, and to minimize the potential for confounding, we chose control periods using the same hours of day and day of week as the case period.

Our findings suggested different effects of PM10 across study centres, in which Stockholm showed no clear association between increased PM10 and ventricular arrhythmias. The reason for these unexpected results is unclear. The use of studded winter tires is more common in Stockholm and road wear constitutes at times up to 90%31 of the local contribution to PM10. Also, the suggested positive association of ventricular arrhythmias with fine particles (PM2.5) in Stockholm may indicate that different particulate pollution mixtures influence the observations and there are indications of differences in the composition of PM2.5 in the two cities.36 It may be suggested that single point urban background measurements better describe population exposure in Gothenburg than in Stockholm due to the greater contribution of long-range transport particles in southern Sweden.37

Conclusions

Ventricular arrhythmias in ICD patients appear to be associated with moderate increases in air pollution already within the 2 preceding hours, although future studies including larger numbers of events are required to confirm these findings. When studying the association between ventricular arrhythmias and air pollution, representative temporal and geographical exposure classification seems to be important.

Funding

This study is funded by Swedish Environmental Protection Agency.

Conflict of interest: none declared.

Acknowledgements

We are greatly indebted to Monika Dahlin, Margareta Dalman, Stina Gustavsson, Karin Hellkvist, Boel Hillegren, Eva-Marie Jansson, Ann-Charlotte Månsson, Katarina Ringdahl, Mats Rosenlund, and Eija Tofangchiha-Vitander for their valuable help.

References

View Abstract