Article Text


Original article
Associations between arrhythmia episodes and temporally and spatially resolved black carbon and particulate matter in elderly patients
  1. Antonella Zanobetti1,
  2. Brent A Coull2,
  3. Alexandros Gryparis3,
  4. Itai Kloog1,4,
  5. David Sparrow5,
  6. Pantel S Vokonas5,
  7. Robert O Wright1,6,
  8. Diane R Gold1,6,
  9. Joel Schwartz1
  1. 1Environmental Epidemiology and Risk Program, Harvard School of Public Health, Boston, Massachusetts, USA
  2. 2Department of Biostatistics, Harvard School of Public Health, Boston, Massachusetts, USA
  3. 3Department of Hygiene, Epidemiology and Medical Statistics, University of Athens Medical School, Athens, Greece
  4. 4The Department of Geography and Environmental Development, Ben-Gurion University of the Negev, Beer Sheva, Israel
  5. 5Department of Medicine, VA Normative Aging Study, Veterans Affairs Boston Healthcare System, Boston University School of Medicine, Boston, Massachusetts, USA
  6. 6Channing Laboratory, Department of Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts, USA
  1. Correspondence to Dr Antonella Zanobetti, Department of Environmental Health, Exposure Epidemiology and Risk Program, Harvard School of Public Health, 401 Park Drive, Landmark Center, Suite 415, P.O. Box 15698, Boston, MA 02215, USA; azanobet{at}


Objectives Ambient air pollution has been associated with sudden deaths, some of which are likely due to ventricular arrhythmias. Defibrillator discharge studies have examined the association of air pollution with arrhythmias in sensitive populations. No studies have assessed this association using residence-specific estimates of air pollution exposure.

Methods In the Normative Aging Study, we investigated the association between temporally resolved and spatially resolved black carbon (BC) and PM2.5 and arrhythmia episodes (bigeminy, trigeminy or couplets episodes) measured as ventricular ectopy (VE) by 4 min ECG monitoring in repeated measures of 701 subjects, during the years 2000–2010. We used a binomial distribution (having or not a VE episode) in a mixed effect model with a random intercept for subject, controlling for seasonality, temperature, day of the week, medication use, smoking, having diabetes, body mass index and age. We also examined whether these associations were modified by genotype or phenotype.

Results We found significant increases in VE with both pollutants and lags; for the estimated concentration averaged over the 3 days prior to the health assessment, we found increases in the odds of having VE with an OR of 1.52 (95% CI 1.19 to 1.94) for an IQR (0.30 μg/m3) increase in BC and an OR of 1.39 (95% CI 1.12 to 1.71) for an IQR (5.63 μg/m3) increase in PM2.5. We also found higher effects in subjects with the glutathione S-transferase theta-1 and glutathione S-transferase mu-1 variants and in obese (p<0.05).

Conclusions Increased levels of short-term traffic-related pollutants may increase the risk of ventricular arrhythmia in elderly subjects.

Statistics from

What this paper adds

  • Defibrillator discharge studies have examined the association of air pollution with arrhythmias in sensitive populations.

  • None of these existing studies used exposure estimates that are specific to the location of each study participant.

  • This is the first study to examine the association between arrhythmias and residence-specific estimates of air pollution exposure.

  • We found a clear association between arrhythmias and increased levels of traffic related pollutants and these associations are higher compared with previously published studies.

  • Moreover, individuals with gene variants related to oxidative stress may be at higher risk from air pollution.


Air pollution has been consistently associated with cardiovascular morbidity, mortality1 and premature death,2 ,3 but the underlying mechanisms are not well understood.

In relation to more specific cardiac events, there is evidence that air pollution may contribute through a wide range of pathways, and several biological mechanisms by which air pollution can elicit cardiovascular morbidity and mortality have been identified, including oxidative stress (OS),4 autonomic dysfunction5–9 and systemic inflammation.10–12

Several studies have observed associations between air pollution and cardiac arrhythmias, based on data from implantable cardioverter-defibrillator (ICDs). These devices, which restore a normal heart rhythm, allow the continuous monitoring of patients and the documentation of type and time of ventricular arrhythmias. These studies have found the associations of air pollution with arrhythmias13–16 in sensitive populations, although results have been mixed. However, less is known for the general population. Further, none of these existing studies have used exposure estimates that are specific to the location of each study participant. Ljungman et al17 found that patients with implantable cardioverter defibrillators also showed evidence of rapid effects of air pollution on the risk of life-threatening ventricular arrhythmias. Among cardiovascular health indicators, irregularities in myocardial repolarisation may be especially important because they can lead to the development of cardiac arrhythmias.

Another study18 examined the effects of air pollutants on repeated measurements of QT interval (QTc), an ECG marker of ventricular repolarisation, in 580 men from the Veterans Affairs (VA) Normative Aging Study (NAS). The authors found an association between QTc and black carbon (BC) measurements from the Harvard supersite. Similarly, in a repeated-measures study of patients with coronary artery disease,19 we found associations between ambient and indoor BC and a report of being in traffic with risk of increase in T-Wave Alternans (TWA), a marker of cardiac electrical instability. In Boston, traffic-related pollutants were also related to the promotion of ST segment depression among elderly subjects.20

These findings suggest a possible biological pathway linking acute effects of air pollution on increased risk of ventricular repolarisation, cardiovascular arrhythmias and cardiac death.

Short-term exposure studies have mostly used stationary monitors to estimate exposure; however, specific components of traffic-related air pollution vary substantially within cities, and traffic variables may contribute to this variation.21–23 This suggests that spatiotemporal prediction of BC levels within the Boston area could substantially improve exposure assessment. An important tool for studying within-city variation in air pollution is the development of geographically based exposure models; however, previous studies have been limited by the lack of high-resolution daily exposure data.

BC is a traffic-related particle and a common surrogate for traffic particles in general, weighted towards diesel particles. We have developed a spatiotemporal land-use regression model for traffic particles based on BC in the greater Boston metropolitan area.24 Predictions from this model have been merged to the geocoded address of each subject of the NAS cohort and have been used to study the effect of traffic pollution with markers of inflammatory and endothelial response,25 blood pressure26 and other outcomes. This model for BC has been now updated and revised to include data from 125 monitoring stations recording BC levels at some point between January 1999 and August 2011.

Growing evidence also suggests that traffic-related components of PM pollution contribute significantly to particle-related cardiovascular effects. For example, a recent chamber study examined the short-term effects of PM2.5 on blood pressure and found that effects were much stronger for the samples collected from a high-traffic area.27

We have recently presented a new method28 of assessing temporally resolved and spatially resolved PM2.5 exposures for epidemiological studies. This approach is an extension of existing land-use models to include satellite-based physical aerosol optical depth (AOD) measurements.24 ,29 We have applied these predictions to study the association between PM2.5 exposure and hospital admissions among elderly across New England, as well as with geocoded birth weight in Massachusetts.30 ,31

In this study, we hypothesised that increases in estimated BC concentrations and estimated PM2.5 concentrations may be associated with increases in ventricular arrhythmias.

We therefore investigated the association between BC and PM2.5 and arrhythmia episodes measured as ventricular ectopy (VE) among participants of the VA NAS residing in the greater Boston area. We also examined whether participant characteristics (obesity, diabetes, statin use, genetic susceptibility) might modify this association.


Our study population consisted of a cohort of elderly men recruited to the NAS, an ongoing longitudinal study of aging established by the Veterans Administration in 1963 from community-dwelling men from the greater Boston area aged 21–80 years at the time of entry, who were free of any known chronic medical conditions and were asked to return for onsite physical examinations and questionnaires every 3–5 years. Study participants provided written informed consent, and the study protocol was approved by the Institutional Review Boards of all participating institutions.32

Eligibility for this study required continued participation as of the time when heart rate variability measurements began in the year 2000. Drop out has been less than 1% per year in the cohort and is predominantly when subjects move out of the study area or mortality. The study population is a self-selected group of people who continued to participate in this ongoing study. Since 1963, every 3–5 years, participants have undergone routine physical examinations, laboratory tests, collection of medical history information and completion of questionnaires on smoking history, educational level, food intake and other factors that may influence health.

Diabetes was defined as either a physician diagnosis of diabetes, use of any diabetes medication or fasting glucose >126 mg/dL, and obesity was defined as a body mass index (BMI) of at least 30. Self-reported data on diabetes status and statin use were updated at each study visit and confirmed by a physician interview. In addition, BMI and obesity were updated based on height and weight measurements at each visit. Thus, these data reflect changes in disease status and medication use over time.

Beginning in November 2000, ECG measurements were also obtained during each participant's regularly scheduled visit.

ECG measurement and analysis

After a 5 min rest, with the subject seated, the ECG was recorded for 5–10 min with a two-channel (five lead) ECG monitor (Trillium 3000; Forest Medical, Inc, East Syracuse, New York, USA) using a sampling rate of 256 Hz per channel. A detailed description of the protocol is provided elsewhere.6 ,33 Briefly, with a PC-based software, beats were automatically labelled and assigned tentative annotations, and then an experienced scanner reviewed the results to correct for any mislabelled beats or artefacts, and the best 4-consecutive-minute intervals were used for the analysis. The normal or supraventricular beats were recorded; 153 measurements were excluded because of ECG recordings that have recording time <3.5 min, or insufficient T-wave amplitude.

In this study, we examined arrhythmia episodes measured as VE. We defined VE as having bigeminy, trigeminy or couplets episodes. We then created an indicator variable for having (=1) or not having (=0) episodes.

We examined 1448 observations of subjects who had either one (n=701), two (n=451), three (n=217) or four (n=79) ECG measurements.

Air pollution and meteorology

PM2.5 (particulate air matter with aerodynamic diameter less than 2.5 mm) and BC concentrations were continuously measured at the stationary ambient monitoring site at the Harvard University Countway Library, which is located <1 km from the clinical laboratory where subjects were examined.

Hourly outdoor BC concentrations were also obtained from two sites operated by the Massachusetts Department of Environmental Protection.

We obtained measurements from specific monitoring campaigns from two sources: beginning in 1999, hourly outdoor and indoor BC concentrations were measured inside and outside of 30 residential homes using aethalometers over 48 h intervals as part of a NIEHS funded study of air pollution and heart rate variability (APAHRV) conducted at the Harvard School of Public Health. Secondly, outdoor 24 h averages of elemental carbon (EC) concentration were obtained from an Environmental Protection Agency (EPA) funded multipollutant exposures study of sensitive individuals during winter and summer of 2000 for 7-day periods in 23 locations. A third source was rotating monitoring sites located at the homes of 50 people, sampling for 1 month each in the hot and cold seasons, with addresses chosen to fill in locations undersampled or not sampled in the previous monitoring campaigns.

Finally, reflectance data were obtained from PM2.5 filters from six monitors located geographically around the Boston area (one in the city, one slightly north of the city, two north and west of the city, and two south and east of the city). These data were obtained daily.

Hourly meteorological measurements, such as mean temperature (°F) and dew point temperature (°C), were obtained from the National Weather Service First Order Station at Logan Airport in East Boston.

BC exposure prediction

We estimated BC exposures using a validated spatiotemporal land-use regression model that provides daily estimates of BC concentrations throughout the greater Boston, Massachusetts, area; details of this model have been published previously.24 We used an updated and revised version of the aforementioned model, with data from 125 monitoring stations that collected data during the period of January 1999 to August 2011, with a total of 14 238 daily averages from 125 monitoring stations.

We used measurements from continuous ambient monitors as well as specific monitoring campaigns as described above. Briefly, daily average BC estimates from 125 monitoring sites were used to develop a BC prediction model, where the majority of sites measured BC continuously using aethalometers and other sites collected particles on a filter over 24 h and measured EC using reflectance analysis. Predictors in the final model included measures of land use for each address (cumulative traffic density within 100 m, land cover within 1 km radius and distance to coast), geographic information system location (latitude, longitude), daily meteorological factors (apparent temperature, wind speed and height of the planetary boundary layer) and other characteristics (day of week, day of season), as well as interaction terms of land-use measures and daily meteorological factors (eg, interaction between the height of the planetary boundary layer and an indicator of the values above the median value of land cover within 1 km radius). The Boston central site monitor was also included as a predictor to reflect average pollutant concentrations over the entire region on each day. Predicted daily concentrations showed a >3-fold range of variation in exposure across measurement sites (adjusted R2=0.83). Out-of-sample cross-validation at 32 monitoring sites showed an average correlation of 0.73 between predicted and observed daily BC levels. As opposed to evaluating the predictive ability of the model for all monitors, which included both residential and urban, non-residential locations, we selected as our test set a subset of monitors that were sited in residential locations. This validation approach provides a better estimate of the predictive performance of the exposure model at study subject residences as opposed to a more general metric that encompasses locations in which no subjects are located.

This predictive performance of the model at any given held out (out-of-sample) location is being compared with the predictive performance when one uses the central site monitor as the value for that given location. The direct comparison of the correlations between the predictions from these two exposure measures and the actual monitored value at any location is directly comparable since the central site reading is effectively an out-of-sample prediction because the data from that monitor are not being used when one uses the central site reading as a surrogate. We used the 24 h average estimates of BC exposure at the geocoded residence of each participant as a surrogate for individual exposure to traffic-related air pollution and average these predicted values to obtain 2-day, 3-day and 4-day moving averages of exposure.

PM2.5 exposure prediction

We generated daily predictions of PM2.5 concentration levels across New England for 2000–2008 at a 10×10 km spatial resolution using satellite-derived AOD measurements. Full details of the modelling framework have been previously published.28 ,30 In brief, we used day-specific calibrations of AOD data using ground PM2.5 measurements from 78 monitoring sites in the EPA and Interagency Monitoring of Protected Visual Environments monitoring network. The model included also land-use regression and meteorological variables (temperature, wind speed, visibility, elevation, distance to major roads, percent of open space, point emissions and area emissions). To estimate PM2.5 concentrations in each grid cell on each day, we started by calibrating the AOD-PM2.5 relationship for each day using grid cells with both monitors and AOD values using mixed models with random slopes for day and nested regions. The first-stage calibrations resulted in high out-of-sample R2 (mean out-of-sample R2=0.85). We then used a second model to address days when AOD measures are not available (due to cloud coverage, snow, etc.), fitting a model with a smooth function of latitude and longitude and a random intercept for each cell that takes advantage of the association of grid cells AOD values with PM2.5 monitoring located elsewhere, and the association with available AOD values in neighbouring grid cells, which provided a mean out-of-sample R2=0.81. PM2.5 exposure data were generated by our prediction models at a 10×10 km spatial resolution. These data were matched with the NAS subject's addresses using ArcGIS and SAS based on spatial location and date. For both BC and PM2.5, we computed averages of two, three and four previous days as our exposure metrics.


Participants’ blood was collected at each visit. Multiplex PCR assays were designed using the Sequenom Spectro DESIGNER software. Assays were genotyped using the Sequenom MassArray matrix-assisted laser desorption-ionization time-of-flight mass spectrometer with semiautomated primer design and implementation of the very short extension method. Assays that failed to multiplex were genotyped using the TaqMan 5′ exonuclease (ABI PRISM V.7900 Sequence Detector).

To reduce multiple comparisons by investigations of multiple gene deletions and polymorphisms, and based on previous literature, we selected only two gene deletions (glutathione S-transferase mu-1 (GSTM1) and glutathione S-transferase theta-1 (GSTT1)) that were related to the genetic susceptibility of a participant to OS. The gene deletions were categorised as 1=no deletion and 0=deletion.

Statistical analyses

We used mixed effects models to examine the associations between ambient pollution and VE. The VE episodes were dichotomised as present or absent, and we therefore applied mixed logistic regression models to account for correlation among measurements on the same subject across different medical visits. This is a standard approach for analysing longitudinal data with repeated measures on the same subject.34 We adjusted for a priori chosen known or plausible confounders, including specific personal and temporal characteristics. We included in all models random subject-specific intercepts and fixed effects for BMI (as a continuous variable), age, cumulative cigarette smoking in pack-years, use of medication such as angiotensin-converting enzyme inhibitors, β-blockers, calcium blockers, statins and having diabetes (yes/no), and an indicator variable for alcohol consumption defined as two or more drinks per day. We adjusted for socioeconomic status at the individual level with average income and years of education (defined as less than 12 years, 12–16 years, over 16 years), and at the neighbourhood level with poverty level for each address as measured by per cent below poverty level of each census block group in the 1999 census, and per cent of population over 25 years of age without a high school diploma as measured by the 1999 census.

We also adjusted for the 24 h means of temperature for seasonality using sine and cosine terms and a linear variable for year to capture long-term changes in both outcome and exposure.

Healthier men are more likely to come back to subsequent visits; therefore, we used inverse probability weighting (IPW) to correct for a potential survival bias.35 The probability at first visit was 1; we then calculated the probability of having a second, third of subsequent visit using logistic regressions given all relevant factors at the previous visit: age, BMI, smoking status and pack-years, hypertension, cholesterol and diabetes. We then used the inverse of the predicted probabilities as the weights in the models.

In the models used to examine effect modification, we included an interaction term that allowed associations between the pollutants and the outcomes to vary among subgroups. Diabetes, obesity and statin use were treated as time-varying covariates, where the status was updated at each visit to reflect changes since the last visit. We also examined effect modification by gene variants GSTM1 and GSTT1. Interaction terms were added to the model one at a time.

Air pollution was examined for the previous 24 h, and for the moving averages of two, three and four previous days.

As sensitivity analysis we included in the model ozone that activates C-fibres and could thereby influence ventricular rhythm; we also examined whether the results changed without using the IPW.

The results are reported as ORs per an IQR increase in the pollutants. The analyses were performed with statistical software R V.2.15.1.


In table 1, we present characteristics of the patients, health and environmental variables among the NAS population during 2000–2010. The age range of the study subjects was 57–100 years, and 189 (26.9%) of them were obese, as defined by a BMI ≥30 at the first visit. In total, 701 subjects had one clinic visit, 451 had two visits, 217 had three visits and 79 had four visits, with a total of 1448 observations.

Table 1

Characteristics of the VA Normative Aging Study subjects at baseline and across all visits and distributions of the weather and outcome variables

Table 2 shows the results of the association between VE and air pollution.

Table 2

ORs and 95% CI of the association between ventricular arrhythmia episodes and air pollution

We found increases in the odds of having ventricular ectopic beats, with an OR of 1.32 (95% CI 1.06 to 1.65) for an IQR (0.39 μg/m3) increase in the same day of BC. The OR associated with an IQR (0.30 μg/m3) increase in the 3-day average of BC was 1.52 (95% CI 1.19 to 1.94).

The association between having ventricular ectopic beats and estimated PM2.5 was positive and significant.

We found increases in the odds of having ventricular ectopic beats with an OR of 1.26 (95% CI 1.05 to 1.52) for an IQR (6.89 μg/m3) increase in the same day of PM2.5, and an OR of 1.39 (95% CI 1.12 to 1.71) for an IQR (5.63 μg/m3) increase in the 3-day average of PM2.5.

We then examined effect modification by obesity, statin use, having diabetes and gene variants (figure 1). We did not find any evidence of modification by personal characteristics and BC, but we found significantly higher effects in obese in the association between estimated PM2.5 and having ventricular ectopic beats with an OR of 1.17 (95% CI 0.95 to 1.43) in non-obese and an OR of 1.80 (95% CI 1.24 to 2.63) in obese for an IQR (6.89 μg/m3) increase in the same day of PM2.5. Subjects having the deletion of GSTM1 and for subjects without the deletion of GSTT1 had also significantly higher effects (figure 1).

Figure 1

OR and 95% CI per an IQR increase in each pollutant. Effect modification by statin, diabetes, obesity and glutathione S-transferase mu-1 and glutathione S-transferase theta-1 genes. The symbol * indicates a significant interaction.

When we included ozone in the model, this was not significant; when included with each of the other pollutants, ozone was not significant and did not change the effects of PM2.5 or BC.

The results without adjusting for the IPW are consistent but less strong.


We found a significant association between having ventricular arrhythmias and both BC concentrations estimated at each participant's home, and temporally resolved and spatially resolved PM2.5 concentrations, in a cohort of elderly men. Our interactions with obesity (only for BC), statin and diabetes were not significant, indicating no effect modification by these individual characteristics, while we found that GSTM1 and GSTT1 significantly modify the association.

This study, therefore, provides evidence that the traffic-related pollution markers are associated with ventricular arrhythmias.

Our results further suggest that individuals with gene variants related to OS may be at higher risk from air pollution.

Our findings are consistent with previous studies based on data from ICDs, which have found the associations of air pollution with arrhythmias13–17 in sensitive populations. In this study in which we examined a more general population and focused on a more accurate exposure defined by estimates of spatially resolved BC and PM2.5, we found a stronger association with ventricular arrhythmias compared with the previously published studies.

Consistent with these findings, ambient traffic-related pollutants’ concentrations have been associated in several Boston area studies of intermediate cardiovascular indicators, including heart rate variability,5 ,7 ,8 ST-segment depression,20 ,36 QTc, an ECG marker of ventricular repolarisation18 and TWA.19

BC is a traffic-related particle and a common surrogate for traffic particles in general, weighted towards diesel particles. We have developed a spatiotemporal land-use regression model for traffic particles based on BC in the greater Boston metropolitan area.24 In the same NAS cohort of elderly men and using the same BC predictions at each home address, we have previously reported associations with several outcomes: markers of inflammatory and endothelial response25 and blood pressure.26

Our observed associations can be explained by plausible pathophysiological mechanisms. Toxicity of traffic particles may have a direct effect on the blood, cardiovascular system and lung receptors.37 ,38

Deposition in the airways and lung alveoli may trigger proinflammatory signalling via a reactive oxygen species (ROS)-dependent mechanism.39 ,40 For example, Ghelfi et al41 have shown that blocking the TRPV1 receptor in the lung prevents particle-induced disturbances in cardiac rhythm. Diesel particles have also been shown to increase OS in endothelial tissue,42 ,43 inducing the production of heme oxygenase-1, a rapid response part of the body's defence system against OS. The viability of cell cultures of microvascular endothelial cells was impaired by diesel particles with an accompanying large increase in induction of heme oxygenase-1.43

Moreover, traffic particles may also cross the pulmonary epithelium and may be able to reach the heart via the vasculature,44 ,45 where they may induce OS and proinflammatory changes in the vasculature and myocardial substrate.38 The generated proinflammatory cytokines and ROS may subsequently affect a variety of health measures, including autonomic cardiac control.38

We did not find effect modification by obesity, diabetes or statin use for BC and only by obesity for PM2.5; this is contrary of what has been previously found for other endpoints.6 ,7

Genetic polymorphisms of the GSTs are common, the GSTM1 gene is deleted in approximately half of the white population (the polymorphic ‘null’ genotype) and lack of the GSTM1 protein has been shown to modify the response to air pollutants; for example, GST deficiency was associated with significant HRV alterations in the general population.46

In the NAS cohort, GSTM1 deletion has been associated with the high-frequency component of heart rate variability,9 and larger effects of BC on soluble vascular cell adhesion molecule were seen in subjects who were GSTM1 null.10 Baja and coauthors found a stronger association between QTc and BC among participants who had higher OS gene scores. Therefore, this is in agreement with our results of a higher risk of having ventricular episodes due to BC in subject GSTM1-null. Ren et al47 found that the GSTT1 modified effects of BC on total plasma homocysteine, with a higher effect in subjects with the deletion, and no effect modification by GSTM1. A case–control study looking at the relationship between occupational exposure to vehicular exhaust and OS in traffic police found a significant increase in urinary 8- hydroxydeoxyguanosine (8-OHdG) levels in null GSTM1 (p<0.01) genotypes, but the genotype frequencies GSTT1 genes did not vary in both exposed and control groups.48 Similar results were found by Ren et al49 in the NAS population. On the other hand, survival of patients with malignant glioma was reduced in subjects with the GSTM1 deletion, but longer in subjects without the GSTT1 deletion.50 GSTs are the most important family of phase II isoenzymes known to detoxify a variety of electrophilic compounds, including carcinogens, and environmental toxins. Individuals with the deletion of GSTM1 or GSTT1 have been shown to reduce GST activity and thus may be unable to eliminate toxins as efficiently when they are exposed to oxidative pollutants.50 A recent review and meta-analysis also found that null genotypes of GSTM1/GSTT1 were associated with increased risk of diabetes.51 Differential findings among studies of effect modification by GSTM1 and GSTT1 variation might be attributable to several issues such as differences in exposure, outcome and population, measurement errors in exposure or phenotype, or chance. Further research is needed in order to understand the biological mechanisms.

One limitation of this study is that these results cannot be generalised to other populations without further research as our study population consisted entirely of elderly men, 97% of whom were Caucasian. Additionally, the study population is a self-selected group of people who continued to participate in an ongoing study for many years and may not be representative of all elderly men in the USA. In particular, there may be survivor bias if the subjects who continue to participate are healthier than other older people, which would bias effect estimates towards the null, so the true effect in the general population may be stronger.

Another limitation is that we do not have the exposure during travel as that would have required personal monitoring, which was not available for a large cohort with measurements over a decade.

In conclusion, these results add support for the hypothesis that traffic-related pollution alters autonomic control in a manner conducive to increased arrhythmia and that OS plays a role in this process.


View Abstract


  • Contributors AZparticipated in the conception and design, analysed and interpreted the data, drafted the article and approved the final version. BAC, DS, PSV, ROW, DRG and JS participated in the conception and design, revised the paper critically for important intellectual content and approved the final version. AG analysed and interpreted the pollution data, revised the paper and approved the final version. IK analysed and interpreted the pollution data creating the prediction models for PM2.5, revised the paper and approved the final version.

  • Funding This work was supported by the National Institute of Aging R21 AG040027-01 and the National Institute of Environmental Health Sciences P01 ES009825, ES-00002, by US Environmental Protection Agency RD-83241601 and RD 83479801 and by a VA Research Career Scientist award to DS. The Veterans Administration's Normative Aging Study is supported by the Cooperative Studies Program/Epidemiology Research and Information Centers of the US Department of Veterans Affairs and is a component of the Massachusetts Veterans Epidemiology Research and Information Center, Boston, Massachusetts, USA. This publication was made possible by USEPA grant RD-83241601 and RD 83479801. Its contents are solely the responsibility of the grantee and do not necessarily represent the official views of the USEPA. Further, USEPA does not endorse the purchase of any commercial products or services mentioned in the publication

  • Competing interests None.

  • Ethics approval IRB at Harvard School of Public Health.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.