Article Text

Download PDFPDF

Original article
Evidence of the role of short-term exposure to ozone on ischaemic cerebral and cardiac events: the Dijon Vascular Project (DIVA)
  1. Jean-Bernard Henrotin1,
  2. Marianne Zeller2,
  3. Luc Lorgis2,
  4. Yves Cottin2,
  5. Maurice Giroud1,
  6. Yannick Béjot1
  1. 1Dijon Stroke Registry (Inserm, InVS), IFR 100 STIC-Santé, University Hospital and Faculty of Medicine of Dijon, EA 4184, University of Burgundy, Dijon, France
  2. 2Observatory of Myocardial Infarction of Côte-d'Or (RICO), Laboratoire de Physiopathologie et Pharmacologie Cardiovasculaire Expérimentale (EA 2979), Faculty of Medicine of Dijon, University of Burgundy, Dijon, France
  1. Correspondence to Dr Yannick Béjot, Dijon Stroke Registry, Department of Neurology, University Hospital, 3 Rue du Faubourg Raines, Dijon 21000, France; ybejot{at}


Objectives To confirm the effects of short-term exposure to ozone (O3) on ischaemic heart and cerebrovascular disease.

Methods Daily levels of urban O3 pollution, the incidence of first-ever, recurrent, fatal and non-fatal ischaemic cerebrovascular events (ICVE) and myocardial infarction (MI) were correlated using a case-crossover design. The authors analysed 1574 ICVE and 913 MI that occurred in Dijon, France (150 000 inhabitants) from 2001 to 2007. Sulfur dioxide (SO2), nitrogen dioxide (NO2), carbon monoxide (CO) and particulate matter with an aerodiameter of ≤10 μg/m3 (PM10) were used to create bi-pollutant models. Using the adjusted OR, the effects of O3 exposure were calculated for every 10 μg/m3 increase in pollutants in multivariate logistic models adjusted for temperature, humidity, flu outbreaks and holidays.

Results The authors found a significant association between exposure to O3 and recurrent ICVE with a 3-day lag (OR=1.115; 95% CI 1.027 to 1.209). The direction and magnitude of the association between exposure to O3 and recurrent MI were similar but not statistically significant. For incident events, the authors detected only a non-significant association for ICVE with a 2-day lag (OR=1.041; 95% CI 0.996 to 1.089). In the subgroup analysis for ICVE, the authors observed an increased association with cardiovascular risk factors (OR=1.523; 95% CI 1.149 to 2.018). For MI, the authors found an association with O3 when hypercholesterolaemia was present (OR=1.111; 95% CI 1.020 to 1.211), and the association became stronger with the number of cardiovascular risk factors. The authors found a marked dose–response relationship.

Conclusion Recurrent ICVE and MI could be triggered by short-term exposure to even low levels of O3, especially among subjects with severe vascular risk factors.

  • Stroke
  • cerebral infarction
  • myocardial infarction
  • epidemiology
  • air pollution
  • acute coronary syndrome
View Full Text

Statistics from


Atmospheric pollution is a major problem of environmental health.1 Ozone (O3) is the main component of this photochemical pollution. Among the expected effects on health, the short-term effects on respiratory function2 and on mortality3 4 induced by O3 pollution are well documented. In contrast, the impact of O3 on cardiovascular disease is less well known. An association between short-term O3 exposure was observed with acute myocardial infarction (MI)5 and ischaemic stroke.6 The objective of the present study was to confirm the observed relationship between O3 exposure and the onset of ischaemic cerebrovascular events (ICVE) and MI in a population-based study, the Dijon Vascular Project (DIVA).7

Material and methods

Our study was based on both the Dijon Stroke Registry, which has been running since 1985,8 and the Registry of MI of Dijon and Côte d'Or (Rico) which was launched in 2001.9 Data from these two registries between January 2001 and January 2007 have been pooled in a retrospective analysis for a collaborative study called the DIjon VAscular (DIVA) project.7

Meteorological data

The city of Dijon, with 150 800 inhabitants in 2007, is characterised by a four-season continental climate. The daily meteorological data (temperature, relative humidity, sunlight, wind) were obtained from the National Meteorological Office. Air pollution data were obtained from Dijon Environmental Monitoring Centre. Exposure measurements during the study were taken at the background monitoring station located in the town centre. The station was remote from the main sources of urban pollution (major roads, industrial areas). The following atmospheric pollutants were measured: particulate matter of aerodynamic diameter ≤10 μm (PM10) measured by β-ray absorption, ozone (O3) by ultraviolet fluorescence, sulfur dioxide (SO2) by ultraviolet fluorescence, nitrogen dioxide (NO2) by chemoluminescence and carbon monoxide (CO) by non-dispersive infrared photometry. We calculated the hourly mean of each pollutant (μg/m3) and then computed their 24 h arithmetic averages (noon to noon). For O3, we used the means calculated over 8 h daytime periods. Also, for O3, we used daily 1 h maximum values to study the impact of O3 peaks ICVE and MI. Data on flu epidemics were obtained from the Sentiweb monitoring network of the region of Dijon.

Ascertainment of stroke and transient ischaemic attack

Continuously since 1985, the Stroke Registry of Dijon has been using a standardised prospective and exhaustive method of ascertainement, for patients of all ages. A detailed description of the Dijon Stroke Registry has been published elsewhere.7 8 To ensure the completeness of case ascertainment, information was provided by six sources: (1) the emergency rooms, and all of the clinical and radiological departments of Dijon University Hospital; (2) the emergency rooms and all of the clinical departments of the three private hospitals of the city and its suburbs; (3) the patient's home with diagnosis assessed by the 250 general practitioners with the help of an outpatient clinic; (4) the three private radiological centres, where the medical records were reviewed to identify missed cases; (5) the ultrasound Doppler centres of the University Hospital and private centres, where medical records were reviewed; and (6) the death certificates with stroke as the underlying cause of death obtained from the local Regional Health Service, which is responsible for registering all deaths in the community. All of the collected death certificates were checked by a member of our team in order to include only patients who died from stroke.

Stroke was defined according to WHO recommendations and according to the International Classification of Disease.7 8 The ischaemic mechanism was identified by a CT scan performed in 98% of cases and/or MRI in 22% of cases. The diagnosis of ICVE was always performed on clinical criteria, cerebral imaging and complementary investigations.7 8 We included both ischaemic stroke and Transient Ischaemic Attack (TIA), defined according to WHO recommendations as sudden signs and symptoms affecting motor, sensory, sensorial, speech, brainstem and cerebellar functions lasting less than 24 h.

ICVE were defined as incident (first-ever ICVE) or recurrent (if known previous ICVE).

Ascertainment of myocardial infarction

The design and methods of the Registry of MI of Dijon and Côte d'Or (Rico) survey have been published.7 9 Briefly, since 1 January 2001, the Rico registry has been collecting data from patients hospitalised with MI in the City of Dijon and in five cities of the Côte d'Or area. To ensure the completeness of case ascertainment, the collaboration of GPs and private cardiologists was also required in order to identify patients who had suffered an MI without hospitalisation in one of the cities. Both ST-elevation MI (STEMI) and non-ST-elevation MI (NSTEMI) were diagnosed according to the European Society of Cardiology and the American College of Cardiology criteria.7 9 For this study, only MI patients living in Dijon were included.7

AMI were defined as incident (first-ever AMI) or recurrent (if known previous AMI).

Vascular risk factors and prior treatments

Vascular risk factors were collected with a common methodology as previously described.7–9 Hypertension was defined by a history of known hypertension (systolic blood pressure ≥140 mm Hg and/or diastolic blood pressure ≥90 mm Hg) or antihypertensive treatment. Diabetes mellitus was recorded if a glucose level of ≥7.8 mmol/l had been reported in the medical record or if the patient was under insulin or oral hypoglycaemic agents. Hypercholesterolaemia was defined by total cholesterol level ≥5.7 mmol/l. We also recorded tobacco use (one cigarette per day, current or former habit) and atrial fibrillation diagnosed from EKG or Holter recordings. For stroke, a history of TIA and previous AMI were recorded, and for AMI, history of TIA and stroke were collected.

Preventive treatments were also recorded. We included anticoagulants (warfarin, acenocoumarol or fluindione), antiplatelet agents (aspirin, clopidogrel, ticlopidine or dipyridamole), antihypertensive treatment and statins.

Statistical analysis

A case-crossover design was adopted to study the relationship between MI, IS and TIA and urban ozone pollution.10 For each subject, the level of exposure during an at-risk day before the health event (‘case day’) was compared with the levels of exposure during ‘control days.’ We used a time-stratified approach to choose the control days.11 12 For each subject, 1 case day (coded as 1) was matched with eight control days (coded as 0). The control days were chosen 7, 14, 21 and 28 days before and after the case day. This design was used to control for seasonality, long-term time trends and days of the week.12

Conditional logistic regression was used to estimate the ajusted ORs and their 95% CIs. Stratified analyses by subgroup according to vascular risk factors were carried out to identify subjects susceptible to the effects of O3 pollution. In our study, we systematically adjusted for temperature, relative humidity, flu epidemics, and public and school holidays.13 14 The O3 variable was entered into the model as a continuous variable. OR were calculated for every 10 μg/m3 change. O3 exposure was tested in models for the day of the cerebral and cardiac event (D-O), from a 1-day lag (D-1), to a 4-day lag (D-4). We evaluated dose–response relationships across four exposure levels of O3. The first quartile was used as the reference group.

We analysed O3 data stratified according to the warm period (from June to September), the intermediate period (from March to November) and the cold period (from December to February). We used separate models to examine the effect of O3 alone and O3 adjusted for each of the other pollutants in two-pollutant models (SO2, NO2, CO and PM10). The data were analysed using Stata 9.0 software (StataCorp LP, College Station, Texas).


The Diva project was approved by the Ethics Committee of the University Hospital of Dijon.


A total of 2487 ischaemic events (1574 ICVE and 913 acute MI) were collected from 2001 to 2007 (table 1). The daily mean number of ICVE, MI and both events together was 0.61 per day, 0.36 per day and 0.97 per day, respectively. Subjects with ICVE were older than subjects with MI (median of age, 78.9 vs 73 years) (p<0.01).

Table 1

Baseline characteristics of patients with ischaemic cerebrovascular events and myocardial infarction

For the whole study period, the daily maximal temperatures ranged from –4.5°C to 39.3°C (table 2). The mean concentrations of SO2, NO2, CO and PM10 during the 6-year period were relatively low: 3.4 μg/m3, 36.0 μg/m3, 456 μg/m3 and 20.5 μg/m3, respectively (table 2). As expected, the highest concentrations of O3 were observed on days with the weakest winds and the highest temperatures, the sunniest and driest days.

Table 2

Distribution of air pollutants and meteorological variables

O3 concentrations correlated strongly and positively with temperature and sunlight (p<0.001) (table 3). A negative association between O3 and other pollutants was also noted.

Table 3

Spearman rank correlation between O3 concentration and air pollutants and meteorological variables

Table 4 summarises the relationship between O3 concentrations for the two categories of exposure (daily 8 h average concentration or daily 1 h maximum value) and daily ischaemic events. These findings were obtained after adjustment for minimum temperature (1-day lag), maximum relative humidity (1-day lag), school and public holidays and flu epidemics (0-day lag). The OR are given with an increment of 10 μg/m3 in O3 concentration. Associations between O3 exposures and recurrent ischaemic events (ICVE, MI) with a 3-day lag (OR=1.074; 95% CI 1.016 to 1.135) and recurrent ICVE with a 3-day lag (OR=1.115; 95% CI 1.027 to 1.209) were noted. The direction and magnitude of association between O3 exposures and recurrent MI were similar but not statistically significant. For incident events, an association between incident ICVE and O3 was found at 1-day lag (OR=1.014; 95% CI 0.965 to 1.065) and at 2-day lag (OR=1.041; 95% CI 0.996 to 1.089) but these were not significant. Using 1 h maximum indicators, the ORs presented lower magnitudes than did the 8 h averaged indicators, suggesting an absence of a direct effect of O3 peaks in this study. Also, we did not observe any cumulative effect of O3 exposure over 2 or 3 days. Of note, the associations observed remained unchanged if maximum rather than minimum temperature was included in the logistic regression model.

Table 4

Relationship between daily ischaemic events and O3 exposure for an increase of 10 μg/m3

In a stratified analysis by vascular risk factors, a 1-day lag association with each factor was observed for both recurrent and incident ICVE (table 5). The strength of association was the highest and most significant when a medical history of diabetes was present (OR=1.138; 95% CI 1.027 to 1.260), for patients with one factor. By contrast, for recurrent and incident MI, a significant association with O3 was found only when hypercholesterolaemia was present (OR=1.111; 95% CI 1.020 to 1.211). Furthemore, for MI and ICVE, we observed an increase in the strength of association with the number of combined factors.

Table 5

1-day lag association between ozone exposure and ischaemic cerebrovascular event or myocardial infarction according to vascular risk factors

The dose–response relationship (with 1-day lag) between O3 quartiles and ischaemic events according to the number of vascular risk factors is shown in figure 1. Using the first quartile as the reference group, we observed a dose–response relationship for subgroups with at least one combined risk factor (figure 1A, test for trend, p=0.04). This dose–response relationship markedly increased with the number of risk factors (figure 1B–D). In contrast, for subgroups without vascular risk factors, no dose–response relationship was noted (figure 1E).

Figure 1

Dose–response relationship (with 1-day lag) between O3 quartiles and ischaemic events according to the number of vascular risk factors. OR, adjusted odds ratio for minimum temperature (1-day lag), maximum relative humidity (1-day lag), school and public holidays.

Several sensitivity analyses were performed. First, the risk of an ischaemic vascular event by season was assessed. The OR was greatest in the warm period (OR=1.137; 95% CI 1.034 to 1.250). Second, after the introduction of other pollutants in two-pollutant statistical models, the ORs were not affected (data not shown).


This study assessed the effects of urban O3 pollution on ischaemic cardiac and cerebrovascular events in a medium-sized city with low levels of pollution. For the whole year, we observed associations between urban O3 pollution and recurrent ICVE and MI over several lagged days and adjusted for potential confounding variables (temperature, humidity, flu epidemics, holidays, days of the week, season, time trend). These associations were statistically significant for recurrent ICVE at the 3-day lag. The direction and magnitude of association between O3 exposures and recurrent MI was similar but not statistically significant. These results suggest that recurrent ICVE and MI could be triggered by short-term exposure to low levels of O3. In stratified analysis by vascular risk factor, an association with all the factors was noted for ICVE. The strength of association was the highest when a medical history of diabetes was present. For MI, an association with O3 was found only when a medical history of hypercholesterolaemia was present. However, for both MI and ICVE, we found both an increase in the strength of association with the number of combined vascular risk factors and a marked dose–response relationship. Hence, the subjects with vascular risk factors could be particularly sensitive to exposure to O3.

Several studies have shown an association between O3 and admissions for ischaemic cardiac disease.5 15 16 For ICVE, data are more scarce. All of the studies concerning admissions for cerebrovascular disease did not report any association with ozone.17 18 By contrast, data on ICVE reported in population or hospital-based studies show a significant association with ozone.6 19 20

For the O3 variable, different methodological approaches (outcomes, designs, modelling choices) and features of local pollution (mix, levels, meteorological data) may explain the different results between studies. Local climatic conditions, environmental situations and levels of O3 pollution may also play a role.21 Furthermore, some of the differences could be due to the different distribution of confounders and effect modifiers between studied populations. Hence, in studies comparing personal exposure and ambient concentrations of O3, the time spent indoors at home with or without air conditioning strongly modified personal exposure to O3.22 Demographic features such as the proportion of elderly residents could also explain differences between cities.21

Our findings suggest that exposure to O3 could increase the risk of ischaemic events, particularly in subjects with one or several vascular risk factors inducing atheromatous lesions. Few studies have reported that pre-existing vascular risk factors increase the risk of ischaemic vascular events due to air pollution.23 The link between O3 pollution and cardiac and cerebral ischaemic events could be the inflammatory reaction induced by O3, suggesting an increase in biological parameters related to oxidative stress, systemic inflammation or procoagulant reactions after O3 exposure.23 24 These observations may be interpreted in the light of some controlled toxicological studies in animals and in humans. These studies indicate an increase in biological parameters related to oxidative stress, systemic inflammation or procoagulant reaction after O3 exposure.23 25 In humans, a recent panel-design study using personal air monitors on students from a university in Tapei found that O3 was associated with increased levels of C-reactive protein, fibrinogen, 8-hydroxy-2′-deoxyguanosine, plasminogen activator inhibitor-1 and decreased heart rate variability.24

Our study has several potential limitations. The question of temperature measurement is a problem because O3 and temperature are expected to be highly correlated. Ambient O3 is produced to a large extent by interactions betweeen exhaust fumes and sunlight. High O3 episodes are typically observed on sunny, hot, dry windless days. Extreme temperature may also be directly associated with morbidity and premature mortality, and thus may be an important confounder. On high temperature days, there is also a possible interaction between the effects of temperature and relative humidity. In our study, we took several precautions. We incorporated temperature and maximal humidity into our models, and the interaction between the two was investigated. We did not detect any significant interaction. We verified the lack of a non-linear effect of the temperature variable for extreme values. The use of linear splines, different transformations (log, quadratic or cubic) or categorical transformation did not markedly improve the fitted models (data not shown). Finally, we verified that our results were not modified by excluding the highest 10% of temperatures.

Analysis of O3 was also complicated by the potential for O3 effects to be confounded by other pollutants and particularly by particles. In our study, in agreement with previous results,3 4 the effect of O3 remained similarly associated with ischaemic vascular outcome after inclusion of PM10 as a covariate in the models. Because PM2.5 monitoring was not available in Dijon, we did not examine the relationship between PM2.5 and O3 pollution. Some authors argue that ambient concentrations of gases could serve as a surrogate for personal exposure to PM2.5.22

For the collection of exposure data, several precautions were taken. We limited our investigations to a small area (40 km2). Only subjects living within the Dijon connurbation were investigated. Thus, the greatest distance between home and the monitoring site was less than 3.5 km. Furthemore, Dijon is not considered a particularly polluted town; the levels of pollutants are relatively low except for ozone (mean=38.2 μg/m3). Also, the assumption that exposure of the study population was relatively homogeneous is based on the following arguments. First, Dijon is located in a bowl-shaped basin which tends to retain the pollution in the city. Second, the population of Dijon on a given day is relatively stable. Indeed, only 7.4% of residents of Dijon regularly work outside Dijon, and this falls to 1.8% for those over 50 years of age. However, exposure misclassification cannot be avoided. The same approximate exposure (‘proxy’) is used for all subjects, but true exposure may vary randomly.

The characteristics of the time series of pollutants, including long-term time trends, seasonality and short-term autocorrelation, expose the case-crossover design to some potential biases. Notably, the implications of referent period selections on risk estimates have been evaluated in great detail.11 12 Based on this work, we chose our control days (same day of the week, month and year as the case day) by using a time-stratified approach.12 This approach is not subject to time-trend biases and ensures unbiased conditional logistic regression estimates.11 12

To conclude, this is the first large population-based study to show an association between O3 concentrations and recurrent ICVE and MI over a long period. These associations depend on the number of vascular risk factors. This study strengthens the hypothesis according to which subjects presenting severe vascular risk factors are at risk from ambient O3 pollution. As a result, preventive strategies could be designed for such subjects.


We thank the University Hospital and the Faculty of Medicine of Dijon, the Univesity of Burgundy, Inserm and InVS, the Centre Départemental de la Côte d'Or de Météo-France, the association Atmosph'air and the URCAM/ARH de Bourgogne, for their contributions to this study, and P Bastable, for reviewing the English.


View Abstract


  • Linked articles 205583.

  • Funding The Dijon Stroke Registry was supported by Inserm (National Institute for Health and Medical Research) and InVS (Institute for Public Health Surveillance).

  • Competing interests None.

  • Ethics approval Ethics approval was provided by the Ethics Committee of the University Hospital of Dijon.

  • 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.

Linked Articles