Background Health effects of ozone have been observed in numerous studies. However, analyses of more cause-specific morbidity or mortality outcomes have rarely been performed. A study was undertaken to determine the short-term associations of ozone with cause-specific cardiorespiratory mortality and morbidity by age groups.
Methods Daily levels of ozone were measured at a background measurement station in 1998–2004 in Helsinki, Finland. All analyses were a priori restricted to the warm season. Daily cause-specific cardiorespiratory mortality and hospital admissions were studied in elderly people (≥65 years) and adults (15–64 years) and associations between ozone and asthma emergency room visits in children (<15 years) were analysed. All models were adjusted for PM2.5 and Poisson regression was used for the analyses.
Results There was a positive association between ozone and admissions for asthma-chronic obstructive pulmonary disease (COPD) in elderly people (9.6%; 95% CI 2.0% to 17.8% at 0-day lag for 25 μg/m3 increase in ozone). Consistent associations were also found between ozone and asthma emergency room visits in children (12.6%; 95% CI 0.8% to 25.1%, 0-day lag). There was a suggestion of an association between ozone and admissions for arrhythmia among elderly people (6.4%; 95% CI 0.63% to 12.5%, 1-day lag), which was slightly confounded by PM2.5.
Conclusions Positive associations were found for ambient ozone with asthma visits among children and with pooled asthma/COPD admissions among elderly people. The evidence for a positive association between ozone and cardiovascular health was weaker.
- air pollution
Statistics from Altmetric.com
Ozone (O3) is commonly known as an irritant air pollutant. Ground level O3 levels are expected to increase as a result of climate change,1 2 and therefore the importance of O3 as a potential risk for health is also increasing. The association between ozone and mortality has been well established in epidemiological multi-city studies.3–5 Respiratory morbidity, including asthma and chronic obstructive pulmonary disease (COPD), has also been linked with the short-term changes in O3 levels in several studies.6–10 However, results on cardiovascular outcomes have been fairly inconsistent so far.11 Only a few studies have provided evidence of the association between cardiovascular health and exposure to O3.12–14
In previous studies determining the health effects of O3, analyses by cause-specific cardiorespiratory outcomes have not been included and therefore there has been little comparison between the effects of O3 on cardiovascular and respiratory health. Epidemiological studies have shown that ambient particulate matter has the strongest effects on health among children and elderly people.7 15 16 However, the differences in the vulnerability to O3 between age groups have rarely been evaluated on a cause-specific level in the same study.
This study provides information about the cardiorespiratory health effects of O3 during the warm season. Because the adverse effects of O3 have been found to be stronger or to only occur during the warm season,8 10 17 the current analyses were performed only for the warm season. Much of the O3 and fine particle mass over Helsinki is long-range transported from other European countries and Russia,18–21 which results in considerable correlations between the two pollutants. Thus, even though the effects of O3 have mainly been found to be independent of particulate pollutants,3 7 we controlled for the possible confounding by adjusting all O3 analyses for particles with an aerodynamic diameter <2.5 μm (PM2.5). The analyses were performed by age groups in order to identify whether some population subgroups are more sensitive to O3 than others.
The study took place in 1998–2004 in the metropolitan area of Helsinki which includes four municipalities: Helsinki, Vantaa, Espoo and Kauniainen (1 million inhabitants, 745 km2). We obtained the daily mortality and acute hospital admission and emergency room visit counts from national registers. For mortality, we used the underlying cause of death and, for morbidity, the primary diagnoses were used. As the cardiovascular outcomes in the analyses we used mortality and morbidity for all cardiovascular diseases (ICD-10 I00–I99), coronary heart diseases (CHD, I20–25), stroke (I60–61, I63–64) and arrhythmia (I46.0, I46.9, I47–I49). As the respiratory outcomes, we used all respiratory diseases (J00–J99), pneumonia (J12–J15, J16.8, J18) and pooled asthma-COPD (J41, J44–J46). These disease groups were identified among elderly people (≥65 years) and adults (15–64 years). In the data, hospital admissions were counted as those including an overnight stay in the hospital. In addition, we determined the effects of O3 on asthma (J45, J46) emergency room visits of children (<15 years). The emergency room visit data were used here because overnight stays in hospital for asthma among children are rare.
The measurements of O3 were performed at an urban background measurement station at 6 m height, located at a distance of 200 m from the closest busy road (14 000 vehicles per day) and 15 km from Helsinki city centre. The O3 levels measured at this site are primarily affected by long-range transport and local traffic. Daily levels of PM2.5 were measured at two central sites for the same time period.22 The measurements for both pollutants were performed according to the WHO recommendations,23 but the sites for O3 and PM2.5 were not the same. Missing data for both pollutants were replaced with measurement values from the closest monitoring station.
We restricted the analyses to the warm season (May–September). The division of the seasons was based on temperature in the study area, which presumably has the largest effect on both the amount of exposure to O3 and the concentrations of O3. During the warm season there were no days when the temperature fell below 0°C.
We first built the core models without the air pollutants using Poisson regression. Penalised thin plate regression splines were used in the generalised additive model (GAM) framework.24 25 Long-term time trend was controlled for by dummy variables (for each year and month) that were always in the model as were a dummy variable for weekday and continuous variables for the current day mean temperature, relative humidity and barometric pressure. We used penalised spline smoothing to examine the shape of the association between weather variables and mortality or morbidity, which all turned out to be linear. The significance of other potential confounders—the mean temperature on the three previous days and relative humidity, dummy variables for influenza epidemics, high pollen episodes and general holidays—were always checked. These confounders were dropped from the model if their p value was >0.20. The same variable selection method was used in our previous analyses16 26 to keep the modelling simple and not to be too strict when dropping the possible confounders. We had weekly data on the influenza counts, and therefore we created a three-level variable for influenza epidemics described earlier.26 Similarly, a two-level variable was created for the pollen episodes that occurred during the pollen period (March–August).26 Regression models were implemented in R27 with the MGCV 1.4-0 procedure.28
The analyses of the associations between O3 and mortality and morbidity were conducted with robust Poisson regression. Robust standard errors of the estimates were used to account for the heteroskedasticity and autocorrelation observed in the residuals. Modelling was implemented using the Zelig package29 30 in R 2.7.1 software, based on the models built using GAM.
In the models we used the 8 h maximum moving average of each day for O3. All the analyses were also adjusted for the 24 h (from midnight to midnight) mean levels of PM2.5 in order to take into account the possible confounding effect of particulate air pollution. In advance, we chose to analyse the effects of exposures during 0–5 days before the event and the average of 0–4 lag days (5-day mean). Lag 0 is defined as the average concentration on the day of death, hospitalisation or emergency room visit, lag 1 as the average concentration on the preceding day, and so on. All the results are provided as percentage change in mortality or morbidity for an interquartile (25th–75th percentile) range increase in pollutant concentration.
A partial autocorrelogram was conducted to define the amount of autocorrelation in the residuals and to avoid over-smoothing. As sensitivity analyses, we ran the models using the same lag for temperature as for O3 and PM2.5 in the models. The effect of high pollutant days was checked by analysing the data using the 98th percentile as the cut-off point for the pollutant levels. Sensitivity analyses had a minor effect on the results.
There were 1071 days during the warm season. The number of deaths among elderly people for all cardiovascular diseases, CHD, stroke and all respiratory diseases are shown in table 1. Among the elderly population, the total number of deaths for pneumonia was 776 (mean of daily deaths 0.7) and for COPD the total number of deaths was 404 (mean 0.4). Among adults the total number of deaths for all respiratory diseases was 167 (mean of daily deaths 0.2), for all cardiovascular diseases the number was 1292 (mean 1), for CHD 674 (mean 0.6) and for stroke 216 (mean 0.2). For mortality from arrhythmias the total number of cases among all age groups was only 296. Cause-specific analyses could be performed only for CHD and stroke mortality among the elderly population where the total number of cases was >1000. A summary of the mortality, morbidity and environmental data is given in table 1. Ozone was not correlated with pollutants (data not shown) other than PM2.5 (table 1).
Among the elderly population we found negative associations between O3 and all cardiovascular mortality at lag days 2 and 3 (table 2). All respiratory mortality did not show any associations with O3. However, the current day PM2.5, when adjusted for O3, was associated with all cardiovascular mortality of the elderly population.
A positive association was observed between O3 and all cardiovascular mortality of adults with 1-day delay (see table 1 in online supplement), but only when the O3 model was adjusted for PM2.5 (13.8%, 95% CI 0.31% to 29.0%). No other associations were observed for mortality among adults.
In the cause-specific mortality analyses among the elderly population (see table 2 in online supplement) we found a negative association between CHD and O3 adjusted for PM2.5 over the 5-day mean (−11.6%, 95% CI −20.2% to −2.14%). Mortality from stroke was not associated with O3.
All cardiovascular hospital admissions in the elderly population (see table 3 in online supplement) had a borderline significant association with O3 at 1-day lag (2.45%, 95% CI −0.40% to 5.39%, only when adjusted for PM2.5). There was also a significant association between all respiratory admissions in the elderly population and current day PM2.5 that was adjusted for O3 (4.33%, 95% CI 1.08% to 7.68%). Among adults, all cardiovascular or all respiratory hospital admissions were not associated with O3 or PM2.5 (data not shown).
In the cause-specific analyses for hospital admissions, three cardiovascular outcomes were studied but few associations were observed. Arrhythmia in the elderly population showed a significant positive association with the previous day O3 level with a 6.4% increase (95% CI 0.63% to 12.5%) in admissions in the single-pollutant model (see table 4 in online supplement). However, when the O3 model was adjusted for PM2.5, the association became non-significant and the estimate for previous day O3 was reduced to 4.81% (95% CI −1.61% to 11.7%) (figure 1). Coronary heart disease admissions among adults were positively associated with PM2.5 that was adjusted for O3, but not with O3 (see table 4 in online supplement). Stroke admissions were not associated with O3 or PM2.5 in either of the age groups.
Consistent associations were found for O3 with hospital admissions for pooled asthma-COPD among the elderly population. A strong and significant association was observed on the current day (9.62%, 95% CI 0.02% to 17.8%) and a borderline significant association over the 5-day mean (10.0%, 95% CI −0.56% to 21.8%) (see table 5 in online supplement). Adjusting the asthma-COPD model for PM2.5 had a minor effect on the association between admissions and O3 (figure 2), with an increase in admissions on the current day of 8.60% (95% CI 0.29% to 17.6%) and in the 5-day mean of 11.8% (95% CI −0.44% to 25.5%). Among adults, no significant associations were observed with cause-specific respiratory hospital admissions (see table 5 in online supplement).
Most of the asthma-COPD admissions in the elderly population were diagnosed as COPD (n=2569). When COPD was analysed separately for elderly people, a positive association was observed in the PM2.5 adjusted analysis at lag day 3 with an increase in admissions of 10.6% (95% CI 1.86% to 20.1%) for an interquartile increase in O3 (figure 2). Because of the low number of COPD cases (n=776), the CIs for the COPD results among adults are wide. Borderline significant associations on the current day were observed between pneumonia admissions in the elderly population and PM2.5 adjusted for O3, and between asthma-COPD admissions for adults and PM2.5 adjusted for O3 (see table 5 in online supplement). Asthma emergency room visits of children increased consistently in association with both O3 and PM2.5 (table 3).
We found positive associations between the warm season O3 levels and asthma-COPD hospital admissions for the elderly population (≥65 years) and between O3 and asthma exacerbations in children (<15 years). In general, the associations between O3 and respiratory diseases were stronger than those observed for cardiovascular outcomes. There was a strong positive association between O3 and asthma-COPD admissions in the elderly population which was also stronger than the association found for PM2.5. No associations were found between O3 and pneumonia, in contrast to some previous studies.10 31 However, there was an association between pneumonia admissions in the elderly population and PM2.5.
We converted here the effect estimates of single pollutant models for O3 into percentage change for a 10 μg/m3 increase in O3 in order to compare our results with those obtained previously. The associations between O3 levels and asthma-COPD admissions in the elderly population (3.74%; 95% CI 0.12% to 6.70%, lag 0) or between O3 and COPD alone (6.35%; 95% CI 0.79% to 12.12%, 5-day mean) are higher than those reported previously in Europe (0.85%; 95% CI 0.4% to 1.3%, lag 0 or 1)6 and in North America (0.27%; 95% CI 0.08% to 0.47%, 2-day mean).10 The reason for the larger CIs in our study is the small number of admissions compared with those in multi-city studies.
The respiratory system of elderly people may be more vulnerable to air pollutants than that of adults because of more commonly existing underlying diseases. As O3 is an irritating and oxidative pollutant, it readily reacts with the compounds in the epithelial lining fluid in the airways. Oxidative processes in the epithelial lining fluid, on the other hand, can be prevented by antioxidants.32 It has been suggested that elderly people have decreased availability of antioxidants, which may in part result in reduction in the defence mechanisms.32 It can be speculated that this also leads to increased susceptibility to O3 and other pollutants.
O3 and PM2.5 were also positively associated with asthma emergency room visits of children. An earlier multi-city study, including Helsinki, investigated the associations between O3 and asthma hospital admissions among children, but the authors did not find as strong an association33 as we did in the current study. However, the emergency room visit data may be more sensitive for determining the associations between air pollutants and asthma exacerbations than hospital admissions data. We found that asthma visits increased in association with O3 levels (4.9%; 95% CI 0.32% to 9.6% per 10 μg/m3 at lag 0), and the association was stronger than those reported from earlier studies on asthma morbidity in children in Paris (2.81%; 95% CI −2.0% to 7.8%, 0-day lag),34 Washington DC (3.2%; 95% CI 1.4% to 5%, 0-day lag)9 and Hong Kong (3.9%; 95% CI 3.0% to 4.9%, 5-day mean).7 As in the elderly population, the association between asthma exacerbation and O3 was stronger than the association observed for PM2.5.
It has been suggested that children are susceptible to the effects of air pollutants because of the developing stage of their respiratory tract, and people with asthma in general have been shown to be more sensitive to the effects of O3 than healthy people.35 Children generally spend more time outdoors than other population groups, and they breathe more in relation to their body mass than older people,36 which is why they probably have higher exposures to ambient pollutants. This may partly explain why the effect of O3 was stronger among children than among other age groups. However, the stronger associations observed for children may also derive from the use of emergency room visit data rather than admissions data that were used for adults and elderly people. The former possibly describes better the acute exacerbation of asthma and is more sensitive, therefore providing greater effect estimates.
We found few associations between O3 and cardiovascular outcomes. This is in accordance with some earlier studies.37–39 However, positive associations between O3 and cardiovascular morbidity have been found earlier in European,14 North American40 and Asian41 studies. The possible mechanism behind the cardiovascular effects may be inflammation or induction of artery vasoconstriction.42
We found a positive association between O3 and all cardiovascular mortality in adults, but the association became significant only when the O3 model was adjusted for PM2.5. In addition, we found a delayed negative effect of O3 on all cardiovascular and coronary heart disease mortality among the elderly population. There is no clear explanation for the negative association.
There was also some suggestion of an association between O3 and arrhythmia admissions in the elderly population, which is in line with recent results where atrial fibrillation increased the vulnerability to death from exposure to O3.43 However, the association we found became weaker when the model was adjusted for PM2.5.
As can be seen from the results discussed above, associations between O3 and cardiovascular health outcomes seem to be somewhat sensitive to adjustment for PM2.5. Some earlier studies have reported that associations between O3 and cardiovascular health are independent of particulate pollutants.14 44 However, the difference between studies is probably due to lower correlations between O3 and particulate matter (PM10) in previous studies (0.1614 and 0.0344) compared with ours (correlation with PM2.5 0.43). As most of the O3 and PM2.5 in Helsinki originates from long-range transport18–21 and therefore they also have similar concentration peaks, it is possible that more confounding between these pollutants is present in Helsinki than in other study areas.
Despite actions to limit the emissions of the precursor gases of O3 such as volatile organic compounds (VOCs) and nitrogen oxides (NOx) in Europe, ozone concentrations have stayed approximately at the same level in the clean background areas of the Helsinki metropolitan area during the last 10 years. On the other hand, in the city centre O3 levels have even increased since local traffic emissions (eg, NO+O3→NO2+O2 reaction) that act as an O3 sink have decreased. Furthermore, some new exhaust cleaning technologies of diesel vehicles lead to high NO2 emissions (high NO2:NOx ratio), resulting in a higher local formation rate of O3.
A limitation of this and other time-series studies on O3 is the validity of exposure assessment for O3, because there is lack of personal O3 measurements.45 The measurements were performed outdoors while people spend most of their time indoors. The central site measurements of PM2.5 have been found to be good proxies of personal exposure in Helsinki,46 however, the validation of personal O3 exposure assessment has rarely been done. Indoor O3 levels tend to be much lower than outdoors. This is due to the lack of indoor sources and rapid conversion of O3 into by-products upon reaction with surfaces and other gases. However, these by-products include irritating substances such as formaldehyde and some carboxylic acids. Secondary organic aerosols also result from O3 reactions. Thus, some of the respiratory health effects of O3 could be due to exposure to these by-products indoors.47
Ventilation has also shown to have an effect on the indoor levels of O3 and other pollutants.48 Buildings with mechanical ventilation, which are usually also more recently built, have lower O3 concentrations because filtering materials of ventilation systems can remove some of the ambient O3.49 However, in Helsinki, most residences are not mechanically ventilated but use gravitational ventilation in which air slowly flows through window edges. Most of the buildings in Helsinki are not air conditioned, which could further reduce infiltration of outdoor pollutants indoors. We therefore believe that changes in the outdoor O3 levels mirror the changes in indoor concentrations, especially during the warm season when the increased opening of windows provides additional ventilation to residential buildings.
In summary, we found that O3 levels during the warm season are associated with the exacerbation of asthma-COPD among elderly people and with exacerbation of asthma in children. The effects of O3 on asthma and COPD were stronger than the effects of PM2.5. Respiratory morbidity of adults was less correlated with O3, possibly because of the lesser degree of vulnerability in this age group. There was no clear effect between O3 and cardiovascular health. It also seems that the health effects of O3 and PM2.5 can sometimes be difficult to separate. However, there is a need for studies investigating personal exposure to O3 among different age groups.
What is already known on this subject
Ambient O3 has been found to be harmful, especially for respiratory health. However, studies on more specific outcomes and comparison between respiratory and cardiovascular outcomes or between age group sensitivity have rarely been done.
What this study adds
Ambient O3 exacerbates asthma and COPD among the elderly population and asthma among children during the warm season, but adults seem to be less sensitive to the effects of O3. No clear associations were found for cause-specific cardiovascular outcomes. In addition, it seems that the effects of O3 and PM2.5 may sometimes be difficult to separate.
The authors thank Statistics Finland, the Finnish Meteorological Institute and the Skin and Allergy Hospital of Helsinki University Central Hospital for providing data used in this study.
Funding The National Technology Fund (TEKES, Contract 40715/01) and the Ministry of Education (Graduate School in Environment Health SYTYKE).
Competing interests None.
Patient consent Obtained.
Provenance and peer review Not commissioned; externally peer reviewed.
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.