Association between modelled traffic-related air pollution and asthma score in the ECRHS

It is well known from time-series and other studies that fluctuations in air pollution levels are associated with short-term effects among asthmatics 1. Evidence of adverse effects of air pollution on asthma outcomes and of a potential asthmogenic role of traffic-related local pollutants are studied far more frequently in children than in adults 2–4. In adults, traffic-related pollutants have been associated with cough, bronchitis and chronic obstructive pulmonary disease (COPD) 5–7, and living close to a major road has been associated with asthma-related symptoms 8. An Italian study found indications of an association between nitrogen dioxide (NO₂) and asthma prevalence in young adults when stratified by different climates 9. US veterans residing near a major road had higher prevalence of persistent wheeze 10. One study found that healthcare use for asthma in adults was associated with traffic volume 11. In a recent study, Mc Creanor et al. 12 reported that exposure to road-traffic air pollution levels was associated with a decrease in lung function and with an increase in biomarkers of neutrophilic inflammation in asthmatics adults. Air pollutants may amplify the inflammatory reactions in the airways but, in addition, promote allergic disease in asthmatics 13–15. Pollutant-induced oxidative stress could promote airway inflammation and, therefore, hyperresponsiveness, which may be one path to the development of asthma.

The European Community Respiratory Health Survey (ECRHS) has found a positive association between NO₂ and asthma incidence in adulthood 16. In that analysis, “asthma incidence” was defined in the traditional way, assuming asthma to be a well-defined dichotomous phenotype with a specified time of onset (“doctor-diagnosed asthma”). However, as previously promoted, asthma can be defined as a continuous trait using a grading scheme based on reported symptoms 17, 18. Markers of asthma severity and major risk factors of asthma showed a significant linear association with this novel asthma score. Using a continuous measure of a complex disease has some appeal as it improves the power to identify risk factors. The score is also not affected by diagnostic practices, which may vary grossly between regions and also over time within the same area 18. The aim of our study was to explore the association between the asthma score and traffic-related air pollution estimated at follow-up. The concept of asthma taken as a continuous trait is rather novel and has not yet been explored in air pollution epidemiology, thus, the utility and possible interpretations of the score in this field will be discussed.

METHODS

RESULTS

The study includes 4,394 participants with both data on asthma and modelled NO₂ concentration, used as a marker of traffic-related ambient air pollution. The distribution of the asthma score was skewed: 70% of the study participants reported none of the asthma symptoms and scored 0. The mean score was 0.52 and the standard deviation 0.98. Table 1⇓ shows the distribution of the score and NO₂ by city ordered from northernmost to southernmost.

Table 2⇓ describes the distribution of the covariates. Of the participants, 55% are females. The distribution of the covariates between ECRHS I and II did not vary except for smoking, with a decrease of smokers in ECRHS II.

The crude association between the covariates and the score is shown in table 3⇓. Season and any exposure to gas, dust or fumes at work were not associated with the score. The RMS for each increase of 10 μg·m⁻³ of NO₂ is 1.15 (95% CI 1.05–1.25).

In the multivariate analysis, the RMS for each increase of 10 μg·m⁻³ of NO₂ was 1.23 (95% CI 1.09–1.38; table 4⇓).

The association was homogeneous among centres (p = 0.169); after excluding Turin, which had very large confidence intervals, the p-value for heterogeneity was still not significant (p = 0.244) (fig. 1⇓). We did not observe heterogeneity across the three geographical areas (p-value of homogeneity 0.947) (fig. 2⇓).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1—

Adjusted ratio of mean asthma score (RMS) comparing the effect of NO₂ per centre (after excluding Turin). □: RMS per centre, where the size of the box is proportional to the sample size of each centre; –––––: 95% CI of the respective RMS; ⋄: combined RMS.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2—

Adjusted ratio of mean asthma score (RMS) comparing the effect of NO₂ per geographical area. □: RMS per region, where the size of the box is proportional to the sample size of each region; –––––: 95% CI of the respective RMS; ⋄: combined RMS.

After stratifying for baseline characteristics, all the associations were always positive, but the RMS was higher in males, nonasthmatics, atopics, participants with BHR and in nonsmokers and ex-smokers (table 5⇓). However, only the p-value for interaction for smoking was significant (RMS in nonsmokers and ex-smokers: 1.30, 95% CI 1.11–1.52; RMS in smokers: 1.07, 95% CI 0.92–1.26; p-value for interaction: 0.005).

In participants with no asthma and no symptoms at baseline, the associations between NO₂ and asthma score were positive (RMS 1.25, 95% CI 1.05–1.50). When comparing participants reporting none or one symptom with those reporting a score ≥2, the effect estimates increased (RMS 1.45, 95% CI 1.03–2.05). The association became even higher when considering participants reporting a score ≥3 (RMS 2.57, 95% CI 1.31–5.04; table 5⇑).

DISCUSSION

We observed significant associations between traffic-related air pollution and the asthma score, with similar findings across Europe.

To put the findings into context, we refer to table 4⇑, which shows that a rather modest change in NO₂ of only 10 μg·m⁻³ resulted in a similar association to that between the asthma score and being a current smoker. A contrast of 20 μg·m⁻³ (a difference well within the range of ambient concentrations within European cities) was associated with a difference in asthma score as strongly as a family history of asthma or atopy, which may be a good proxy for genetic factors. Our estimates are not easily comparable with other studies, as the use of this particular score is rather novel and the ratio of the mean score estimated with negative binomial regression is not directly comparable with the odds ratios obtained with logistic regression. Nevertheless, our estimates for well-known risk factors of asthma symptoms, such as smoking and a family history of asthma or atopy, are in the same range as those described in the literature. In a study among young adults, the odds ratio for asthma-like symptoms varied from 1.74 (shortness of breath) to 3 (wheezing) for smoking 28. In a study conducted in Northern Europe, the odds ratio of presenting wheezing in the last 12 months ranged from 1.5–3.6 depending on the intensity of smoking. In the same study, the odds ratio for wheezing in the last 12 months was 2.24 for family history of asthma 29. Regarding air pollution, an increase of almost 20 μg·m⁻³ in NO₂ levels was moderately associated with tightness in the chest (OR 1.11) and wheeze (OR 1.11) in an Italian study 9. In a Swedish study, NO₂ values >19 μg·m⁻³ measured at the address were associated with a moderate increase in asthma symptoms (OR 1.21) 30. Our air pollution estimates seem higher than those previously described, but the association between air pollution and asthma symptoms in adults has not been frequently studied and comparability with the score is limited.

The score comes with several advantages as compared with a dichotomous definition of asthma. As previously shown, it may be a valuable instrument to reduce misclassification bias due to dichotomisation of asthma 17, 18. The use of a continuous measure increases the power to detect risk factors, a notion supported by our findings. The use of a score is a novel approach not only to characterise asthmatic symptoms, but also as an approach to “asthma severity” and to investigate related risk factors 17, 18. Furthermore, the incidence of doctor-diagnosed asthma is relatively low, requiring large studies, and the time of onset of asthma is in fact hard to define and subject to diagnostic attitudes of physicians. Moreover, asthma is a complex phenotype and the clinical severity (or symptoms) is indeed relevant for patients. Thus, to study associations between risk factors and a complex disease with all these features, the use of continuous traits is appealing. Asthma is a disease that is probably the result of a long process and it is neither easy to define the presence or absence of the disease at some definite point in time nor to identify the time of “onset” 31. These difficulties could lead to misclassifications; the use of the symptoms-based score resolves some of the challenges faced with a dichotomous definition of asthma.

In air pollution epidemiology, the continuous score is appealing as associations with air pollution are usually low across the observed ranges of pollution, thus power is limited for disease incidence (i.e. it is either present or not). Resolving the problem of geographical variation in diagnostic patterns encountered with doctor-diagnosed asthma is particularly attractive in air pollution research where the pooling of studies, cities and countries is appealing not only to increase the power but to also investigate susceptibility factors.

However, the disadvantage of the score lies in the challenges of its interpretation, which may be the subject of continued debates, in particular in the context of air pollution research. The question emerges, as to whether the score, as used in our analyses, reflects acute effects of air pollution or its long-term chronic contribution to the underlying pathology leading to asthma incidence. We will discuss these two perspectives in more detail.

On the one hand, when asking about symptoms experienced during the last 12 months in a cross-sectional survey of the general population, the score may reflect the cumulated “acute conditions” triggered by air pollution. Among asthmatics, symptoms reflect the inherently variable phenotype of asthma; thus, the score may capture the course and severity/control of asthma, and our findings may be interpreted as a cumulative summary (last 12 months) of acute effects of air pollution on asthma symptoms, which may be a proxy of asthma severity. In fact, in a recent study investigating the cross-sectional relationships between asthma severity and background annual concentrations of air pollutants 32, this score was used as a marker of the activity of asthma along with another asthma severity score, both being assessed for the last 12 months. Both scores were associated with ambient home outdoor ozone concentrations, whereas no association was observed with NO₂. This inconsistency may, in part, be explained by the large spatial scale of the NO₂ model used in that analysis (grids of 4 km²) and, thus, spatial heterogeneity may not have been adequately captured.

With this interpretation of “cumulative summary”, our findings are in line with the body of evidence suggesting the acute effects of air pollution 1. The score may, thus, be used efficiently in future studies to evaluate whether acute effects of pollution on asthma performance may change in response to changes in treatments and/or in the composition of pollutants due to changes in engine technology, fuel formulation or the composition of the vehicle fleets.

On the other hand, the asthma score may also be considered as a tool to identify the incidence of asthma of various levels of severity, if we accept the notion of asthma being a continuous trait. With this interpretation, results ought to be compared with traditional “incidence studies”, such as our own analysis of the ECRHS data. In that previous work, we used the dichotomised definition of new onset of doctor-diagnosed asthma. An increase of 10 μg·m⁻³ of NO₂ was associated with asthma incidence (OR 1.43, 95% CI 1.02–2.01) 16. The use of the dichotomous definition, however, presented limitations that could be overcome with the score. In adults, the definition of “asthma” remains a challenge and the score clearly complements attempts to better understand the aetiology of this disease, independent of secular trends in the labelling of “asthma” by the community of physicians. In a previous publication within the ECRHS, Chinn et al. 33 have shown that, from ECRHS I to II, the prevalence of asthma increased while the prevalence of symptoms did not, suggesting a change over time in the diagnosis or treatment of asthma. The results shown here using the asthma score are in the same direction as those we found in the previous analysis. Given the inherent variability in symptom occurrence in both asthmatics and nonasthmatics, the mere “onset” of only one symptom among those free of symptoms and asthma at baseline is unlikely to reflect incidence of asthma. This notion may be more acceptable among those presenting a high score at follow-up. Our last analysis was restricted to participants with neither asthma nor symptoms at baseline. The particularly strong effects among those presenting at least two symptoms at ECRHS II would be in line with the interpretation of an effect of pollution on onset of asthma in adults. This “incidence interpretation” of the score comes with a caveat, as Sunyer et al. 18 have shown substantial changes in the score, with many participants losing or gaining one or more symptom. As a consequence, it may be questionable to consider those with a change from no disease at baseline to a very low asthma score as “incident cases”. However, a “high score” phenotype has been shown to be particularly strongly associated with doctor-diagnosed asthma 18; thus our related finding may indicate a role of pollution in new onset of asthma in adults. This interpretation is consistent with our previous finding based on the more traditional definition of “asthma incidence”, using asthma at follow-up among those free of the disease at baseline 16. That latter approach is less ambiguous in the interpretation regarding acute and chronic effects as a “doctor's diagnosis” does reflect a “chronic condition” rather the accumulated acute expression of the disease. Further research on the use and interpretation of the score as a measure of the course, and also the incidence, of this chronic disease will be useful to enhance our aetiological understanding of asthma.

We characterised exposure to pollution for approximately the same period as the assessment of the score. While a strength in the assessment and interpretation of “acute effects”, the availability of air pollution measurements only for the follow-up period is a limitation that may affect the incidence interpretation of this analysis. Pollution levels may have changed differently across cities and participants who moved to another residence would have inherently larger misclassification of exposure. Although trends in air pollution are often spatially correlated across areas, this simplification might be less true across the very large European geographical area. However, based on network data available in several cities, one can assume that changes in air quality differed across European areas, which may bias toward null findings.

It is important to note that the association between NO₂ levels and the asthma score were rather homogenous for all the cities included in this analysis. The homogeneity was even stronger when taking into account geographical area instead of city. This supports a causal interpretation of our findings.

We found a stronger effect of NO₂ in nonsmokers and ex-smokers compared with smokers. Smoking might be considered as an effect modifier, but previous studies have not been consistent and it remains unclear whether smokers are more or less susceptible to the effects of air pollution than nonsmokers. In the Swiss Air Pollution And Lung Disease In Adults (SAPALDIA) study, Zemp et al. 7 found that smokers reported more asthma-like symptoms than nonsmokers, suggesting an effect modification by smoking status. It was also suggested, however, that smokers have altered lung function and an increase in mucus, which decreases the air pollutant amount in some regions of the lungs, thereby decreasing the susceptibility of smokers to air pollution 34.

The strengths and the limitations of the exposure assessment have been discussed previously 16; by geocoding home addresses of ECRHS participants we were able to assign an ambient NO₂ concentration derived from the APMoSPHERE map to each participant. Nevertheless, it is also important to take into account that the APMoSPHERE map has a spatial resolution of 1 km², which may not capture all the variability, especially in cities with high population density.

In conclusion, NO₂ is associated with an asthma score suggesting that traffic-related air pollution causes asthma symptoms in adults. The stronger effects among those free of asthma and symptoms at baseline may also indicate a role of pollution in the onset of asthma. This needs further investigation. The use of the asthma score offers very attractive alternatives to investigate the aetiology of asthma and the course of this disease in adults.

Support statement

B. Jacquemin benefits from an Inserm post-doctoral fellowship (“poste vert”). The co-ordination of ECRHS II was supported by the European Commission as part of their Quality of Life programme. A full list of bodies that funded the local studies in ECRHS II included in this paper can be found in the Acknowledgements section.

Statement of interest

None declared.