The aim of this study was to investigate the association between traffic-related air pollution and asthma and COPD in adults. The outcomes investigated were prevalence of; 1) asthma diagnosis 2) COPD diagnosis 3) asthma symptoms last 12 months, and 4) chronic bronchitis symptoms, in relation to residential traffic exposure.

The study area was the most southwestern part of Sweden (figure 1), the most populated part of the county of Scania. The study area contains 840 000 of Sweden's total population of 8.9 million, and has a population density of 170 inhabitants per km² (data from 2000). The majority of the population live in six of the communities, the largest of which is Malmö, the third largest city in Sweden, with a population of 260 000. A detailed regional description has previously been given 27. In the geographical stratification of the present study, "Malmö" refers strictly to the city boundaries of Malmö, not the larger municipality.

The climate in the region is homogenous. Although pollutant levels in the region are low in an European context, they are higher than in the remainder of Sweden 28, due to long-range transport of pollutants from the continent and extensive harbor and ferry traffic.

In 2000, a questionnaire was sent to a total of 11933 individuals aged 18–77, of whom 9319 (78%) answered. The study population originated from two different study populations, with 5039 (response rate: 71%) from a new random selection, and 4280 (response rate: 87%) constituting a follow-up group from an earlier selection 29.

The questionnaire dealt with respiratory symptoms, potential confounders such as smoking habits and occupation, and exposures such as living close to a road with heavy traffic 29. An external exposure assessment was also obtained by geocoding the residential addresses (as of 2000) of both respondents and non-respondents. This was achieved by linking each individual's unique 10-digit personal identity codes to a registry containing the geographical coordinates of all residential addresses.

Non-respondents had a higher mean of NO_x than respondents; 14.7 μg/m³ versus 13.5 μg/m³. To a large extent this was due to a lower response rate in the more polluted city of Malmö (73% vs. 80% in the remaining region).

The following outcomes were investigated, as obtained by the postal questionnaires:

• Diagnosis of asthma. "Have you been diagnosed by a doctor as having asthma?"

• Diagnosis of COPD/CBE (Chronic Bronchitis Emphysema). "Have you been diagnosed by a doctor as having chronic bronchitis, emphysema, or COPD?"

• Asthma symptoms during the last 12 months. "Have you had asthma symptoms during the last 12 months, i.e. intermittent breathlessness or attacks of breathlessness? The symptoms may exist with or without cough or wheezing."

• Chronic bronchitis symptoms. "Have you had periods of at least three months where you brought up phlegm when coughing on most days?", and if so, "Have you had such periods during at least two successive years?"

The questionnaire has been published previously 29. No information regarding year of disease onset was available.

Exposure to traffic-related air pollution was assessed at each participant's residential address in 2000, using three different proxies:

1. Self-reported exposure to traffic. This was obtained from the survey. Exposure was defined as a positive answer to the question "Do you live close to a road with heavy traffic?"

2. Traffic intensity on the heaviest road within 100 m. GIS-based registers from The Swedish National Road Database 30 provided information about traffic intensity for all major roads in the county (figure 2). To assess exposure to traffic, we identified the road with the heaviest traffic intensity within 100 m of the residence. Traffic intensity was categorized as 0–1 cars/min, 2–5 cars/min, 6–10 cars/min, and >10 cars/min, based upon 24-hour mean levels.

3. Modeled exposure to NO_x (figure 3). Annual mean concentrations of NO_x were acquired from a pollutant database, based on the year 2001 2531. Emission sources included in the model were: road traffic, shipping, aviation, railroad, industries and larger energy and heat producers, small scale heating, working machines, working vehicles, and working tools. Meteorological data were also included. A modified Gaussian dispersion model (AERMOD) was used for dispersion calculations; a flat two-dimensional model which did not adjust for effects of street canyons or other terrain, but which did take the height of the emission sources into consideration. Concentrations of NO_x were modeled as annual means on a grid with a spatial resolution of 250 × 250 m. Bilinear interpolation was used to adjust individual exposure with weighted values of neighboring area concentrations. Concentrations modeled with this spatial resolution have been validated and found to have a high correlation with measured values in the region 2531.

A categorical classification of NO_x was used in order to allow analysis of non-linear associations with outcomes. To determine the category limits, the subjects (n = 9274) were divided into NO_x-quintiles. The five exposure groups used were 0–8 μg/m³, 8–11 μg/m³, 11–14 μg/m³, 14–19 μg/m³, and >19 μg/m³.

For all measures of exposure, subgroup analyses were made for Malmö and the remaining region. Relative risk was not estimated in exposure groups with fewer than 50 individuals. As few individuals in Malmö had a low exposure to NO_x, the middle exposure group was used as the reference category for NO_x, in Malmö. Because of this, NO_x was also used as a continuous variable for trend analysis using logistic regression. A p-value < 0.05 was regarded as evidence of a trend. Stratified analyses were performed for sex, age, smoking, geographical region (Malmö vs. remaining region), and study population (new random selection vs. follow-up group). Sensitivity analyses of the associations with traffic were also performed by restricting the groups to those with asthma but not COPD, and COPD but not asthma, to exclude confounding by comorbidity. This process was also followed for symptoms.

Relative risk was estimated using Odds Ratios (ORs) with 95% Confidence Intervals (CI). Odds Ratios and tests of trends were obtained by binary logistic regression, using version 13.0 of SPSS.

Sex, age (seven categories), and smoking (smokers/ex-smokers vs. non-smokers) are known risk factors for asthma, and were therefore adjusted for in the model. Socio-Economic Indices (SEI codes, based on occupational status 32) and occupational exposure (ALOHA JEM 33) were tested as confounders, using the "change-in-estimate" method 34, where a change in the OR of 10% would have motivated an inclusion in the model.

Neither occupational exposure nor Socio-Economic Indices fulfilled the predetermined confounder criteria, or had any noticeable impact on the risk estimates, and were thus not included in the model.

Asthma diagnosis and asthma symptoms in the last 12 months were associated with self-reported traffic exposure (table 2). These results were consistent in a geographical stratification (tables 3, 4).

COPD diagnosis was associated with self-reported traffic exposure, both for the whole region (table 5) and when geographically stratified (table 6). Chronic bronchitis symptoms were not associated with self-reported traffic exposure (tables 5, 7).

Asthma diagnosis and asthma symptoms were associated with traffic intensity (table 2), with higher prevalence of asthma symptoms among those living next to a road with at least 6 cars/minute, and higher prevalence of asthma diagnosis among those exposed to at least 10 cars/minute, compared with the group having no road within 100 m. The effects seemed consistent, although statistically non-significant, across geographical region (tables 3, 4).

COPD and chronic bronchitis symptoms were associated with traffic intensity (table 5). However, when stratified geographically, the effect estimates indicated that chronic bronchitis symptoms were not associated with traffic intensity in Malmö (table 7).

Asthma symptoms, but not asthma diagnosis, were associated with NO_x in the trend tests (table 2). However, effects were only seen in the highest quintile of >19 μg/m³. A geographical stratification showed that it was only in Malmö that high exposure was associated with asthma; no association was found in the region outside (tables 3, 4).

COPD diagnosis and chronic bronchitis symptoms were associated with NO_x(table 5). After geographical stratification, associations were seen only in Malmö, and not in the region outside (tables 6, 7).

The major strength of this study was the use of three different proxies of exposure, which may have different intrinsic strengths and weaknesses. The strengths of the NO_x model are firstly that it reflects total traffic density in the area, and secondly the fact that the dispersion model has been validated, with a resolution of 250 × 250 m showing a high correlation with measured background concentrations 25. Nevertheless, street-level concentrations may vary on a much smaller scale. High peak concentrations are often found in so-called "street canyons" in urban areas, where pollutants are trapped between high buildings 23. Since the dispersion model did not take account of this kind of accumulation effect, the true exposure among people living in these surroundings might have been underestimated. This may partly explain why effects from NO_x were seen in the urban city of Malmö but not in the surrounding area.

The proportion of NO_x that originates from traffic is also dependent on geographical area. In urban areas of southern Sweden, local traffic contributes approximately 50–60% of total NO_x, while in the countryside such traffic is responsible for only 10–30% of total NO_x (S. Gustafsson, personal communication, 2007-10-17). This difference was also reported by the SAPALDIA study, which found that local traffic accounted for the majority of NO_x in urban but not rural areas 35. This indicates that our model of NO_x is a good proxy for exposure to traffic-related air pollution in an urban area, but may not be sensitive enough to capture individual risk in the countryside, where traffic contributes to a lower proportion of total concentrations.

Self-reported living close to a road with heavy traffic, and traffic intensity on the heaviest road within 100 m, are simple proxies; they do not reflect, for example, whether someone lives at a junction. Still, they have the advantage that they are less limited by aggregation in space than the NO_x model. In the present study, both of these variables showed fairly consistent associations, at least with asthma, despite large differences in the level of NO_x that they corresponded to in Malmö and the region outside (table 10); this may indicate that adverse effects from traffic pollutants are mainly seen in close proximity to traffic. High traffic intensity, however, may not only correlate with high total number of vehicles, but also with a higher proportion of heavy vehicles, an additional factor which could affect the outcome, since diesel exhaust from heavy vehicles might have more adverse respiratory effects 36.

It should be noted that the distribution of exposure is not comparable between the proxies. While NO_x was divided into quintiles, with 20% in the highest exposure category, only 6% of the population lay in the highest traffic intensity category. Thus, the different proxies are complementary rather than comparable in this study.

One limitation of all three proxies of exposure was that traffic-related air pollution was only estimated by residential address. Lack of individual data about work address and time spent commuting could have biased the exposure assessments, particularly for people living in areas with low exposure to traffic-related air pollution, where individual differences in daily activities outside the residential area translate to a large proportion of total exposure 37. In Los Angeles, 1 h commuting/day contributes approximately 10–50% of people's daily exposure to ultrafine particles from traffic 38. While only 20% of the working population living in Malmö commute to work outside Malmö, the majority of the population in smaller municipalities in the remaining region commute to work outside their own municipality 39.

Another limitation was the cross-sectional nature of the study; we had no information about disease onset or years living at current address, making it hard to establish a temporal relationship between cause and effect. However, since asthma and COPD are known to be exacerbated by traffic-related air pollution, subjects with disease may have been more likely to move away from traffic, rather than towards it, and so a migrational bias would mainly be expected to dilute the effects.

Areas with high levels of exposure to traffic-related air pollution were mainly located in the city of Malmö (table 4 and figure 2), while low exposure was found in more sparsely populated areas. It is a well recognized problem that the different exposure levels in rural and urban environments are also accompanied by large differences in lifestyle factors, and even if confounders are controlled for, unmeasured factors may remain. Since NO_x was limited by its spatial resolution, it is also the measure that was most susceptible to ecological bias. The lack of association seen with NO_x, in the region outside Malmö might thus reflect that the individual risk from traffic is being overridden by some other factor correlating with low exposure levels. The existence of independent risk factors correlating with low exposure is given some support by a Swedish study which found a tendency to higher adult asthma incidence in rural areas, after adjustment for exposure to traffic 11.

The most important risk factors from a validity standpoint, however, are factors that could correlate with high exposure to traffic-related air pollution, and thus cause a false positive relationship, such as socio-economic and occupational risk factors. However, the present study, which used individual-level data, found no confounding effects for either socio-economic status or occupational exposure. A recently developed and validated JEM was used to adjust for occupational exposure 33. In a JEM, people are assigned the statistically average level of exposure in their occupation; this is an aggregated form of exposure assessment that can suffer from misclassification bias, although non-differential to disease. Since we only had information on the participants' current occupations, we cannot rule out the possibility of a "healthy worker effect". Lack of information about occupational history may be a limitation, especially in relation to the prevalence of COPD/chronic bronchitis.