It is widely known that exposure to outdoor air pollution is associated with acute and chronic changes in expiratory lung volumes. However, most of these studies have assessed exposure to air pollutants using monitors that primarily measure background levels. While this method may be useful in estimating exposure to ozone, it may not provide an adequate estimate of exposure to other pollutants that display small area spatial variations such as ultrafine particulate matter, elemental carbon, polycyclic aromatic hydrocarbons and nitrogen dioxide.^1–5 Evidence that there is considerable spatial variability in the concentration of traffic related pollutants has sparked interest in assessing the health effects associated with vehicular emissions. Several studies have found that exposure to traffic related emissions are associated with higher rates of adverse respiratory health outcomes including increased respiratory symptoms,⁶ reduced lung expiratory volumes,⁷ increased airway inflammation⁸ and higher rates of healthcare utilisation.⁹ This phenomenon may be partly explained by the properties of the pollutant mix near vehicular emission sources such as a larger concentration of fine and ultrafine particles and higher carbon monoxide and nitrogen dioxide levels.^10,11 What level of proximity to vehicular emissions is associated with adverse health effects? Is there a boundary between exposure to vehicular emissions and background levels? Is there a traffic density threshold associated with health effects? Answers to these and other important questions are unknown, although most studies have found that the rate of adverse respiratory health effects appear to increase proportionately in relation to road proximity and with increasing traffic density.¹² However, whether or not close proximity exposure to vehicular emissions is more harmful than background air pollutants remains controversial as several studies have failed to show significant associations.^13–15

In this issue of Thorax Kan et al¹⁶ evaluated the association between traffic related exposures and changes in exhaled lung volumes in participants from the Atherosclerosis Risk in Communities (ARIC) cohort study, which included 13 972 subjects from four US counties with a mean (SD) age of 54.2 (5.8) years (see page 873). Lung function measures were obtained during visit 1 (1987–9) and visit 2 (2000). To determine traffic exposure, researchers used a geographic information system (GIS) to geocode the residential address of each participant and determined the distance to the nearest road (including road type) and traffic density using a decay function that approximates the potential influence of activity weighted mobile source emissions at varying distances from a roadway. The use of GIS related traffic variables to determine spatial variability of air pollutants has been validated in several studies.^17,18 Traffic density estimates were directly available for the year 2000 but were extrapolated for years 1987–92 using changes in population density. Exposure to background air pollutants was estimated by interpolating the distance of each participant’s residence to the corresponding Environmental Protection Agency monitoring stations.

Based on a positive interaction term between sex and traffic exposures, the authors presented their results in a stratified analysis. In women there was a significant trend for reduced forced expiratory volume in 1 s (FEV₁) and forced vital capacity (FVC) with increasing quartiles of traffic density. Women who resided within 150 m of a major road were also more likely to have reduced FEV₁ and FVC. No significant associations were observed in men. Modification of the association between traffic exposures and lung function by sex has not previously been described.

Several explanations could be offered to account for this unexpected finding. First, it could be that women are in fact more susceptible. Although there is no direct evidence to support this assertion, one study showed that women had more severe respiratory symptoms then men when exposed to air pollution from burning agricultural residue.¹⁹ Several studies have found that women may be more susceptible to the respiratory effects of cigarette smoke.²⁰ Further, in population studies of bronchial hyperresponsiveness, women exhibit a greater response to methacholine than men even when adjusting for sex differences in lung volume sizes.²¹

A second possibility may be related to different patterns of residential exposure to vehicular emissions. Because the study evaluated the exposure to traffic at the residence level and a larger proportion of women were more likely to stay at home, it is possible that exposure misclassification in men could account for the lack of significant associations in this group.

Finally, the possibility for residual confounding is a significant concern, especially in cross-sectional studies evaluating exposure to vehicular traffic emissions. Lower socioeconomic position has been associated with road proximity (it was also in this study) and with increased vulnerability to the effects of air pollutants.²² Although the trend of traffic density quartiles and lung function remained significant after adjusting for confounders, the individual beta coefficients in each quartile of traffic density were attenuated. Furthermore, the multivariable models of the association between proximity to major roads and FEV₁ and FVC were not significant. These results are evidence of the important influence of confounding by socioeconomic position in cross-sectional studies that use traffic related measures as the exposure.

In summary, using traffic density and road proximity as proxy measures of exposure, Kan et al¹⁶ have shown that vehicular emissions are associated with chronic lung function changes in a large sample of adult subjects. A study of this magnitude, evaluating the respiratory effects of traffic related exposures in adults, has not been done previously and is a welcome addition to the field. Whether or not women are more susceptible remains to be elucidated and should be determined in a longitudinal study.