Discussion

In this cohort of participants followed from 43 to 53 years of age, based in Australia where pollution levels are generally low, we observed that in males, higher exposure to ambient air pollution was associated with increased risk of prevalent eczema (both atopic and non-atopic). In contrast, for females, higher exposure to NO2 was associated with a paradoxical protective effect of prevalent eczema (both atopic and non-atopic). Additionally, there was evidence that increased levels of PM2.5 exposure were associated with increased risk of allergic sensitisation, in both males and females.
There are a limited number of previous studies that have considered the association between ambient air pollution exposure and eczema prevalence or incidence, and the results were inconsistent. Two paediatric studies19,20 showed no association between ambient air pollution and increased risk of eczema, while other studies suggested increased eczema risk from ambient air pollution in children7 and adults.21,22 A study by Kim et al23 found similar results in that NO2 was associated with the prevalence of AE in male but not in female, however the study population were children. The reason for these sex specific effects is not clear, but may be partly be due to differences in skin morphology, occupational exposures in adults and behaviour.24As has been reported previously in this age group,25 women tend to spent more time indoors compared to men, with men reported spending almost twice the amount of time outside compared to women (supplementary file 9). Additionally, women are more likely to care about skin issues and avoid exposure to irritants compared to men.26 Therefore, residual confounding may have introduced the sex-specific interaction in the association estimates. On the other hand, Gilmour et al.27 observed that not all oxidative stress responses on the epithelial barrier are injurious and a lower level of oxidative stress might be paradoxically protective. To further elucidate the possible reasons for these sex specific effects, we recommend focused exploration of the physiological, barrier function and immune responses to ambient air pollution at low levels in men and women.
We were unable to replicate the results from the SALIA cohort of elderly women study.6 However, we saw similar non-significant trends of associations between ambient air pollution markers at follow-up and NAE prevalence which may warrant further investigation. There are several reasons why we may not have observed the same associations. First, we used SPT results and hay fever rather than blood IgE levels and hay fever, to determine atopic sensitisation. However given the strong association between SPT and IgE28this would be unlikely to explain the differences in results between these studies. Second, our cohort was younger (baseline at 43 years followed-up to 53 years) than those in SALIA (baseline at 53 years followed-up to 73 years of age), making this a different time in women’s reproductive lives, and changes in sex hormone levels may help explain the differences in results seen between these studies. Third, we included both sexes in our main analyses, whereas the SALIA study included only women. However, we did not observe increased risk in women with higher ambient air pollution exposures. Finally, the sources of ambient PM2.5 16 are different between Australia (mainly from wood heaters, power stations and off-road sources) and Germany (mainly on-road traffic) where the study was conducted.6
Our findings suggest an association between PM2.5 and aeroallergen sensitisation, agreeing with other studies.29,30 In a study with adult participants; living close to busy roads was associated with a higher risk of sensitisation to pollen.29 Furthermore, a previous cross-sectional analysis using data from this cohort reported, in this low pollution setting, that increased levels of ambient air pollution conferred an increased risk of atopy.30 Of two studies that have not observed an association between ambient air pollution and an increased risk of allergic sensitisation, one study had a relatively small sample size and low power.31 Another, in an adult population, reported a cross sectional association with DMR and NO2 and aeroallergen sensitisation, but did not assess PM2.5.32
It has been proposed that air pollutants may lead to eczema and sensitisation via inflammatory oxidative stress leading to skin barrier dysfunction.3 Ambient air pollution may drive these effects either through direct percutaneous absorption or indirectly through inhalation and subsequent systemic inflammation.3 These air pollutants produce reactive oxygen species (ROS) and nitrogen species that lead to damage of proteins, lipids, and DNA.33 Air pollutants can also act as irritants and immunomodulators leading to elevated levels of serum IgE.34 Specifically, PM2.5 may activate the aryl hydrocarbon receptor to promote cell metabolism and inflammation.34 Other proposed mechanisms are by altering trans-epidermal water loss, increasing inflammatory signals and modifying the skin pH and microbiome.35 Our results support an effect of ambient air pollution on immune function, even in this low ambient air pollution setting, with relatively higher exposure levels being associated with eczematous skin symptoms in males and overall increased aeroallergen sensitisation risk in the sample population.
Our study has both strengths and limitations. Strengths include access to a large population-based prospective cohort study with long follow-up allowing for a 10 year assessment period, well-characterized definitions of eczema with objective measures of SPT and land-use regression models. While the questions and definitions used were well validated,36 a limitation is the reliance on self-reporting. Unfortunately, there were no data on the frequency or severity of symptoms and the aero-allergen SPT data were only available at follow-up, which may lead to some misclassification of atopy in the eczema subgroups. It is possible that residual confounding could have been an issue. As such, further replication of these findings is required. Given the exploratory nature of these data with multiple associations being assessed, we have attempted to interpret the pattern of associations, rather than relying on any arbitrary p value threshold to draw conclusions. Finally, as almost all the participants were Caucasian and Anglo-Celtic, the findings may not be generalizable to other ethnicities.