Methods

Study population

The Tasmanian Longitudinal Health study (TAHS) is a population-based prospective cohort study, that has followed participants since 1968 when 7-year-old children (98.7%, n=8583) attending schools in the Australian state of Tasmania were recruited.10 Several follow-up surveys have subsequently been conducted and study methodology has been reported in detail elsewhere.10 The data for this analysis came from participants of the 2002 and 2012 proband studies when participants were 43 and 53 years old respectively. Participants completed a self-administered postal survey that collected the following baseline (2002) and follow-up (2012) data: sociodemographic characteristics, occupation, residence, health service use, medical diagnoses, smoking, reproductive histories, and symptoms. Only participants who completed both assessments and had valid skin prick tests (performed at the 2012 follow-up) were included in the analysis. The study was approved by the Human Research Ethics Committee of the University of Melbourne; all participants provided written informed consent.

Exposures

Distance to a major road (DMR)

Straight-line distances from each participant’s residence to the nearest major road in 2002 and 2012 proband studies were calculated using ArcGIS 10.1 software (Redlands, CA). Major roads were defined using public sector mapping agencies, and Australian transport hierarchy codes 301 and 302.11 Participants were categorized into two groups: (i) living <200m; or (ii) living ≥200m from a major road. Major traffic pollutant concentrations tend to decay as the distance to major roads (DMR) increases, with most components of TRAP reaching near background concentrations at approximately 200m.12

Nitrogen dioxide (NO2)

A satellite-based land-use regression model was used to assign mean annual NO2 exposures for the 2002 and 2012 proband studies.13 Briefly, the land-use regression model-predicted mean annual NO2 levels were based on tropospheric NO2 columns derived from satellite observations in combination with other predictors such as land use and roads, to estimate ground-level NO2 across Australia.13 As more than half of the ambient NO2 is attributed to on-road sources, NO2 is a reasonable proxy for TRAP.14The model’s development and validation are described in detail elsewhere, and it explained 81% of spatial variability in measured annual NO2 at all regulatory monitoring sites in Australia.13 Mean annual residential exposures to outdoor NO2 were estimated and assigned based on participants’ geocoded addresses at baseline (2002 proband study) and follow-up (2012 proband study).

Fine particulate matter with an aerodynamic diameter <2.5 µm (PM2.5)

The methods are explained in more detail elsewhere.15In brief, satellite-based estimates for Australia of ground-level PM2.5 were used as a land-use regression predictor, with other spatial predictors of PM2.5 . This model explained 63% of spatial variability in measured annual PM2.5(RMSE: 1 µg/m3).15 The mean annual residential exposures to outdoor PM2.5 were estimated and assigned based on participants’ geocoded addresses at baseline and follow-up. In Australia, traffic-related sources of PM2.5 are estimated to account for only 17% of ambient PM2.5 mass, while the majority of ambient PM2.5 is from other anthropogenic sources (i.e. wood heaters, power stations and non-road combustion).16

Outcomes

Prevalent eczema at 53 years

Prevalent eczema at age 53 (2012 proband study) was determined using the International Study of Asthma and Allergies in Childhood (ISAAC) definition of eczema.17 Participants were classified as having prevalent eczema if they reported “yes” to all three questions: ‘have you had an itchy rash in the past 12 months?’, ‘Have you ever had an itchy rash coming and going for at least six months?’, and ‘Has this itchy rash at any time affected any of the following places: the folds of the elbows, behind the knees, in front of the ankles, under the buttocks, or around the neck, ears or eyes?’

Incident current eczema at 53 years

Incident current eczema at 53 years was defined as eczema newly arising between the two proband studies i.e. between 43 and 53 years. The participants were classified as having Incident eczema if they answered “no” to “Have you ever had eczema or any skin allergy?” at baseline (2002 proband study), but reported eczema based on the ISAAC definition17 and having eczema for the first time after baseline at “How old were you when you first had this itchy rash?” at the follow-up (2012 proband study.

Persistent current eczema

Persistent current eczema was defined as prevalent eczema at baseline that persisted to follow-up. Participants were classified as having persistent current eczema if they answered “yes” to “Have you ever had eczema or any skin allergy?” at baseline and reported eczema based on the ISAAC definition17 at follow-up.

Atopic status

Subclassification as AE or NAE was based on skin prick testing (SPT) results at age 53 years. 10 In the 2012 proband study, SPTs were performed for eight aeroallergens: Dermatophagoides pteronyssinus , cat pelt, Cladosporioides , Alternaria tenuis , Penicillium mix, Aspergillus fumigatus , mixed grass pollen No. 7 (which included Kentucky bluegrass, orchard, redtop, Timothy, sweet vernal grass, meadow fescue and perennial ryegrass). Histamine was used as the positive control and normal saline as the negative control. After 10 to 15 mins, the wheal diameters were measured in two perpendicular directions in millimetres and an average was derived. A valid SPT was determined by a positive control or allergen wheal equal to or greater than 3 mm in size and a negative control wheal equal to or less than 3 mm in size. A positive SPT was defined as a wheal size of at least 3 mm greater than the negative control and was considered to indicate sensitisation to that allergen.18 Atopy was defined as sensitization to at least one of the allergens tested.

Statistical Analysis

Associations between markers of ambient air pollution and the following outcome measures were assessed: 1) Prevalent eczema, 2) Incident current eczema, 3) Persistent eczema, 4) Prevalent and incident eczema sub grouped by atopic status (neither, atopy alone, non-atopic and atopic eczema) and 5) Sensitisation (regardless of eczema status). Second, in accordance with the eczema and atopy classification used in the SALIA cohort study 6, we examined NAE incidence and prevalence using three increasingly restricted subgroups: 1) all participants, 2) participants without hay fever ever, 3) participants without hay fever ever and negative SPT.
Logistic regression and multinomial models were fitted to estimate the associations between baseline ambient air pollution and each outcome. The coefficients represented the estimated effect per interquartile range (IQR) increase of air pollutant exposure and were expressed as odds ratios (ORs) with 95% confidence intervals (CIs). A directed acyclic graph (DAG) (supplementary figure 1) was developed to specify the hypothesized causal relationships and to determine which confounders to include in the model (supplementary table 1). The potential presence of non-linearity of these associations was assessed using Stata’s “fracpoly” command; no evidence of non-linearity of associations between ambient air pollution markers and eczema was identified. Potential effect modification by sex was explored using likelihood ratio tests and a p value < 0.1 was considered as significant. A sensitivity analysis was performed, where associations were assessed only in the participants who did not change residential address during follow-up period. All analyses were carried out using the statistical software Stata (release 16; Stata Corporation, College Station, TX).