Childhood Exposure to Ambient Air Pollutants and the Onset of Asthma: An Administrative Cohort Study in Québec

Background: Although it is well established that air pollutants can exacerbate asthma, the link with new asthma onset in children is less clear. Objective: We assessed the association between the onset of childhood asthma with both time of birth and time-varying exposures to outdoor air pollutants. Method: An open cohort of children born in the province of Québec, Canada, was created using linked medical–administrative databases. New cases of asthma were defined as one hospital discharge with a diagnosis of asthma or two physician claims for asthma within a 2 year period. Annual ozone (O3) levels were estimated at the child’s residence for all births 1999–2010, and nitrogen dioxide (NO2) levels during 1996–2006 were estimated for births on the Montreal Island. Satellite based concentrations of fine particles (PM2.5) were estimated at a 10 km × 10 km resolution and assigned to residential postal codes throughout the province (1996–2011). Hazard ratios (HRs) were assessed with Cox models for the exposure at the birth address and for the time-dependent exposure. We performed an indirect adjustment for secondhand smoke (SHS). Results: We followed 1,183,865 children (7,752,083 person-years), of whom 162,752 became asthmatic. After controlling for sex and material and social deprivation, HRs for an interquartile range increase in exposure at the birth address to NO2 (5.45 ppb), O3 (3.22 ppb), and PM2.5 (6.50 μg/m3) were 1.04 (95% CI: 1.02, 1.05), 1.11 (95% CI: 1.10, 1.12), and 1.31 (95% CI: 1.28, 1.33), respectively. Effects of O3 and PM2.5 estimated with time-varying Cox models were similar to those estimated using exposure at birth, whereas the effect of NO2 was slightly stronger (HR = 1.07; 95% CI: 1.05, 1.09). Conclusions: Asthma onset in children appears to be associated with residential exposure to PM2.5, O3 and NO2. Citation: Tétreault LF, Doucet M, Gamache P, Fournier M, Brand A, Kosatsky T, Smargiassi A. 2016. Childhood exposure to ambient air pollutants and the onset of asthma: an administrative cohort study in Québec. Environ Health Perspect 124:1276–1282; http://dx.doi.org/10.1289/ehp.1509838


Supplemental Material
Childhood Exposure to Ambient Air Pollutants and the Onset of

Table of Contents
Indirect adjustment for second hand smoke Table S1. Distributions of estimated annual average concentrations of PM 2.5 and O 3 at both the annual and the birth address in the Montreal sub-cohort Table S2. Associations between asthma onset and air pollutant levels in Quebec, per increase of an interquartile range in air pollutant levels, without regions of the province where health services may be under-reported Table S3. Associations between asthma onset and time-varying air pollutant levels, per increase in interquartile range, restricted to non-movers Table S4. Associations between asthma onset and air pollutant levels, per increase of an interquartile range in air pollutant levels, with reconfirmation of onset when it occurred before the age of five Table S5. Associations between asthma onset and air pollutant levels, per interquartile range increase in air pollutant levels, stratified by sex Table S6. Associations between asthma onset and air pollutant levels in Quebec, per increase in interquartile range of pollutant levels, stratified by region Figure S1. Distribution of the potential bias introduced by second hand smoke on the association between onset of asthma on the island of Montreal and air pollutant levels of: NO 2 (for the years 1996 to 2006) B) PM 2.5 (for the years 1996 to 2006) and C) O 3 at the birth address (for the years 1999 to 2006)

Indirect adjustment for second hand smoke
Individual information on secondhand smoke (SHS) exposure where not available in this study, thus we could not controlled for this potential confounder. However we performed an indirect adjustment for SHS for the Montreal sub-cohort by using a strategy proposed by Steenland and  quintiles. For these quintiles of exposure, we estimated the proportion of children exposed at home to SHS. Area-specific (i.e. postal code) prevalence of at home childhood exposure to SHS was retrieved from a survey conducted in 2006 in Montreal (Deger et al. 2010). We also retrieved from a meta-analysis (Tinuoye et al. 2013) a rate ratio representing the association between childhood asthma and SHS. The bias associated with SHS for each quintile was then estimated using the following formula for indirect adjustment (Equation 1): where P e,q is the prevalence of exposure to SHS at quintile q, P e,l is the prevalence of exposure to SHS at the lowest quintile and RR e is the rate ratio of childhood asthma associated with exposure to SHS. I 0 in this equation is the incidence rate of asthma among children unexposed to SHS.
However this equation could be simplified to discard the I 0 term. As proposed by Villeneuve et al. 2011, we performed a linear regression in order to present estimation of indirectly adjusted point estimates for a continuous scale. In this model, the estimated bias for each quintile (Bias q ) was considered as the dependent variable whereas the independent variable was a random sample of a uniform distribution of air pollutant levels per quintile. The slope of this regression represents an estimation of the potential confounding induced by SHS exposure. Finally, the 4 indirectly adjusted HRs were calculated by dividing the Montreal sub-cohort HRs, adjusted for sex and the Pampalon deprivation index, by the exponentiation of the bias slope.
In order to compute the uncertainty around our indirectly adjusted point estimates, we used Monte Carlo sampling with 100,000 replications to repeatedly sample from distributions of rate ratios as well as prevalence exposures to SHS. The distribution the rate ratio linking SHS and childhood asthma was based on data from a meta-analysis (n=20) (Tinuoye et al. 2013), we assumed a normal distribution with a mean value equal to the natural logarithm of the rate ratio and standard deviations equal to the standard errors. Identical assumptions were made to construct the distribution around our association between childhood asthma and outdoor air pollutants that were derived from this study. The distribution of the prevalence exposure to SHS was assessed using data from the aforementioned survey (Deger et al. 2010). We presumed a normal distribution with a mean equal to the logit of the proportion of exposed. Adjusted HRs and their 95% Monte Carlo confidence interval were calculated for each pollutant.