Air Pollution and Chronic Lung Allograft Dysfunction

Methods

Patients

The cohort consisted of all bilateral lung transplant recipients from 1996 to 2009 who were permanent residents of Ontario, Canada, at the time of transplantation, survived more than 3 months posttransplant and had full pulmonary function data.

Exposure Assessment

We assessed long-term exposure to TRAP using two metrics of distance from major traffic roads (highways, primary urban roads and arterial roads). First, we computed the shortest distances between the patients' residential addresses at the time of transplantation and major traffic roads. Distances were categorized as 0–100, 101–200, 201–1000 and >1000 m. Second, we calculated the total length of major roads that fell within circular buffer regions of a series of radii (200, 300, 500 and 1000 m) from the patients' home addresses. These two metrics have been used as surrogates of exposure to TRAP in epidemiological studies of chronic effects of air pollution.

We estimated exposure to nitrogen dioxide (NO₂) at subjects' homes using satellite-based estimates of surface concentrations of NO_2. The satellite-based concentrations were derived using 2005–2009 data from the Ozone Monitoring Instrument (OMI) onboard the Earth Observing System (EOS) Aura satellite, which passes over Ontario near midday each day. Variation in annually averaged ambient NO₂ in the province of Ontario is primarily spatial in nature rather than temporal; thus, it is valid to use mean NO₂ estimates from 2005 to 2009 to characterize NO₂ exposure over our retrospective cohort period. This inferred NO₂ exposure was expressed in terms of the interquartile range (IQR) of the 5-year average vertical column density (4.7 × 10 molecules/cm). Values for days without excessive cloud cover were re-gridded onto ~0.1° × 0.1° grid squares to yield exposure estimates with a spatial resolution of ~10 km × 10 km. The OMI data were converted to vertical column densities through the Dutch OMI NO₂ (DOMINO) Product provided by the Tropospheric Emissions Monitoring Internet Service (http://www.temis.nl/index.php).

Exposure to fine particulate matter (particles with aerodynamic diameter <2.5 μm [PM_2.5]) and ozone (O₃) at the patient's home were determined using measured concentrations at nearby National Air Pollution Surveillance (NAPS) fixed-site monitoring stations where concentrations were measured for the entire study period. Air quality standards in Canada are based on PM_2.5 rather than PM₁₀, thus much more extensive PM_2.5 than PM₁₀ data were available for Ontario over the study period. Long-term exposures were calculated by averaging daily concentrations of PM_2.5 and O₃ between 1996 and 2009 and assigned to subjects' homes using inverse distance weighted interpolation method. A subset of the patients living residing within 10 km of a NAPS monitoring station was also examined.

Outcome Measures

Patients had FEV₁ measurements weekly to monthly for at least the first 2 years posttransplant, and subsequently every 1 to 3 months thereafter. Baseline FEV₁ was determined as an average of the two best FEV₁ values posttransplant that were taken at minimum 3 weeks apart.

CLAD was defined as an irreversible decline in FEV₁ to less than 80% of baseline, measured on two separate occasions at least 3 weeks apart. Irreversibility was determined after appropriate treatment for infection, rejection or both. The first date of an irreversible decline in FEV₁ was recorded as the date of onset for CLAD.

Deaths were reported directly to the Lung Transplant Program by the patient's most responsible physician, family or both. Cause of death was determined by chart review, discussion with most responsible physician, death certificate, postmortem examination or a combination of the foregoing.

Calculations of overall and post-CLAD survival were based on start dates of transplant and date of CLAD diagnosis, respectively, and end dates of April 30, 2011.

Covariates

From patient's charts and the Lung Transplant Program Research Database, we extracted the age, gender, clinical indication for transplant, type of transplant (double-lung vs. heart-lung), spirometry, diagnosis of CLAD, chronic azithromycin use for greater than 1 month, and incidence of death for each lung allograft recipient. Determination of donor age and gender was extracted from our database based on data provided by the Trillium Gift of Life Network.

We created three contextual variables: (1) proportion of population aged >15 years with less than high school education; (2) unemployment rate and (3) mean household income from the Canadian Census in 2001 using census tracts. In addition, we created a variable of urban/rural residence. Urban areas are those continuously built-up areas having a population of >1000 and a population density of 400 or more per square kilometer based on the 2001 census.

To control for large-scale spatial patterns in CLAD that might be caused by factors other than pollution, we created an indicator variable classifying Ontario into southern and northern regions according to the 14 Local Health Integrated Networks of Ontario.

Statistical Methods

We used a stratified Cox proportional hazards model with strata defined as region (south/north). Follow-up time was measured in days, calculated from the date of transplantation through the last date of the pulmonary function test or the date of last clinical visit, whichever came later. CLAD and death were examined separately as outcomes. For CLAD analyses, subjects were censored if they died without CLAD.

In separate Cox proportional hazards models, we modeled exposure to TRAP as: (1) proximity to roadways (0–100, 101–200 and 201–1000 m), with the cohort residing >1000 m from a major road or highway considered the reference group for comparison and assigned a hazard ratio (HR) of 1; (2) lengths of major roads within 200, 300, 500 and 1000 m radii from the patient's home addresses; (3) satellite-based estimates of NO₂ and (4) annual concentrations of PM_2.5 and O₃ averaged over the period 1996 to 2010. In all models, we adjusted for age at baseline, sex, pretransplant diagnosis, age and gender of donor, sex matching between donor and recipient, year of transplantation, rural/urban residency, as well as neighborhood-level unemployment, education and average household income. Because azithromycin was generally prescribed in response to a drop in lung function, we did not adjust our analyses for azithromycin use. With the exception of proximity to major roads, we reported adjusted HR and 95% confidence interval (CI) for other exposure metrics per increment of IQR.

For all the Cox models, we tested deviations from the proportional hazards assumption by assessing whether the cross-product of each variable with the natural logarithm of the time variable was statistically important. We also investigated the presence of effect modification by age, sex, sex-matching and rural/urban residency.

As a sensitivity analysis, we restricted analysis to subjects who lived within 10 km of monitoring stations and excluded subjects whose residential addresses did not contain any major road within radii of 200, 300, 500 and 1000 m, respectively. We further adjusted for an additional neighborhood-level variables and neighborhood percentage of landed immigrants.

Ethics Approval

Our study was approved by the Research Ethics Board of the University Health Network.