You are here

Identifying the shape of the association between long-term exposure to low levels of ambient air pollution and the risk of mortality: An extension of the Canadian Census Health and Environment Cohort using innovative data linkage and exposure methodology

Principal Investigator: 

University of British Columbia, Canada

This study will investigate mortality effects of low levels of air pollution in Canada using Canadian Census data from about 6 million people. The shape of the exposure-response function will be characterized using newly developed flexible non-linear exposure-response functions. See also this Program Summary of HEI's research program on low levels of air pollution. 

Funded under
Abstract for the 2017 HEI Annual Conference
Identifying the Shape of the Association Between Long-Term Exposure to Low Levels of Ambient Air Pollution and the Risk of Mortality: An Extension of the Canadian Census Health and Environment Cohort Using Innovative Data Linkage and Exposure Methodology
Michael Brauer1, Jeffrey R. Brook2, Richard T. Burnett3, Daniel L. Crouse4, Anders Erickson1, Lauren Pinault5, Randall V. Martin6, Michael Tjepkema5, Scott Weichenthal7.
1The University of British Columbia, Vancouver, British Columbia, Canada; 2Environment Canada, Toronto, Ontario, Canada; 3Health Canada, Ottawa, Ontario, Canada; 4University of New Brunswick, Fredericton, New Brunswick, Canada; 5Statistics Canada, Ottawa, Ontario, Canada; 6Dalhousie University, Halifax, Nova Scotia, Canada; 7McGill University, Montreal, Quebec, Canada.
Background Fine particulate matter (PM2.5) is generally accepted as a causal mortality risk factor.  However, the range in concentration for which this association is present is not known. Since nearly the entire population of Canada lives in areas with ambient concentrations below 12 µg/m3, and studies repeatedly demonstrate associations with mortality in this population, it is an ideal environment to study the relationship between mortality and low concentrations of PM2.5.
Objectives To apply novel satellite-based estimates of exposure to PM2.5 to several large population-based cohorts, to characterize the shape of the relationship between PM2.5 exposure with all cause and cause-specific mortality.
Methods We developed novel satellite-based PM2.5 exposure estimates at 1 km by 1 km resolution for each year from 1998 to 2012 across Canada. The estimates were based on a combination of remote sensing based aerosol optical depth (AOD), translation of AOD to surface PM2.5 concentrations using the chemical transport model GEOS-Chem, and integration of these concentrations with land use and ground monitoring data. Estimates will eventually be back-casted to 1981 using available historical ground monitoring data. Further refinements will be made after incorporating new information on the relationship between AOD and PM2.5 based on measurements of PM2.5 at 5 sites across Canada where AOD is measured with sun photometers.
We applied these exposure estimates to four large, population-based, cohorts: 1) ~2.5 million subjects who completed the 1991 census long form; 2) ~3.5 million subjects who completed the 1996 census long-form; 3) ~3.5 million subjects who completed the 2001 census long-form; 4) 389,000 subjects who participated in the Canadian Community Health Survey (CCHS) 2001, 2003, 2005, and 2007/2008 panels. All subjects were linked to annual mortality and tax records until 2011, to establish residential histories. The potential confounding influence on the PM2.5-mortality association due to behavioral risk factors not recorded in the census/tax cohorts was examined using the CCHS and indirect adjustment methods.
Using several exposure-time windows, we characterize the shape of the concentration-mortality association using newly developed Shape Constrained Health Impact Function (SCHIF) models. We will examine the sensitivity of the shape of the association to age, sex, socio-economic position, ozone and NO2 exposure, behavioral and contextual risk factors. Both relative and additive risk models will be examined.


Preliminary Results In the 2001 census cohort, PM2.5 was associated with increased risk for natural-cause mortality (HR=1.17 per 10 μg/m3 increase in concentration, 95% CI: 1.14, 1.20), cardiovascular diseases (HR=1.24, 95% CI: 1.18, 1.29), respiratory diseases (HR=1.21, 95% CI 1.11, 1.31), and lung cancer (HR=1.15, 95% CI: 1.07, 1.25). The shape of the relationships for all-cause and specific causes of death were supra-linear with upper uncertainty bounds on threshold concentration estimates below 5 μg/m3.


Ambient PM2.5, O3, and NO2 Exposures and Associations with Mortality over 16 Years of Follow-up in the Canadian Census Health and Environment Cohort

Dan L. Crouse,1, 2 Paul A. Peters,2 Perry Hystad,3 Jeffrey R. Brook,4, 5 Aaron van Donkelaar,6 Randall V. Martin,6 Paul J. Villeneuve,7 Michael Jerrett,8 Mark S. Goldberg,9 C Arden Pope III,10 Michael Brauer,11 Robert D. Brook,12 Alain Robichaud,13 Richard Menard,13 Richard T. Burnett.1

1. Environmental Health Science and Research Bureau, Health Canada, Ottawa, Ontario, Canada; 2. Department of Sociology, University of New Brunswick, Fredericton, New Brunswick, Canada; 3. College of Public Health and Human Sciences, Oregon State University, Corvallis, Oregon, USA; 4. Air Quality Research Division, Environment Canada, Downsview, Ontario, Canada; 5. Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada; 6. Department of Physics and Atmospheric Science, Dalhousie University, Halifax, Nova Scotia, Canada; 7. Institute of Health: Science, Technology and Policy, Carleton University, Ottawa, Ontario, Canada; 8. School of Public Health, University of California, Berkeley, California, USA; 9. Department of Medicine, McGill University, and Division of Clinical Epidemiology, Research Institute of the McGill University Hospital Centre, Montreal, Quebec, Canada; 10. Department of Economics, Brigham Young University, Provo, Utah, USA; 11. School of Population and Public Health, University of British Columbia, Vancouver, British Columbia, Canada; 12. Division of Cardiovascular Medicine, University of Michigan, Ann Arbor, Michigan, USA; 13. Atmospheric Science and Technology Directorate, Environment Canada, Dorval, Quebec, Canada

Background and Objectives Few studies examining associations between long-term exposure to ambient air pollution and mortality have considered multiple pollutants simultaneously, while also assessing changes in exposure due to residential mobility patterns during follow-up.  We investigated associations between cause-specific mortality and ambient concentrations of fine particulate matter (PM2.5), ozone (O3), and nitrogen dioxide (NO2) – jointly and independently – in a national cohort of about 2.5 million Canadians.

Methods and Approach This study was conducted with the 1991 Canadian Census Health and Environment Cohort; a nationally-representative sample of ~2.5 million Canadian adults.  This cohort was linked to the Canadian mortality database and to annual income tax filings through 2006.  Subjects provided information on education, income, employment status, and immigrant status, among other topics.  The tax files provide annual residential postal codes, allowing us to track mobility patterns.  We assigned estimates of annual exposures to these pollutants to subjects’ annual postal codes for each year of follow-up.  We estimated hazard ratios for each pollutant separately and adjusted for the other pollutants.  We also estimated the product of the three hazard ratios as a measure of the cumulative association with mortality for several causes of death.  We estimated the hazard ratios per increment of the mean minus the 5th percentile of each pollutant, namely: 5.0 μg/m3 for PM2.5, 9.5 ppb for O3, and 8.1 ppb for NO2.

Results All three pollutants were associated with non-accidental and cause-specific mortality in single-pollutant models.  Assuming additive associations, the estimated hazard ratio for non-accidental mortality corresponding to a change in exposure from the mean to the 5th percentile for all three pollutants together was 1.075 (95% confidence interval: 1.067 - 1.084).  Accounting for residential mobility had only a limited impact on the association between mortality and PM2.5 and O3, but increased associations with NO2, which had been modelled at a much finer spatial resolution than had the other two pollutants.

Conclusions In this large, national-level cohort, we found positive associations between several common causes of death and exposure to PM2.5, O3, and NO2.  We found that exposure to PM2.5 alone was not sufficient to fully characterize the toxicity of the atmospheric mix, or to fully explain the risk of mortality associated with exposure to ambient pollution.


Poster by Crouse et al, 2017 Annual Conference

Associations between fine particulate matter and mortality in the 2001 Canadian Census Health and Environment Cohort (CanCHEC)

Lauren L Pinault1, Scott Weichenthal2,3, Daniel L Crouse4, Michael Brauer5, Anders Erickson5, Aaron van Donkelaar5, Randall V Martin6,7, Perry Hystad8, Hong Chen9, Philippe Finès1, Michael Tjepkema1, Richard T Burnett3

1Health Analysis Division, Statistics Canada, Ottawa, ON, Canada; 2McGill University, Montreal, QC, Canada; 3Air Health Effects Science Division, Health Canada, Ottawa, ON, Canada;  4University of New Brunswick, Fredericton, NB, Canada; 5University of British Columbia, BC, Canada;  6Dalhousie University, Halifax, NS, Canada; 7Harvard-Smithsonian Center for Astrophysics, Cambridge, MA, USA; 8Oregon State University, Corvallis, OR, USA; 9Public Health Ontario, Toronto, ON, Canada


Background Large cohort studies have been used to characterise the association between long-term exposure to fine particulate matter (PM2.5) air pollution and mortality from natural causes, and from specific cardiovascular and respiratory causes. However, there remains no consensus as to the shape of the association between concentration and response.

Methods To examine the shape of this association, we developed a new cohort based on respondents to the 2001 Canadian long-form Census. The 2001 Canadian Census Health and Environment Cohort (CanCHEC) is a linkage product of the 2001 census long-form questionnaire, the Canadian Mortality Database, and tax files. We followed 2.4 million census respondents who were non-institutional non-immigrants aged 25-90 years, over a 10-year follow-up period for mortality. We developed new annual PM2.5 concentration surfaces for Canada at a 1 km spatial resolution from 1998 to 2011. Exposures were assigned as a 3-year mean moving average of the 3 years prior to the follow-up year. We used income tax files to track subjects’ annual mobility patterns through annual residential postal codes, and a probabilistic imputation program to impute missing records in the tax data. Cox survival models were used to determine cause-specific mortality hazard ratios (HRs). Shape Constrained Health Impact Functions (SCHIF) were estimated for specific causes of death, in addition to standard threshold models.

Results In fully adjusted models stratified by age, sex, airshed, and population centre size, HR estimates for natural-cause mortality were HR=1.17 (95% CI: 1.14 to 1.20) per 10 μg/m3 increase in concentration. Higher HRs were observed for ischemic heart disease, (HR=1.35; 95% CI: 1.27 to 1.43), cardio-metabolic disease, (HR=1.26; 95% CI: 1.20 to 1.31), and COPD mortality (HR=1.23; 95% CI: 1.20 to 1.31). Non-significant associations were observed for cerebrovascular disease and pneumonia. For causes of death examined, the shape of the concentration-response curve was supra-linear.

Conclusions We overcame some previous limitations and expect to reduce exposure misclassification by using a finer-scale (~1 km2 grid) PM2.5 surface, following respondent mobility using tax data, imputing missing postal codes for residential mobility, and by assigning exposures based on annual, rather than longer-term average exposures. The association between ambient concentrations of fine particulate matter and both natural and specific causes of death was supra-linear, with no evidence of a threshold.


Poster by Pinault et al, 2017 Annual Conference