Spatial Estimation and Long-term Health Effects of Exposure to Air Pollution in Central Scotland
Christina Yap (1), Chris Robertson (2), Iain Beverland (2), MATHEW HEAL (3), Raymond Agius (4), David Hole (deceasd), Deborah Henderson (2), Geoff Cohen (5), George Morris (6).
(1) University of Glasgow, Glasgow, UK, (2) University of Strathclyde, Glasgow, UK, (3) University of Edinburgh, Edinburgh, UK, (4) University of Manchester, Manchester, UK, (5) Emmes Corporation, Maryland, USA (6) Health Protection Scotland, Glasgow, UK.
Abstract Number: 251
Preference: Poster Presentation
Last modified: November 9, 2009
Working Group: sq1
Long-term exposure to air pollution is widely believed to have significant public health burden. The use of community-average air pollution concentrations to represent an individual’s exposure in cohort studies may lead to significant error in the calculation of associations between health outcomes and air pollution. In this study, three different methods for estimating individuals’ exposure to air pollution were applied to a combined cohort of >21,000 subjects recruited between 1970-6 in central Scotland, whose residential postcodes spanned the range from small villages to the centre of the Glasgow conurbation. Air pollution predictors included daily black smoke concentrations from up to 180 monitoring sites operational in the 1970s plus the local environmental predictors (LEP) of altitude, household density within 250 m buffer, distance to nearest major road and distance to urban boundary. Missing daily black smoke data were imputed using a log-linear model, allowing for day of week, seasonality and a linear time trend. The three exposure models were: (1) inverse-squared distance weighting of measured geometric mean black smoke; (2) multivariate spatial smoothing of black smoke using a semiparametric additive model with the LEPs; (3) multilevel spatio-temporal modelling of monthly black smoke with the LEPs. The resulting estimated 1970s decadal geometric mean exposure to black smoke for individual cohort members were in the ranges 5.5 - 70 microgram/m3, 5.9 - 49 microgram/m3 and 4.6 - 55 microgram/m3 for the three estimation methods, respectively.
The magnitude of associations between estimated exposure and mortality were derived using Cox’s proportional hazard regressions. The potential for confounding and effect modification by both individual and aggregate level factors (including smoking, deprivation, occupation, educational attainment, prior ill health, physiological factors, and gaseous co-pollutants) was examined. Using the more sophisticated multilevel exposure model the following significant mortality hazard ratios per 10 microgram/m3 increment in black smoke were identified, for follow-up to 1998: ischaemic heart disease mortality 1.07 (95% CI: 1.01-1.15), cardiovascular mortality 1.07 (1.00-1.13) and all-cause mortality 1.05 (95% CI: 1.01-1.09). Hazard ratios of 1.11 and 1.09 for respiratory mortality and lung cancer were not significant (p = 0.13 and 0.15, respectively). Interaction tests failed to provide compelling evidence that the range of risk factors investigated modified the observed black smoke effects. Hazard ratios varied with exposure model used, highlighting the crucial importance of reliable estimation of intra-urban variations in exposure. The findings are broadly consistent with previous evidence and hypotheses of how long-term exposure to air pollution may affect human health. However, the dissimilarities noted between exposure models suggest that accurate exposure classification will continue to be one of the most pressing issues in air pollution epidemiological research.
This work was funded by the UK Department of Health.