Air pollution exposures from the 1970s confound associations between PM10 in 2001 and mortality 2001-2009 in a long-running English cohort
Ghosh, R.E., Blangiardo, M., Morris, C., Vienneau, D., Gulliver, J., Lee, K., Briggs, D. & Hansell, A. (2013) International Society for Environmental Epidemiology Conference, Basel, Switzerland. 19 - 23 August 2013 [ONS LS]
Other information: Poster presentation
Background: Long-term air pollution exposure contributes to mortality but there are few studies examining effects of very long term (>20 years) exposures. Aim: This study investigates air pollution exposures from 1971 on mortality to 2009 in the ONS Longitudinal Study (LS), a long-running cohort comprising an anonymised 1% representative sample from the census of England and Wales. Methods: Land use regression techniques were used to model annual average exposures to PM10 in 2001 and black smoke (BS) and SO2 in 1971. Logistic regression was used to investigate the association of BS or SO2 exposure in 1971 and PM10 in 2001 with later mortality in 367,658 members of the LS present in 1971, adjusting for age, sex, social class of individual, area-level deprivation, population density, urban land cover and geographical region. Results: PM10 exposure in 2001 was weakly correlated with BS in 1971 (Spearman correlation r=0.20) and moderately with SO2 in 1971 (r=0.41). BS and SO2 in 1971 were highly correlated (r=0.73) The association between all-cause mortality 2002-2009 and PM10 exposure in 2001 was OR 1.28 (1.18-1.38). Exposure to BS and SO2 in 1971 reduced observed associations of mortality 2002-9 with PM10– when BS 1971 was included in the model, the PM10 OR reduced to 1.02 (1.00-1.04) and when SO2 1971 was included in the model the PM10 OR reduced to 1.00 (1.00-1.01). Similar patterns were seen for cardiovascular and respiratory disease mortality. (All ORs expressed per 10?g/m3.) Conclusions: Historic air pollution exposures appear to confound associations of PM10 with mortality in more recent years. Residual effects from unobserved covariates such as smoking cannot be excluded as limited covariate information is available from the Census.
Available online: Link
Output from project: 0301018