Particulate Pollution and Cause Specific Mortality in London and Hong Kong Using Changes in Pollutant Level between Successive Days as Exposure Metric

Abstract

Background Most time series studies in air pollution epidemiology correlate absolute daily pollution concentration to health outcome without considering the time order of exposure. However, change or rate of change in exposure may be important because physiological systems often tend to respond to changes in environmental stimuli as well as absolute concentrations. This has been investigated in temperature studies but not yet for air pollution exposure. Methods Using data from London (2000-2005) and Hong Kong (2002-2008) on PM10 we computed daily changes (delta=lag1 PM10-lag0 PM10) and relative daily changes (relative delta=delta/lag1 PM10). We used Poisson generalized additive models to study associations of PM10, delta and relative delta with mortality controlling for time trend, seasonality, day of the week and weather. Results Delta for both cities and relative delta in London were negatively related with non-accidental mortality when used as single exposure metrics. For Hong Kong, excess risk (ER, with 95% CI) in non-accidental mortality for interquartile range increase in PM10 before and after controlling for delta contrasted as 1.93 (1.19, 2.68) and 2.44 (1.61, 3.28) at lag0 and 1.97 (1.23, 2.73) and 2.44 (1.61, 3.28) at lag1 respectively. But such associations were less evident for cardiovascular and respiratory causes and results were sensitive to model specification. In London, ER before and after controlling for delta compared as 0.37 (-0.04, 0.78) and 0.83 (0.37, 1.3) at lag0 and 0.91 (0.5, 1.32) and 0.83 (0.37, 1.3) at lag1 respectively. Yet delta metrics were not significant and contrary to Hong Kong, ER at lag1 declined after controlling for delta. Conclusion While the delta metrics could provide a convenient interpretation biologically, further investigations with robust modelling approaches are needed to justify discrepancies in risk estimates across different locations and mortality causes.