Comparing types of exposure metrics for identifying the effects of air pollution: an application in a panel of COPD patients

Abstract

BACKGROUND AND AIM: Measured or modelled concentrations used in epidemiological studies can differ from true exposures, a difference known as measurement error. To account for this, personal exposure measurements can be regarded as the “gold-standard” method but do not differentiate between the effects of indoor- and outdoor-generated pollution. We apply a statistical apportionment method to identify and separate the effects of personal exposure to air pollution from indoor and outdoor sources on chronic obstructive pulmonary disease (COPD) patients in London. METHODS: A panel of 130 patients was followed up for an average period of 128 days during which each patient carried sensors measuring PM2.5, PM10, NO2, NO, CO and O3 at one-minute resolution. Total personal exposures (PExT) were separated into those from indoor- (PExI) and outdoor-generated pollution (PExO) using GPS data and matching ambient concentrations from the London reference ambient air quality monitoring network. Each patient recorded daily respiratory symptoms, peak expiratory flow (PEF) and exacerbation of symptoms. The associations of these outcomes with PExT, PExI, PExO and ambient pollution were assessed with mixed-effects models. RESULTS:Preliminary results for PM2.5 have shown that while total personal exposure was not associated with any outcome, a unit increase in PExO was found to increase the odds of exacerbation, cough and sputum by approximately 1.0%. Using ambient concentrations, which is the equivalent error-prone proxy to PExO, these odds were decreased to 0.6-0.7% for exacerbation and cough, while for sputum the finding was not statistically significant. PEF was negatively associated only with PExI. Work is ongoing for the other pollutants. CONCLUSIONS:Our findings suggest that the exposure metric used in identifying the health effects of air pollution is important in epidemiological studies. Using proxies can introduce measurement error bias. For decision-making, it is important to separate the effects of indoor- and outdoor-generated pollution as different policies apply to each. KEYWORDS: Air pollution, Respiratory outcomes, Methodological study design, Exposure assessment, Measurement error