Effects of exposure measurement error on particle matter epidemiology: a simulation using data from a panel study in Baltimore, MD

Abstract

Ascertaining the true risk associated with exposure to particulate matter (PM) is difficult, given the fact that pollutant components are frequently correlated with each other and with other gaseous pollutants; relationships between ambient concentrations and personal exposures are often not well understood; and PM, unlike its gaseous co-pollutants, does not represent a single chemical. In order to examine differences between observed versus true health risk estimates from epidemiologic studies, we conducted a simulation using data from a recent multi-pollutant exposure assessment study in Baltimore, MD. The objectives of the simulation were twofold: (a) to estimate the distribution of personal air pollutant exposures one might expect to observe within a population, given the corresponding ambient concentrations found in that location and; (b) using an assumed true health risk with exposure to one pollutant, to estimate the distribution of health risk estimates likely to be observed in an epidemiologic study using ambient pollutant concentrations as a surrogate of exposure as compared with actual personal pollutant exposures. Results from the simulations showed that PM 2.5 was the only pollutant where a true association with its total personal exposures resulted in a significant observed association with its ambient concentrations. The simulated results also showed that true health risks associated with personal exposure to O 3 and NO 2 would result in no significant observed associations with any of their respective ambient concentrations. Conversely, a true association with PM 2.5 would result in a significant, observed association with NO 2 (b ¼ 0.0115, 95% confidence interval (CI): 0.0056, 0.0185) and a true association with exposure to SO 4 2À would result in an observed significant association with O 3 (b ¼ 0.0035, 95% CI: 0.0021, 0.0051) given the covariance of the ambient pollutant concentrations. The results provide an indication that, in Baltimore during this study period, ambient gaseous concentrations may not have been adequate surrogates for corresponding personal gaseous exposures to allow the question to be investigated using central site monitors. Alternatively, the findings may suggest that in some locations, observed associations with the gaseous pollutants should be interpreted with caution, as they may be reflecting associations with PM or one of its chemical components.

Estimating risk of emergency room visits for asthma from personal versus fixed site measurements of NO2

“…Recent studies have documented the potential for measurement error to bias risk estimates toward the null for pollutants with high spatial variability (Suh and Zanobetti, 2010;Van Roosbroeck et al, 2008) and previous analyzes of exposure measurement error in timeseries studies of mortality suggest that risk estimates based on fixed-site ambient air pollution data are smaller than those estimated from personal measures (Schwartz et al, 2007;Zeger et al, 2000). In this study we examined the potential impact of exposure measurement error for short-term (24-h) exposure to NO 2 on risk estimates derived from a case-crossover analysis using existing data from Windsor, Ontario (Lavigne et al, 2012;Wheeler et al, 2011).…”

“…Here we focus on the classical model, which typically leads to greater bias. Moreover, if fixed-site monitors systematically over or underestimate personal exposures, a scaling factor is required to further adjust model coefficients, since incremental changes in ambient concentrations translate into changes of a different magnitude at the personal level (Schwartz et al, 2007). Unfortunately, paired personal and fixed-site exposure data are rarely available in practice.…”

Panel discussion review: session two — interpretation of observed associations between multiple ambient air pollutants and health effects in epidemiologic analyses

“…These results, therefore, suggest that multipollutant models using ambient concentrations of gases may be a form of source apportionment of the effects of particle phase pollution and may provide limited information about the effects of gases. A simulation study conducted by Schwartz et al (2007) using the Baltimore data showed that a significant association with ambient O 3 is much more likely to result from a true association with sulfate than from a true association with exposure to O 3 . However, this may not be true in all cities.…”

Windsor, Ontario Exposure Assessment Study: Design and Methods Validation of Personal, Indoor, and Outdoor Air Pollution Monitoring

“…[1][2][3] The impact of air pollutants on the health of susceptible populations such as the elderly, 4 diabetics, 5 children, 6 and asthmatics [7][8][9][10] indicate that exposure sources and baseline health are important factors for guiding regulatory decisions related to ambient or indoor air quality guidelines or standards.…”