Does personalization in exposure assessment change ambient air pollution exposure-response relationships? A panel study

Accurate exposure assessment is crucial to understand linkages between ambient air pollution and cardiopulmonary disease. Air quality monitors (AQM) are widely used, but do not account for personal behaviors. We compare the exposure-response relationships between ambient air pollution (PM2.5 and O3) and cardiopulmonary biomarkers in a panel study using both stationary AQM and Exposure Model for Individuals (EMI). Participants (n = 28) underwent 3–5 sessions totaling 134 visits. Participants underwent spirometry and blood sampling. PM2.5 and O3 concentrations were calculated for each visit (lag0) and 4 preceding days (lag1–4) using AQM and EMI. A mixed-effects model was applied to examine the associations between exposure and outcomes. AQM and EMI were strongly correlated for PM2.5 (ρ = 0.89) and moderately correlated for O3 (ρ = 0.46). Exposure-response relationships for PM2.5 were similar, with PM2.5 associated with increased oxLDL at lag1 (12.2 % (95 %CI: 4.5, 20.2) AQM, 17.9 % (95 %CI: 8.1, 27.8) EMI), increased vWF at lag0 (4.27 % (95 %CI: 0.15, 8.39) AQM, 7.12 % (95 %CI: 2.57, 11.67) EMI) and decreased vWF at lag3 − 6.5 % (95 %CI: − 11.4, − 1.6) AQM, − 5.6 % (95 %CI: − 10.6, − 0.7) EMI) and lag4 (-5.4 % (95 %CI: − 10.2, − 0.7) AQM, − 6.7 % (95 %CI: − 12.1, − 1.3) EMI). O3 showed more variability, with positive associations with vWF at lag0 (12.9 % (95 %CI: 6.1, 19.7) AQM, − 2.77 % (95 %CI: − 8.1, 2.6) EMI) and D-dimer at lag1 27.0 % (95 %CI: 0.9, 53.0) AQM, − 6.86 % (95 %CI: − 26.3, 12.6) EMI), for AQM only, and negative associations with tPA at lag3 for EMI only (-10.0 % (95 %CI: − 21.5, 1.4) AQM, − 11.2 % (95 %CI: − 19.6, − 2.8) EMI). Our findings suggest that exposure-response associations to short-term PM2.5 and oxLDL and markers of coagulation are consistent between the AMQ and EMI methods, implying increased risk for cardiovascular disease. For O3, AQM and EMI were less consistent, highlighting the challenges of estimating and modeling O3 exposure.