Particle inhalation rate as a metric for ambient air pollution exposure

摘要

Aim: Many air pollution epidemiology studies use ambient residential air pollutant concentrations to represent exposure. Nevertheless, ambient concentrations may not accurately reflect even relative exposure levels since factors such as age, sex, weight, and physical activity differentially affect intake of air pollutants. We developed and tested a new exposure metric to try to better approximate the biologically relevant dose. Methods: Using data from a longitudinal study of 812 adults (69% female) in Boston (MA, USA), we first modeled participants' annual residential average exposure to ultrafine particulate matter (UFP, <0.1 urn diameter, measured as particle number concentration or PNC). We then multiplied PNC estimates (particles/L) by hourly respiratory volume (L of air inhaled/hr) for each participant to obtain the average particle inhalation rate (PIR, particles inhaled/hr). Respiratory volume was estimated using published estimates of sex-, age-, and physical activity-adjusted ventilation rates together with data on how many hours per day participants engaged in defined levels of physical activity. We compared the distributions of PNC and PIR, considered whether associations between UFP exposure and cardiovascular disease (CVD) risk factors differed for PNC and PIR, and examined how sensitive the PIR effect estimates were to factors such as physical activity and respiratory medication use. Results: While the PNC distribution was slightly left skewed (mean=23,000, median=24,000, min=10,000, max=32,000 particles/cc), the PIR distribution was slightly right skewed and had greater variability (mean=13x10^9, median=12x10^9, min=3.7x10^9, max=54x10^9 particles inhaled/hr). Distributions were stable over the five year study period. By design, distributions for the PIR strongly reflected physical activity patterns (r=0.7 p<0.001) even in this population with generally low physical activity levels. Notably, among those with highest PIR (greater than 90th percentile), 6% had low PNC exposure (<10th percentile). As may be expected based on the different distributions, PNC and PIR showed different associations with CVD risk factors. We found that PNC was more strongly associated with increases in systolic blood pressure, pulse pressure, and high sensitivity C-reactive protein (a biomarker of systemic inflammation) while the PIR was more strongly associated with increases in diastolic blood pressure. Our PIR results did not change substantially when we conducted sensitivity analyses excluding participants taking respiratory medications (25% of participants) or excluding participants reporting the highest physical activity levels. Conclusions: Our findings suggest that adjusting ambient UFP exposure estimates for inhalation rate affects the shape and variability of the exposure distribution and alters effect estimates for the association with CVD risk factors.