Short-Term Effects of Airport-Associated Ultrafine Particle Exposure on Lung Function and Inflammation in Adults with Asthma

Abstract

Background:

Exposure to ultrafine particles (UFP, particles with aerodynamic diameter less than 100nm) is associated with reduced lung function and airway inflammation in individuals with asthma. Recently, elevated UFP number concentrations (PN) from aircraft landing and takeoff activity were identified downwind of the Los Angeles International Airport (LAX) but little is known about the health impacts of airport-related UFP exposure.

Methods:

We conducted a randomized crossover study of 22 non-smoking adults with mild to moderate asthma in Nov-Dec 2014 and May-Jul 2015 to investigate short-term effects of exposure to LAX airport-related UFPs. Participants conducted scripted, mild walking activity on two occasions in public parks inside (exposure) and outside (control) of the high UFP zone. Spirometry, multiple flow exhaled nitric oxide, and circulating inflammatory cytokines were measured before and after exposure. Personal UFP PN and lung deposited surface area (LDSA) and stationary UFP PN, black carbon (BC), particle-bound PAHs (PB-PAH), ozone (O₃), carbon dioxide (CO₂) and particulate matter (PM_2.5) mass were measured. Source apportionment analysis was conducted to distinguish aircraft from roadway traffic related UFP sources. Health models investigated within-subject changes in outcomes as a function of pollutants and source factors.

Results:

A high two-hour walking period average contrast of ~ 34,000 particles.cm⁻³ was achieved with mean (std) PN concentrations of 53,342 (25,529) and 19,557 (11,131) particles.cm⁻³ and mean (std) particle size of 28.7 (9.5) and 33.2 (11.5) at the exposure and control site, respectively. Principal components analysis differentiated airport UFPs (PN), roadway traffic (BC, PB-PAH), PM mass (PM_2.5, PM₁₀), and secondary photochemistry (O₃) sources. A standard deviation increase in the ‘Airport UFPs’ factor was significantly associated with IL-6, a circulating marker of inflammation (single-pollutant model: 0.21, 95% CI=0.08 – 0.34; multi-pollutant model: 0.18, 0.04 – 0.32). The ‘Traffic’ factor was significantly associated with lower Forced Expiratory Volume in 1 second (FEV₁) (single-pollutant model: −1.52, −2.28 – −0.77) and elevated sTNFrII (single-pollutant model: 36.47; 6.03 – 66.91; multi-pollutant model: 64.38; 6.30 – 122.46). No consistent associations were observed with exhaled nitric oxide.

Conclusions:

To our knowledge, our study is the first to demonstrate increased acute systemic inflammation following exposure to airport-related UFPs. Health effects associated with roadway traffic exposure were distinct. This study emphasizes the importance of multi-pollutant measurements and modeling techniques to disentangle sources of UFPs contributing to the complex urban air pollution mixture and to evaluate population health risks.

INTRODUCTION

Exposure to ultrafine particles (UFP, particles with aerodynamic diameter < 100 nm) in ambient air is associated with decreased lung function and increased airway inflammation in individuals with asthma (Buonanno et al. 2013; Heinzerling et al. 2016; McCreanor et al. 2007). While fresh fuel combustion and roadway traffic sources have long been recognized as major primary sources of UFPs (Hofman et al. 2016; Kukkonen et al. 2016), only recently have measurement campaigns shown aircraft traffic activity to be a significant source of UFPs, with elevated particle number (PN) concentrations in close proximity to runways (Hsu et al. 2012; Hsu et al. 2013; Westerdahl et al. 2008) and further downwind of airports (ACI Europe 2012; Choi et al. 2013; Hsu et al. 2014; Hudda et al. 2014; Hudda and Fruin 2016; Hudda et al. 2016; Keuken et al. 2015). In Los Angeles, CA, Hudda et al. (2014) showed that PN concentrations downwind of the Los Angeles International Airport (LAX) are at least twice as high as background during most hours of the day with a 4- to 5-fold increase up to 10 km under typical westerly wind conditions.

Inflammation and oxidative stress are thought to be the main pathways of UFP toxicity. Because of their smaller size and diffusion-driven behavior in the lungs once inhaled, UFPs deposit efficiently in the alveolar region (Delfino et al. 2005). Once there, they can evade macrophage clearance, enter lung cells, cross the epithelial barrier into the blood and lymphatic circulation, elicit systemic effects and reach other organs (Elder et al. 2006; Geiser 2010; Nemmar et al. 2004; Samet et al. 2009). They can also damage airway epithelial cells and macrophages via reactive oxygen species production from redox reactions occurring in the mitochondria (Li et al. 2003; Nel 2005). UFPs are also retained very effectively in the lungs and can remain there for long periods of time (Araujo and Nel 2009). Surface coating is important in determining mucus penetration potential and retention time in the lungs, where biodegradable, hydrophilic or negatively charged UFPs can evade adhesive interactions with the mucus mesh or diffuse through pores, reach the adherent mucus layer and evade rapid clearance (Lai et al. 2009; Schuster et al. 2013; Xu et al. 2013). Möller et al (2008) showed that most inhaled carbon UFPs are retained in the lung periphery and conducting airways without substantial systemic translocation 48 hours after exposure. In addition, their large surface area to mass ratio and ability to carry reactive oxygen generating species such as metals (Vitkina et al. 2016) and PAHs (Delfino et al. 2010) on their surface (redox potential) makes them more toxic than larger particles such as PM_2.5 (particles with aerodynamic diameter < 2.5 μm) on an equal mass basis (Ayres et al. 2008; Cho et al. 2005; Gong et al. 2014; Nel et al. 2001; Sioutas et al. 2005). Weichenthal et al (2007) provide an excellent review of in vitro, in vivo and population studies of UFPs, their composition and mode of action.

Epidemiological evidence of UFP health effects is limited compared to PM_2.5, likely due to their highly dynamic and variable nature in space and time which complicates exposure assessment (2013). Wichmann and Peters (2000) provide a review of the epidemiological evidence on short-term health effects of UFPs and explain the potentially independent physiological pathways by which UFPs induce toxicity compared to PM_2.5 also demonstrated in Gong et al (2014) Generally longer exposure-response lag times are observed in panel studies for UFPs, possibly related to their longer retention time in the lungs. Buonanno et al. (2013) found daily UFP alveolar-deposited surface area dose to be associated with exhaled nitric oxide, a marker of pulmonary inflammation, in asthmatic children. Delfino et al. (2009) found “quasi-UFPs” (particles with aerodynamic diameter < 0.25 μm) to be significantly associated with the inflammation markers IL-6 and soluble TNF-α. Roadway traffic studies also suggest that fresh combustion products in exhaust - of which UFP is a large component - play a major role in asthma attacks and chronic bronchitis (Brauer et al. 2002; Kunzli et al. 2000), cause acute decreases in lung function that is more pronounced in asthmatics (McCreanor et al. 2007), and may be a cause of asthma (Brauer et al. 2002; Gauderman et al. 2005; McConnell et al. 2006). Knibbs et al (2011) reviewed 10 studies of commuter exposure in-transit and found UFP exposure during commuting can elicit acute effects in both healthy and health-compromised individuals. Lanzinger et al (2016) found 0–5 day lag central site UFP levels were associated with respiratory mortality independent of particle mass in five central European cities.

Cardiovascular effects have also been reported especially in individuals with existing metabolic or cardiovascular conditions. Lag 4-day PN was associated with total and cardio-respiratory mortality in Germany (Stolzel et al. 2007). Thrombogenic effects and platelet activation were seen in patients with coronary heart disease (Ruckerl et al. 2006). An increase in pulse wave velocity and augmentation index was seen in individuals with chronic obstructive pulmonary disease (Sinharay et al.) and immediate changes in heart rate variability were found in diabetics or people with impaired glucose metabolism (Peters et al. 2015).

However, very few studies have investigated the effects of UFPs resulting from aviation activity on asthma and respiratory health. Children living in 17 Massachusetts communities within a 5-mile radius of the Boston Logan International Airport were 3 to 4 times more likely to experience respiratory symptoms indicative of undiagnosed asthma compared to low exposure areas (Massachusetts Department of Public Health 2014). Schlenker and Walker (2011) estimated that one standard deviation increase in daily air pollution levels attributable to runway congestion at the 12 largest airports in California leads to an additional $1 million in hospitalization costs for respiratory and heart related admissions, for the 6 million individuals living within 10km. However, these studies relied on spatially coarse estimates of residential exposure that suffer from exposure measurement error in estimating personal exposures.

To our knowledge, no studies to date have assessed personal exposure to real-life, airport-related UFPs, distinctly from roadway traffic-related UFPs, and investigated their effect on acute respiratory health in asthmatics. To this end, we conducted a quasi-experimental panel study designed to capture the high UFP plume downwind of LAX reported in Hudda et al. (2014). We hypothesized that short-term exposure to LAX-related UFPs results in acute decreased pulmonary function and increased pulmonary and systemic inflammation in adult asthmatics following mild walking activity.

METHODS

Study Design

We conducted a randomized crossover study of 22 adults in two phases, Nov-Dec 2014 and May-July 2015, modeled after the McCreanor et al. (2007) quasi-experimental design. Eligibility criteria included the following: Non-current smokers (zero cigarettes smoked in the last month, regardless of earlier smoking history), English-speaking (individuals who can speak and understand English for the sake of communicating with study staff and answering questions, since it was not feasible to translate study materials into other languages), and adults aged 18 years or older with mild to moderate asthma as defined by symptoms-based National Heart, Lung and Blood Institute (NHLBI) criteria. Participants were mainly recruited as a convenience sample by advertising to University of Southern California (USC) staff and students.

Participants conducted mild, scripted walking activity for two hours, resting every 15 minutes, on two occasions in two public parks inside and outside of the high LAX UFP zone reported in Hudda et al. (2014). We selected Jesse Owens Park as the ‘exposure’ site because of its location downwind of LAX, ~10 km to the east along the dominant daytime westerly wind direction (supplement Figure S1). Jesse Owens is in a dense urban area near busy, major roadways (W Century Blvd to the south and S Western Ave to the east). We selected Kenneth Hahn State Recreational Area as the ‘control’ site, ~9 km northeast of LAX, as it is located on a hill at the periphery of the high UFP plume, surrounded by greenness and further away from immediate traffic. The order of the visits to the control and exposure sites was randomized, and the visits were separated by at least one week to minimize carryover effects.

We transported participants to and from the walking sites in a 2015 Toyota Prius hybrid car, under recirculating air and closed window conditions, along pre-designated routes to minimize UFP exposure from traffic. To ensure maximum LAX UFP impacts, we visited the exposure site on days with stable midday westerly wind conditions, to the extent logistically possible. We conducted all walking exposures midday (~ 12–2PM) to control for diurnal variations and ensure maximum wind direction stability. The USC Institutional Review Board approved all study procedures (IRB protocol number HS-14–00504), and all participants provided written informed consent and were compensated for their contribution to the study.

Health Outcomes Assessment

Participants reported to the USC Health Sciences Campus in the morning on both study days. In the first visit, we collected detailed demographics, medical history, environmental conditions at the residence, and commuting and time activity patterns using an interviewer-administered questionnaire. We measured height (stadiometer), weight and body composition (Tanita scale) and resting heart rate at baseline. In addition, on each visit, we administered a questionnaire asking about the prior week’s activities, asthma control and severity, as well as their morning commute and dietary intake on the day of the visit.

Respiratory testing and blood draws were performed on each visit before and after exposure at generally similar, consistent times visit-to-visit for each person and across participants (~10.30 AM and 4.00 PM). We conducted multiple flow exhaled nitric oxide testing (FeNO) using our previously developed protocol at 30, 50, 100 and 300 ml/s expiratory flow rates using the EcoMedics CLD88-SP with DeNOx (Linn et al. 2009). Immediately prior to each maneuver, the participant breathed through a DeNOx scrubber for ≥ 2 tidal breaths followed by inhalation to total lung capacity and exhalation at the target flow rate. Analyzer zero checks against air drawn through a zero-NO filter (Sievers Division, GE Analytical Instruments, Boulder, CO) were done twice daily. A Morgan SpiroAir-LT rolling seal spirometer was used for pulmonary function testing (forced vital capacity (FVC), forced expiratory volume in 1 second (FEV₁), peak expiratory flow rate (PEFR), and maximum mid-expiratory flow (MMEF)) and calibrated twice daily with a 3L syringe and tested for leaks. Each participant was asked to perform seven maximum effort maneuvers per test.

An Immunocap antigen-specific IgE panel (Quest Diagnostics, Inc.) for the 16 most common Southern California upper respiratory allergens was conducted using the first blood sample to determine atopic status at baseline. A complete blood count was also obtained using the morning blood draw on each visit. In addition, pre- and post-exposure blood samples on both visits were analyzed for the following inflammatory cytokines and pro-thrombotic clotting factors: high-sensitivity Interleukin 6 (IL-6) and soluble tumor necrosis factor receptor II (sTNFrII) using ELISA kits (R&D Systems, HS600B and DRT200 respectively), and von Willebran factor (vWF) and fibrinogen using the Millipore Luminex magnetic bead panel (HCVD3MAG-67K).

Air Pollution Exposure Assessment

During transport to and from the parks, we measured ultrafine particle number (PN) concentrations using a DiscMini diffusion charger (Testo AG) and condensation particle counter (CPC 3007, TSI Inc) to verify low traffic-related UFP exposure conditions inside the vehicle. During the walking exposure period at the parks, we measured ‘personal’ PN, particle size and lung deposited surface area (LDSA) using the DiscMini and PN using the CPC carried by the research assistant walking alongside the participants. Relative humidity and temperature were measured using an Onset HOBO data logger. We also used a mobile monitoring platform to measure PN (CPC 3007, TSI Inc), black carbon (BC, AE51, Magee Scientific), particle-bound PAHs (PB-PAH, PAS 2000, EchoChem Analytics), ozone (O₃, Model 205, 2B Technologies), carbon dioxide (CO₂, Li-820, LI-COR Biosciences) and particulate matter mass in four size fractions (PM₁, PM_2.5, PM₄ and PM₁₀, DRX 8534, TSI Inc) at each park in a stationary location to obtain more detailed characterization of the air pollution mixture. All exposures were continuously logged at a 10 second time resolution. The DiscMini was considered the primary source of personal PN exposure data as it also provided particle size and LDSA data. Unless otherwise stated, all subsequent references to PN correspond to DiscMini data. Agreement between the personal DiscMini and CPC measurements in terms of PN by particle size bins are shown in the Figure S2.

Statistical Analysis

Air Pollution Exposures

We inspected all air pollutant measurement data for outliers and errors at the original 10 second time resolution and averaged up to one minute for use in source apportionment analyses (described below). We then calculated average concentration for the transport periods to and from the park (inside the vehicle) and the walking period at the parks (exposure time) for use in health models.

Because of the highly correlated multi-pollutant nature of the data, we conducted a source apportionment analysis on the one-minute, walking-period data (shown in red in Figure 1) to disentangle the impact of the airport from other major sources of UFPs contributing to the complex air pollution mixture in this urban area (mainly traffic). We used principal components analysis (PCA) with an oblique (promax) rotation in SAS 9.4 (SAS Institute Inc., NC). Ten variables were included in the PCA (PM₁, PM_2.5, PM₁₀, BC, PB-PAH, CO₂, PN (personal DiscMini), PN (stationary CPC), particle size, and O₃). Four distinct ‘source factors’ were resolved based on their eigenvalues (profiles), physical interpretability and least factor smearing. Walking-period average PCA-derived factor scores (eigenvectors) were then calculated for each day and used as the main exposures of interest in the health models, in addition to the measured pollutants.

Figure 1.

Ultrafine particle number concentrations (PN, particles.cm⁻³) on study days grouped by exposure scenario and colored by transport (blue) and walking exposure (red) period.

Spirometry and Exhaled Nitric Oxide

Pulmonary function test indices (FVC, FEV₁, PEFR, MMEF) were assigned based on criteria described in the 2005 ATS/ERS (Miller et al. 2005). Age, height, gender and race specific percent predicted values were calculated based on equations from Knudson et al (1983).

FeNO data processing was based on the ATS/ERS guidelines for FeNO at 50 ml/s (ATS/ERS 2005) and an airway turnover search window (Puckett et al. 2010) similar to previous studies (Eckel et al. 2016). FeNO₅₀ and FeNO₃₀₀ were calculated as the average of reproducible maneuvers at 50 ml/s and 300 ml/s, respectively. Multiple flow FeNO data were input to nonlinear mixed effects models (based on the deterministic, steady-state two compartment model of NO in the lower respiratory tract) to estimate parameters quantifying airway (D_awNO – airway wall tissue diffusing capacity (pl(s·ppb)⁻¹), C_awNO – airway wall concentration (ppb)) and alveolar (C_ANO –alveolar region concentration (ppb)) sources of NO and to predict FeNO₅₀ (Eckel and Salam 2013; Eckel et al. 2014). We used predicted FeNO₅₀ rather than measured FeNO₅₀ in health models to minimize the number of missing observations.

Health Models

Single-, two- and multi-pollutant ANCOVA (Analysis of Covariance) models examining within-subject changes in outcome related to the exposures were fit as follows:

$$Y_{\mathrm{ij},\mathrm{POST}}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1\mathrm{*}{Y}_{\mathrm{ij},\mathrm{PRE}}+{\mathrm{\beta }}_2\mathrm{*}\mathrm{Exposure}{(\mathrm{s})}_{\mathrm{ij}}+U_i+{\mathrm{\epsilon }}_{\mathrm{ij}}$$ eq. (1);

; where Y_ij,POST is the outcome measured post-exposure for participant i on day j, Y_ij,PRE is the outcome measured pre-exposure, Exposure(s)_ij is one or more continuous measure(s) of the walking-period average air pollution concentration or source factor contribution on day j, U_i is a fixed intercept for every participant, and ε_ij is a normally distributed random error term with variance σ² (ε_ij ∼ N(0,σ²)). β₀ is a fixed intercept, β₁ is the parameter estimate capturing visit-to-visit variability in the baseline outcome, and β₂ is the main parameter of interest capturing the effect of air pollution exposure(s) (Metcalfe 2010).

Outcomes were examined for normality and log-transformed where appropriate (FeNO parameters). Multi-pollutant models of measured concentrations were adjusted for PN, BC, PM_2.5 and O₃ – the key source tracers identified in the source apportionment modeling. Whereas multi-pollutant models of sources were adjusted for all four modeled source factors. All reported effect sizes are scaled to a standard deviation (SD) increase in the exposure of interest.

Outliers were examined and excluded as appropriate for the different sets of health outcomes (1 to 3 data points depending on outcome). The model focuses on within-participant changes in health outcomes and includes an intercept for each participant, thus there is no need to adjust for time-constant individual-level covariates such as age or gender. Given the limited sample size, a list of binary variables was selected a priori based on the literature, with at least 40% of participants in a cell, to investigate interactions with the main exposures of interest (PN, LDSA and Airport UFPs factor) in single- and multi-pollutant models: asthma control, allergic status (reported or measured using specific IgE panel), race and ethnicity, physical activity levels, body mass and composition and commuting patterns (further details in Table S1). Models with significant interaction terms were reported. For all hypothesis tests, the threshold of statistical significance was defined as p-value<0.05; analyses were conducted in SAS 9.3 (SAS Institute Inc., Cary, NC, USA).

RESULTS

The majority of the 22 participants in the study were female (16, 73%), white (9, 43%) and Hispanic (9, 43%). The mean age was 27 years (range 18–60) and mean BMI 24.8 kg/m² (17.4–46.7). The average Asthma Control Test (ACT) score was 18.7 (11–22) at recruitment and 20.6 (11–25) on the day of the first visit. All participants reported a doctor diagnosis of asthma at mean age of 13 years (3–58) (Table 1).

Table 1.

Participant characteristics (N=22)

		N (%)
Gender	Female	16 (73%)
Race	White	11 (50%)
	African-American	3 (14%)
	Asian	3 (14%)
	American Indian	1 (5%)
	Other	4 (18%)
Ethnicity	Hispanic	9 (43%)
		Mean	Std Dev	Min	Max
Age		27	9.5	18	60
Age at asthma diagnosis		13	12.7	3	58
ACT* Score (At recruitment)		18.7	3.2	11	22
ACT* Score (On day of visit)		20.6	3.8	11	25
Body Mass Index (kg/m2)		24.8	6.1	17.4	46.7

^*ACT = Asthma Control Test

The top 5 most common upper airway allergens as measured with a specific IgE response were dust mites (d1 and d2), followed by dog (e5) and cat (e1) dander and Bermuda grass (g2), respectively. Baseline levels of cytokines, spirometry and FeNO parameters are shown in Table 2 with average change in post-exposure value compared to pre-exposure at each of the sites. Predicted FeNO₅₀ was highly correlated with measured FeNO₅₀ (r=0.99).

Table 2.

Distribution of health outcomes at baseline (morning assessment on first visit) and change (post-pre) in outcomes following walking exposure period at the two study sites.

	Baseline Level						Change (Post-Pre)
							Control			Exposure
Outcome	N	Mean	Std Dev	Min	Median	Max	N	Mean	Std Dev	N	Mean	Std Dev
Cytokines (pg/ml)
IL-6	18	1.7	2.8	0.4	0.8	12.3	20	0.3	1.7	18	0.4	0.4
sTNFrII	18	1,083.1	922.9	146.2	940.6	2,384.0	20	−79.0	131.7	18	−85.6	87.8
vWF	17	0.5	0.2	0.3	0.5	0.8	18	0.0	0.2	18	0.0	0.1
Fibrinogen	17	78.4	29.6	45.4	67.4	127.6	18	0.6	19.7	18	−0.5	19.9
% Predicted Spirometry
FEV1	22	105.0	14.5	72.3	105.7	132.8	21	0.0	4.2	22	−1.2	3.4
FVC	22	108.9	11.3	80.5	108.4	129.0	20	−0.8	2.8	22	−0.7	3.1
MMEF	22	99.9	30.4	34.2	96.9	153.5	20	−0.3	9.5	22	−0.7	9.2
PEFR	22	107.4	24.1	59.6	105.6	152.6	21	3.0	8.4	22	2.1	10.3
Exhaled Nitric Oxide*
log(FeNO50,pred )	22	3.30	0.8	2.2	3.0	4.6	21	0.0	0.1	22	−0.1	0.1
CANO	22	1.1	0.9	−0.7	1.2	2.5	21	0.0	0.2	22	−0.2	0.3
log(CAWNO)	22	4.0	0.9	2.7	3.8	5.6	21	0.0	0.2	22	−0.1	0.3
log(DAWNO)	22	3.6	0.6	2.4	3.5	4.8	21	0.0	0.2	22	0.1	0.3

^*Exhaled nitric oxide units as follows: log(FeNO₅₀) in log(ppb); CANO and log(C_AWNO) in log(ppb); log(D_AWNO) in log(pl(s·ppb)⁻¹)

Table 3 shows the distribution of air pollutant concentrations during the walking period at the two sites. UFP PN (stationary and personal) was significantly higher at the exposure site per study design, with an average two-hour walking period PN contrast of ~34,000 particles.cm⁻³ between the two sites. Figure 1 shows the time-resolved personal PN (DiscMini) measurements for each study day grouped by site. Figure S3 shows the distribution of PN and LDSA inside the vehicle during participant transport to and from the exposure sites. Particle size was lower at the exposure site (28.7 vs 33.2 nm) and LDSA was higher (64.8 vs 28.8 cm²) consistent with the smaller particle size and greater lung deposition efficiency of airport-related UFPs. Particle mass concentrations in the 1, 2.5, 4 and 10 μm size fractions were slightly but not significantly higher, while the combustion-related pollutants BC, CO₂ and PB-PAHs were significantly higher at the exposure site. No differences in O₃ concentration or meteorological parameters were observed (Table 3). The second phase of the study (May-July 2015) was characterized by breezier conditions and warmer temperatures compared to the first phase (Nov-Dec 2014) and generally more stable and predictable wind direction patterns (Figure S4).

Table 3.

Distribution of air pollution exposure and meteorology parameters during the walking periods.

	Overall (n=43)		Control (n=21)		Exposure (n=22)		Pearson t-test
Pollutants (units)	Mean	Std Dev	Mean	Std Dev	Mean	Std Dev	p-value
PN (particles.cm−3, personal DiscMini)	36,842.2	26,016.4	19,556.6	11,131.0	53,342.1	25,528.5	3.97E-06
Particle Size (nm)	30.9	10.6	33.2	11.5	28.7	9.5	1.67E-01
LDSA (cm2)	47.2	27.1	28.8	13.0	64.8	25.4	1.59E-06
PN (particles.cm−3, stationary CPC)	23,013.6	14,062.5	13,036.0	4,491.7	32,537.6	13,480.1	8.66E-07
PN (particles.cm−3, personal CPC)	31,705.0	18,589.5	19,066.1	6,879.7	43,769.4	18,271.3	2.60E-06
PM1 (μg/m3)	4.7	3.6	3.9	2.7	5.5	4.2	1.56E-01
PM2.5 (μg/m3)	12.0	7.6	10.1	5.8	13.7	8.8	1.17E-01
PM4 (μg/m3)	14.8	8.8	12.7	6.7	16.9	10.2	1.24E-01
PM10 (μg/m3)	30.1	22.1	27.4	12.3	32.6	28.7	4.42E-01
BC (ng/m3)	523.5	291.9	410.0	207.3	631.9	322.9	1.09E-02
CO2 (ppb)	407.8	13.3	401.4	9.3	413.9	13.8	1.28E-03
PB-PAH (μg/m3)	3.2	1.5	2.6	0.6	3.8	1.9	8.07E-03
O3 (ppb)	45.9	14.4	44.9	12.0	46.7	16.7	6.89E-01
Source Factors
Airport UFPs	0.06	0.74	−0.32	0.49	0.42	0.77	5.91E-04
PM Mass	−0.05	0.46	−0.14	0.33	0.04	0.55	1.85E-01
Traffic	−0.14	0.86	−0.53	0.58	0.23	0.92	2.45E-03
Secondary Photochemistry	−0.04	0.80	−0.31	0.62	0.21	0.88	3.06E-02
Meteorology
Temperature (°C)	27.05	2.74	26.3	2.5	27.7	2.8	9.58E-02
Relative Humidity (%)	45.02	9.32	46.5	8.1	43.6	10.3	3.21E-01

PN=Ultrafine Particle Number, LDSA=Lung-deposited surface area, PM₁, PM_2.5, PM₄, PM₁₀=Particle Mass in the 1, 2.5, 4 and 10 μm size fraction, BC=Black Carbon, CO₂=Carbon Dioxide, PB-PAH=Particle-bound Polyaromatic Hydrocarbons, O₃=Ozone.

The overall standard deviation is used to scale reported health effect estimates.

The source apportionment analysis resolved four distinct source factors characterized by the following species in their loading profiles in parentheses: Airport UFPs (personal and stationary PN, smallest particle size) consistent with jet emissions (Shirmohammadi et al. 2017), PM Mass (PM₁, PM_2.5 and PM₁₀ mass) consistent with heavier particles and wind-blown dust, Traffic (BC, CO₂, PB-PAH and lowest O₃) consistent with fresh combustion emissions and O₃ quenching, and secondary photochemistry (PM_2.5 mass and O₃) consistent with secondary formation. The contributions of these modeled source factors were all significantly higher at the exposure site except for ‘PM Mass’ (Table 3). The ‘Secondary Photochemistry’ and ‘PM Mass’ factors were most highly correlated (Table S2). The average contributions of the ‘Airport UFPs’ and ‘Secondary Photochemistry’ factors were higher in the second phase while ‘Traffic’ was higher in the first phase of the study likely due to cooler temperatures and less vertical mixing (Figure S5).

Single- and multi-pollutant health analysis results are reported in Tables 5 and 6, respectively, while two-pollutant results are included in Supplement Table S3. Adjustment for day-level, time-varying potential confounders such as relative humidity and temperature was explored but did not have any influence on the magnitude of main effects in PN and ‘Airport UFPs’ models.

Table 5.

Single-pollutant model associations of measured concentrations and modeled source factor contributions with cytokines, exhaled nitric oxide and spirometry outcomes.

POLLUTANTS
		PN			LDSA			BC			PB-PAH			O3
Outcome	N	Est	95% CI	p-value	Est	95% CI	p-value	Est	95% CI	p-value	Est	95% CI	p-value	Est	95% CI	p-value
Cytokines
IL-6	36	0.13	−0.03, 0.29	0.100	0.07	−0.11, 0.25	0.407	−0.09	−0.31, 0.12	0.373	0.05	−0.11, 0.20	0.523	0.03	−0.20, 0.27	0.751
Fibrinogen	35	−0.01	−9.83, 9.82	0.999	−0.98	−11.21, 9.25	0.841	−3.07	−14.23, 8.09	0.565	−2.99	−11.76, 5.77	0.476	4.64	−7.36, 16.64	0.421
sTNFrII	36	10.57	−24.06, 45.19	0.525	21.03	−14.18, 56.24	0.222	49.38	10.18, 88.59	0.017	30.23	1.55, 58.92	0.040	−52.30	−91.09, −13.51	0.012
vWF	34	−0.01	−0.06, 0.04	0.612	−0.01	−0.07, 0.05	0.703	−0.00	−0.06, 0.05	0.940	−0.00	−0.05, 0.05	0.873	0.00	−0.07, 0.07	0.975
% Predicted Spirometry
FEV1	40	−0.52	−1.54, 0.50	0.293	−0.99	−1.98, −0.00	0.050	−1.60	−2.68, −0.51	0.006	−0.85	−1.76, 0.06	0.064	1.34	−0.05, 2.73	0.058
FVC	39	0.14	−0.74, 1.02	0.746	0.24	−0.71, 1.19	0.604	0.23	−0.92, 1.38	0.679	0.03	−0.84, 0.89	0.952	0.28	−1.09, 1.66	0.667
MMEF	39	0.27	−2.72, 3.25	0.851	−0.25	−3.47, 2.97	0.871	−1.41	−5.25, 2.42	0.444	1.03	−1.84, 3.89	0.457	0.48	−3.97, 4.92	0.822
PEFR	42	0.49	−3.33, 4.31	0.791	0.61	−3.39, 4.61	0.752	0.03	−5.09, 5.16	0.989	−1.62	−5.52, 2.29	0.396	2.75	−2.64, 8.13	0.298
Exhaled Nitric Oxide
log(FeNO50,pred)	41	−0.02	−0.06, 0.02	0.299	−0.02	−0.06, 0.02	0.377	0.00	−0.05, 0.05	0.974	0.00	−0.04, 0.04	0.945	−0.02	−0.07, 0.03	0.483
CANO	41	−0.07	−0.14, 0.01	0.082	−0.07	−0.15, 0.00	0.062	−0.05	−0.15, 0.05	0.336	−0.02	−0.10, 0.07	0.684	−0.05	−0.16, 0.06	0.357
log(CAWNO)	41	−0.07	−0.13, −0.00	0.044	−0.06	−0.13, 0.01	0.103	0.01	−0.08, 0.10	0.775	0.01	−0.06, 0.08	0.868	−0.06	−0.15, 0.02	0.144
log(DAWNO)	41	0.09	−0.02, 0.19	0.090	0.06	−0.05, 0.18	0.250	−0.05	−0.19, 0.08	0.418	−0.02	−0.12, 0.08	0.702	0.09	−0.04, 0.21	0.175
		PM1			PM25			PM4			PM10
Outcome	N	Est	95% CI	p-value	Est	95% CI	p-value	Est	95% CI	p-value	Est	95% CI	p-value
Cytokines
IL-6	36	−0.14	−0.29, 0.02	0.075	−0.19	−0.35, −0.04	0.015	−0.22	−0.37, −0.06	0.008	−0.31	−0.53, −0.08	0.010
Fibrinogen	35	−2.94	−11.71, 5.84	0.485	−4.45	−13.74, 4.84	0.322	−4.58	−14.21, 5.05	0.325	−9.54	−23.64, 4.55	0.169
sTNFrII	36	14.57	−19.19, 48.33	0.372	23.89	−11.45, 59.23	0.170	26.67	−9.73, 63.08	0.139	25.22	−31.18, 81.63	0.356
vWF	34	−0.00	−0.04, 0.04	0.956	−0.00	−0.05, 0.04	0.876	−0.01	−0.06, 0.04	0.766	−0.02	−0.10, 0.05	0.504
% Predicted Spirometry
FEV1	40	−1.28	−2.04, −0.51	0.003	−1.45	−2.26, −0.64	0.002	−1.51	−2.35, −0.67	0.002	−2.02	−3.43, −0.61	0.008
FVC	39	0.50	−0.41, 1.40	0.260	0.43	−0.54, 1.40	0.358	0.42	−0.58, 1.43	0.386	0.46	−1.06, 1.99	0.527
MMEF	39	−2.90	−5.38, −0.41	0.025	−2.98	−5.76, −0.21	0.037	−3.08	−5.98, −0.19	0.038	−5.56	−9.83, −1.29	0.014
PEFR	42	2.49	−1.10, 6.08	0.163	2.22	−1.82, 6.27	0.263	2.17	−2.07, 6.42	0.296	3.30	−3.08, 9.68	0.292
Exhaled Nitric Oxide
log(FeNO50,pred)	41	0.01	−0.03, 0.04	0.723	−0.00	−0.04, 0.04	0.910	−0.01	−0.05, 0.04	0.764	−0.02	−0.08, 0.04	0.481
CANO	41	−0.02	−0.10, 0.05	0.497	−0.05	−0.13, 0.03	0.229	−0.06	−0.14, 0.03	0.173	−0.10	−0.22, 0.02	0.107
log(CAWNO)	41	0.00	−0.06, 0.07	0.913	0.01	−0.06, 0.08	0.767	0.01	−0.07, 0.08	0.799	0.02	−0.10, 0.13	0.755
log(DAWNO)	41	−0.03	−0.13, 0.07	0.541	−0.06	−0.17, 0.05	0.289	−0.06	−0.18, 0.05	0.282	−0.12	−0.29, 0.05	0.150
SOURCE FACTORS
		PM Mass			Traffic			Airport UFPs			Secondary Photochemistry
Outcome	N	Est	95% CI	p-value	Est	95% CI	p-value	Est	95% CI	p-value	Est	95% CI	p-value
Cytokines
IL-6	36	−0.15	−0.31, 0.01	0.071	−0.06	−0.24, 0.11	0.457	0.21	0.08, 0.34	0.003	−0.20	−0.39, −0.01	0.044
Fibrinogen	35	−3.28	−12.52, 5.96	0.459	−2.60	−11.95, 6.74	0.560	1.74	−7.73, 11.21	0.699	−5.28	−16.36, 5.81	0.325
sTNFrII	36	13.77	−22.13, 49.68	0.426	36.47	6.03, 66.91	0.022	−5.21	−40.30, 29.89	0.756	27.83	−14.21, 69.87	0.179
vWF	34	−0.00	−0.05, 0.04	0.888	−0.00	−0.05, 0.05	0.933	−0.01	−0.05, 0.04	0.768	−0.01	−0.06, 0.05	0.777
% Predicted Spirometry
FEV1	40	−1.31	−2.14, −0.49	0.004	−1.52	−2.28, −0.77	0.001	0.37	−0.62, 1.36	0.438	−1.55	−2.55, −0.54	0.005
FVC	39	0.52	−0.44, 1.48	0.270	0.25	−0.68, 1.19	0.574	−0.04	−0.91, 0.83	0.920	0.45	−0.68, 1.58	0.413
MMEF	39	−3.12	−5.75, −0.48	0.024	−1.48	−4.50, 1.53	0.311	1.49	−1.18, 4.17	0.252	−2.95	−6.43, 0.52	0.090
PEFR	42	2.70	−1.08, 6.47	0.151	0.39	−3.77, 4.56	0.845	−0.19	−4.08, 3.70	0.920	2.36	−2.26, 6.97	0.298
Exhaled Nitric Oxide
log(FeNO50,pred)	41	0.01	−0.03, 0.04	0.772	−0.00	−0.05, 0.04	0.873	−0.01	−0.05, 0.02	0.514	−0.01	−0.06, 0.03	0.568
CANO	41	−0.03	−0.11, 0.05	0.446	−0.05	−0.13, 0.04	0.244	−0.03	−0.10, 0.05	0.454	−0.08	−0.16, 0.01	0.073
log(CAWNO)	41	0.00	−0.07, 0.07	0.934	−0.00	−0.08, 0.07	0.947	−0.05	−0.11, 0.01	0.090	−0.01	−0.09, 0.08	0.884
log(DAWNO)	41	−0.03	−0.14, 0.07	0.518	−0.03	−0.14, 0.09	0.615	0.09	−0.00, 0.18	0.062	−0.05	−0.18, 0.08	0.464

Exposures: PN=Ultrafine Particle Number; LDSA=Lung-deposited surface area; PM₁ PM_2,5, PM₄, PM₁₀=Particle Mass in the 1, 2.5, 4 and 10 μm size fraction; BC=Black Carbon; PB-PAH=Particle-bound Polyaromatic Hydrocarbons; O3=Ozone

Outcomes: Cytokines: IL6 = High-sensitivity Interleukin-6; sTNFrll = Soluble TNF receptor II; vWF = Von Willebrand Factor. Exhaled Nitric Oxide: FeNO50 = Predictedexhaled nitric oxide at 50ml/s flow rate; C_ANO = Distal Alveolar Nitric Oxide; C_awNO = Airway Wall Nitric Oxide; D_awNO = Diffusivity. Lung Function (% predicted): FEV₁ = Forced Expiratory Volume in 1 second; FVC = Forced Vital Capacity; MMEF = Maximum Mid-expiratory Flow; PEFR = Peak Expiratory Flow Rate.

All reported effect estimates are scaled to one standard deviation change in the exposure of interest.

Table 6.

Multi-pollutant model associations of measured concentrations and modeled source factor contributions with cytokines, exhaled nitric oxide and spirometry outcomes.

POLLUTANTS
		PN			PM25			BC			O3
Outcome	N	Est	95% CI	p-value	Est	95% CI	p-value	Est	95% CI	p-value	Est	95% CI	p-value
Cytokines
IL-6	36	0.13	−0.02, 0.27	0.087	−0.28	−0.52, −0.05	0.023	0.13	−0.23, 0.50	0.438	0.05	−0.18, 0.28	0.659
Fibrinogen	35	−2.18	−15.17, 10.81	0.719	−9.56	−28.69, 9.56	0.294	10.07	−17.98, 38.13	0.446	7.31	−9.83, 24.44	0.368
sTNFrII	36	−3.11	−38.21, 31.99	0.850	−9.14	−69.37, 51.08	0.746	44.07	−48.54, 136.69	0.320	−33.28	−84.40, 17.83	0.181
vWF	34	−0.02	−0.10, 0.05	0.529	−0.02	−0.13, 0.09	0.659	0.03	−0.13, 0.20	0.646	0.02	−0.10, 0.14	0.702
% Predicted Spirometry
FEV1	40	−0.75	−1.55, 0.05	0.065	−1.92	−3.13, −0.70	0.005	1.12	−0.71, 2.94	0.209	1.51	0.27, 2.75	0.021
FVC	39	0.24	−0.94, 1.43	0.661	0.75	−1.21,2.72	0.418	−0.40	−3.25, 2.46	0.767	0.38	−1.56, 2.31	0.679
MMEF	39	−0.87	−4.19, 2.46	0.581	−5.31	−10.37, −0.25	0.041	4.25	−3.27, 11.77	0.242	1.42	−3.86, 6.71	0.569
PEFR	42	1.32	−3.92, 6.57	0.599	5.48	−2.14, 13.10	0.146	−5.47	−17.54, 6.61	0.350	1.00	−6.59, 8.60	0.782
Exhaled Nitric Oxide
log(FeNO50,pred)	41	−0.03	−0.08, 0.03	0.347	−0.02	−0.10, 0.06	0.617	0.03	−0.09, 0.15	0.596	0.00	−0.07, 0.07	0.995
CANO	41	−0.07	−0.18, 0.03	0.168	−0.08	−0.22, 0.07	0.295	0.05	−0.17, 0.28	0.610	−0.01	−0.15, 0.13	0.856
log(CAWNO)	41	−0.08	−0.16, 0.01	0.085	−0.02	−0.15, 0.10	0.697	0.06	−0.12, 0.24	0.501	−0.02	−0.13, 0.09	0.649
log(DAWNO)	41	0.10	−0.03, 0.23	0.129	−0.02	−0.21, 0.17	0.816	−0.06	−0.34, 0.23	0.678	0.05	−0.11, 0.21	0.506
SOURCE FACTORS
		PM Mass			Traffic			Airport UFPS			secondary photochemistry
Outcome	N	Est	95% CI	p-value	Est	95% CI	p-value	Est	95% CI	p-value	Est	95% CI	p-value
Cytokines
IL-6	36	0.05	−0.20, 0.31	0.652	0.09	−0.14, 0.31	0.423	0.18	0.04, 0.32	0.017	−0.29	−0.64, 0.07	0.102
Fibrinogen	35	1.24	−20.42, 22.90	0.902	2.63	−17.06, 22.31	0.775	0.26	−12.07, 12.58	0.964	−9.09	−38.13, 19.96	0.505
sTNFrII	36	−33.72	−97.38, 29.93	0.271	64.38	6.30, 122.46	0.033	−16.80	−51.64, 18.05	0.314	−6.45	−95.94, 83.04	0.878
vWF	34	0.00	−0.11, 0.11	0.981	0.01	−0.10, 0.12	0.785	−0.01	−0.07, 0.05	0.707	−0.03	−0.17, 0.12	0.709
% Predicted Spirometry
FEV1	40	−0.63	−2.04, 0.78	0.353	−1.35	−2.80, 0.10	0.066	0.14	−0.76, 1.03	0.746	0.37	−1.58, 2.32	0.686
FVC	39	0.83	−1.13, 2.79	0.376	−0.46	−2.48, 1.55	0.626	0.33	−0.87, 1.52	0.564	0.27	−2.30, 2.84	0.821
MMEF	39	−3.48	−8.81, 1.84	0.180	1.80	−3.70, 7.31	0.489	−0.42	−3.89, 3.05	0.795	−1.66	−8.95, 5.62	0.628
PEFR	42	5.00	−1.87, 11.87	0.142	−5.57	−12.95, 1.80	0.128	2.02	−2.36, 6.41	0.341	3.32	−5.70, 12.34	0.445
Exhaled Nitric Oxide
log(FeNO50,pred)	41	0.02	−0.05, 0.10	0.539	0.01	−0.07, 0.10	0.756	−0.01	−0.06, 0.04	0.686	−0.05	−0.14, 0.04	0.298
CANO	41	0.01	−0.13, 0.15	0.878	0.04	−0.11, 0.19	0.595	−0.04	−0.13, 0.05	0.315	−0.13	−0.31, 0.04	0.120
log(CAWNO)	41	−0.04	−0.17, 0.09	0.494	0.04	−0.10, 0.17	0.546	−0.07	−0.15, 0.01	0.075	−0.02	−0.18, 0.13	0.758
log(DAWNO)	41	0.08	−0.13, 0.28	0.440	−0.05	−0.25, 0.15	0.580	0.11	−0.01, 0.23	0.077	−0.03	−0.28, 0.21	0.768

Exposures: PN=Ultrafine Particle Number; LDSA=Lung-deposited surface area; PM2 5=Particle Mass in the 2.5 pm size fraction; BC=Black Carbon; O3=Ozone

Outcomes: Cytokines: IL6 = High-sensitivity Interleukin-6; sTNFrII = Soluble TNF receptor II; vWF = Von Willebrand Factor. Exhaled Nitric Oxide: FeNO50 = Predicted exhaled nitric oxide at 50ml/s flow rate; CaNO = Distal Alveolar Nitric Oxide;CawNO = Airway Wall Nitric Oxide; DawNO = Diffusivity. Lung Function (% predicted): FEV1 = Forced Expiratory Volume in 1 second; FVC = Forced Vital Capacity; MMEF = Maximum Mid-expiratory Flow; PEFR = Peak Expiratory Flow Rate

All reported effect estimates are scaled to one standard deviation change in the exposure of interest.

The strongest evidence for associations were for the ‘Airport UFPs’ source with IL-6, PM_2.5 and ‘Traffic’ with FEV₁, and ‘Traffic’ with sTNFrII. The ‘Airport UFPs’ source – characterized by high PN and low particle size, our main hypothesized exposure of interest - was significantly associated with IL-6 in all models (0.18, 0.04–0.32 in multi-pollutant model) and was robust to all adjustments. The correlation between DiscMini and CPC PN measurements varied by particle size (Figure S2), and health model results were slightly different by instrument (Table S5) with generally stronger IL-6 effects seen with the CPC. Contrary to what we expected, IL-6 had a stronger association with PN than LDSA. None of the other systemic or pulmonary inflammation or lung function metrics were positively associated with PN or the ‘Airport UFPs’ source in our study.

For lung function, measured PM (PM₁, PM_2.5, PM₄ and PM₁₀) and the modeled ‘PM Mass’ source were all associated with lower FEV₁ and MMEF in single-pollutant models. For example, a 1 SD increase in PM_2.5 (7.6 μg/m³) was associated with 1.45% and 2.98% drop in % predicted FEV₁ and MMEF, respectively. Effect estimates were even larger for PM₁₀ (2.02% and 5.56%, respectively). Similarly, in multi-pollutant models, PM_2.5 was associated with 1.92% and 5.31% drop in % predicted FEV₁ and MMEF, respectively. Measured PM_2.5 was more strongly associated with lower FEV₁ and MMEF compared to the modeled ‘PM Mass’ source in all models. FEV₁ was also negatively associated with BC (−1.60, −2.68 – −0.51) in single-pollutant models and the modeled ‘Traffic’ source in the single-pollutant model (−1.52, −2.28 – −0.77).

sTNFrII had consistent and significant positive associations with the modeled ‘Traffic’ source factor in single- and multi-pollutant models, and with measured BC and PB-PAH in single-pollutant models. In single-pollutant models, sTNFrII increased by: 36.5 pg/ml (95% CI 6.0 – 66.9) per SD increase in ‘Traffic’, 49.4 pg/ml (10.2 – 88.6) per SD (292 ng/m³) increase in BC, and 30.2 pg/ml (1.6 – 58.9) per SD (1.5 μg/m³) increase in PB-PAHs. In multi-pollutant models, the ‘Traffic’ effect increased to 64.4 pg/ml (6.3 – 122.5).

Less consistent associations were observed with the other measured pollutants or modeled source factors and other health outcomes. A significant negative association of PM_2.5 mass with IL-6 was found in single- and multi-pollutant models; however, the ‘PM mass’ source factor and IL-6 association was marginally significant (negative) in single-pollutant models but positive and non-significant in multi-pollutant models. PN exposure was associated with decreased log(C_AWNO) in single- and two-pollutant models; however, this association became non-significant in multi-pollutant models. Finally, O₃ exhibited results that were contrary to the expected direction in single- and two-pollutant models with FEV₁ and sTNFrII and with FEV₁ in multi-pollutant models. Similarly, ‘Secondary Photochemistry’ exhibited associations in the opposite direction of what is expected for IL-6 (single-pollutant model) and C_ANO (adjusted for ‘Airport UFPs’); however, all associations became non-significant in multi-pollutant models.

Models with significant interaction terms (p<0.05) are reported in Figure S6 and Table S4. Given the limited sample size, multiple tests, and underpowered statistical analysis of interactions, these results should only be interpreted qualitatively. While interaction results were generally inconsistent, Hispanic ethnicity was associated with poorer % predicted PEFR following ‘Airport UFPs’ exposure compared to non-Hispanic ethnicity; whereas, being non-Hispanic was associated with higher log(D_AWNO) response following PN exposure. Finally, having high muscle mass (> median 45.1kg) and being sick in the last month were ‘protective’ following ‘Airport UFPs’ and PN exposure, respectively.

DISCUSSION

We conducted a crossover panel study with a quasi-experimental design modeled after the McCreanor et al. (2007) study to investigate the effects of real-life exposure to airport-related UFPs on acute respiratory and systemic outcomes in 22 adults with asthma. Air pollution measurements and modeled source factor contributions reflected expected patterns at the two sites, and across both seasons of the study. We found significant increases in markers of systemic inflammation associated with ‘Airport UFPs’ (IL-6) and ‘Traffic’ (sTNFrII) exposure and a significant decrease in FEV₁ associated with measured PM and BC and modeled ‘Traffic’ exposure. The robust IL-6 effects we found with the ‘Airport UFPs’ source, which would have been masked by considering PN alone, suggest that some characteristic of the airport-related air pollution mixture as a whole might be more important for IL-6 response than particle number concentration. This could be the smaller particle size and alveolar deposition potential of airport-related UFPs (compared to overall PN which comingles airport and traffic contributions) or other gaseous, volatile or non-volatile components of the mixture that we did not measure or account for. To our knowledge, this is the first study to document acute systemic inflammation following airport-related UFPs exposure.

Most previous studies have investigated total or traffic-related personal UFP exposures. Buonanno et al. (2013) conducted personal monitoring for two days and found daily UFP alveolar-deposited surface area dose to be associated with increased exhaled nitric oxide and decreased FEV₁ (−0.0025 ± 0.0012 % per 100 mm² alveolar deposited surface area dose) in children with asthma and children with house dust mite allergies but no asthma. However, these children’s daily UFP dose was dominated by indoor microenvironments (15% indoor home, 19% sleeping and 18% school) with a likely substantially different composition due to indoor UFP sources (Deffner et al. 2016; Gu et al. 2015; Vu et al. 2017; Wallace 2006; Weichenthal et al.) as compared to our study.

Steenhof et al (2013) exposed 31 healthy volunteers to air pollution for 5 hours while exercising at 3 of 5 sites in the Netherlands (2 traffic, 1 underground train station, 1 farm and 1 urban background site) and found NO₂ effects on proinflammatory cytokines measured in nasal lavage but no PN effects, while Janssen et al (2015) found significant associations between measures of oxidative potential from 3 a-cellular assays with increased eNO and IL-6 in nasal lavage 2 hours post exposure at all four outdoor sites (not including the underground metal-rich site). While not directly comparable to our study, these findings support the role of oxidative stress in acute inflammatory response following urban air pollution exposures and highlight the importance of considering composition.

In a panel study of 29 elderly subjects with coronary artery disease, Delfino et al. (2008) found a 7,337 particles.cm⁻³increase in outdoor PN was significantly associated with 0.50 pg/ml increase in IL-6 and 153.24 pg/ml increase in sTNFrII. PN and PM_0.25 (PM mass in the quasi-ultrafine size fraction, <0.25 μm) were also more strongly associated with IL-6 and sTNFrII than PM_0.25–2.5 mass (Delfino et al. 2009). A 0.56 ng/m³ increase in outdoor total PAHs was associated with 135 (45 – 225) pg/ml increase in sTNFrII and 0.27 (0.10 – 0.44) pg/ml increase in IL-6 (Delfino et al. 2010). However, PN in this study was mainly traffic-related (0.5 correlation with elemental carbon) and more closely resembled our ‘Traffic’ source with loadings of PN, BC and PB-PAHs. When taking particle composition into account, Delfino et al (2010) found that PM_0.25 associations with IL-6 and sTNFrII were completely confounded by PAHs. The high correlation (0.85) between BC and PB-PAHs in our study meant that we could not include them in the same model; however, the ‘Traffic’ source captured their combined effect on sTNFrII. In general, higher effects were seen in the Delfino et al. studies for IL-6 and sTNFrII compared to our study, and this could be due to the differences in the composition and oxidative potential of the exposure mixtures (Delfino et al. 2011), or differences in susceptibility of asthmatics compared to elderly participants with a history of coronary artery disease.

In the McCreanor et al. (2007) study, walking for 2 hours in a diesel vehicular traffic zone with elevated PM_2.5, UFP, EC and NO₂ levels on Oxford Street, London, resulted in up to 6.1% and 5.4% decrease in FEV₁ and FVC compared to baseline, respectively, in asthmatics. Similarly, we found a 1.6% and 1.52% drop in % predicted FEV₁ two hours post BC and ‘Traffic’ exposure, respectively. In addition, we found that measured PM_2.5 was more strongly associated with reduced FEV₁ and MMEF than the modeled ‘PM Mass’ source, and that the PM₁₀ size fraction had the largest effect on these lung function outcomes, suggesting that the actual PM mass or amount inhaled plays a role in worsening lung function, potentially related to increased burden on the lungs to clear particles from the airways.

FeNO₅₀ and airway NO source parameters were not associated with PN in our study, although associations have been previously reported in the literature (Buonanno et al. 2013; Strak et al. 2012). We also did not find any fibrinogen or vWF associations as previously reported in patients with chronic obstructive pulmonary disease (Hildebrandt et al. 2009).

Strengths of our study include: a randomized cross-over within-person, semi-experimental design; a susceptible study population (adults with asthma); participants who performed moderate-light activity to increase ventilation rates; randomized assignments to control and exposure scenarios with a 1+ week washout period in-between; exposures to real-life airport emissions; and the high exposure contrasts achieved at the two exposure locations. Using multi-pollutant measurements and source apportionment modeling, we distinguished the contribution of aviation activities at LAX from traffic, another major source of UFPs in this urban area. In addition, the use of personal monitoring accurately captured exposures in the breathing zone, while the DiscMini diffusion charger provided more detailed particle size and lung deposited surface area. Limitations of our study include a short follow up time, with only one health assessment ~2 hours immediately after the walking exposure period, and the limited sample size in this pilot study that reduced statistical power. We were also unable to adjust for the variable inhalation rates across subjects due to varying levels of fitness, age, etc.. but ensured an almost identical walking pace on all study days.

One of the biggest sources of uncertainty in estimating acute and chronic health effects of UFPs in epidemiological studies lies in the exposure assessment as noted by a European expert panel (Hoek et al. 2010). Specifically, for future airport-related UFP health investigations, it is important to consider the entire source to receptor pathway to accurately assess exposures and estimate health effects, starting from emissions, composition, fate and transport, exposures and confounding factors in the population of interest.

At low power conditions (thrust <30%), commercial aircraft gas turbine engine emissions are dominated by organics –a variety of unburned hydrocarbons (ethylene, formaldehyde, acetaldehyde, and benzene) and lubrication oils. Whereas, higher power conditions are dominated (~80%) by soot or elemental carbon particles referred to as the non-volatile PM fraction (nvPM, the regulated fraction), which directly correlates with the fuel sulfur content (Onasch et al. 2009). As the plume cools downstream of the exhaust, volatile PM forms by two main processes: nucleation of exhaust gases such as SO_x creating new particles (<20nm, high PN and low mass) or condensation of gases onto existing soot particles (see Whitefield et al. (2011; 2008) for a detailed overview). Nucleation typically outnumbers condensation by a factor of 10 to 100 and is also dependent on fuel sulfur content (Lobo et al. 2007; Timko et al. 2010; Timko et al. 2013; Wong et al. 2015). Secondary organic aerosol formation in the aging plume likely exceeds primary organic aerosol emissions (Herndon et al. 2008; Presto et al. 2011). This is why measurements taken at the point of exit from the engine typically underestimate particle mass downwind by a factor of 5 to 10 (Timko et al. 2013).

As for composition, emitted nucleation mode particles are rich in carbon, oxygen, sulfur and chlorine (Mazaheri et al. 2013), and the oxidative reactivity of emitted soot particles is inversely proportional to thrust (Liati et al. 2014). Lubrication oil and incomplete combustion products are the primary sources of organics in emitted particles (Timko et al. 2010). Cross et al. (2013) resolved aliphatic, aromatic and oxygenated organics in aircraft emissions, mainly from unburned fuel at idling and from pyrolysis products at higher power. Timko et al. (2014) identified two lubrication oil factors, two aliphatic factors - one related to soot emissions and another to mixing with ambient organic aerosol – and a fifth factor related to benzene emissions at low thrust using the Positive Matrix Factorization (PMF) model.

Several modeling approaches have been used to predict the fine spatial and temporal variability in PN and separate the contribution of aircraft flight activity from other outdoor important UFP sources - namely traffic, fuel combustion, and secondary formation - ranging from statistical regression approaches (Diez et al. 2012; Hsu et al. 2012) to source-oriented and receptor-oriented source apportionment models. Source-oriented models include simple dispersion models such as a AERMOD that might perform well near the source but do not handle the complicated UFP particle dynamics and chemical transformations that are crucial determinants of the volatile PM fraction (Levy et al. 2015). More sophisticated source-oriented models include chemical transport models such as the Community Multiscale Air Quality (CMAQ) model that generally have lower spatial resolution but account for all sources and emissions in an urban area and fully model fate and transport with proper treatment of chemistry and particle dynamics and typically larger spatial domains that can capture communities further downwind (Arunachalam et al. 2011; Kukkonen et al. 2016; Levy et al. 2008; Levy et al. 2015; Levy et al. 2012). Receptor-oriented source apportionment models such as PMF or PCA used in our study have proven valuable for determining source impacts at affected communities and disentangling the airport signal from other potentially correlated UFP sources in the air pollution mixture (Masiol et al. 2016).

For all modeling efforts, detailed meteorological data and multiple pollutant measurements, including gases, semi-volatiles and particulate matter characteristics (composition, size distribution, particle number concentration, etc.) are recommended to characterize the mixture and obtain the best performance, especially in receptor models. While particle size and PN ratios relative to BC have been used to separate aircraft from traffic signals (Riley et al. 2016), an inert and unique chemical tracer of aircraft emissions would be ideal to facilitate source separation and minimize factor smearing in receptor models - possibly from the jet fuel formulation, lubrication oil additives or other compounds uniquely emitted by aircraft engines. The property of non-reactivity or known chemical reactivity where the species is conserved would facilitate the separation of aircraft impacts in fresh emissions as well as in more aged plumes downwind of airports.

Outdoor exposure estimates should be combined with information on individuals’ time-activity patterns and UFP infiltration efficiency indoors to disentangle indoor- from outdoor-generated UFPs and isolate aviation/airport contributions to total personal UFP exposure. Cooking, smoking, burning wood, candles or incense, and cleaning are some of the indoor UFP sources (Habre et al. 2014; Vu et al. 2017; Wallace 2006; Wallace et al.). UFPs are generally less efficient at penetrating indoors compared to PM_2.5, with infiltration factors (F_inf) ranging from around 0 (particles < 10nm) to 0.3 (particles between 80 and 100 nm) with windows closed and from 0 to 0.6 with one window open in a test house (Rim et al. 2010). Kearney et al. (2014) found large variability in UFP F_inf both within and between homes in Edmonton, with the majority of indoor UFPs being of indoor origin (contrary to indoor PM_2.5). Confounding from co-occurring exposures such as noise or socioeconomic factors related to health disparities should also be adjusted for in epidemiological studies of aviation-related UFP exposures. Finally, recent advances in miniaturization of personal UFP monitors combined with detailed time-activity and geolocation tracking to capture individuals’ behaviors and time spent in various microenvironments can prove crucial in estimating the contribution of aviation-related sources to total personal UFP exposure, especially in heavily exposed occupational subgroups such as baggage handlers (Moller et al. 2014; Moller et al. 2017).

In conclusion, and up to our knowledge, our study is the first to demonstrate increased acute systemic inflammation following exposure to airport-related UFPs. These effects were distinct from traffic-related exposures. Further research is needed to replicate these findings in different susceptible populations and at longer time lags to determine downstream health effects, especially in communities heavily impacted by multiple environmental exposures. This study also emphasizes the importance of multi-pollutant measurements and modeling techniques to disentangle sources of UFPs contributing to the complex urban air pollution mixture and to evaluate population health risks.

Table 4.

Loading profiles (eigenvalues) of air pollution source factors resolved by principal components analysis.

	SOURCE FACTORS
Pollutant	Airport UFPs	PM Mass	Traffic	Secondary Photochemistry
PN (personal DiscMini)	0.72	0.00	0.16	0.20
PN (stationary CPC)	0.71	−0.02	0.35	0.19
Particle Size	−0.81	0.01	0.23	0.15
PM1	−0.04	0.93	0.07	0.06
PM2.5	−0.10	0.63	0.09	0.47
PM10	0.07	0.98	−0.07	−0.08
BC	0.05	0.17	0.76	−0.14
CO2	−0.03	−0.10	0.83	−0.10
PB-PAH	0.19	0.07	0.59	−0.21
O3	0.19	0.06	−0.63	0.68