Article Text

Download PDFPDF

Original article
Relations of exhaled nitric oxide and FEV1 to personal endotoxin exposure in schoolchildren with asthma
  1. Ralph J Delfino1,
  2. Norbert Staimer1,
  3. Thomas Tjoa1,
  4. Daniel L Gillen2
  1. 1Department of Epidemiology, School of Medicine, University of California, Irvine, California, USA
  2. 2Department of Statistics, School of Information and Computer Sciences, University of California, Irvine, California, USA
  1. Correspondence to Dr Ralph J Delfino, Department of Epidemiology, School of Medicine, University of California, Irvine, 224 Irvine Hall, Irvine, CA 92697-7550, USA; rdelfino{at}uci.edu

Abstract

Background Asthma prevalence and acute exacerbations have been associated with endotoxin exposure. However, there are limited data on relations between acute asthma outcomes in children and personal exposure to endotoxin or whether this relation is modified by personal air pollution exposures.

Methods We made repeated measurements of the fractional concentration of exhaled NO (FeNO), forced expiratory volume in 1 s (FEV1) and personal endotoxin exposures in patients with persistent asthma aged 9–18 years, each of whom was followed for 10 consecutive days in Riverside and Whittier, California. Endotoxin was measured in PM2.5, and simultaneously we measured personal exposure to air pollutants: NO2 and PM2.5 mass, elemental carbon and organic carbon. Endotoxin exposure–response relations and interactions between endotoxin and air pollutants were analysed with mixed models controlling for personal temperature, humidity and the 10-day period.

Results Neither percent-predicted FEV1 nor FeNO was associated with personal endotoxin overall; however, endotoxin was associated with FEV1 among patients with average percent-predicted FEV1<80%. When NO2 was above its median, FeNO increased by 2.2% (95% CI −0.8% to 5.2%) for an interquartile increase in personal endotoxin, whereas FeNO was lower by −1.8% (95% CI −4% to 0.5%) when NO2 was≤its median. However, this is out of 12 interaction tests between personal endotoxin and a binary air pollutant for each outcome (FEV1 and FeNO), and there were no interactions with any continuous-scaled pollutant.

Conclusions Personal endotoxin exposure was not associated with acute daily changes in FeNO or FEV1 in a cohort panel of schoolchildren with asthma, except for decreased FEV1 among patients with more severe asthma (percent-predicted FEV1<80%). There was limited evidence of effect modification of endotoxin by personal exposure to air pollution.

  • exhaled nitric oxide
  • endotoxin

Statistics from Altmetric.com

What this paper adds

  • There are limited data on relations between acute asthma outcomes in children and personal exposure to endotoxin or whether this relation is modified by personal air pollution exposures.

  • This is the first report on the relation between a biomarker of airway inflammation (exhaled nitric oxide) and personal endotoxin exposure in schoolchildren with asthma.

  • Personal endotoxin was not associated with daily changes in exhaled nitric oxide.

  • Personal endotoxin was not associated with daily changes in expiratory lung function except for patients with more severe asthma.

  • There was little evidence of an interaction of endotoxin with higher versus lower personal air pollution in relation to the outcomes.

Introduction

Endotoxin is a cell wall component (lipopolysaccharide) of the outer membrane of Gram-negative bacteria that has toxic and immunogenic effects.1 It is carried on particles from sources related to pets, farm and other animals, and agricultural activities. Endotoxin is found in inhalable particles <10 µm in diameter (PM10) and also in respirable particles <2.5 µm in diameter (PM2.5). The PM2.5 fraction is more relevant than coarser particles to lower airway dose and inflammation. Increased systemic inflammation and airway inflammation with activation and migration of neutrophils has been demonstrated with experimental inhalation of endotoxin in humans.2 ,3 Population exposures to endotoxin have been associated with increased asthma prevalence,4 and with acute exacerbation of asthma,1 ,5 although many studies have been in occupational settings with levels of exposure higher than in a general urban population. There is also evidence that endotoxin exposure early in life is instead protective against allergic asthma development through a promotion of T helper cell immune responses.6

A previously reported birth cohort study observed that endotoxin concentrations in settled house dust measured in infancy positively interacted with estimated long-term exposure to traffic-related air pollution in relation to risk of persistent wheeze at age 3 years.7 Matsui et al8 analysed relations of asthma outcomes to indoor endotoxin and air pollutants measured up to five times over 1 year in 146 paediatric patients with persistent asthma. They found increased risk of acute visits for asthma exacerbations and asthma symptoms in homes with high versus low endotoxin only when indoor NO2 was below the overall median concentration. This is an unexpected negative interaction with NO2 because a substantial fraction of NO2 is most likely infiltrated from traffic-related sources, although the source of NO2 was unclear in this study.

Experimental data9–11 and human observational data12 support the hypothesis that both endotoxin and traffic-related air pollutants induce non-allergic (neutrophil-mediated) asthma.13 Given this evidence, combined high exposures to both endotoxin and air pollution could lead to additive or even synergistic effects on acute asthma expression. Experimental data confirm this, showing that the pulmonary effect of endotoxin may be enhanced with coexposure to diesel exhaust particles.11 ,14 ,15

There is evidence that house dust endotoxin has marked room-to-room and temporal variability,16–18 and it is expected that further variability in exposures occurs outside the home. Patient-specific personal exposures to airborne endotoxin are expected to be inaccurately represented by house-dust samples or isolated indoor air samples, and thus respiratory effects may not be detected.19 However, there are limited data on relations between acute asthma outcomes in children and personal endotoxin exposure. Rabinovitch et al20 conducted a panel study of 24 schoolchildren with persistent asthma (164 person-days) and found that personal PM10 endotoxin exposures were significantly associated with increased asthma symptoms and decreased FEV1. Both this study and our previous exposure analysis21 observed that personal endotoxin exposure was consistently higher on average than indoor home/school or outdoor home endotoxin and that correlations between personal and fixed site measurements were non-significant.

The present study sought to add to the limited data in the literature on the relations between personal endotoxin and acute asthma outcomes. To the best of our knowledge, this is the first report to evaluate the relationship between repeated measures of the fractional concentration of exhaled nitric oxide (a biomarker of airway inflammation) and personal airborne endotoxin exposure in paediatric patients with asthma. We also assess the potential modification of effects from personal endotoxin exposures on FEV1 and exhaled nitric oxide by concurrent personal exposure to air pollutants, including particles and NO2.

Material and methods

Population and study design

The present cohort panel study included 10 consecutive days of repeated measurements of both health outcomes and exposures in schoolchildren with a confirmed diagnosis of persistent asthma. Patients were ages 9–18 years (average 14.3 years), non-smoking and unexposed to environmental tobacco smoke in the home. Eligible patients had to have mild to moderate persistent asthma, evidenced by requiring at least one of the following: (1) as-needed bronchodilators ≥2 days/week, (2) taking oral or inhaled anti-inflammatory medications, regardless of recent symptom frequency, or (3) <80% predicted FEV1 from baseline spirometry. The overall cohort included two subcohorts. The first group was from Riverside, California and followed from August until early October 2003, and the other group was from Whittier, California in eastern Los Angeles County and followed from July until November 2004. In general, the Riverside region tends to be warmer, drier and mostly downwind of air pollution sources, whereas Whittier is nearest to vehicular emission sources and closer to the Pacific Ocean.

The study protocol was approved by the Institutional Review Board of the University of California, Irvine. Informed written consent was obtained from all patients and one of their legal guardians. We recruited patients via notification of parents about the study by their local public schools.

We followed 45 patients in four 10 day periods in Riverside, California (N=13 patients) and eight 10-day periods in Whittier, California (N=32 patients). Out of the 45 patients, two patients did not have sufficient personal endotoxin data, leaving 43 patients available for the analysis (table 1).

Table 1

Patient characteristics

Health outcomes

FeNO was measured using recommended methods for offline testing,22 with additional validated improvements.23 We asked patients to refrain from exercising, performing spirometry or consuming food or beverages for 1 h before the FeNO procedure. As described elsewhere in an online supplement,24 instances of these events were relatively infrequent (<3% for exercising, food or beverage, and 20% for spirometry) and they did not confound the associations of FeNO with air pollutants. Sampling was conducted by patients under the direction of field technicians who visited the home daily. Patients inhaled orally to total lung capacity and this was followed immediately with an exhalation (slow vital capacity manoeuvre) at a target flow rate of 100 mL/sec25 into an offline device attached to a 1.5 L Mylar reservoir bag (Ionics Inc, Boulder, Colorado, USA). Approximately 200 mL of dead-space air was vented prior to sampling to reduce contamination from upper airway NO.26 To control for inspired ambient NO, we attached an NO/NO2 chemisorbent filter at the air intake of the device and patients breathed through it for two or more tidal breaths before the sampling manoeuvre. Additionally, patients wore a nasal clip to prevent unfiltered air from entering the lungs. We also collected an indoor air sample to assess the effect of indoor NO on FeNO and found that there was no association.24 Patients provided two breath samples at the same time of day each day (late afternoon to early evening ranging between 15:00 and 21:00 h) over the 10-day follow-up period. Mylar bags were sealed, refrigerated at 6°C after collection, and analysed for NO within 20 h using a chemiluminescence NO analyser (NOA 280i Sievers, GE Analytical Instruments, Boulder, Colorado, USA). For the analysis, we used reliable sample pairs identified a priori where differences were ≤3 ppb or ≤10% of the larger reading. This included 356 reliable pairs (84%) in the 43 patients. The regression slope of the reliable sample pairs was nearly 1.0 (R2=0.99). For 68 pairs that did not meet the reliability criteria, six patients dominated the unreliable pairs, suggesting a pattern of non-compliance with the proper procedure. The regression slope of the unreliable sample pairs was 0.7 (R2=0.58).

The spirometry methods and results of a validation study for the present data are presented in Thompson et al.27 Spirometry was self-administered by patients in their home using the ndd EasyOne Frontline Spirometer (ndd Technologies, Chelmsford, Massachusetts, USA). We trained patients during a 5-day run-in period (not analysed). Patient spirometry sessions were completed in the morning, afternoon and evening prior to the use of inhaled β2-agonist bronchodilator medications.

During each session, patients were allowed up to six chances to perform three good spirometry manoeuvres to meet acceptability and repeatability criteria on the ndd device. The device gave visual instructions to patients based on the success or type of error of each attempt. Each curve was evaluated for acceptability (by criteria and visually) and for repeatability.27 Compliance was high (90%) and 87% had two acceptable and reproducible FEV1 manoeuvres. We excluded 5 out of the 43 patients who consistently did not meet the manoeuvre quality thresholds, leaving 38 patients with valid FEV1 data (619 observations). We selected for analysis the highest FEV1 from two acceptable and reproducible manoeuvres. We analysed percent-predicted normal FEV1 based on each patient's height, age, gender and race-ethnicity to standardise measurements across patients and to provide a clinically meaningful outcome.28

Every two waking hours, patients completed a personal digital assistant (PDA) diary to report their use of asthma medications and symptoms. We did not analyse asthma symptoms due to the rarity of reported symptoms sufficient to interfere with daily activities or sleep. We mitigated missed PDA diary prompts with paper diaries and daily technician-administered questionnaires. Self-reported medication use included as-needed inhaled β2-agonist bronchodilators and daily controller (preventive) medications. In addition, we logged daily puff counts from patients’ as-needed inhalers with a pressure-actuated recording device (Doser, Meditrack Products, Easton, Massachusetts, USA).

Exposures

Each patient carried a personal exposure monitor (PEM) for 10 consecutive days during one of 12 exposure assessment periods when we followed 3–4 patients. We followed up patients daily in their residence to exchange PEMs and download PDA and spirometer data. PEMs were carried during waking hours in a backpack and were kept at all times in close proximity but off the ground when patients were unable to carry it. The PEM air inlets were close to the breathing zone while being carried by placing them over the backpack shoulder strap.

The sampler used was the personal DataRAM model 1200 (MIE Inc, Bedford, Massachusetts, USA). Personal measurements included 24 h average PM2.5 endotoxin, EC and OC collected on quartz filters (Whatman Inc, Florham Park, New Jersey, USA) using an attached filter cassette behind a 2.5 μm sharp-cut cyclone that was attached downstream of a PM2.5 nephelometer (flow rate, 4 L/min). We also measured 24 h average personal NO2 using a miniaturised diaphragm pump to sample air at 0.1 L/min through triethanolamine-treated molecular sieve sorbent tubes (SKC, Fullerton, California, USA). NO2 was then measured in our laboratory based on Method 6014 of the National Institute for Occupational Safety and Health. Attached loggers collected personal temperature and relative humidity with (Onset Computer Corp, Pocasset, Massachusetts, USA). We have previously validated both the personal PM2.5 sampler29 and the personal NO2 active sampler.30 Results regarding ambient (regional) air pollution data are presented elsewhere.24 ,31 ,32

Data from an attached motion logger (Onset Computer Corp, Pocasset, Massachusetts, USA) was checked daily to assure compliance. If there was no motion at the expected times such as going to and from school, then participants were not given monetary compensation for that day and the whole day's data were excluded (<6% of data). Air sampler malfunction or problems with filters occurred on 10% of days and filters were not assayed for endotoxin.

Before sampling, we baked quartz filters to remove organic carbon. Following sampling, around 10% of the filter surface area was punched out using heat-sterilised instruments for measurements of OC and EC. Particulate carbon was speciated into OC and EC using the thermal manganese dioxide oxidation technique.33 This procedure left enough filter media for the personal endotoxin assays. For the personal endotoxin measurements, we calculated the remaining surface area for each filter to estimate particle mass using mass data from the 24 h average PEM PM2.5.

For the endotoxin assay, we extracted particles from quartz PM2.5 filters by homogenising the quartz filter membranes with conventional sonication using a high-speed reciprocating instrument (FastPrep, MP Biomedicals, Inc, Solon, Ohio, USA).21 Quartz filters were first placed into pyrogen-free extraction tubes with 4 mL pyrogen-free water and then loaded into the FastPrep and sonicated at 6.5 m/s for 60 s. Extraction tubes were then rotated for 30 min (Dynal Biotech, speed 36), followed by 15 min sonication and clearing of the aqueous extracts of quartz fibres and particles by centrifugation (at 4000 rpm for 5 min, 4°C). We used the undiluted supernatants for endotoxin measurements using the Limulus Amoebocyte Lysate kinetic chromogenic assay according to the manufacturer's protocol (Pyrochrome Associates of Cape Cod, Falmouth, Massachusetts, USA). Field blanks of quartz filters were used as negative controls and extracted and analysed with each set of air samples. We found detectable endotoxin concentrations in 356 daily personal PM2.5 filters analysed for associations with outcomes (median 0.61, range 0.002–25.3 endotoxin units (EU)/m3). All 52 personal field blank filters showed low or non-detectable endotoxin (median 0.004 EU/filter). The detection limit for the overall method was estimated at 0.004 EU/m3 air (non-detects were set to halve this). Within-subject coefficients of variation for personal endotoxin ranged from 69% to 224% (median 116%).

Statistical analysis

Relations between health outcomes and air pollution exposures were analysed using a linear mixed effects regression model originally considered by Janes et al.34 Briefly, let the index i indicate the patient (i=1,…,45) within session j, and t indicate the daily outcome measurement (t=1,…,10). Then outcome measurement, Yi,j,t was modelled asEmbedded Image where ai,j is the random patient intercept nested in session, Zi,j is a vector of patient characteristics specifying such covariates as medication use, Embedded Image is the average exposure for patient i (between-subject (bs) component), Embedded Image is the within-subject (ws) component of interest, which is the assigned exposure at outcome measurement time t for patient i minus the average exposure for the patient, and εi,j,t denotes random within-person error in the outcome measurement. We assumed an autoregressive-1 correlation structure given the observed variability using the SAS V.9.2 procedure Mixed (SAS Institute Inc, Cary, North Carolina, USA). This assumption was assessed via empirical variograms and moment-based estimates of the correlation structure. Regression parameters were estimated via restricted maximum likelihood estimation (REML).

Interactions between personal endotoxin and personal air pollutants were tested in product-term models including both continuous-scaled exposures. We used 2-day average personal air pollutant concentrations because this averaging time was most strongly associated with the outcomes.24 ,31 Stratified results from product term models were also tested for continuous endotoxin with air pollutants dichotomised at the median concentration. Additional product term models were tested to assess whether indicators of asthma severity modified the response to endotoxin. These potential effect modifiers included the use of inhaled corticosteroids and classification of patients with an average during follow-up of <80% predicted FEV1 (N=9) versus ≥80% predicted FEV1. Patients taking inhaled corticosteroids and those with <80% predicted FEV1 are considered to have more severe asthma. Interaction was considered significant where product term p values were at the nominal 0.1 level.

FeNO was not normally distributed and was log-transformed. As the log-transformed distribution of FeNO was approximately symmetric, results are presented as a percentage change in median FeNO (the exponentiated coefficients from the regression model may be interpreted as relative changes in the median of the outcome distribution). A priori adjustments were made for potential time-variant confounders (personal temperature and relative humidity) based on our previous findings.24

For FEV1, we fit transitional models by adjusting for the previous FEV1 measurement to adjust for sinusoidal circadian rhythms that were observed. On the basis of our previous findings,31 a priori adjustments included personal temperature and relative humidity, cumulative daily use of rescue inhalers during the previous day from a pressure-actuated recording device (Doser, Meditrack, S Easton, Massachusetts, USA), and diary-reported recent use of rescue inhalers but excluded observations where patients reported use of rescue inhalers in the 4 h before FEV1 measurements.

We expressed all results as the relative change in median FeNO or predicted FEV1 per IQR (25th–75th centile) increase in each pollutant to standardise effect estimates.

Results

Table 2 shows univariate distributions for daily endotoxin and other exposures. All 52 personal field blank filters showed low or non-detectable endotoxin (median 0.004 EU/filter). Within-subject coefficients of variation for daily personal endotoxin ranged from 69% to 224% (median 116%). This shows that exposure varies sufficiently within each patient from day-to-day to allow for the possibility that acute effects could be detected. Endotoxin was log-normally distributed (mean 2.04±3.71, median 0.61 EU/m3) and was log transformed before analysis.

Table 2

Daily measurements of exposures to personal endotoxin, air pollution and weather

Linear mixed effects regression models in figure 1 show that FeNO was not associated with personal endotoxin. Figure 1 also shows models evaluating interactions between personal endotoxin and air pollutant exposures in relation to FeNO. There was a stronger positive association of FeNO with lag 1 day personal endotoxin when 2-day average personal NO2 was high (interaction p value <0.05; figure 1A). Specifically, on days when NO2 was above the median, FeNO increased by 2.2% (95% CI −0.8% to 5.2%) for an interquartile increase in log-transformed personal endotoxin (roughly equivalent to 2.19 EU /m3 on an untransformed scale), whereas FeNO was lower by −1.8% (95% CI −4.0% to 0.5%) when NO2 was less than or equal to the median concentration. However, none of the individually stratified estimates for this NO2 model were statistically significant. Also, binary personal PM2.5 mass, EC and OC did not modify associations of FeNO with endotoxin.

Figure 1

Associations of FeNO with personal exposure to PM2.5 endotoxin and differences by personal air pollution exposures. Person-days were dichotomised at greater than and less than or equal to the median concentration of 2-day average personal air pollution exposure, including NO2, PM2.5 mass, PM2.5 EC and PM2.5 OC (table 2). Results are from product term models for personal endotoxin and personal air pollution exposure. The percent change in FeNO is in relation to a one-IQR change in the concentration of endotoxin (2.19 endotoxin units/m3), adjusted for personal temperature, personal relative humidity and exposure run. *p<0.05 for a difference in association with FeNO between low versus high personal NO2 exposure.

Figure 2 shows mixed effects models for FEV1, and this outcome was also not significantly associated with personal endotoxin. Figure 2 also shows models evaluating interactions between personal endotoxin and air pollutant exposures in relation to FEV1. Dichotomous personal NO2, PM2.5 mass, EC and OC did not modify associations of FEV1 with endotoxin. Finally, there were no significant interactions (including at p<0.1) between any air pollutants expressed on a continuous scale and log personal endotoxin in relation to outcomes (not shown).

Figure 2

Associations of FEV1 with personal exposure to PM2.5 endotoxin and differences by personal air pollution exposures. Person-days were dichotomised at greater than and less than or equal to the median concentration of 2-day average personal air pollution exposure, including NO2, PM2.5 mass, PM2.5 EC and PM2.5 OC (table 2). Results are from product term models for personal endotoxin and personal air pollution exposure. The change in percent-predicted FEV1 is in relation to a one-IQR change in the concentration of endotoxin (2.19 endotoxin units/m3), adjusted for personal temperature, personal relative humidity, exposure run, cumulative rescue inhaler use during the previous day (device-recorded) and diary-reported recent use of rescue inhalers.

In the analysis of effect modification by classification of patients with an average of <80% (N=9) versus ≥80% predicted FEV1 (N=29) over the course of the study, there was a nominally significant interaction with lag one personal endotoxin (p<0.06) and a significant interaction with 2-day average endotoxin (p<0.005). Patients with low FEV1 experienced a significant 7.7% drop in percent-predicted FEV1 (95% CI −12.3% to −3%) for an interquartile increase in personal endotoxin, whereas patients with a more normal FEV1 showed a non-significant 0.5% drop in percent-predicted FEV1 (95% CI −2.2% to 1.2%). However, for the FeNO outcome, there were no significant interactions (product term p values >0.52) between the percent-predicted FEV1 classification and personal endotoxin (lag 0, lag 1 and 2-day average). Also, there were no significant interactions between inhaled corticosteroid use and personal endotoxin (lag 0, lag 1 or 2-day average) for either FeNO or FEV1 (product term p values>0.22).

Discussion

Personal endotoxin exposure was not associated with acute daily changes in FeNO or FEV1. There are no other comparable studies that have evaluated FeNO and personal endotoxin. The null FEV1 finding contrasts the study by Rabinovitch et al20 that found a significant decrease in FEV1 with personal endotoxin exposure. This was a comparable study to the present one using basically the same design (a cohort panel) and target population, schoolchildren with persistent asthma. However, there may have been differences in susceptibility since that study had a majority of patients who were African-American (75%) compared with only 12% in this study, which had a majority of Hispanic patients (56%). Differences in statistical power do not explain differences since this study had around twice the sample size.

We also found little evidence of interaction with personal air pollution, although there was a suggestion of possible effect modification by higher NO2 on associations between FeNO and the previous day's personal endotoxin exposure. We found that when NO2 was above its median, FeNO increased in relation to personal endotoxin, whereas FeNO was lower in relation to personal endotoxin when NO2 was less than or equal to its median. Since EC is another exposure marker for products of fossil fuel combustion like NO2, and results for personal EC showed no evidence of interaction, it is possible that the NO2 results were spurious under the assumption that it is similarly acting as a surrogate for traffic-related air pollutants (this may not be entirely the case as discussed below). Also, out of 12 product term models for each of the two health outcomes, there was only one product term that reached the p<0.1 significance level. These results are generally consistent with the findings of Matsui et al,8 which had a majority of patients who were African-American (91%). They showed no interactions between indoor endotoxin and indoor NO2 in relation to measurements of FeNO or FEV1 measured up to five times over 1 year in 146 paediatric patients with persistent asthma. In contrast, Matsui et al8 did show unexpected negative interactions between indoor measurements of NO2 and endotoxin on acute visits for asthma exacerbations and asthma symptoms. However, no effect of indoor endotoxin on any outcome for the full cohort was reported, consistent with the results of this study for personal endotoxin. The source of NO2 in that study was not identified, but could have included indoor gas appliances and outdoor air pollutant sources such as traffic. Concentrations of NO2 in that study were low (median 20 ppb), similar to this study (median 27 ppb), and are not expected to have effects on respiratory responses. This view is based on consensus manuscripts on clinical experimental data35 ,36 showing that adverse effects in the airways are only observed at concentrations of NO2 that are a magnitude higher than that observed in this study and that of Matsui et al.8 Nevertheless, personal NO2 exposure may have some direct toxic effect in particular patients with persistent asthma who are more susceptible to longer term exposures than the typical experimental patient who has no or mild asthma and who is exposed for only short durations of 1–6 h.37 Epidemiological studies have frequently shown positive associations of short-term ambient NO2 with respiratory outcomes in patients with asthma.38 However, NO2 may be acting to some degree as a surrogate of exposure to a complex mixture of air pollutants from fossil fuel combustion that instead may be the primary causal components.37 Experimental results show an enhancement in responses related to airway oxidative stress and inflammation with coexposure to diesel exhaust particles and endotoxin.11 ,14 ,15 Therefore, assuming that NO2 is a reasonably good surrogate of products of fossil fuel combustion, the results of this study are nominally coherent with the experimental data, whereas the results of Matsui et al8 are unexpected.

We analysed two factors reflecting asthma severity that may modify the effects of endotoxin, inhaled corticosteroid use and percent-predicted FEV1 below 80%. This included 12 product term models involving the two outcomes and three lags and lag averages for personal endotoxin. There were no significant interactions for inhaled corticosteroid use, but there were two significant interactions between study-average percent-predicted FEV1 below 80% and lag 1 day and 2-day average personal endotoxin in relation to the continuous daily FEV1 outcome. This suggests that patients with more severe asthma based on lung function are at greater risk of acute lung function deficits from increases in personal endotoxin exposure. Although children taking inhaled corticosteroids also have more severe asthma, the lack of findings for them may be attributed to the possibility that inhaled corticosteroids attenuate endotoxin effects.39 Most patients taking inhaled corticosteroids (70%) had ≥80% predicted FEV1, but nearly all patients with <80% predicted FEV1 (89%) were taking inhaled corticosteroids. It is possible that effect modification was observed because patients with lower lung function were a more severe subset who were still responsive to endotoxin.

A key limitation of this study is that we measured endotoxin only in the PM2.5 fraction of PM10 but not in the coarse PM fraction from 2.5 to 10 µm in diameter (PM2.5–10), which is enriched in endotoxin. We did so because the respirable PM2.5 fraction is more relevant to a lower airway dose and thus airway inflammation. However, lower endotoxin concentrations in PM2.5 and the potential presence of a threshold may have limited our ability to detect an overall association. Behbod et al2 reported results of a clinical trial of patients exposed to concentrated ambient particles and found that peripheral blood leucocyte increases in relation to an interquartile increase in endotoxin (5.4 ng/m3) did not differ by particle size. However, elsewhere they reported lower methylation of a toll-like receptor (TLR)-4 by the coarse PM fraction than by the fine PM fraction.40 This could conceivably lead to greater respiratory responses with coarse PM exposures because of an enhanced activation of TLR-4 by endotoxin that would result in greater inflammatory responses.

Another limitation is that the panel includes a small subject number (not measurement) sample size that may not be sufficiently representative of the at-risk population. The measurement sample size of repeated measures in the 43 patients may have been underpowered, leading to the observed lack of overall associations with endotoxin. We performed the statistical power analysis based on a simulation model that we developed for this purpose.41 Power was simulated for the covariate-adjusted main effects of personal endotoxin on FEV1 and FeNO for chosen β’s and for a null hypothesis of no change in outcome per IQR increase in endotoxin, or β=0.0. From this analysis, we conclude that the available sample sizes have adequate power (≥87%) to detect changes of <13.5% in median FeNO for an IQR increase in endotoxin exposures. We also conclude that there was adequate power (≥84%) to detect mean changes of<4.5 percent-predicted FEV1. We also tested the power to detect interaction of endotoxin with a binary air pollutant variable. We estimated the change in outcomes associated with the interaction term (interpretable as the differential increase in the effect of endotoxin attributable to an air pollutant at the upper half relative to the lower half of its distribution). Incorporating the observed SD of 0.142% and within-subject exchangeable correlation of 0.28 for percent-predicted FEV1, we estimate via simulation that the current study design would yield 80% power to detect an interaction coefficient of 8%-predicted FEV1 for an interquartile change in endotoxin. Similarly, incorporating the observed SD of 3.22 log-ppb and within-subject exchangeable correlation of 0.286 for FeNO, we estimate that the current study design would yield 80% power to detect an interaction coefficient of 16.5% FeNO associated with an interquartile change in endotoxin.

Conclusions

Personal endotoxin exposure was not associated with acute daily changes in FeNO or FEV1 in the overall cohort panel of children with persistent asthma and there was limited evidence of effect modification of this relation by personal exposure to air pollution. However, among patients with more severe asthma (study-average percent-predicted FEV1 <80%), we found that daily FEV1 was significantly lower with personal endotoxin exposure. It is possible that adverse effects of endotoxin on asthma are manifest over longer durations of exposure than that measured in this study and this could affect the chronicity of asthma and/or propensity for exacerbations. Longitudinal studies of patients with asthma with repeated measurements of personal endotoxin exposure over long durations of time may be needed to address this issue.

Acknowledgments

The authors thank staff in the Department of Epidemiology and the General Clinical Research Center (National Institutes of Health grant MO1-RR00827), University of California Irvine.

References

Footnotes

  • Contributors All authors have contributed to the planning, conduct and reporting of the work.

  • Funding National Institute of Environmental Health Sciences, U.S. National Institutes of Health (R01 ES11615 and R21 ES019711), and South Coast Air Management District, through the University of California Los Angeles Asthma and Outdoor Air Quality Consortium (Contract No. UCLA-35692).

  • Competing interests None declared.

  • Patient consent Obtained.

  • Ethics approval University of California, Irvine, Institutional Review Board.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Data are available by email request from the corresponding author.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles