Exposure to nitrogen dioxide (NO2), a byproduct of combustion and a respiratory irritant,1,2 can occur both indoors and outdoors. Gas appliances such as gas cooking stoves are primary sources indoors, where children spend large amounts of time. Gas stoves are used by ~39% of U.S. households.3 Indoor levels where NO2 sources are present can be much higher than outdoors, where the primary source of NO2 is traffic. Exposure to NO2 continues to be a public health concern, especially with regard to the respiratory health of children with asthma.
A randomized controlled trial conducted in Australia has provided compelling evidence for an association between indoor NO2 exposure and adverse respiratory outcomes among children with asthma.4 The study, which involved replacing unflued gas heaters in selected schools with flued or electric heat, found improved average asthma morbidity over a 12-week period among students in the intervention schools. Two recent reviews of indoor environmental influences on asthma in children included NO2 as an important potential trigger of asthma morbidity.5,6 Both reviews summarized key studies dating back to the 1980s and concluded that there is limited but suggestive evidence of associations between indoor NO2 exposure and asthma morbidity in children. Asthma morbidity measures used in studies of NO2 exposure include number of symptom days or nights (wheeze, persistent cough, shortness of breath, chest tightness), frequency of rescue medication use, peak expiratory flow, upper and lower respiratory tract symptoms, limited speech, and forced expiratory volume.4,7–11 Many of these outcomes (especially symptoms and medication use) have limitations because they tend to be associated with access to healthcare and other socioeconomic factors. Confounding by these factors may account for some of the persistent inconsistency of asthma morbidity associations in the indoor NO2 exposure literature.
We previously conducted a study of 728 asthmatic children and associations of symptoms with measured indoor NO27 and found increased risks of wheeze and chest tightness associated with increased levels of NO2. Risks were confined, however, to children living in multifamily homes, a study characteristic associated with lower socioeconomic status, higher proportion of gas stove use, and smaller proportion of asthma maintenance medication use. Analysis was based on a single NO2 measurement per child and did not account for other important factors such as atopic status or indoor allergen exposure.
The current analysis characterizes the relationship between measured indoor NO2 and concurrent asthma severity in a repeated measures analysis of a diverse population while considering some common mediating factors, such as atopy, allergen exposure, seasonality, and socioeconomic status.
The Study of Traffic, Air quality and Respiratory health was a prospective, 1-year follow-up study of school-aged children with asthma. From 2006 through 2009, the study enrolled 1401 children recruited through flyers distributed to schools in 23 cities and towns with gas lines in Connecticut and western Massachusetts. Volunteer families contacted the office and were screened (n = 2,175) via telephone. Eligible children (n = 1,642) were age 5–10 years, had a caregiver who spoke English, and had active asthma defined as two or more of the following: physician diagnosis; asthma symptoms within the past 12 months (wheeze, persistent cough, chest tightness, shortness of breath); use of prescription asthma medication within the past 12 months (short-acting rescue medications and maintenance medications including inhaled steroids, systemic steroids, cromolyn, and leukotriene inhibitors). The race/ethnicity distribution of children enrolled (ie, those who completed a home interview and provided a blood sample) was similar to that of the towns where the children resided. Children (n = 1,342) who had complete information for health outcome measures and successful concurrent monitoring of indoor NO2 were included in this analysis.
At the time of enrollment, a research assistant visited the home, obtained consent, and interviewed the mother or primary caregiver (respondent) to obtain demographic data (age, sex, race/ethnicity, mother’s education), and medical history of the child. The research assistant also observed and recorded housing type (single family or multifamily) and cooking appliance (gas or electric) of the enrollment residence. The mother was given a calendar to record daily symptoms and medication use.
At the end of each of the four, month-long monitoring periods, a research assistant phoned the respondent to obtain reports of daily symptoms and medication use and data on smoking in the home during the monitoring period. Sampling seasons were defined by winter and summer solstice and vernal and autumnal equinox. The midpoint of the observation period was used to assign the observation to a season.
At the end of 1 year, an exit interview was conducted via telephone. At this time, a detailed address history was collected and the respondent provided housing characteristics such as housing type and type of cooking stove in each residence during the study. Housing type was later confirmed for all addresses with publicly available tax-assessor records.
At the enrollment visit, the research assistant placed passive monitors (Palmes tubes)12 to measure NO2 in rooms where the child spent the most time awake (dayroom) and asleep (bedroom). After 1 month, the respondent was contacted via telephone and instructed to cap the NO2 monitors and return them in a prepaid mailing envelope provided. Additional monitors were sent at 3-month intervals for repeat sampling.
Palmes tubes were analyzed for NO2 concentration.12 Duplicate samples and field blanks were used for quality control. Regression analysis of duplicate samples (n = 183) produced an adjusted R2 = 0.91 with a slope = 0.96 and intercept = 0.84. Coefficients of variation for the dayroom, dayroom duplicates, bedroom, and bedroom duplicates were 95.3, 94.5, 120.4, and 116.8, respectively. Dayroom and bedroom concentrations of NO2 were highly correlated (r = 0.89). In the present analysis, indoor NO2 concentrations are defined as the average of the two indoor measurements per home for each monitoring period. Measurements matching monitoring periods with complete health data were used for analysis (n = 4,499). Quintile concentration boundaries (in ppb) were ≤4.02, >4.02–6.02, 6.03–8.88, 8.89–14.32, and >14.32.
Environmental Sampling and Allergy Testing
At the enrollment visit, the research assistant collected dust from the main living area for the measurement of common allergens, using a protocol described previously.7,13,14 Dust samples were assayed by enzyme-linked immunosorbent assay for detectable levels of dust mite allergens (Der p 1 ≥ 0.10 µg/g and Der f 1 ≥ 0.10 µg/g), cat allergen (Fel d 1 ≥ 0.12 µg/g), dog allergen (Can f 1 ≥ 0.12 µg/g), and cockroach allergen (Bla g 1 ≥ 0.60U/g).
Using blood samples collected at the time of enrollment, serum for allergy testing was analyzed using the UniCAP system to determine total IgE and specific sensitivity to a panel of 10 allergens. Atopy was defined as a sensitivity to any of the specific allergens or as total IgE exceeding age-adjusted levels.15 For each allergen (Der p 1, Der f 1, Can f 1, Fel d 1, Bla g 1), a binary variable was used that included allergen-specific sensitization and allergen-specific exposure14: for this analysis, “1” indicated a specific sensitivity and detectable allergen in the home and “0” indicated no sensitization to the specific allergen or no detectable allergen in the home.
An asthma severity score based on the Global Initiative for Asthma guidelines16 was constructed for each observation period. The score was composed of two components: a symptom step and a medication step. We defined symptom steps as (0) no symptoms, (1) 1–3 symptom days and 0–2 nights OR 0 days and any nights, (2) 4–19 symptom days OR 1–3 symptom days and 3 or more nights, (3) 20 or more symptom days OR 4–19 days and 5 or more nights, and (4) >20 symptom days and 10 or more nights. Medication steps were defined as (0) no asthma medication use, (1) rescue medication use only, (2) use of one controller medication, (3) simultaneous use of two controller medications, and (4) simultaneous use of three or more controller medications.
Symptom and medication steps were combined to determine overall asthma severity for each child in each monitoring period. A composite severity score of 0 was possible only if no symptoms were experienced and no asthma medication was used (symptom and medication step combination of (0, 0)). A score of 1 (mild transient) was assigned for symptom and medication step combinations of (1, 1), (0, 1), or (1, 0), respectively. A score of 2 (mild persistent) was assigned for symptom and medication step combinations of (2, 0), (2, 1), (0, 2), or (1, 2), respectively. Symptom and medications step combinations of (3, 0), (3, 1), (2, 2), (0, 3), and (1, 3), respectively, were assigned a score of 3 (moderate persistent). Finally, a score of 4 (severe persistent) was assigned if either the symptom or medication step was a 4 OR with symptom and medication step combinations of (3, 2), (3, 3), and (2, 3) (see Fig. 1 in the article by Gent et al).14
Additional outcomes of interest included frequency of wheeze, night symptoms, and use of rescue medication. For analysis, we classified these into categories corresponding to symptom steps for the severity score: “0,” “1–3,” “4–19,” and “more than 19” days per month.
Descriptive statistics and unadjusted associations between health outcomes, quintiles of NO2 exposure, and covariates were computed with SAS version 9.2 (Cary, NC). We examined both unadjusted and adjusted associations with ordered logistic regression (proportional odds model). The proportional odds assumption for all outcomes was tested using NLMIXED in SAS in unadjusted models with quintiles of NO2 exposure.
To allow for repeated measures of the health outcomes and exposure, we used a hierarchical ordered logistic model with a random term for subject. We assumed a normal distribution with unknown variance for subject effects. Associations between health outcomes and NO2 exposures, both unadjusted and adjusted for covariates, were examined using a Bayesian approach with a Markov Chain Monte Carlo strategy implemented in OpenBUGS.17 Bayesian estimates of model parameters were obtained by drawing samples from the posterior distribution using uninformative prior distributions (normal with mean zero and precision 1.0×10–6) for model parameters in the linear predictor, flat priors with ordered ranges for the ordinal parameters, and a gamma prior (with shape = 0.001 and scale = 0.001) specified for precision for the random-subject effect. Estimates for final models were based on a sample of 10,000 iterations with thinning of 20 after burn-in of 20,000 iterations.
Initially, unadjusted models were constructed with exposure represented as quintiles of NO2 concentration. We explored the shape of the exposure-response relationships between health outcomes and NO2 using a natural spline function of the natural log (ln) of NO218 specifying five knots (at NO2 concentrations representing the 10th, 25th, 50th, 75th, and 95th percentiles of the distribution). Posterior means at exposure levels corresponding to the knots indicated that a threshold model would fit the data well and that the threshold was near the boundary of the second and third quintiles of the NO2 distribution. Thus, in adjusted models we combined the bottom two exposure quintiles. Linear trends above the threshold were examined in a fully adjusted model using ln NO2 concentration as a continuous variable. Adjusted models for asthma severity score included age, sex, atopy, season of monitoring, race/ethnicity, mother’s education, smoking in the home, and all five variables for combined specific sensitization and exposure to indoor allergens (Der p 1, Der f 1, Fel d 1, Can f 1, Bla g 1). Models for wheeze, night symptoms, and rescue medication included age, sex, atopy, season of monitoring, and all five variables for combined specific sensitization and exposure to indoor allergens (Der p 1, Der f 1, Fel d 1, Can f 1, and Bla g 1), as well as maintenance medication use (which represents a critical aspect of disease severity not included in these outcome measures). Because of colinearity with maintenance medication use, race/ethnicity, mother’s education, and smoking in the home were excluded from models for wheeze, night symptoms, and rescue medication.
Each monitoring period was 4 weeks long, and all symptom and medication-use day counts were standardized to 28 days. The mean monitoring length was 33 (standard deviation [SD] = 7) days; median = 30 days; and mode = 28 days. This analysis used NO2 concentrations and health outcomes measured concurrently during 4499 monitoring period observations contributed by 1342 subjects. Of these, 870 (65%) subjects contributed complete asthma symptom, medication use, and concurrently measured indoor NO2 data for all monitoring periods; 202 (15%), 143 (11%), and 127 (9%) contributed data for 3, 2, and 1 monitoring periods, respectively. Of 4499 monitoring periods, 1,163 (26%) took place in summer, 1,092 (24%) in fall, 1,117 (25%) in winter, and 1,127 (25%) in spring.
Table 1 describes the enrollment characteristics of the study population. Just over half of children were age 5–7 years (52%) and male (59%). Two-thirds of the population were considered atopic (66%) and used maintenance medication at some point during the year of follow-up (66%). The population was 40% white, 19% African American, and 36% Hispanic. Only 16% of the mothers had less than a high school education, whereas 29% were college graduates. At the time of enrollment, 10% of respondents reported having a smoker in their home. For four of the five allergens, less than one-third of the population was both sensitized and exposed (Der p 1 26%, Der f 1 29%, Fel d 1 29%, Can f 1 27%). Only 7% of children were both sensitized and exposed to cockroach (Bla g 1).
The mean daily indoor NO2 level over all observations was 10.6 (SD = 9.4) ppb, with interquartile range 4.5–12.5 ppb. Table 2 shows the distribution of all indoor NO2 measurements (by quintile) over subject characteristics. White respondents were predominantly in the lower exposure quintiles, whereas African American and Hispanic families fell in the higher quintiles. Among women who did not complete high school, 7% are in the lowest exposure categories, whereas 37% are in the highest exposure categories. Among women who completed college, the distribution is reversed. Nonsmokers were distributed fairly evenly across exposure quintiles, whereas smokers were more often in the heavily exposed category. Indoor NO2 measurements in the highest concentration quintile are most likely in the winter and least likely in the summer. For allergens Der p 1, Der f 1, Can f 1, and Fel d 1, 17% of observations contributed by sensitized and exposed respondents fall into the highest NO2 exposure categories compared with 34% of those contributed by respondents sensitized and exposed to Bla g 1.
Table 3 shows the distribution of asthma severity scores across subject characteristics. The most common level of symptoms was mild persistent (25%), and the least common was mild transient (10%). Atopic children were slightly less likely to be categorized as having no symptoms or medications during a monitoring period than nonatopic participants, but were no more likely to be categorized as severe. There were minor differences by ethnicity. Asthma severity scores were generally lower in the summer months and higher in the fall. Children who were both sensitized and exposed to Der p 1, Der f 1, Fel d 1, and Can f 1 were less likely to be in the severity score category 0 than nonsensitized or unexposed children.
Figure 1 displays the seasonal distributions of health outcomes. A comparison of Figure 1A with Figure 1B–D reveals a flat distribution of scores across asthma severity categories compared with the skewed distributions for categorized days of wheeze and night symptoms and somewhat less skewed distribution for rescue medication use. In general, summer is the season with lowest asthma severity (for all outcomes).
Figure 2 shows distributions of asthma severity score, wheeze, and both rescue and maintenance medication use stratified by mother’s education. The distributions for wheeze (Fig. 2A) and rescue medications (Fig. 2B) are similar: subjects whose mother did not complete high school were more likely to report wheeze (41%) and rescue medication use (54%) compared with children of mothers who completed high school (wheeze 35%, rescue medication use 46%) or college (wheeze 31%, rescue medication 45%). However, children of mothers who completed college were more likely (58%) to report use of maintenance medication compared with children of mothers who did not complete high school (46%) or college (47%) (Fig. 2C). Figure 2D shows that the asthma severity score, which incorporates both symptoms and medication use, is not associated with mother’s education. Because of collinearity between maintenance medication and all socioeconomic variables, models for wheeze, night symptoms, and rescue medication included maintenance medication use (an important indicator of disease status), but did not include race/ethnicity, mother’s education, or smoking in the home.
The proportional odds assumption was satisfied for all outcomes in unadjusted models using quintiles of NO2 exposure. Table 4 presents the results of Bayesian cumulative logistic regression models of associations between health outcomes and NO2 exposure. In unadjusted models, compared with the lowest quintile of exposure (Table 4, unadjusted Model 1), the odds ratios (ORs) for severity score imply a protective effect for exposure to NO2 levels in the second two quintiles and an increased risk for exposure in the higher quintiles. A similar pattern is seen for night symptoms and rescue medication use and suggests a threshold for health effect. Unadjusted models using the combined lowest two quintiles as the reference group are shown in Table 4, unadjusted Model 2.
Figure 3 illustrates, for fully adjusted models, the exposure-response relationships between NO2 and health outcomes using a constrained, natural spline function of ln NO2 and 95% confidence limits as well as threshold functions for each outcome. In adjusted models of NO2 exposure as quintiles (Table 4), levels >14.3 ppb compared with the reference level (≤6 ppb, the threshold value) resulted in an increased risk of a one-level increase in asthma severity score (OR = 1.43 [95% confidence interval [CI] = 1.08–1.88]). These same exposures were also associated with increased risks of wheeze (1.53 [1.16–2.02]), night symptoms (1.59 [1.24–2.01]), and rescue medication use (1.74 [1.34–2.26]). In the fully adjusted threshold models, every 5-fold increase in NO2 exposure >6 ppb was associated with a dose-dependent increase in asthma severity score (1.37 [1.01–1.89]) and asthma morbidity measured by wheeze (1.49 [1.09–2.03]), night symptoms (1.52 [1.16–2.00]), and rescue medication use (1.78 [1.33–2.38]).
In this study of school-aged children, we observed an association of increasing NO2 concentration in the home with asthma severity assessed by a five-level score and with asthma morbidity measured by days of wheeze, night symptoms, and rescue medication use. Analyses were based on repeated measures of both NO2 and asthma outcomes controlling for atopic status and common household allergen exposures.
These associations are consistent with findings in the literature suggesting an association between NO2 exposure at both relatively low and high levels and increased asthma severity and morbidity.4,7,9–11,19 The mean indoor NO2 level over all 4499 observations was 10.6 (SD = 9.4) ppb and was 15.6 (10.4) ppb among observations from homes with gas stoves. Figure 3D (rescue medication use) displays a histogram of NO2 levels measured in all subjects’ homes and in homes with gas stoves. In our previous study, the mean indoor NO2 for all observations was 14.5 (SD = 15.2) ppb and was 25.8 (SD = 18.1) ppb in homes with gas stoves. Figure 1 in that publication7 describes the distribution of NO2 with respect to both stove type and housing type. The lower NO2 levels in this study reflect the expanded use of high-efficiency gas appliances, which can reduce residential gas usage by up to 30%.20 Differences among studies in NO2 distributions also can be attributed to variations in recruitment strategies. We enrolled both urban and suburban children residing in homes with either electric or gas stoves and found a wide distribution of household NO2 exposures.
In our previous study of children with asthma,7 indoor NO2 was associated with respiratory symptoms but only among children in multifamily housing (an indicator of lower socioeconomic status). To compare the two studies, we explored associations between housing type and respiratory symptoms in this study and found that children living in multifamily housing were 75% more likely to wheeze, 68% more likely to have night symptoms, and twice as likely to use rescue medication (data not shown). However, we did not find a differential effect of housing type on the asthma severity score.
An important confounder of the association of indoor NO2 exposure with asthma morbidity is socioeconomic status. Higher NO2 concentrations were found in homes of minority children and children whose mothers reported the fewest years of education (Table 2). These children also reported less use of maintenance medication (Fig. 2). Three of our four outcome measures (frequency of wheeze, night symptoms, and rescue medication use) represent only part of a child’s disease status. For example, a child reporting no wheeze who is also not taking controller medication will have less severe asthma than a child with no wheeze who is taking maintenance medication. To control for this aspect of disease severity (which is not included in the outcome measure), we included maintenance medication use as a covariate in models exploring associations between symptoms and NO2 exposure. Because use of maintenance medication is also associated with socioeconomic status, we did not include additional socioeconomic status variables in the adjusted models for these outcomes. When these additional variables are added, the ORs for the association with NO2 exposure are attenuated and the CIs widen (for wheeze, OR = 1.03 [95% CI = 0.75–1.42]; night symptoms, 1.16 [0.87–1.54]; and rescue medication use, 1.24 [0.91–1.68]).
A strength of our study is that one of our outcome measures, the asthma severity score, incorporates both symptom frequency and medication use. The asthma severity score is not associated with the socioeconomic status variables (Table3) included as covariates in adjusted models.
In the Inner City Asthma Study10 among nonatopic children, those with high NO2 exposure were more likely to have more than four symptom days in a 2-week period and more likely to have peak flow values <80% of predicted values. That study found no association between NO2 exposure and symptoms or peak flow among atopic children. In our study, atopic children were no less likely to experience an increased risk of asthma morbidity associated with increased NO2 than their nonatopic counterparts. This finding is in agreement with the Baltimore Indoor Environment Study of Asthma in Kids,9 which found that atopy did not modify the association between NO2 and asthma symptoms.
Strengths of this study include large sample size, seasonal repeated measurements of NO2 concurrent with measurements of asthma symptoms and medication use, and an asthma severity score not associated with socioeconomic variables. Associations between NO2 and asthma were consistent across all outcome measures. Allergy testing and household allergen sampling at the time of enrollment permitted inclusion of additional important household asthma triggers.14 In addition, the hierarchical analysis permitted estimates of associations between, rather than within, subjects, across homes with different levels of exposure.
The focus of our analysis was on the health effects of indoor exposure to NO2 measured with passive monitors placed in a child’s home where they spend the major portion of their time. One limitation of the passive monitoring method is that it results in an integrated average NO2 concentration and does not allow for the measurement of peak exposures. Sources of NO2 were not part of the statistical model, and in homes without indoor sources (such as gas appliances), the only source of NO2 would be outside the residence. This study included passive monitors placed outside of the residence.21 It remains for future analyses to model the complex relationship between outdoor and indoor levels of NO2 and health effects. For example, when outdoor levels are added as a variable to the adjusted, threshold model for asthma severity score (Table4), the OR for indoor NO2 exposure became 1.21 (0.88–1.67) and 1.31 (0.95–1.83) for outdoor NO2 exposure. One could argue that indoor levels of NO2 already account for a child’s home exposure to outdoor NO2 and adding NO2 concentrations measured outside of a residence results in overcontrolling for indoor levels. An alternative model might be one that adds only “residual” amounts above what is measured indoors. In this alternative model, where only “extra” NO2 not accounted for in the indoor measurement is added, the OR for indoor NO2 exposure on the asthma severity score is 1.52 (1.06–2.18), and the OR for outdoor NO2 exposures is 1.20 (0.98–1.46). The child’s exposure away from home was not assessed either through personal monitoring or by taking measurements in other environments such as school. We would not expect children to be exposed to sources of NO2 (eg, gas stoves, unvented gas heaters) in schools or other nonresidential environments in our study area. Other limitations include the lack of biological measures of asthma (eg, peak flow or spirometric measures) and lack of control for viral respiratory illness (another known trigger of asthma exacerbations with possible potentiating effects on NO2 exposure in asthmatic children).8
Our results contribute to a growing body of literature associating low levels of NO2 exposure with adverse respiratory outcomes in asthmatic children. Furthermore, the apparent threshold for these effects in asthmatic children (6 ppb indoors) was comparable to the 10th percentile of mean levels measured outdoors22—far below the U.S. EPA 53 ppb standard—and with increasing risk of adverse respiratory morbidity above that level.
We gratefully acknowledge Rashele Yarborough and the staff of the Center for Perinatal, Pediatric and Environmental Epidemiology for their contributions to the collection and analysis of study data.
1. Utell M, Frampton MWRoth RA. Oxides of nitrogen. In: Comprehensive Toxicology: Oxides of Nitrogen. 1997 Cambridge, UK Cambridge University Press:303–312
2. Samet JM, Utell MJ. The risk of nitrogen dioxide: what have we learned from epidemiological and clinical studies? Toxicol Ind Health. 1990;6:247–262
3. US Department of Housing and Urban Development and US Census Bureau. . American Housing Survey for the United States: 2009. Available at: http://www.census.gov/prod/2011pubs/h150-09.pdf
Accessed 1 April 2011
4. Pilotto LS, Nitschke M, Smith BJ, et al. Randomized controlled trial of unflued gas heater replacement on respiratory health of asthmatic schoolchildren. Int J Epidemiol. 2004;33:208–214
5. Breysse PN, Diette GB, Matsui EC, Butz AM, Hansel NN, McCormack MC. Indoor air pollution and asthma in children. Proc Am Thorac Soc. 2010;7:102–106
6. Sharma HP, Hansel NN, Matsui E, Diette GB, Eggleston P, Breysse P. Indoor environmental influences on children’s asthma. Pediatr Clin North Am. 2007;54:103–120, ix
7. Belanger K, Gent JF, Triche EW, Bracken MB, Leaderer BP. Association of indoor nitrogen dioxide exposure with respiratory symptoms in children with asthma. Am J Respir Crit Care Med. 2006;173:297–303
8. Chauhan AJ, Inskip HM, Linaker CH, et al. Personal exposure to nitrogen dioxide (NO2
) and the severity of virus-induced asthma in children. Lancet. 2003;361:1939–1944
9. Hansel NN, Breysse PN, McCormack MC, et al. A longitudinal study of indoor nitrogen dioxide levels and respiratory symptoms in inner-city children with asthma. Environ Health Perspect. 2008;116:1428–1432
10. Kattan M, Gergen PJ, Eggleston P, Visness CM, Mitchell HE. Health effects of indoor nitrogen dioxide and passive smoking on urban asthmatic children. J Allergy Clin Immunol. 2007;120:618–624
11. Gillespie-Bennett J, Pierse N, Wickens K, Crane J, Howden-Chapman PHousing Heating and Health Study Research Team. . The respiratory health effects of nitrogen dioxide in children with asthma. Eur Respir J. 2011;38:303–309
12. Palmes ED, Gunnison AF, DiMattio J, Tomczyk C. Personal sampler for nitrogen dioxide. Am Ind Hyg Assoc J. 1976;37:570–577
13. Gent JF, Belanger K, Triche EW, Bracken MB, Beckett WS, Leaderer BP. Association of pediatric asthma severity with exposure to common household dust allergens. Environ Res. 2009;109:768–774
14. Gent JF, Kezik JM, Hill ME, Tsai E, Li DW, Leaderer BP. Household mold and dust allergens: exposure, sensitization and childhood asthma morbidity. Environ Res. 2012;118:86–93
15. Mayo Clinic. . Immunoglobulin E (IgE), Serum. 2006 Available at: http://mayomedicallaboratories.com/test-catalog/pring.pht?unit_code=8159
Accessed 5 April 2011
16. US Department of Health and Human Services. . Global Initiative for Asthma: Global Strategy for Asthma Management and Prevention. 2002 Available at: http://www.ginasthma.org/Guidelines/guidelines-2002-original%3a-workshop-report%2c-global-strategy-for-asthma-management-and-prevention.html
Accessed 1 January 2011
17. MRC Biostatistics Unit. . The BUGS Project. Available at: http://www.mrc-bsu.cam.ac.uk/bugs
Accessed 8 July 2011
18. Durrleman S, Simon R. Flexible regression models with cubic splines. Stat Med. 1989;8:551–561
19. Nitschke M, Pilotto LS, Attewell RG, et al. A cohort study of indoor nitrogen dioxide and house dust mite exposure in asthmatic children. J Occup Environ Med. 2006;48:462–469
21. Skene KJ, Gent JF, McKay LA, Belanger K, Leaderer BP, Holford TR. Modeling effects of traffic and landscape characteristics on ambient nitrogen dioxide levels in Connecticut. Atmos Environ. 2010;44:5156–5164
22. US EPA. . Air Trends. 2009 Available at: http://www.epa.gov/airtrends/nitrogen.html#nonat
Accessed 1 April 2011