The best magazine
Methods
Data Source and Study Sample
We used data from the Chicago Initiative to Raise Asthma Health Equity (CHIRAH) study for which the design and procedures have been previously described. Between 2004 and 2005, CHIRAH enrolled a cohort of 561 asthmatic children aged 8–14. Participants and their caregivers completed an inperson interview to determine household sociodemographic and psychosocial characteristics, the home environment, and participants' asthma symptoms. Participants underwent anthropometric and spirometry measurements and had their saliva sampled. We excluded 76 participants missing key data (valid cotinine level, detailed exacerbation data, race/ethnicity, home location and/or body mass index (BMI)), and four children with exceptionally high numbers (≥25) of caregiver-reported asthma exacerbations that were likely errors in data collection or entry. We additionally excluded participants with a salivary cotinine level of >10 ng/ml (n=15, or 2.6% of the CHIRAH cohort) as extremely high cotinine values are more likely to represent active smoking. Our final sample included 466 children.
This study was approved by the institutional review boards of the Ann and Robert H. Lurie Children's Hospital and Northwestern University.
Outcomes
We used two outcome measures, asthma exacerbation frequency and asthma severity. Asthma exacerbation frequency was assessed by caregiver report to the prompt, 'Total number of asthma exacerbations requiring hospitalisation, ER [emergency room], or same day medical care in the past 12 months.' Asthma severity was assessed by the participant's classification based on the National Heart, Lung, and Blood Institute/National Asthma Education and Prevention Program guidelines which divided asthma severity into intermittent or mild-, moderate-, or severe-persistent categories on the basis of symptoms, short-acting β2-agonist use, interference with normal activity, exacerbations requiring systemic corticosteroids in the prior year and, if available, spirometry measurements, available for 71% of the CHIRAH cohort.
Exposure Assessment
We measured exposure to ETS by caregiver report of smoking in the household and by measured salivary cotinine. Household smoking was assessed by caregiver report to the question, 'Does anyone in your house smoke?' Saliva was collected with a Quanitsal collection pad (Immunalysis, Pomona, California, USA). Cotinine level was determined using a high-sensitivity immunoassay (Salimetrics, Inc., State College, Pennsylvania, USA). For comparisons of tobacco exposure by report versus by biomarker, we considered participants with salivary cotinine levels greater than or equal to 1 to be exposed to ETS, consistent with prior studies.
Covariates
Potential confounders were identified from caregiver reports and included child age, sex, race/ethnicity, household income (greater than or less than or equal to $50 000 per year) and access to any asthma controller medication. Because salivary cotinine is negatively associated with BMI, we included participant BMIs, categorised into normal or underweight, overweight, and obese (BMI <85th, 85–94th, and ≥95th percentiles, respectively).
Analysis
Our outcomes were: (1) total number of caregiver-reported asthma exacerbations in the previous year and (2) asthma severity in four ordered categories (intermittent, mild persistent, moderate persistent, severe persistent). Our primary predictors were (1) caregiver-reported household smoking as a dichotomous variable and (2) salivary cotinine as a continuous variable. Because cotinine levels were not normally distributed they were log-transformed.
For each outcome, we estimated two multivariable models, one with caregiver-reported smoking as the primary predictor and one with log-transformed cotinine level as the primary predictor, with age, sex, race/ethnicity, household income, controller medicine access and BMI category as covariates. Exacerbation frequency was modelled using negative binomial regressions because exacerbation count data were overdispersed. Asthma severity was modelled using ordered logit models after we verified that the model met the proportional odds assumption.
We used a two-sided Type I error <5% level to determine statistical significance. STATA SE V.12 (StataCorp LP, College Station, Texas, USA) was used for all data analyses.
Source: ...