Statistical analysis:
Bivariate analyses comparing baseline demographic and clinical characteristics of children in the “no wheezing” vs. “any wheezing” groups were performed using Fisher’s exact test or t tests as appropriate. Also, bivariate analysis for wheezing characteristics at baseline and at follow-up between the two treatment arms were conducted. Multivariable analyses were then performed using logistic regression to evaluate the association between wheezing characteristics at follow-up and arm of treatment (AT vs. WWSC), conservatively adjusting for variables with a p value < 0.10 in the bivariate analysis: gender, race, rhinitis, eczema, second hand-smoke exposure, premature, number of siblings, sibling with asthma, maternal level of education and family income. In addition, a priori selected potential confounder factors, such as age, BMI z-score, AHI, and parental asthma were also included. These were retained in the final model if their coefficients were significant, altered the significance of the wheezing characteristics, or contributed significantly to the best model fit. A p < .05 was considered statistically significant. The R version 4.1.3 ® software was used (www.r-project.org).