Statistical analysis
PEAO was defined as WCS > 4. Post-extubation, patients were treated with adrenaline nebulization if WCS was> 4, and reintubated if so required. We wanted to make sure that 95% confidence interval (CI) of the low dose arm does not overlap with 95% CI of the previously described placebo arm of a trial which was done in a setting similar to ours. In a previous small placebo-controlled trial 11, incidence of PEAO sufficient to require adrenaline nebulization was 69% (22/32; exact binomial 95% CI, 50-83.9%) in placebo group compared to 13% (4/31; exact binomial 95% CI, 3.6-29.8%) in the HD arm (0.5mg/Kg/dose). If we arbitrarily assume an incidence of 15% in the LD arm with a non-inferiority margin of 12% from the baseline incidence of HD arm, a sample size of 478 patients (239 in each treatment arm) would be required to prove non-inferiority of LD (0.25mg/Kg/dose) treatment regime over the HD (0.5mg/Kg/dose) treatment regime with 80% power and 5% one-sided α-error. Considering attrition of 10%, 526 patients would be required (263 patients in each treatment arm).
Per-protocol analysis was performed with SPSS version 22 (IBM®). Patients with self-extubation, deferred extubation and those who were reintubated due to non-PEAO reasons werea priori deemed un-assessable and were excluded from analysis. Continuous and categorical variables were analyzed by Mann-Whitney U test and Chi-square test (and Fisher exact test if needed) respectively. Relative risk (RR) and their 95% confidence interval (CI) were estimated for the primary endpoint. Two-way mixed model ANOVA was used for intergroup comparisons of Westley’s Croup Score during 24 hours post-extubation. Time to recovery from PEAO (i.e., time to achieve WCS< 2 irreversibly) was compared by Kaplan-Meier Curve after censoring those who could not achieve WCS< 2 during 24 hours of observation. To identify independent risk factors for PEAO and changes over last decade, post hoc univariate analysis followed by multivariate analysis were performed. Multivariate analysis was done with ENTER selection procedure incorporating variables with p-value of <0.05 in univariate analysis. A p-value of <0.05 indicated significance.