Page 122 - Comprehensive treatment of patients with glucocorticoid-dependent severe asthma

Chapter 7
Data analysis (SPSS 15.0) was carried out according to a pre-established analysis-plan and no interim analysis was done. Between-group comparisons of normally and non-normally distributed data were performed with the use of Student’s t-test; and the Mann-Whitney U test, respectively. The repeated measurements of the oral corticosteroid usage were analyzed with mixed- effects regression models including time, treatment strategy and their interaction as covariates, and with random patient-effects for the intercept and the regression weight of time on oral corticosteroid treatment. The effect of time was modelled with a linear model and the goodness of fit was checked by introducing higher order polynomial terms of time into the model. The fit of the covariance structure of the residuals was checked by extending the model with an autoregressive structure. Histograms of the residuals were further inspected for normality. The main trend in the average change with time of the averaged corticosteroids usage was summarized with the change-rate estimated from the linear model.
The repeated measurements of the ACQ and AQLQ were analyzed in similar fashions, but the dependent variables were the change from baseline ACQ and AQLQ, respectively. The effect of time on average change of ACQ and AQLQ was modelled non-parametrically and summarized with the change from baseline ACQ and AQLQ averaged over all repeated measurements during follow-up.
Results
Ninety-five patients were randomized from November 2007 to October 2008. Five patients in the conventional management group withdrew consent before the study had started and one patient was excluded from the study because of poor adherence to the trial protocol (Figure 1).
Eighty-nine patients were included in the intention-to-treat analysis (Table 2). Fifty-one patients and thirty-eight patients were randomized to the internet-based and conventional management, respectively. Randomization was stratified without using allocation blocks, which explained the unequal number of patients in each study group.
120