2.4-Statistical analysis
Patient characteristics were presented as frequencies and percentages for categorical data and means (SD) or medians (IQR) for continuous data. The primary analysis of all outcomes followed the intention-to-treat (ITT) principle. Between groups, we used the Mann-Whitney-U test for comparing the time of success for HR, RR, and CRS (SPSS Inc., version 25.0, Chicago, IL, USA). The longitudinal analysis part of the study was performed non-parametrically with the Brunner-Langer model for CRS conducted with a linear mixed model in a parametric way for HR and RR. The Brunner-Langer model (F1-LD-F1) was applied using web-based software (R software, version 3.5.2, package: nparLD, R Foundation for Statistical Computing, Vienna, Austria;http://r-project.org). The PROC MIXED procedure in the SAS software was used (Version 9.3; SAS Institute, Cary, NC, USA) to perform linear mixed models in which subjects were included as random effects. When there was a significant interaction effect in the longitudinal analysis (P < .1), the time effect was analyzed in each group, and groups were compared at baseline and following time points (the baseline was subtracted from these time points). The proportions of the treatment failure, rescued patient, ICU admission, and AE were analyzed using Fisher’s Exact test and, given with odds ratios (ORs) and 95% Confidence Intervals (CIs). A 2-sided P < .05 defined statistical significance.