Harmen Beurmanjer

82 Chapter 6 Analysis Demographics were calculated using descriptive statistics and compared between groups using univariate analysis of variance (ANOVA) and chi-square analyses. Lapse, relapse and relapse including drop-out rates in each group were compared using chi-square analyses. In contrast to the original protocol publication(Kamal, Schellekens, et al., 2015), we only analysed primary outcomes using both intention to treat (ITT) and per protocol (PP) analyses, due to the limited influx of patients receiving baclofen after a prolonged inclusion period (n=37 instead of n=80). In the more conservative ITT analyses all participants receiving baclofen were compared to TAU. In the PP analyses only those participants receiving baclofen according to the protocol were compared to TAU. Though a historical control group was available for comparison (Kamal, Loonen, et al., 2015), only the current control group was included in the analyses. First, the current control group is substantially larger that the intervention group, making addition of an extra control group redundant. Second, relapse rates in the current control group were substantially lower compared to our historical control group (50% versus 65% respectively). Finally, the current control group was more comparable to the baclofen group in terms of received TAU. Therefore, adding a historical control group to the analyses was considered of no added value. All analyses were carried out in SPSS version 21, with alpha <.05 considered significant. Results Demographics Patients in the baclofen group were more often male than in the TAU group, but gender was not related to treatment outcome. There were no other differences in demographics, GHB use or psychiatric comorbidity, see Table 1. Of the 37 patients receiving baclofen (included in ITT analysis), 13 received baclofen according to protocol (included in PP analysis). Effectiveness ITT analysis showed no difference in lapse rates (χ²=0.20, p=.885) and relapse rates excluding drop-out (χ²=3.29, p=.069) in the baclofen-treated group, compared to TAU, see Table 2a. In the baclofen group relapse rates including drop-out as relapse were lower compared to TAU (χ²=6.59, p=.010). PP analysis showed no difference in lapse rates (x²=1.99 p=.158), but lower relapse rates in the baclofen group when drop-out rates were not included (x²=3.97, p=.046) and included as relapse (x²=5.31, p=.021), compared to TAU, see Table 2b.