Statistical methods
In order to achieve a power of 90% to detect 25% improvement in fatigue from baseline to month 12 when performing a paired Student's t-test, with the assumption that the standard deviation (SD) was close to 1.0, the sample size was estimated to be 168. The statistical analyses were based on pooled datasets from Sweden, Norway, Denmark and Austria, except SF-12, ESS, CES-D, PASAT and 6MWT, which were not performed in Denmark.
In general, descriptive statistics for continuous variables are presented with Mean, Standard Error (SE), Median, Minimum and Maximum values. Descriptive statistics for discrete variables are presented as percentages. All statistical tests were carried out as two-sided on a 5% level of significance unless otherwise stated.
The primary efficacy analysis was based on an analysis of variance (ANOVA) method on the change from baseline to month 12 for the FSMC total score. Baseline FSMC value was included as a continuous fixed effect.
For the secondary efficacy variables, mixed linear models with repeated measures were done for the FSMC total score, motor score, cognitive score, SF-12, ESS, CES-D, 6MWT and PASAT scores with change from baseline values as response, and baseline value as a continuous fixed effect, and visits as a categorical fixed effect. The covariance matrix was assumed unstructured. Since the EDSS step score was not normally distributed, a non-parametric model and a Hodges-Lehmann confidence interval (CI) with associated p–value from the Wilcoxon-Signed Rank test, were used to assess the change from baseline. The SDMT and Step Counter scores were analysed in a generalised linear mixed model. All statistical analysis and programming were done using SAS v9.2.