The null hypothesis for comparison of initial IFNB-1b versus initial placebo treatment was: H0: The survival functions (i.e., the probability that time to confirmed EDSS progression is ≥ t) are not identical for both treatment arms for some points in time t>0. The two-sided alternative hypothesis was: H1: The survival functions (i.e., the probability that time to confirmed EDSS progression is ≥ t) are identical for both treatment arms for some points in time t>0.

A 2-sided error level of 0.0253 was used for analyses at month 36 and 60 in order to keep the study-wise error level at 0.05. A conditional sequential testing approach was used for the family of null hypotheses of the primary efficacy variables.

Time to confirmed EDSS progression was modelled by a Cox proportional hazards regression model with the following covariates: treatment group and volume of T2 lesions on screening MRI. The null hypothesis was: Initial IFNB-1b and initial placebo do not differ with regard to the hazard for confirmed EDSS progression, i.e. hazard ratio = 1.

The variable used as additional covariate adjustment was volume of T2 lesions on screening MRI.

The direction of comparison is initial IFNB-1b (numerator) versus initial placebo (denominator), i.e. the Hazard Ratio represents the relative risk of initial IFNB-1b treatment compared to initial placebo treatment.

The null hypothesis H0: "The distribution of FAMS TOI at month 60, adjusted for baseline FAMS TOI, is identical for both treatment arms" was tested against the alternative hypothesis HA: "The distribution of FAMS TOI at month 60, adjusted for baseline FAMS TOI, is not the same in the two treatment arms" using a non-parametric analysis of covariance (ANCOVA).

Variable used as covariate: FAMS TOI measured at baseline

A 2-sided error level of 0.0253 was used for analyses at month 36 and 60 in order to keep the study-wise error level at 0.05. A conditional sequential testing approach was used for the family of null hypotheses of the primary efficacy variables.

The null hypothesis H0: "The mean FAMS TOI at month 60, adjusted for baseline FAMS TOI, is identical for both treatment arms" was tested against the alternative hypothesis HA: "The mean FAMS TOI at month 60, adjusted for baseline FAMS TOI, is not the same in the two treatment arms" using a parametric analysis of covariance (ANCOVA).

Variable used as covariate: FAMS TOI measured at baseline

The null hypothesis for comparison of initial IFNB-1b versus initial placebo treatment was:

H0: The survival functions (i.e., the probability that time to McDonald MS is ≥ t) are identical for both treatment arms for all points in time t>0.

The two-sided alternative hypothesis was:

H1: The survival functions (i.e., the probability that time to McDonald MS is ≥ t) are not identical for both treatment arms for some points in time t>0.

The two-sided error level for the exploratory analysis of the secondary outcome measures was of 0.05.

Time to McDonald MS was modelled by a Cox proportional hazards regression model with the following covariates: treatment group, steroid use during first event (yes vs. no), onset of disease (monofocal vs. multifocal) and number of T2 lesions at screening (categories: 2-4, 5-8 and at least 9 T2 lesions). The null hypothesis was: Initial IFNB-1b and initial placebo do not differ with regard to the hazard for McDonald MS, i.e. hazard ratio = 1.

covariate adjustment: steroid use during first event, onset of disease, categorized number of T2 lesions at screening

The direction of comparison is initial IFNB-1b versus initial placebo, i.e. the Hazard Ratio represents the relative risk of initial IFNB-1b treatment compared to initial placebo treatment.

The time to recurrent relapses was modelled by an extension of Cox's PH regression model (Andersen-Gill Model) for recurrent events with the following covariates: treatment group, steroid use during first event (yes vs. no), onset of disease (monofocal vs. multifocal) and number of T2 lesions at screening (2-4, 5-8 and at least 9 T2 lesions). The null hypothesis was: Initial IFNB-1b and initial placebo do not differ with regard to the hazard for recurrent relapses, i.e. hazard ratio = 1.

Covariate adjustment: steroid use during first event, onset of disease, categorized number of T2 lesions at screening

The two-sided error level for the exploratory analysis of the secondary outcome measures was of 0.05.

The direction of comparison is initial IFNB-1b (numerator) versus initial placebo (denominator), i.e. the Hazard Ratio represents the relative risk of initial IFNB-1b treatment compared to initial placebo treatment.

Relapse rate was analyzed by a generalized linear Poisson regression model with individual relapse counts as dependent variable, covariates: treatment arm, steroid use during first event, onset of disease and categorized number of T2 lesions at screening and offset variable natural log of time (in years) as difference between last clinical visit and baseline visit. The treatment effect on the relapse rate was of primary interest.

covariate adjustment: steroid use during first event, onset of disease, categorized number of T2 lesions at screening

The two-sided error level for the exploratory analysis of the secondary outcome measures was of 0.05.

The direction of comparison is initial IFNB-1b (numerator) versus initial placebo (denominator), i.e. the Risk Ratio represents the relative risk of initial IFNB-1b treatment compared to initial placebo treatment.

The null hypothesis H0: "The distribution of MSFC Z-scores at month 60 adjusted for baseline MSFC Z-score is identical for both treatment arms" was tested against the alternative hypothesis HA: "The distribution of MSFC Z-scores at month 60 adjusted for baseline MSFC Z-score is not the same in the two treatment groups" based on a non-parametric analysis of covariance (ANCOVA).

Variable used as covariate: MSFC Z-scores measured at baseline

The two-sided error level for the exploratory analysis of the secondary outcome measures was of 0.05.

The null hypothesis H0: "The mean MSFC Z-scores at month 60 adjusted for baseline MSFC Z-score are identical for both treatment arms" was tested against the alternative hypothesis HA: "The mean MSFC Z-scores at month 60 adjusted for baseline MSFC Z-score are not the same in the two treatment groups" based on a non-parametric analysis of covariance (ANCOVA).

Variable used as covariate: MSFC Z-scores measured at baseline

The two-sided error level for the exploratory analysis of the secondary outcome measures was of 0.05.