Magnetic Resonance Imaging (MRI) was used to measure presence of new or reduction in number of gadolinium enhancing T1 lesions. Each MRI scan was previewed by a local neuroradiologist. The quality of each scan performed was assessed by a central MRI reading center and evaluated for quality, completeness and adherence to the protocol. A nonresponder imputation (NRI) for missing data approach was applied. NRI assumes that a participant was a treatment failure, i.e. non-responder, if they did not have a valid Month 12 MRI assessment, or if they discontinued the study prematurely and did not have a valid Month 12 MRI assessment.

Magnetic Resonance Imaging (MRI) was used to measure presence of new or reduction in number of gadolinium enhancing T1 lesions. Each MRI scan was previewed by a local neuroradiologist. The quality of each scan performed was assessed by a central MRI reading center and evaluated for quality, completeness and adherence to the protocol. A sensitivity analysis of the primary endpoint was performed based on an observed data approach.

ARR is the number of confirmed relapses in a year calculated based on cumulative number of relapses by patient (adjusted for time-in-study by patient). Confirmed relapses are those accompanied by a clinically relevant change in the expanded disability status scale (EDSS). ARR was estimated from fitting a negative binomial regression model with log-link, and adjusted for prior MS therapies as a factor, number of relapses in previous year, baseline EDSS, baseline number of T1 Gd-enhancing lesions and the subject's age at baseline as covariates. The primary analysis describes the ARR with one-sided 95% confidence bound and test for null hypothesis (H0): ARR \>=0.18 versus alternative hypothesis (H1): ARR\<0.18.

Total number of Gd-enhancing T1 lesions across scans at Week 12, 16, 20, and 24 were adjusted for the different numbers of scans. This was calculated as a rate for population, rather than at patient level, using a negative binomial regression model with log link. The model included each patient's total number of Gd-enhancing T1 lesions as the response variable, and treatment, region, and subgroup of baseline number of Gd-enhancing T1 lesions (0 or \>=1) as explanatory variables, and logarithm of the patient's number of scans as the offset variable.