Real-life clinical use of natalizumab and fingolimod in Austria.

The generalized linear model (GLM) for relapse count as Poisson distributed dependent variable and propensity score as covariate showed a statistically significant reduction for the mean relapse count in the natalizumab group compared with fingolimod. This effect was smaller in the analyses of log-transformed ARR with propensity score matching, loosing statistical significance although showing the same direction for the effect. We assume that the GLM was the more sensitive model analysing this question.