Background: Therapeutic decisions in RRMS have become complex as many drugs with different benefit/risk ratios are available. Method: Development of PHREND (Predictive Healthcare with Real world Evidence in Neurological Disorders) is based on 3320 therapy cycles of adult RRMS patients with an initial EDSS < 6.5 in the NeuroTransData MS registry from 2009 onwards with therapies initiated later than 6 months after diagnosis of RRMS. Parameter of high predictive value are age, gender, duration of RRMS, previous therapy and its duration, indicator if one of the two previous therapies was second line, EDSS total score, number of relapses within last 12 months, time since last relapse. The predictive mathematical models are based on the assumption, that EDSS progressions follow a binomial and the number of relapses a negative binominal distribution. Generalized linear models are employed for both efficacy responses using Bayesian inference and integrating cluster effects of multicentric studies and variable duration of therapies in the data base. Models were tested by 10-fold cross-validation. Mean square error of the forecast (Brier score) and Harrell’s concordance-index mark quality of prediction. Comparative prognostic models based on relapse rate and EDSS progression only were implemented for bench marking. Results: The predictive model based on the individual clinical situation shows a higher accuracy for all treatment scenarios compared to prediction based on frequencies in each therapeutic cohort alone: the concordance index increases from 0.55 to 0.64 for relapse probability, from 0.55 to 0.56 for EDSS progression. The Brier score for probability of relapse is 0.17, for EDSS progression 0.03 (both equal benchmark). After entering individual patient information personalized probabilities for freedom of relapse and of EDSS progression in relation to each available therapy are provided for a specific time period (2-4 years). Details of the quality of each prediction are shown. The application is CE certified. Conclusion: PHREND is a personalized treatment decision tool which assists individual treatment decisions by prodiving predictive data of individual treatment response. This model needs additional and ongoing data to improve quality of prediction and to include new treatment options.