The therapeutic armamentarium for autoimmune diseases of the central nervous system, specifically multiple sclerosis and neuromyelitis optica, is steadily increasing, with a large spectrum of immunomodulatory and immunosuppressive agents targeting different mechanisms of the immune system. However, increasingly efficacious treatment options also entail higher potential for severe adverse drug reactions. Especially in cases failing first-line treatment, thorough evaluation of the risk–benefit profile of treatment alternatives is necessary. This argues for the need of algorithms to identify patients more likely to benefit from a specific treatment. Moreover, paradigms to stratify the risk for severe adverse drug reactions need to be established. In addition to clinical/paraclinical measures, biomarkers may aid in individualized risk–benefit assessment. A recent example is the routine testing for anti-John Cunningham virus antibodies in natalizumab-treated multiple sclerosis patients to assess the risk for the development of progressive multi-focal leucoencephalopathy. Refined algorithms for individualized risk assessment may also facilitate early initiation of induction treatment schemes in patient groups with high disease activity rather than classical escalation concepts. In this review, we will discuss approaches for individiualized risk–benefit assessment both for newly introduced agents as well as medications with established side-effect profiles. In addition to clinical parameters, we will also focus on biomarkers that may assist in patient selection.