OBJECTIVES: The purpose of this study was to review cost-effectiveness models in multiple sclerosis (MS) to identify accepted methods, key challenges, and best practices.  METHODS: We searched MEDLINE, Embase, the Health Economics Evaluation Database (HEED), the Cochrane library, and recent HTA agencies’ (NICE and SMC) decisions  for studies published prior to March 7, 2014. Following duplicate removal, 100 studies were identified. Studies were excluded if they did not estimate cost-effectiveness in MS, were duplicates, or weren’t published in English, resulting in a total of 26 studies sourced. A data extraction form was developed to capture information about model characteristics, patient natural history progression, utility estimates, and the author’s comments on the modelling methods RESULTS: All studies were published after 2000, with most focused on first-line USA and EU patient populations. The majority of models utilised a cohort Markov model approach, with health states defined by patient expanded disability status scale (EDSS) scores. Health states included either individual or aggregate EDSS scores (0-9.5) for relapsing-remitting MS (RRMS) and secondary progressive MS (SPMS), as well as a death state. Transition probabilities were sourced from trial data for low score EDSS states, while transitions for higher score EDSS states were sourced from a longitudinal study of Canadian MS patients, due to insufficient clinical trial data for patients with advanced disease. Key challenges identified in recent HTA decisions include modelling EDSS score improvement early in disease natural history, patients’ initial distributions across EDSS states, extended benefits of therapy, and patient treatment adherence. CONCLUSIONS: Established modelling best practice in MS utilises a cohort Markov model approach with health states simulating patient populations via EDSS scores in RRMS and SPMS. Future studies and HTA submissions should focus on more accurately reflecting patient natural history in the early stages of disease.