OBJECTIVES: Explore how MS DMT-persistence can be modelled, compare model’s performance and assess the independent drivers for DMT-persistence. METHODS: Analysis was based on 1638 DMT uses (1.1.1991–31.12.2010, time at risk 4009 years) in incident MS-patient register from Tampere, Vaasa and Seinäjoki regions, Finland. DMT-persistence = DMT end-day - DMT initiation day. Cox, exponential, generalized gamma, Gompertz, log-logistic, log-normal, and Weibull regression survival models were used to model DMT-persistence. Models were compared based on goodness-of-fit statistics (Akaike and Bayesian information criteria). RESULTS: Mean follow-up from first MS-symptoms and age at first DMT-initiation were 13 and 36 years, respectively. 73% of patients were female. Based on the data exploration of all known covariates, three DMT-persistence approaches with different interpretations, selected covariates and data needs were modelled: 1] sex, birth year, time from symptoms to DMT, age, DMT line (1^st, 2^nd, 3^rd, 4^th, 5-7^th), DMT (interferon-β-1a and -1b,  glatiramer acetate, other) at DMT initiation; 2] approach 1] + Expanded Disability Status Scale (EDSS) at DMT initiation; and 3] approach 2] + DMT-discontinuation reason (pregnancy plan, flu-like symptoms, injection-site reactions, ineffectiveness, antibodies, other/unknown). There was no gold standard survival model for DMT-persistence, and some models accommodated higher number of covariates and associated dependencies better. For approaches 1] and 2] Weibull and for 3] Gompertz model provided the best goodness-of-fit. Based on all three models (one per approach 1]–3]) with best goodness-of-fit, higher EDSS or higher age at DMT initiation, treatment line (3rd and later), and incidence of intolerable adverse events (AE) or ineffectiveness were independently associated with shorter DMT persistence. In the approach 3], flu-like symptom and injection site AEs had the highest hazard ratio for shorter DMT-persistence. CONCLUSIONS: AEs, EDSS, age, treatment line and ineffectiveness were strong predictors for DMT-persistence. Flu-like symptoms and injection site AEs showed the highest hazard for DMT-discontinuation.