Objective:To investigate discriminative ability and clinical utility of the Timed Up and Go under single- and dual-task conditions between fallers and non-fallers in multiple sclerosis (MS).Design:Prospective cohort study.Setting:Neurology service in a tertiary hospital.Subjects:Participants were 101 people with MS and Expanded Disability Status Scale score of 3–6.5. One participant withdrew after the baseline assessment, and hence the data were analysed for 100 participants.Interventions:No specific intervention.Main Measures:Timed Up and Go and Timed Up and Go-Cognitive. Three-month prospective diaries recorded falls.Results:Mean age was 52.6 (SD 10.7) and 66 were female. Majority of the participants had progressive MS (72) and 73 used a walking aid; 56 participants recorded 791 falls. The area under the receiver operating characteristic curve values for Timed Up and Go and Timed Up and Go-Cognitive in distinguishing fallers (person with ⩾1 fall) from non-fallers are 0.60 and 0.57, respectively, and in distinguishing multiple fallers (⩾2 falls) the values are 0.46 and 0.43. A Timed Up and Go score of ⩾9 seconds has a sensitivity of 0.82 and a specificity of 0.34 to identify fallers and a sensitivity of 0.79 and a specificity of 0.27 to identify multiple fallers. A Timed Up and Go-Cognitive score of ⩾11 seconds has a sensitivity of 0.77 and a specificity of 0.30 to identify fallers and a sensitivity of 0.71 and a specificity of 0.26 to identify multiple fallers.Conclusion:The Timed Up and Go and Timed Up and Go-Cognitive do not demonstrate sufficient clinical utility or discriminative ability for assessing falls risk in MS.