Objective: To assess the efficacy of virtual reality (VR)-based vs conventional balance training on the improvement of balance and reduction of falls in people with multiple sclerosis (PwMS).Design: Single-blinded, randomized, controlled trial.Setting: Musculoskeletal Rehabilitation Research Center, Ahvaz Jundishapur University of Medical Sciences. Participants: PwMS (N=39), randomized into VR (n=19) and control (n=20) groups.Intervention: The VR group performed exergames using Kinect, while the control group accomplished conventional balance exercises. Both groups received 18 training sessions for 6 weeks.Main Outcome Measures: Limits of stability, timed Up and Go (TUG) test, and 10-m walk tests with and without cognitive task and their dual-task costs (DTCs), Berg Balance Scale, Multiple Sclerosis Walking Scale-12, Fall Efficacy Scale-International, Activities-specific Balance Confidence Scale, and fall history were obtained pre- and post intervention and after a 3-month follow-up.Results: At both post intervention and follow-up, TUG and DTCs on the TUG were significantly lower and the 10-m walk was significantly higher in the VR group. At follow-up, reaction time and the number of falls demonstrated significant differences favoring the VR group, whereas the directional control revealed significant difference in favor of the control group (P