Aim. To compare reactive balance in people with multiple sclerosis (MS) with healthy controls and to examine the ability of people with MS to adapt their reactive balance and retain training effects. Data Sources. Electronic databases (PubMed, EMBASE, PsychINFO) and reference lists of included articles from inception to February 25, 2019. Study Selection. Case-control and intervention studies that assessed reactive balance using mechanical perturbations in people with a confirmed diagnosis of MS. Results. Meta-analyses of 9 studies (n = 342) showed that people with MS have significantly worse reactive balance than healthy controls (standardized mean difference [SMD] 0.78, 95% CI 0.44-1.11, P < .0001, I2 = 47%). Specifically, people with MS have greater center of mass displacements (SMD 0.41, 95% CI 0.05-0.77, P = .02, I2 = 9%) and longer response times (MD (ms) 31.45, 95% CI 19.91-42.98, P < .0001, I2 = 75%) in response to standing perturbations than healthy controls. Subsequent meta-analyses revealed training comprising repeated exposure to perturbations improved response times ( P < .001) and training effects on response times could be retained for 24 hours ( P < .001) in people with MS. Conclusions. Reactive balance assessments can highlight functional impairments related to falls in people with MS, and perturbation training can acutely improve reactive balance control and such improvements can be retained for 24 hours in this population. Systematic review registration number: CRD42019126130.