Background: Plasma exchange (PE) has usually to be considered as a rescue therapy when intravenous corticosteroids is insufficient in acute attacks of neuromyelitis optica spectrum disorders (NMOSD). The efficacy of PE has not been quantified. This system review and meta-analysis was aimed to evaluate the efficacy of PE therapy in acute attacks of NMOSD. Methods: Studies evaluating the efficacy of PE in patients with NMOSD were identified from PubMed and Embase. Changes of Expanded Disability Status Scale (EDSS) score between before and after PE therapy, and the rate of response to PE, were defined as the main efficacy outcomes. Meta-regression was performed to identify the sources of heterogeneity. Subgroup meta-analysis were performed based on the interval of initiation PE after attack onset and AQP4-IgG serostatus of patients. Results: Twenty-four studies containing 528 patients with NMOSD were included in this meta-analysis. As a rescue therapy when patients failed to respond to intravenous corticosteroids (PE rescue), PE treatment resulted in a reduction in the mean EDSS score by 1.69 (95% CI: 0.88-2.50), with a response rate of 75%(95%CI: 66%-83%). As a first-line therapy being used alone or simultaneously with intravenous corticosteroids (PE first-line), PE resulted in a reduction in the mean EDSS score by 2.34 (95% CI: 1.69-2.98), with a response rate of 71%(95% CI: 44%-93%). Overall, PE resulted in a reduction in the mean EDSS score by 1.83 (95% CI: 1.19-2.47), with a response rate of 74% (95%CI: 66%-82%). Subgroup analysis suggested that earlier PE initiation and AQP4-IgG seronegative patients seemed to be associated with a superior response to PE therapy. Conclusion: Plasma exchange, whether used as rescue or as first-line therapy, is an effective therapeutic method in patients during acute attacks of NMOSD.