BACKGROUND: Severe optic neuritis (ON) is an acute inflammatory attack of the optic nerve(s) leading to severe visual loss that may occur in isolation or as part of a relapsing neuroinflammatory disease, such neuromyelitis optica spectrum disorder (NMOSD), myelin oligodendrocyte glycoprotein antibody associated disease (MOGAD), or more rarely multiple sclerosis (MS). In cases of first-ever severe ON of uncertain etiology best treatment strategies remain unclear. METHODS: We reviewed records of all patients with a documented diagnosis of ON between 2004 and 2019 at Mass General Brigham (MGB) and Johns Hopkins University (JHU) hospitals. Out of 381 patients identified, 90 (23.6%) satisfied the study criteria for severe ON with visual acuity (VA) equal to or worse than 20/200 (logMAR=1) at nadir in the affected eye and had sufficient follow-up data. Treatment strategies with corticosteroids only or treatment escalation with therapeutic plasma exchange (PLEX) after steroids were compared and evaluated for differences in visual outcomes at follow-up. RESULTS: Of the 90 patients with severe optic neuritis, 71(78.9%) received corticosteroids only, and 19 (17.0%) underwent PLEX following corticosteroids. Of the 71 patients who received steroids without escalation to PLEX, 30 patients (42.2%) achieved complete recovery (VA 20/20 on the affected eye), whereas 35 (49.3%) had a partial recovery and 6 (8.4%) had no recovery. Among the 19 corticosteroid non-responders patients who underwent escalation treatment, 13 (68.4%) made complete recovery, 6 (31.6%) had partial visual recoveries (p=0.0434). The median delta logMAR of patients who underwent escalation of care was -1.2 compared with 2.0 for the ones who did not (p=0.0208). A change of delta logmar 2.0 is equivalent of going from hand motion to light perception and the positive delta value refers to intra-attack worsening. Other than not responding to steroids, patients who underwent PLEX tended to have more severe ON with significantly worse nadir visual acuity compared with those who received corticosteroids alone (logMAR 3.12 (min 2.0 - max 5.0) vs. 2.17 (min 1.3 - max 3.0); p=0.004). CONCLUSION: In our cohort of first-ever severe optic neuritis of unknown etiology, patients that did not respond adequately to corticosteroids benefited from treatment escalation to PLEX, followed in most cases by Rituximab, regardless of final etiology. Randomized controlled trials are needed to confirm the best treatment strategies.