Abstract
BACKGROUND AND OBJECTIVES: Neuromyelitis optica spectrum disorder (NMOSD) is an autoimmune inflammatory disease of the central nervous system characterized by simultaneous or consecutive episodes of acute optic neuritis and transverse myelitis. Attacks of NMOSD can result in the accrual of severe visual disability over time. This study aimed to develop and validate prognostic models for visual disability risk within 1, 3, and 5 years. METHODS: Medical records of NMOSD patients were retrospectively analyzed. The least absolute shrinkage and selection operator (LASSO) regression algorithm and univariate and multivariate Cox regression analyses were performed to select predictors of visual disability. Two models predicting the probability of visual disability in 1, 3, and 5 years were developed based on different selections and displayed as nomograms. Risk scores were calculated for every patient, and a cut-off point was obtained to recognize patients at high risk. RESULTS: In total, 161 (25.2%) patients developed visual disabilities during the follow-up period. Four visual disability-related factors were selected using LASSO regression: optic neuritis (ON) onset, higher annual relapse rate (ARR) before maintenance therapy, no maintenance immune suppression therapy (IST), and initial severe attack. Three additional predictors were determined using multivariate Cox regression: male sex, age at first onset, and positive AQP4-IgG serology. Discrimination and calibration were satisfied, with concordance indexes (C-index) close to 0.9 in both models. Decision curve analysis showed good clinical usefulness in both models, and Kaplan-Meier curves showed satisfactory discrimination between patients with high risk and low risk by the cut-off points. CONCLUSION: This study reported predictors of visual disability and generated nomograms. High-risk patients need more active treatment and management to avoid unfavorable outcomes.