Early Clinical and Cerebrospinal Fluid Predictors of 1-Year Recurrence in Autoimmune GFAP Astrocytopathy

Autoimmune glial fibrillary acidic protein astrocytopathy (GFAP-A) is an inflammatory central nervous system disorder with variable outcomes. Relapse occurs in a subset of patients, but early predictors remain unclear. We aimed to identify admission-available features associated with 1-year recurrence and develop an interpretable risk stratification model.

We retrospectively included 156 patients with CSF GFAP-IgG-positive GFAP-A between January 2021 and February 2025. Patients were classified by recurrence within 1 year. Early demographic, clinical, CSF, serological, neuroimaging, and treatment-related variables were compared between groups. Candidate predictors were screened using elastic net regression with repeated cross-validation, followed by multivariable logistic regression and internal validation.

Thirty-five patients (22.4%) relapsed within 1 year. Patients with and without recurrence within 1 year were broadly comparable in age, sex, baseline severity, neuroimaging findings, and treatment exposure. Patients with recurrence within 1 year had lower CSF white blood cell counts (42.0 [10.0-104.0] vs. 107.0 [34.0-200.0] × 10⁶/L; p = 0.002) and more frequent movement disorders (71.4% vs. 43.8%; p = 0.004). The final model included CSF white blood cell count, movement disorders, CSF chloride, and CSF GFAP-IgG titer category. Lower CSF white blood cell count, movement disorders, and higher GFAP-IgG titer category were associated with increased recurrence risk. The model showed acceptable optimism-corrected discrimination (AUC, 0.726), reasonable calibration, and potential clinical utility. Machine-learning models did not outperform logistic regression.

In GFAP-A, 1-year recurrence was associated with distinct early clinical and CSF features rather than baseline severity or treatment exposure alone. This admission-based model may support recurrence risk stratification and follow-up planning.