Clinical features associated with cerebrospinal fluid Epstein-Barr virus positivity in autoimmune GFAP...

Autoimmune glial fibrillary acidic protein astrocytopathy (GFAP-A) is an inflammatory disorder of the central nervous system. The clinical implications of Epstein-Barr virus (EBV) sequences detected in cerebrospinal fluid (CSF) by metagenomic next-generation sequencing (mNGS) remain uncertain.

This multicenter retrospective cohort study enrolled 60 patients with GFAP-A screened from electronic medical records of two hospitals (Jan 2019-Dec 2025). Patients were stratified into CSF EBV-positive (n = 10) and CSF EBV-negative (n = 50) groups based on CSF mNGS results. Demographic characteristics, clinical manifestations, laboratory findings, MRI features, treatment, and outcomes were compared. Bonferroni correction was applied to three clinically prioritized endpoints (CSF white blood cell count, altered consciousness, CSF chloride) selected based on prior GFAP-A literature to limit multiplicity bias; the corrected two-sided significance threshold was set at α = 0.05/3 ≈ 0.0167. An exploratory Firth penalized multivariable logistic regression model fitted via R (v4.3.1, logistf package) was used to examine the adjusted associations between CSF mNGS-detectable EBV sequences and acute altered consciousness, adjusting for age and sex to mitigate small-sample bias. All other unadjusted baseline comparisons were considered exploratory and were not adjusted for multiplicity. Spearman's rank correlation was used to assess the relationships of raw EBV mNGS read counts with peak pretreatment modified Rankin Scale (mRS) score and CSF WBC count among EBV-positive patients.

CSF WBC count was higher in the CSF EBV-positive group than in the CSF EBV-negative group [median, 120 (interquartile range, 105-220) vs. 33 (16-97) × 10⁶/L; raw p = 0.004; Bonferroni-adjusted p = 0.012]. Altered consciousness was more frequent in the EBV-positive group (80.0% vs. 42.0%; crude odds ratio [OR] = 5.52, 95% confidence interval [CI] 1.06-28.71; raw p = 0.039), and CSF chloride was numerically lower [111.50 (104.00-117.00) vs. 115.70 (113.00-120.00) mmol/L; raw p = 0.048]; however, neither comparison met the Bonferroni-corrected criterion (adjusted p = 0.117 and 0.144, respectively). Firth penalized multivariable logistic regression fitted in R was adjusted for age and sex. CSF EBV positivity exhibited a non-significant trend toward higher odds of acute altered consciousness (OR = 4.10, 95% CI: 0.94-24.84, p = 0.060). The extremely wide confidence interval indicated high estimation uncertainty limited by the small EBV-positive subgroup (n = 10). Female sex was associated with lower estimated odds of altered consciousness (OR = 0.17, 95% CI: 0.04-0.58, p = 0.004), while age showed no independent association (OR = 0.98, 95% CI: 0.95-1.02, p = 0.312). No statistically detectable between-group differences were observed in GFAP-IgG characteristics, MRI findings, relapse, or functional outcomes. Among the 10 EBV-positive patients, raw EBV read counts were not statistically correlated with peak pretreatment mRS score (r _s = 0.118, p = 0.745) or CSF WBC count (r _s = -0.073, p = 0.841).

CSF WBC count was the only clinically prioritized comparison that remained statistically significant after Bonferroni correction, suggesting greater CSF pleocytosis and possibly increased intrathecal inflammatory activity in GFAP-A patients with mNGS-detectable CSF EBV. Differences in altered consciousness and CSF chloride remained inconclusive after multiplicity adjustment. Within the EBV-positive subgroup, raw EBV sequencing reads showed no correlation with inflammatory or disability measures. Current findings do not establish CSF EBV positivity as an independent pathogenic or prognostic biomarker for GFAP-A, and the biological origin of the detected EBV sequences remains uncertain.