Blood Glial Fibrillary Acidic Protein in MRI‐Negative Attack‐Like Episodes of Neuromyelitis Optica Spectrum Disorder

ABSTRACT

Objectives

Magnetic resonance imaging (MRI) is widely used to support relapse diagnosis in neuromyelitis optica spectrum disorder (NMOSD), yet some patients present with attack‐like symptoms without corresponding lesions. We assessed serum glial fibrillary acidic protein (sGFAP) levels in MRI‐negative attack‐like episodes.

Methods

We retrospectively analyzed aquaporin‐4‐IgG–positive NMOSD patients presenting with relapse‐like symptoms but without corresponding new MRI lesions. sGFAP was measured from stored serum collected at presentation and converted to demographic‐adjusted z‐scores. A predefined attack‐associated threshold (z‐score ≥ 3.0) was based on prior work.

Results

We identified 39 MRI‐negative attack‐like episodes from 34 patients. Most episodes were clinically mild or subjective, with objective worsening observed in 23%. Eight episodes (21%) showed elevated sGFAP levels (median z‐score 4.56 [1748 pg/mL]). Compared with episodes without elevated sGFAP, these episodes occurred closer to prior attacks (8 vs. 33 months) and more frequently met objective worsening criteria (50% vs. 16%). In all episodes with elevated sGFAP, remission samples showed normalization below the threshold.

Discussion

Marked sGFAP elevation in a subset of MRI‐negative episodes suggests that some events may represent relapse not captured by conventional imaging. sGFAP may serve as a complementary biomarker to support attack identification when clinical and MRI findings are inconclusive.

1. Background

In neuromyelitis optica spectrum disorder (NMOSD), relapses are the principal drivers of long‐term disability, yet their identification remains challenging. Symptoms such as weakness, sensory disturbance, and bladder dysfunction overlap with noninflammatory worsening, limiting the reliability of clinical assessment alone [1]. Although magnetic resonance imaging (MRI) is widely used as an objective correlate [2, 3], it is not universally sensitive. In the N‐MOmentum trial, some investigator‐identified attacks were not adjudicated because of absent MRI lesions, whereas some adjudicated attacks lacked corresponding MRI abnormalities [4]. These observations highlight the need for additional biomarkers.

Serum glial fibrillary acidic protein (sGFAP), a biomarker of astrocytic injury, increases during NMOSD attacks and declines during remission [5, 6, 7]. In our previous study, sGFAP peaked within 1 week of symptom onset, and sGFAP z‐score ≥ 3.0 showed high specificity for acute attacks [8]. This suggests that marked sGFAP elevation may support relapse identification even without corresponding MRI lesions. We examined MRI‐negative attack‐like episodes in NMOSD, defined by new or worsening neurological symptoms without corresponding MRI lesions, and evaluated serum sGFAP levels in samples collected during the acute symptomatic phase.

2. Methods

2.1. Study Design and Participants

We conducted a retrospective single‐center study of AQP4‐IgG–positive NMOSD patients followed at the National Cancer Center, Korea (March 2005–December 2025). MRI‐negative attack‐like episodes were defined as new or worsening neurological symptoms lasting ≥ 24 h without new T2‐hyperintense or gadolinium‐enhancing lesions in the clinically relevant region. All suspected optic neuritis episodes underwent orbital MRI with fat‐saturated sequences, and all suspected myelitis episodes underwent spinal MRI with axial T2‐weighted and short tau inversion recovery sequences. Episodes were included if stored serum obtained at symptom onset was available.

Clinical data, including symptom onset, phenotype, Expanded Disability Status Scale (EDSS), Visual Functional Score (VFS), MRI findings, treatment, and outcomes were extracted from electronic medical records. Objective worsening at the index episode was defined by worsening in either the EDSS or VFS. EDSS worsening was defined according to N‐MOmentum trial criteria [9], whereas VFS worsening was defined using prespecified criteria. EDSS worsening was defined as a ≥ 1.5‐, ≥ 1.0‐, or ≥ 0.5‐point increase from baseline EDSS of 0, 1.0–5.0, or > 5.0, respectively, whereas VFS worsening was defined as a ≥ 2‐point increase from a baseline of 0 or a ≥ 1‐point increase from a baseline ≥ 1.

sGFAP and serum neurofilament light chain (sNfL) were measured using the Simoa Neurology 2‐plex B assay (Quanterix). Concentrations were converted to demographic‐adjusted z‐scores using predefined thresholds (sGFAP ≥ 3.0; sNfL ≥ 2.1) and previously established optimal sampling windows (≤ 7 days for sGFAP; 7–56 days for sNfL) [8, 10, 11]. Given the descriptive design and sample size, no inferential statistical testing was performed.

3. Results

We identified 39 MRI‐negative attack‐like episodes from 34 patients with AQP4‐IgG–positive NMOSD (Table 1). Episodes generally presented with mild or subjective symptoms, most commonly paresthesia, blurred vision, or ocular pain, often superimposed on preexisting disability; objective worsening was observed in 9 episodes (23%). Blood sampling and MRI were performed a median of 5 days after symptom onset (IQR 2–8).

TABLE 1.

Demographic and clinical characteristics of MRI‐negative attack‐like episodes in aquaporin‐4‐IgG–positive NMOSD.

Characteristic	Value
Episodes (patients), n	39 (34)
Female, n (%)	31 (91)
Age at disease onset (years), median (IQR)	30 (20–41)
Age at episode (years), median (IQR)	39 (31–46)
Disease duration at episode (years), median (IQR)	10 (5–14)
Suspected phenotype, n (%)
Myelitis	27 (69%)
Optic neuritis	12 (31%)
EDSS before episode, median (IQR)	3.5 (2.0–4.5)
EDSS at episode, median (IQR)	3.5 (2.5–4.5)
Objective EDSS or visual score change, n (%)	9 (23)
EDSS worsening	6 (15)
VFS worsening	3 (8)
Number of prior attacks, median (IQR)	6 (4–10)
Interval from last attack (months), median (IQR)	30 (14–50)
Immunosuppressive therapy at the time of the episode
Rituximab, n (%)	26 (67)
Mycophenolate mofetil, n (%)	7 (18)
Tocilizumab, n (%)	1 (2.5)
Satralizumab, n (%)	1 (2.5)
None (Naïve or discontinued), n (%)	4 (10)
Acute management
Intravenous methylprednisolone, n (%)	16 (41)
Plasma exchange, n (%)	1 (3)
Interval from symptom onset to sampling (days), median (IQR)	5 (2–8)
Within optimal sGFAP window (≤ 7 days), n (%)	29 (74)
Within optimal sNfL window (7–56 days), n (%)	14 (36)
sGFAP z‐score ≥ 3.0, n (%)	8 (21)
sNfL z‐score ≥ 2.1, n (%)	7 (18)
Both biomarkers elevated, n (%)	3 (8)

Note: Values are median (interquartile range) for continuous variables and number (percentage) for categorical variables, expressed at the episode level (n = 39).

Abbreviations: EDSS, Expanded Disability Status Scale; IQR, interquartile range; NMOSD, neuromyelitis optica spectrum disorder; sGFAP, serum glial fibrillary acidic protein; sNfL, serum neurofilament light chain; VFS, visual functional score.

Using predefined demographic‐adjusted cutoffs, sGFAP z‐scores were ≥ 3.0 in 8 episodes (21%), whereas sNfL z‐scores were ≥ 2.1 in 7 (18%); both biomarkers were elevated in 3 episodes. In all sGFAP‐elevated episodes, remission samples obtained before and/or after the index episode showed sGFAP levels below the attack‐defining threshold (Table 2). Four episodes, all presenting as myelitis‐like events, showed isolated sNfL elevation without concurrent sGFAP elevation. In 2 episodes, blood sampling occurred 6–7 months after the preceding attack, whereas the remaining 2 occurred in patients with advanced disability (EDSS 6.5 and 7.0). Notably, 3 of the 4 episodes neither met objective worsening criteria nor received IVMP.

TABLE 2.

Clinical and biomarker characteristics of the eight MRI‐negative attack‐like episodes with elevated serum glial fibrillary acidic protein (sGFAP z‐score ≥ 3.0).

Characteristic	Case 1	Case 2	Case 3	Case 4	Case 5	Case 6	Case 7	Case 8
Age group	30–39	30–39	20–29	50–59	40–49	20–29	30–39	40–49
Disease duration (y)	15	15	1	11	10	2	1	9
Prior attacks, n	12	4	4	6	9	4	3	12
Interval from last attack (mo)	98	6	3	6	35	2	10	14
Baseline immunotherapy	Rituximab	MMF	Rituximab	Tocilizumab	Rituximab	MMF	MMF	Rituximab
Suspected phenotype	Myelitis	Myelitis	Optic neuritis	Myelitis	Myelitis	Myelitis	Myelitis	Myelitis
Presenting symptoms	Back pain, bilateral thigh hypesthesia, gait disturbance	Trunk paresthesia	Ocular pain and reduced visual acuity in the right eye	Trunk paresthesia, left leg weakness with gait abnormality	Bilateral leg heaviness with right leg weakness	Paresthesia and urinary disturbance	Trunk paresthesia	Bilateral lower‐extremity weakness and hypesthesia
EDSS pre → at episode	3.5 → 4.5	2.5 → 2.5	1.5 → 2.0	4.0 → 4.0	4.5 → 5.0	1.5 → 2.5	4.0 → 4.0	6.0 → 8.0
Objective worsening criterion met	Yes	None	Yes (VFS [0 → 2])	None	None	Yes	None	Yes
CSF findings	—	—	—	—	—	—	—	WBC 65/μL, protein 116 mg/dL
Sampling day after onset	3	3	7	7	6	3	4	1
sGFAP, pg/mL (z‐score)	43,415 (5.49)	1619 (4.61)	1876 (4.61)	3031 (4.60)	2481 (4.53)	1201 (4.40)	638 (4.29)	287 (3.49)
sGFAP before index (remission), pg/mL (z‐score)	133 (2.12) [3 mo before]	72 (0.05) [2 mo before]	92 (1.41) [4 mo before]	105 (0.50) [3 mo before]	113 (1.28) [3 mo before]	69 (1.28) [3 mo before]	80 (1.25) [12 mo before]	208 (2.45) [2 mo before]
sGFAP after index (remission), pg/mL (z‐score)	168 (2.5) [15.3 mo after]	79 (0.3) [4.0 mo after]	72 (0.05) [8 mo after]	115 (0.73) [6.0 mo after]	—	74 (1.44) [15 mo after]	102 (1.85) [4 mo after]	184 (2.25) [15 mo after]
sNfL, pg/mL (z‐score)	8.4 (0.64)	5.3 (0.64)	18.8 (2.46)	23.9 (1.98)	15.9 (1.91)	15.5 (2.75)	16.5 (2.41)	11.3 (0.71)
Acute therapy	IVMP	IVMP	IVMP	None	None	IVMP	None	IVMP + PLEX

Note: Objective worsening was defined by N‐MOmentum trial criteria: a ≥ 1.5‐point EDSS increase from a baseline of 0, ≥ 1.0‐point increase from a baseline of 1.0–5.0, ≥ 0.5‐point increase from a baseline > 5.0, or a ≥ 2‐point Visual Functional Score increase from a baseline of 0 or a ≥ 1‐point increase from a baseline ≥ 1.

Abbreviations: CSF, cerebrospinal fluid; EDSS, Expanded Disability Status Scale; IVMP, intravenous methylprednisolone; MMF, mycophenolate mofetil; mo, months; PLEX, plasma exchange; sGFAP, serum glial fibrillary acidic protein; sNfL, serum neurofilament light chain; VFS, visual functional score.

Clinically, 7 of the 8 episodes with elevated sGFAP (88%) presented as myelitis‐like events with anatomically localizable symptoms, whereas the remaining episode was optic neuritis–like with reduced visual acuity (Table 2). Compared with episodes without elevated sGFAP, those with elevated sGFAP occurred closer to the preceding attack (median 8 months [IQR 5–25] vs. 33 months [IQR 17–54]) and more frequently met objective worsening criteria (50% vs. 16%). Acute treatment with intravenous methylprednisolone (IVMP) was also more common in episodes with elevated sGFAP (62% vs. 35%). Additional evidence supporting inflammatory activity was observed in several sGFAP‐elevated episodes. In one episode, cerebrospinal fluid analysis showed pleocytosis (WBC 65/μL; protein 116 mg/dL), and treatment was escalated from IVMP to plasma exchange because of inadequate steroid response. In another episode, MRI obtained 2 days after symptom onset showed no definite enhancement, whereas follow‐up MRI performed 1 month later demonstrated a clearly enhancing cervical spinal cord lesion (C2–C6).

Of the 31 episodes without elevated sGFAP, 5 (16%) met objective worsening criteria; 3 were treated with IVMP, whereas in the remaining 2 the treating neurologist favored a non‐relapse etiology. Among the remaining 26 episodes without objective worsening, 8 (31%) nevertheless received IVMP based on subjective symptoms and clinical concern for relapse, whereas the other 18 were managed conservatively and subsequently recovered.

4. Discussion

In this retrospective cohort study, we identified a subset of MRI‐negative attack‐like episodes in AQP4‐IgG–positive NMOSD characterized by marked serum GFAP elevation. Despite the absence of new lesions on conventional MRI and often minimal or subjective clinical worsening, these episodes showed sGFAP levels above the previously reported attack‐defining threshold, which showed high specificity (~96%) for distinguishing attacks from remission [8]. Normalization of sGFAP in paired remission samples further supports that this transient elevation reflects an event‐related biological activity rather than chronic injury, consistent with prior longitudinal data [12]. These findings suggest that some MRI‐negative presentations may represent true relapse not captured by conventional imaging.

Compared with episodes without elevated sGFAP, those with elevated sGFAP showed features more suggestive of relapse, including shorter intervals from prior attacks and more frequent objective worsening and more frequent IVMP treatment. Two illustrative cases—one with cerebrospinal fluid pleocytosis and another with delayed lesion enhancement on follow‐up MRI—further support this interpretation. In contrast, isolated sNfL elevation without concurrent sGFAP elevation occurred in settings potentially reflecting non‐acute processes, including recent prior attacks or advanced disability.

Our cohort underscores the diagnostic uncertainty of MRI‐negative presentations: most episodes were clinically subtle, and objective worsening was observed in a minority. This is consistent with prior observations that clinical features alone are often insufficient to distinguish relapse from noninflammatory worsening in NMOSD [1]. Several factors may contribute to the absence of MRI correlates, including preexisting structural abnormalities obscuring small superimposed lesions and attenuation of inflammatory activity under maintenance immunotherapy. In this setting, diagnostic uncertainty often prompts empirical treatment: 31% of episodes without elevated sGFAP and without objective worsening received IVMP. A similar proportion (35%) of pseudo‐relapse episodes reportedly received steroids in prior studies, highlighting overtreatment as a recurring challenge in NMOSD [1]. In this context, sGFAP may help distinguish relapse from noninflammatory symptom worsening, particularly in clinically mild or equivocal presentations; although a negative result does not exclude relapse, it may support cautious observation and deferral of immediate IVMP. Conversely, elevated sGFAP in MRI‐negative presentations may prompt further evaluation with CSF analysis and repeat MRI to obtain additional objective evidence, as confirming relapse may guide acute treatment and modification of maintenance therapy. In suspected optic neuritis, optical coherence tomography showing acute retinal nerve fiber layer swelling or subsequent retinal thinning may provide complementary structural evidence of optic nerve injury.

This study has limitations, including its retrospective design, single‐center setting, and small sample size. MRI acquisition was primarily focused on the clinically suspected region, which may have contributed to underdetection of lesions outside the imaged area. Despite these limitations, these findings support a complementary role for serum GFAP in the evaluation of MRI‐negative attack‐like episodes and suggest that it may provide objective evidence to support attack identification when conventional clinical and MRI assessment is inconclusive.

Author Contributions

Su‐Hyun Kim: conceptualization, funding acquisition, formal analysis, investigation, writing – review and editing, writing – original draft, resources. Jae‐Won Hyun: methodology, writing – review and editing, resources. Ho Jin Kim: investigation, resources, writing – review and editing, supervision. Jieun Chung: methodology, resources, writing – review and editing. Rosah May Palermo Payumo: methodology, writing – review and editing, resources.

Funding

This work was supported by the National Cancer Center (Grant No. 2510580).

Ethics Statement

The study was approved as exempt by the Institutional Review Board of the National Cancer Center and all participants provided written informed consent.

Conflicts of Interest

S.‐H.K. has received honoraria for consulting, advisory board participation, and lectures from Bayer Schering Pharma, Biogen, Genzyme, Merck Serono, AstraZeneca, Daewoong, Mitsubishi Tanabe Pharma, and Roche, and research funding from the National Cancer Center of Korea.H.J.K. has received research funding from the Korea‐US Collaborative Research Project, the National Research Foundation of Korea, AprilBio, Eisai, Good T Cells, and UCB; honoraria for consulting, advisory board participation, and lectures from Alexion, Altos Biologics, AstraZeneca, Biogen, Daewoong Pharmaceutical, Eisai, GC Pharma, Kaigene, MDimune, Merck, Mitsubishi Tanabe Pharma, Roche, Samsung Bioepis, and Sanofi; and serves as Co‐Editor of Multiple Sclerosis Journal and Associate Editor of Journal of Clinical Neurology. J.C., R.M.P.P., and J.‐W.H. declares no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.