Key Points
Question
Can optimizing serum glial fibrillary acidic protein (sGFAP) and serum neurofilament light chain (sNfL) assessment based on temporal patterns enhance clinical utility in neuromyelitis optica spectrum disorder (NMOSD)?
Finding
In this cohort study of 181 patients with NMOSD (78 discovery cohort patients and 103 validation cohort patients), optimal time frames for evaluating attacks were within 1 week for sGFAP and 1 to 8 weeks for sNfL. Demographic-adjusted z score cutoffs effectively distinguished between attack and remission.
Meaning
Optimizing sGFAP and sNfL timing and demographic-adjusted cutoffs could improve accuracy in assessing NMOSD disease activity.
Abstract
Importance
The temporal dynamics of serum glial fibrillary acidic protein (sGFAP) and serum neurofilament light chain (sNfL) as biomarkers of disease activity for neuromyelitis optica spectrum disorder (NMOSD) remain underexplored.
Objective
To determine optimal timing for assessing sGFAP and sNfL, establish cutoff values differentiating between attacks and remissions in NMOSD, and evaluate these findings across independent cohorts.
Design, Setting, and Participants
This retrospective, longitudinal, multicenter cohort study was conducted among patients with aquaporin-4 antibody (AQP4-IgG)–positive NMOSD. Patients with available stored serum samples were included, totaling 181 patients with 625 samples. Discovery cohort samples were collected from February 2008 to October 2023 and validation cohort samples were collected from January 2013 to October 2023. A combined analysis of both cohorts was conducted from November 2023 to March 2024.
Exposures
sNfL and sGFAP concentrations, measured by a single-molecule array assay.
Main Outcomes and Measures
The primary outcomes were the optimal timing of assessing sGFAP and sNfL and the adjusted cutoff values for evaluating disease activity in NMOSD.
Results
The discovery cohort consisted of 366 samples from 78 Korean patients (median [IQR] age, 35 [30-42] years; 73 female patients [95%]), while the validation cohort included 190 samples from 34 German patients (median [IQR] age, 54 [39-61] years; 32 female patients [94%]) and 69 samples from 69 Brazilian patients (median [IQR] age, 46 [35-55] years; 62 female patients [90%]). Six-month postattack temporal biomarker dynamics were analyzed in 202 samples from 74 patients in the discovery cohort: sGFAP levels peaked within the first week and sNfL levels peaked at 5 weeks postattack. The optimal time frames for evaluating attacks were within 1 week for sGFAP and from 1 to 8 weeks for sNfL, with remission defined as at least 6 months postattack. z Score cutoffs of 3.0 for sGFAP and 2.1 for sNfL effectively distinguished between attack and remission phases, indicated by area under the curve values of 0.95 (95% CI, 0.88-1.02) and 0.87 (95% CI, 0.82-0.91), respectively. The discovery cohort time frames and cutoff values were applied to the validation cohort, achieving 71% sensitivity and 94% specificity for sNfL and 100% sensitivity and specificity for sGFAP in the German and Brazilian cohorts.
Conclusions and Relevance
This longitudinal cohort study established optimal timing and thresholds for sGFAP and sNfL, which were consistent in independent cohorts, supporting these biomarkers’ effectiveness in distinguishing NMOSD attacks from remission.
This longitudinal cohort study determines the optimal timing for assessing serum glial fibrillary acidic protein and serum neurofilament light chain and establishes cutoff values differentiating between attack and remission phases in patients with aquaporin-4 antibody (AQP4-IgG)–positive neuromyelitis optica spectrum disorder across independent cohorts.
Introduction
Advancements in neuromyelitis optica spectrum disorder (NMOSD) treatment have shifted research toward identifying blood-based biomarkers for monitoring disease activity, attack severity, and treatment efficacy. Notably, glial fibrillary acidic protein (GFAP), an intermediate filament of astrocytes, increases in blood and cerebrospinal fluid during NMOSD attacks, which is consistent with the pathophysiology of aquaporin-4 antibody (AQP4-IgG)–mediated astrocytopathy.1,2,3,4,5,6,7 Similarly, neurofilament light chain (NfL), a neuron-specific cytoskeletal protein, is released upon neuroaxonal injury and also exhibits higher levels during attacks compared to remission.2,3,8,9
However, the clinical application of serum GFAP (sGFAP) and serum NfL (sNfL) in NMOSD remains complex, particularly in establishing consistent cutoff levels to distinguish between attack and remission phases. Previous studies have established a range of GFAP cutoff levels from 90 to 337 pg/mL to identify elevated biomarker levels during the remission phase and to predict impending relapses.3,6,10,11 Additionally, there is debate about whether biomarkers increase during remission and whether such an increase indicates ongoing disease activity.1,2,3,6,9,10 Some studies have found elevated sGFAP levels during remission, suggesting ongoing subclinical injury in NMOSD,3,6,9 while others have reported no increase compared to healthy controls.1,2,11 These discrepancies may arise from variations in the timing of attack and remission assessments, often arbitrarily set at 2 or 3 months postattack, without considering the temporal dynamics of the biomarkers. Furthermore, overlooking potential confounding demographic variables, such as age, sex, and body mass index (BMI), could contribute to misinterpreting biomarker data.3,6,9,10,11,12,13,14,15
Our research aimed to elucidate the dynamics of sGFAP and sNfL following attacks to determine the most appropriate timing for evaluating attack and remission phases in NMOSD. Additionally, we aimed to establish thresholds adjusting for demographic features to differentiate between attack and remission in NMOSD.
Methods
Study Design and Patients With NMOSD
This study analyzed 366 blood samples from 78 Korean patients with NMOSD, all confirmed to be AQP4-antibody positive, which were collected at the National Cancer Center, Korea, and formed the discovery cohort. For external validation, 69 samples from 69 Brazilian patients and 190 samples from 34 German patients were used, all verified AQP4-antibody positive. All patients met the 2015 diagnostic criteria for NMOSD. AQP4-antibody status was determined through a cell-based assay conducted at each respective center.16,17,18 Samples were collected between February 2008 and October 2023, with data analysis conducted from November 2023 to March 2024. Demographic data, including age at sampling, sex, and BMI, as well as clinical data, such as date of disease onset, date of last attack before blood sampling, type of attack, attack severity, Expanded Disability Status Scale (EDSS) score, and type of immunosuppressive therapy (IST) at sampling, were collected. The severity of attacks was assessed using a modified version of the opticospinal impairment scale, which classifies attacks as major or minor based on alterations in domain-specific neurological function scores.5,19 Patient data were anonymized and shared according to the individual centers’ local ethical requirements, and the study was approved by the institutional review boards or research ethics committees of each participating center. All participants provided written informed consent. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.
Measurement of sNfL and sGFAP Concentrations
The collected samples were immediately separated by centrifugation and frozen at −80 °C until analysis. sNfL and sGFAP concentrations were measured using ultrasensitive single-molecule array (Neurology 2-plex B assay, Simoa) technology (Quanterix) as per the manufacturer’s instructions. All samples were analyzed in Basel, Switzerland, and subsequently validated in Goyang, Korea, for Korean patients. The intra-assay coefficient of variation of the assays was less than 10%.
Statistical Analysis
Continuous variables were described by median and interquartile ranges (IQRs) or ranges, while categorical variables were described by counts and percentages. Mann-Whitney U tests were used to compare continuous variables. sNfL and sGFAP z scores, adjusted for age, sex (for sGFAP only), and BMI, were calculated using a previously described normative healthy control database.15,20 Spline terms in generalized additive mixed models were used to track longitudinal changes in sNfL and sGFAP levels postattack, treating each patient’s trajectory as a random effect to address interindividual variability. Receiver operating characteristics (ROC) curve analysis and the Youden index were used to determine the optimal z score cutoffs for distinguishing attack and remission phases. At predefined time intervals following onset of attack symptoms, determined by the biomarker dynamics, we calculated diagnostic performance metrics, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy, to determine the optimal timing for assessing the attack phase. Additionally, it was assumed that values exceeding a z score of 2.0 during remission were abnormally elevated, and the percentage of such elevations was evaluated.6 Multivariable Cox regression analyses were performed to assess the association between biomarker levels during remission and the risk of future relapses. The models were adjusted for age at sampling, time between the most recent attack and sampling, type of IST, EDSS scores, and z scores of biomarkers during remission. All P value tests were 2-tailed. An adjusted P value less than .05, incorporating multiple comparisons when necessary, was considered statistically significant. Analyses were performed using GraphPad version 10.2.3 (PRISM) and R version 4.1.2 (R Foundation).
Results
Temporal Dynamics of sGFAP and sNfL Following Attacks
The flowchart of the study design is presented in eFigure 1 in Supplement 1. Our analysis began by examining the temporal patterns of biomarker levels in 202 serial blood samples collected from 74 patients (out of a total of 78 patients with 366 samples in the discovery cohort), with samples available within 6 months postattack. The characteristics of the 74 patients are detailed in the eTable in Supplement 1. All patients received intravenous methylprednisolone treatment at a median (IQR) of 2 (1-4) days after symptom onset, and 6 patients (8%) underwent plasma exchange at a median (IQR) of 8 (7-21) days after the onset of symptoms. The median z score of sGFAP peaked at 4.5 (95% CI, 4.1-5.0) in the first week following symptom onset (week 0), then rapidly declined over the following 12 weeks (Figure 1A). In contrast, sNfL levels gradually increased, peaking at 5 weeks after symptom onset, with the peak z score value of 2.8 (95% CI, 2.4-3.2; Figure 1B). Following this peak, the z scores for sNfL declined more slowly than those of sGFAP over the next 20 weeks. The temporal trajectories of sGFAP and sNfL across individual cases are illustrated in eFigure 2 in Supplement 1.
Figure 1. Temporal Biomarker Profiles Postattack Analyzed Using Generalized Additive Mixed Models Based on the Discovery Cohort.
For both serum glial fibrillary acidic protein (sGFAP) (A) and serum neurofilament light chain (sNfL) (B), the estimated marginal effects of time on serum biomarker concentrations are illustrated with black lines. The 95% confidence intervals surrounding these estimates are depicted as contiguous shaded bands. Data points representing individual serum biomarker measurements are denoted by black dots. For comparative purposes, the average biomarker levels observed in the healthy control groups are overlaid with blue lines (z score: 0).15,20
Establishing Optimal Time Frames and Cutoffs for Differentiation of Attacks From Remission
Considering the temporal dynamics of sGFAP and sNfL, diagnostic performance metrics were assessed across various intervals following the onset of attack symptoms throughout the entire discovery cohort. This analysis included an initial 202 samples from 74 patients and was extended to include an additional 164 samples collected more than 6 months postattack from a total of 78 patients (the original 74 patients and an additional 4 patients). Characteristics of the 78 patients in the discovery cohort are summarized in Table 1. The evaluation during the attack phase focused on peak timings, while the remission phase assumed that biomarker levels returned to normal postattack. Based on AUC results, it was established that the optimal time frames for assessing sGFAP and sNfL levels during an attack are within the first week and from 1 to 8 weeks following the onset of attack symptoms, respectively (Table 2). Optimal effectiveness in assessing sGFAP and sNfL levels during remission was observed at least 6 months following attacks. The optimal cutoff for the sGFAP z score for distinguishing attack from remission was 3.0 (99.9th percentile), with an AUC of 0.95 (95% CI, 0.88-1.02; Table 2; eFigure 3 in Supplement 1). This cutoff demonstrated a sensitivity of 85%, a specificity of 96%, a PPV of 87%, an NPV of 96%, and an overall accuracy rate of 94%. For sNfL, the optimal cutoff was identified at a z score of 2.1 (98.3th percentile), with an AUC of 0.87 (95% CI, 0.82-0.91). This cutoff provided a sensitivity of 71%, specificity of 89%, PPV of 77%, NPV of 85%, and an accuracy of 83%. Upon applying the optimal assessment time frames, the median (IQR) z scores for sGFAP during attacks were significantly higher than during remission (4.7 [3.9-5.2] vs 1.0 [0.3-1.6]; P < .001). Median (IQR) z scores of sNfL also showed a similar trend, being higher during attacks (2.8 [1.8-3.3]) compared to remission (0.8 [−0.2 to 1.6]); P < .001; Figure 2B). Also following this pattern were absolute median (IQR) levels of sGFAP (5076 pg/mL [964-22 872] vs 103 pg/mL [78-133]; P < .001) and sNfL (27 pg/mL [13-46]) vs 8 pg/mL [6-12]; P < .001), which were markedly higher during the attack and lower in remission (Figure 2A).
Table 1. Demographic and Clinical Characteristics of Patients From Korea, Brazil, and Germany.
| Characteristic | Patients, No. (%) | ||
|---|---|---|---|
| Korea (n = 78) | Brazil (n = 69) | Germany (n = 34) | |
| Samples, No. | 366 | 69 | 190 |
| Race, %a | Asian 100% | African 23%; Asian 3%; Caucasian 36%; and mixed race 38% | African 6%; Arabic 3%; Asian 6%; Caucasian 85% |
| Age at disease onset, median (IQR), y | 30 (20-38) | 32 (23-44) | 41 (26-51) |
| Age at sampling, median (IQR), y | 35 (30-42) | 46 (35-55) | 54 (39-61) |
| Sex | |||
| Female | 73 (95) | 62 (90) | 32 (94) |
| Male | 5 (5) | 7 (10) | 2 (6) |
| BMI, median (IQR)b | 21.5 (19.5-23.9) | 26.3 (25.3-26.3) | 23.2 (22-26.4) |
| Disease duration at sampling, median (IQR), mo | 75 (33-137) | 113 (81-219) | 102 (64-158) |
| EDSS score at the time of sampling, median (IQR) | 3.5 (2-4.0) | 4 (3.0-5.0) | 3.5 (2.0-4.0) |
| Immunosuppressive therapy at sampling | 332 (91) | 67 (95) | 155 (79) |
| Azathioprine | 19 (5) | 32 (46) | 38 (19) |
| Mycophenolate mofetil | 61 (17) | 7 (10) | 5 (3) |
| Rituximab | 239 (65) | 26 (37) | 106 (54) |
| Other therapies | 13 (4) | 2 (2) | 6 (3) |
Abbreviations: BMI, body mass index; EDSS, Expanded Disability Status Scale.
Race reported based on information documented in patients' medical records, which were reviewed and categorized by the researchers for the purposes of this study.
Calculated as weight in kilograms divided by height in meters squared.
Table 2. Comparative Analysis of Area Under the Curve (AUC) for Serum Glial Fibrillary Acidic Protein and Serum Neurofilament Light Chain z Scores in Distinguishing Attack vs Remission Across Different Time Frames.
| Time frame for attack and remission | No. of attack samples/No. of remission samples | AUC (95% CI) | Cutoff z score | % | ||||
|---|---|---|---|---|---|---|---|---|
| Sensitivity | Specificity | PPV | NPV | Accuracy | ||||
| Serum glial fibrillary acidic protein | ||||||||
| Within 1 wk post–symptom onset vs at least 6 mo after the attack symptoms | 46/164 | 0.95 (0.88-1.02) | 3.0 | 85 | 96 | 87 | 96 | 94 |
| Within 2 wk post–symptom onset vs at least 6 mo after the attack symptoms | 60/164 | 0.93 (0.86-1.01) | 2.9 | 85 | 93 | 79 | 96 | 91 |
| Within 3 wk post–symptom onset vs at least 6 mo after the attack symptoms | 81/164 | 0.90 (0.82-0.98) | 2.8 | 77 | 92 | 79 | 91 | 88 |
| Within 1 wk post–symptom onset vs at least 3 mo after the attack symptoms | 46/208 | 0.95 (0.87-1.02) | 3.0 | 85 | 96 | 81 | 97 | 94 |
| Within 2 wk post–symptom onset vs at least 3 mo after the attack symptoms | 60/208 | 0.93 (0.88-0.97) | 2.9 | 85 | 93 | 79 | 96 | 91 |
| Within 3 wk post–symptom onset vs at least 3 mo after the attack symptoms | 81/208 | 0.90 (0.85-0.95) | 2.8 | 75 | 92 | 79 | 92 | 88 |
| Serum neurofilament light chain | ||||||||
| Within 1-6 wk post–symptom onset vs at least 6 mo after the attack symptoms | 70/164 | 0.86 (0.80-0.91) | 2.1 | 71 | 89 | 74 | 88 | 84 |
| Within 6 wk post–symptom onset vs at least 6 mo after the attack symptoms | 111/164 | 0.78 (0.72-0.84) | 2.1 | 59 | 89 | 78 | 76 | 77 |
| Within 1-8 wk post–symptom onset vs at least 6 mo after the attack symptoms | 83/164 | 0.87 (0.82-0.91) | 2.1 | 71 | 89 | 77 | 85 | 83 |
| Within 8 wk post–symptom onset vs at least 6 mo after the attack symptoms | 124/164 | 0.79 (0.74-0.84) | 2.1 | 61 | 89 | 81 | 75 | 77 |
Abbreviations: NPV, negative predictive value; PPV, positive predictive value.
Figure 2. Comparison of Serum Glial Fibrillary Acidic Protein (sGFAP) and Serum Neurofilament Light Chain (sNfL) Levels Between Attack and Remission.
Comparison of sGFAP and sNfL levels between attack and remission based on the discovery cohort data: absolute values (A) and z scores (B). The data consist of medians and interquartile ranges. P values were calculated using the Wilcoxon signed rank test.
Clinical Factors Associated With sGFAP and sNfL Levels During Attack and Remission
In the discovery cohort, 83 attacks were identified, comprising 67 cases of myelitis, 10 cases of optic neuritis (ON), 4 attacks involving the brain and/or brainstem, and 2 multifocal attacks. Comparing myelitis to ON, myelitis showed a higher median (IQR) sGFAP z score (4.9 [4.0-5.2]) compared to ON (3.1 [1.9-4.3]) (P = .047), whereas the sNfL z score did not differ. Major attacks were associated with higher median (IQR) sNfL z scores (2.95 [1.47-3.54]) compared to minor attacks (1.82 [0.85-2.73]; P = .003), while sGFAP z scores did not differ significantly between major and minor attacks according to the opticospinal impairment scale. During remission (at least 6 months after attacks), 24 of 164 sGFAP samples (16%) and 20 of 164 sNfL samples (12%) exhibited elevated z scores greater than 2. No significant differences in disease duration, IST type, EDSS score, or time since the last relapse were found between patients with and without elevated sGFAP or sNfL levels during remission. In the multivariable Cox regression analysis, of the 164 remission samples (at least 6 months after the attack) analyzed, 54 were followed by a relapse. The remaining samples were censored at the last follow-up due to the absence of relapse. In these 54 samples, the median (IQR) time to the next relapse was 12 months (3-91), with 18 patients (33%) experiencing a relapse within 6 months. The median (IQR) interval from blood sampling to either relapse or the last follow-up was 107 months (24-155). Higher EDSS scores at remission were linked to an increased risk of early relapse (hazard ratio [HR], 1.26; 95% CI, 1.04-1.54). Conversely, compared to oral drugs, rituximab therapy was associated with a reduction in the risk of early relapse (HR, 0.14; 95% CI, 0.07-0.29). However, sNfL and sGFAP z scores in remission were not associated with early relapse risk (Table 3).
Table 3. Multivariable Cox Regression Analysis of Factors Associated With Time to Next Relapse.
| Variable | Univariable | Multivariable | ||
|---|---|---|---|---|
| HR (95% CI) | P value | HR (95% CI) | P value | |
| Age at sampling, y | 1.02 (0.91-1.06) | .16 | 1.02 (0.98-1.06) | .31 |
| EDSS score at sampling | 1.22 (1.02-1.46) | .03 | 1.26 (1.04-1.54) | .02 |
| Time interval from the most recent attack to sampling, mo | 0.99 (0.97-1.02) | .53 | 0.98 (0.96-1.01) | .26 |
| Type of IST | ||||
| MMF or AZA | 1 | 1 | ||
| Rituximab | 0.19 (0.10-0.37) | <.001 | 0.14 (0.07-0.29) | <.001 |
| z Score | ||||
| sGFAP in remission | 1.16 (0.89-1.49) | .27 | 1.11 (0.78-1.57) | .56 |
| sNfL in remission | 1.34 (1.05-1.71) | .02 | 1.19 (0.89-1.61) | .24 |
Abbreviations: AZA, azathioprine; EDSS, Expanded Disability Status Scale; HR, hazard ratio; IST, immunosuppressive therapy; MMF, mycophenolate mofetil; sGFAP, serum glial fibrillary acidic protein; sNfL, serum neurofilament light chain.
External Validation of Optimal Time Frame and Biomarker Thresholds in Brazilian and German NMOSD Cohorts
The characteristics of the Brazilian and German cohorts are detailed in Table 1. In the Brazilian cohort of 69 patients, only 1 sNfL sample per patient was included due to biodegradation issues with sGFAP data, resulting in a total of 69 sNfL data points. In the German cohort, serial blood sampling was conducted for 28 of 34 patients, resulting in a total of 190 sGFAP and sNfL data points, with a median of 7 samples collected per patient. The limited number of samples collected within 6 months postattack (only 11 for NfL and 5 for GFAP) in the validation cohort hindered analysis of temporal dynamics and ROC curves. Consequently, the cutoffs and optimal time frames established in the discovery cohort were applied to assess sensitivity and specificity in the validation cohort. Among the entire validation cohort, 7 NfL samples were collected within the optimal attack time frame (1 to 8 weeks postattack), and 245 were collected during remission (at least 6 months after the last attack). For GFAP, only 1 sample was available within the 7-day postattack window, while 184 samples were collected during remission. Ultimately, 189 sGFAP samples and 252 sNfL samples were used for validation analysis. The single sample from an attack phase exhibited an sGFAP z score of 5.0, surpassing the established cutoff value of 3.0. Conversely, all 196 samples from the remission phase showed sGFAP values below this cutoff, achieving 100% sensitivity and 100% specificity (eFigure 4 in Supplement 1). Using a z score cutoff of 2.1 to differentiate attacks from remission, sNfL demonstrated a sensitivity of 71%, a specificity of 94%, a PPV of 26%, an NPV of 99%, and an overall accuracy of 94%. During the remission phase, an elevation in z scores beyond 2 was noted in 5 of the sGFAP samples (3%) and 11 of the sNfL samples (4%).
Discussion
This study reports distinct dynamics of sGFAP and sNfL following attacks in patients with NMOSD. sGFAP levels reached a peak within the first week after symptom onset. In contrast, sNfL increased more slowly than sGFAP during attacks, with some patients initially showing normal levels within 1 week of symptom onset, peaking at 5 weeks. The decline rate in sNfL levels was slower than that observed in sGFAP. Considering these distinct temporal profiles, this analysis elucidates the optimal timing for these biomarkers’ assessment in NMOSD attack: within the first week postattack for sGFAP and between 1 to 8 weeks postattack for sNfL. To sufficiently rule out lingering effects of a previous attack, it is recommended that tests be conducted for sGFAP and sNfL levels in remission at least 6 months following attacks. Implementing the defined time frame for attack and remission in this study revealed a significantly larger disparity in median sGFAP concentrations between the attack and remission phases (5076 pg/mL vs 103 pg/mL). This markedly exceeds earlier findings of 184 to 541 pg/mL for attack and 109 to 153 pg/mL for remission.2,3,4,9,10
Concentrations of sNfL and sGFAP are influenced by factors including patients’ age, BMI, and sex (for sGFAP), necessitating adjustments for accurate interpretation.14,15,20 In this study, using demographic features–adjusted z scores enabled more precise identification of abnormal biomarker levels, facilitating personalized assessments. High sensitivity and specificity were achieved in differentiating attacks from remission phases by measuring biomarkers during their optimal periods and applying z scores. The thresholds were validated in external international cohorts, confirming their robustness and applicability across different populations. In the N-MOmentum trial, the increase of sGFAP during remission was defined by elevation above the mean plus 2 SD cutoff of 170 pg/mL in healthy individuals, wherein the elevation of sGFAP at baseline was linked to impending attacks.6 However, in this study’s cohort, the absolute level corresponding to a z score of 2.0 (mean + 2 SD) for sGFAP varied from 130 to 276 pg/mL depending on demographic factors. This suggests that assessing elevation solely based on absolute values without adjusting for demographic factors may pose limitations. Furthermore, the N-MOmentum study’s requirement of only a 4-week stability period following the last attack suggests that elevated baseline sGFAP levels in some participants could be residual from a recent attack.5,6 In this study, multivariable analysis revealed that sGFAP and sNfL levels during remission, measured at least 6 months postattack, did not predict the next relapse. Instead, higher EDSS scores were associated with a greater likelihood of early relapse, reflecting more active disease. The use of rituximab significantly reduced the risk of relapse compared to oral ISTs, aligning with existing literature.21,22,23 Moreover, upon analyzing collective data from patients in Korea, Brazil, and Germany, over 90% of remission samples collected more than 6 months postattack showed sGFAP and sNfL levels that were not abnormally elevated, with 319 of 348 samples for sGFAP (92%) and 378 of 409 samples for sNfL (92%) showing z scores below 2. While these high proportions of sGFAP levels within expected thresholds may suggest minimal overt astrocytic damage during remission in NMOSD,1,7,24,25,26,27 the sensitivity of sGFAP to detect all forms of astrocytic damage is not fully established. Emerging studies using various methodologies suggest potential subclinical injuries,28,29 yet their clinical significance, particularly in relation to disability progression, remains uncertain.27 Further research is crucial to validate the role of sGFAP in detecting subclinical activity and to assess its clinical relevance in NMOSD.
The pattern of sGFAP levels peaking early and sNfL levels peaking later during NMOSD attacks may be attributed to both the pathophysiology, AQP4 antibody–mediated astrocytopathy, followed by neuronal injury,30 as well as the distinct biological half-lives of these biomarkers. sGFAP increases within 1 hour following an injury, with a half-life of 24 to 48 hours.31 Initial intense astrocytopathy may lead to a significant rise in sGFAP within the first week following symptom onset, followed by a rapid decline, indicating that elimination outpaces production during the subsequent period. In contrast, sNfL, with its longer half-life of 20 to 40 days,32 can accumulate during the neuroinflammation phase of the attack, leading to delayed peak timing. The peak in sNfL levels around 5 weeks postattack correlates with worsening neurological conditions, underscoring its utility in assessing neuroaxonal damage. Attack severity correlated with sNfL levels but not sGFAP levels, which is consistent with a recent biomarker study conducted as part of the N-MOmentum trial,5 which found that sNfL levels at the time of attack were superior to sGFAP in predicting disability worsening during attacks.
Limitations
Due to its retrospective design, this study has inherent limitations, including potential sampling bias and reporting biases. Additionally, the heterogeneity of sampling frequency poses challenges in accurately determining standardized elimination half-lives for these biomarkers after peak concentrations. Furthermore, in the validation cohorts, the limited number of samples within the first 6 months postattack restricted the evaluation of biomarker temporal dynamics. Finally, as this study evaluated sGFAP and sNfL in patients receiving conventional ISTs or without any IST, further research is needed to assess the cutoffs in patients treated with the latest US Food and Drug Administration–approved monoclonal antibodies for NMOSD, particularly considering the less-pronounced increases in sGFAP levels during attacks in patients treated with inebilizumab.6
Conclusions
This longitudinal cohort study benefited from an international, multicenter collaboration, using the largest dataset of samples across diverse racial and ethnic groups published to date. This collaboration enhances the generalizability and robustness of these findings. sGFAP may help differentiate genuine attacks from pseudoattacks, particularly in challenging clinical scenarios with preexisting T2 signal abnormalities or limited gadolinium enhancement on MRI.33,34 Furthermore, sNfL’s correlation with neuronal damage severity in NMOSD attacks offers insights into predicting recovery after an attack. In conclusion, this study significantly advances our comprehension of sNfL and sGFAP as biomarkers for NMOSD. By refining their predictive accuracy and enhancing their reliability, our research contributes to improving the use of these biomarkers in clinical settings, facilitating more informed and precise management of patients with NMOSD.
eFigure 1. The Flow Chart of the Study
eFigure 2. Longitudinal Profiles of z Scores of Serum Glial Fibrillary Acidic Protein (GFAP) (A) and Serum Neurofilament Light Chain (NfL) (B) Over 6 Months Following an Attack in Individual Patients From the Discovery Cohort
eFigure 3. ROC Curves for the z Score of Serum Glial Fibrillary Acidic Protein (A) and Serum Neurofilament Acidic Protein (B) in Distinguishing Between Attack and Remission
eFigure 4. Comparison of Serum Glial Fibrillary Acidic Protein (sGFAP) and Serum Neurofilament Light Chain (sNfL) Levels Between Attack and Remission Phases
eTable. Characteristics of the 74 Korean Patients From the Discovery Cohort, Who Were Included in the 6-Month Post-Attack Temporal Dynamics Analysis of sGFAP and sNfL
Data Sharing Statement
References
- 1.Hyun JW, Kim Y, Kim SY, Lee MY, Kim SH, Kim HJ. Investigating the presence of interattack astrocyte damage in neuromyelitis optica spectrum disorder: longitudinal analysis of serum glial fibrillary acidic protein. Neurol Neuroimmunol Neuroinflamm. 2021;8(3):3965. doi: 10.1212/NXI.0000000000000965 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Kim H, Lee EJ, Kim S, et al. Longitudinal follow-up of serum biomarkers in patients with neuromyelitis optica spectrum disorder. Multiple Scler. 2022;28(4):512-521. doi: 10.1177/13524585211024978 [DOI] [PubMed] [Google Scholar]
- 3.Watanabe M, Nakamura Y, Michalak Z, et al. Serum GFAP and neurofilament light as biomarkers of disease activity and disability in NMOSD. Neurology. 2019;93(13):e1299-e1311. doi: 10.1212/WNL.0000000000008160 [DOI] [PubMed] [Google Scholar]
- 4.Chang X, Huang W, Wang L, et al. Serum neurofilament light and GFAP are associated with disease severity in inflammatory disorders with aquaporin-4 or myelin oligodendrocyte glycoprotein antibodies. Front Immunol. 2021;12:647618. doi: 10.3389/fimmu.2021.647618 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Aktas O, Hartung HP, Smith MA, et al. ; N-MOmentum study investigators . Serum neurofilament light chain levels at attack predict post-attack disability worsening and are mitigated by inebilizumab: analysis of four potential biomarkers in neuromyelitis optica spectrum disorder. J Neurol Neurosurg Psychiatry. 2023;94(9):757-768. doi: 10.1136/jnnp-2022-330412 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Aktas O, Smith MA, Rees WA, et al. ; N-MOmentum scientific group and the N-MOmentum study investigators . Serum glial fibrillary acidic protein: a neuromyelitis optica spectrum disorder biomarker. Ann Neurol. 2021;89(5):895-910. doi: 10.1002/ana.26067 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Schindler P, Aktas O, Ringelstein M, et al. Glial fibrillary acidic protein as a biomarker in neuromyelitis optica spectrum disorder: a current review. Expert Rev Clin Immunol. 2023;19(1):71-91. doi: 10.1080/1744666X.2023.2148657 [DOI] [PubMed] [Google Scholar]
- 8.Kim S, Lee JJ, Park JS, Kang M, Seok HY. Neurofilament light chain as a biomarker in neuromyelitis optica spectrum disorder: a comprehensive review and integrated analysis with glial fibrillary acidic protein. Neurol Sci. 2024;45(3):1255-1261. doi: 10.1007/s10072-023-07277-8 [DOI] [PubMed] [Google Scholar]
- 9.Liu C, Lu Y, Wang J, et al. Serum neurofilament light chain and glial fibrillary acidic protein in AQP4-IgG-seropositive neuromyelitis optica spectrum disorders and multiple sclerosis: a cohort study. J Neurochem. 2021;159(5):913-922. doi: 10.1111/jnc.15478 [DOI] [PubMed] [Google Scholar]
- 10.Yang S, Zhang C, Zhang TX, et al. A real-world study of interleukin-6 receptor blockade in patients with neuromyelitis optica spectrum disorder. J Neurol. 2023;270(1):348-356. doi: 10.1007/s00415-022-11364-9 [DOI] [PubMed] [Google Scholar]
- 11.Schindler P, Grittner U, Oechtering J, et al. Serum GFAP and NfL as disease severity and prognostic biomarkers in patients with aquaporin-4 antibody-positive neuromyelitis optica spectrum disorder. J Neuroinflamm. 2021;18(1):105. doi: 10.1186/s12974-021-02138-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Lee EJ, Lim YM, Kim S, et al. Clinical implication of serum biomarkers and patient age in inflammatory demyelinating diseases. Ann Clin Translational Neurol. 2020;7(6):992-1001. doi: 10.1002/acn3.51070 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Yalachkov Y, Schäfer JH, Jakob J, et al. Effect of estimated blood volume and body mass index on GFAP and NfL levels in the serum and CSF of patients with multiple sclerosis. Neurol Neuroimmunol Neuroinflamm. 2022;10(1):e200045. doi: 10.1212/NXI.0000000000200045 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Meier S, Willemse EAJ, Schaedelin S, et al. Serum glial fibrillary acidic protein compared with neurofilament light chain as a biomarker for disease progression in multiple sclerosis. JAMA Neurol. 2023;80(3):287-297. doi: 10.1001/jamaneurol.2022.5250 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Benkert P, Meier S, Schaedelin S, et al. ; NfL Reference Database in the Swiss Multiple Sclerosis Cohort Study Group . Serum neurofilament light chain for individual prognostication of disease activity in people with multiple sclerosis: a retrospective modelling and validation study. Lancet Neurol. 2022;21(3):246-257. doi: 10.1016/S1474-4422(22)00009-6 [DOI] [PubMed] [Google Scholar]
- 16.Wingerchuk DM, Banwell B, Bennett JL, et al. ; International Panel for NMO Diagnosis . International consensus diagnostic criteria for neuromyelitis optica spectrum disorders. Neurology. 2015;85(2):177-189. doi: 10.1212/WNL.0000000000001729 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Kim Y, Kim G, Kong BS, et al. Large-scale in-house cell-based assay for evaluating the serostatus in patients with neuromyelitis optica spectrum disorder based on new diagnostic criteria. J Clin Neurol. 2017;13(2):175-180. doi: 10.3988/jcn.2017.13.2.175 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Waters P, Reindl M, Saiz A, et al. Multicentre comparison of a diagnostic assay: aquaporin-4 antibodies in neuromyelitis optica. J Neurol Neurosurg Psychiatry. 2016;87(9):1005-1015. doi: 10.1136/jnnp-2015-312601 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Wingerchuk DM, Hogancamp WF, O’Brien PC, Weinshenker BG. The clinical course of neuromyelitis optica (Devic’s syndrome). Neurology. 1999;53(5):1107-1114. doi: 10.1212/WNL.53.5.1107 [DOI] [PubMed] [Google Scholar]
- 20.Canellas AR. Educational session 2 (MAGNIMS): imaging features of MOGAD and differential diagnosis with MS. Multiple Sclerosis J. 2023;29(3)(suppl):4-136. doi: 10.1177/13524585231196191 [DOI] [Google Scholar]
- 21.Huang W, Wang L, Xia J, et al. ; Pan-Yangtze River Delta Alliance for Demyelinating Disease . Efficacy and safety of azathioprine, mycophenolate mofetil, and reduced dose of rituximab in neuromyelitis optica spectrum disorder. Eur J Neurol. 2022;29(8):2343-2354. doi: 10.1111/ene.15355 [DOI] [PubMed] [Google Scholar]
- 22.Dresser L, Chaar WA, Reder AT, Abuaf AF, Cipriani VP, Javed A. Effectiveness of rituximab versus oral immunosuppressive therapies in neuromyelitis optica spectrum disorder in a racially diverse cohort of subjects: a single-center retrospective study. Multiple Sclerosis Related Disord. 2023;74:104718. doi: 10.1016/j.msard.2023.104718 [DOI] [PubMed] [Google Scholar]
- 23.Jeong IH, Park B, Kim SH, Hyun JW, Joo J, Kim HJ. Comparative analysis of treatment outcomes in patients with neuromyelitis optica spectrum disorder using multifaceted endpoints. Multiple Sclerosis. 2016;22(3):329-339. doi: 10.1177/1352458515587752 [DOI] [PubMed] [Google Scholar]
- 24.Papadopoulou A, Oertel FC, Chien C, et al. Lateral geniculate nucleus volume changes after optic neuritis in neuromyelitis optica: a longitudinal study. Neuroimage Clin. 2021;30:102608. doi: 10.1016/j.nicl.2021.102608 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Lee MY, Yong KP, Hyun JW, Kim SH, Lee SH, Kim HJ. Incidence of interattack asymptomatic brain lesions in NMO spectrum disorder. Neurology. 2020;95(23):e3124-e3128. doi: 10.1212/WNL.0000000000010847 [DOI] [PubMed] [Google Scholar]
- 26.Rogaczewska M, Michalak S, Stopa M. Optical coherence tomography angiography of peripapillary vessel density in multiple sclerosis and neuromyelitis optica spectrum disorder: a comparative study. J Clin Med. 2021;10(4):609. doi: 10.3390/jcm10040609 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Molazadeh N, Filippatou AG, Vasileiou ES, Levy M, Sotirchos ES. Evidence for and against subclinical disease activity and progressive disease in MOG antibody disease and neuromyelitis optica spectrum disorder. J Neuroimmunol. 2021;360:577702. doi: 10.1016/j.jneuroim.2021.577702 [DOI] [PubMed] [Google Scholar]
- 28.Oertel FC, Havla J, Roca-Fernández A, et al. Retinal ganglion cell loss in neuromyelitis optica: a longitudinal study. J Neurol Neurosurg Psychiatry. 2018;89(12):1259-1265. doi: 10.1136/jnnp-2018-318382 [DOI] [PubMed] [Google Scholar]
- 29.Motamedi S, Oertel FC, Yadav SK, et al. Altered fovea in AQP4-IgG-seropositive neuromyelitis optica spectrum disorders. Neurol Neuroimmunol Neuroinflamm. 2020;7(5):e805. doi: 10.1212/NXI.0000000000000805 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Lucchinetti CF, Guo Y, Popescu BF, Fujihara K, Itoyama Y, Misu T. The pathology of an autoimmune astrocytopathy: lessons learned from neuromyelitis optica. Brain Pathol. 2014;24(1):83-97. doi: 10.1111/bpa.12099 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Papa L, Brophy GM, Welch RD, et al. Time course and diagnostic accuracy of glial and neuronal blood biomarkers GFAP and UCH-L1 in a large cohort of trauma patients with and without mild traumatic brain injury. JAMA Neurol. 2016;73(5):551-560. doi: 10.1001/jamaneurol.2016.0039 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Bavato F, Barro C, Schnider LK, et al. Introducing neurofilament light chain measure in psychiatry: current evidence, opportunities, and pitfalls. Mol Psychiatry. 2024;29(8):2543-2559. doi: 10.1038/s41380-024-02524-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Weinshenker BG. MRI of the relevant domain should be performed to confirm whether clinical symptoms represent an attack of NMOSD: commentary. Multiple Sclerosis. 2024;30(1):23-24. doi: 10.1177/13524585231222131 [DOI] [PubMed] [Google Scholar]
- 34.Kessler RA, Mealy MA, Levy M. Early indicators of relapses vs pseudorelapses in neuromyelitis optica spectrum disorder. Neurol Neuroimmunol Neuroinflamm. 2016;3(5):e269. doi: 10.1212/NXI.0000000000000269 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eFigure 1. The Flow Chart of the Study
eFigure 2. Longitudinal Profiles of z Scores of Serum Glial Fibrillary Acidic Protein (GFAP) (A) and Serum Neurofilament Light Chain (NfL) (B) Over 6 Months Following an Attack in Individual Patients From the Discovery Cohort
eFigure 3. ROC Curves for the z Score of Serum Glial Fibrillary Acidic Protein (A) and Serum Neurofilament Acidic Protein (B) in Distinguishing Between Attack and Remission
eFigure 4. Comparison of Serum Glial Fibrillary Acidic Protein (sGFAP) and Serum Neurofilament Light Chain (sNfL) Levels Between Attack and Remission Phases
eTable. Characteristics of the 74 Korean Patients From the Discovery Cohort, Who Were Included in the 6-Month Post-Attack Temporal Dynamics Analysis of sGFAP and sNfL
Data Sharing Statement
