Abstract
Background and Objectives
In patients with myelin oligodendrocyte glycoprotein (MOG) antibody-associated disease (MOGAD), acute disease activity is generally identified through medical history, neurologic examination, and imaging. However, these may be insufficient for detecting disease activity in specific conditions. This study aimed to investigate the dynamics of serum neurofilament light chain (sNfL) and serum glial fibrillary acidic protein (sGFAP) after clinical attacks and to assess their utility in discriminating attacks from remission in patients with MOGAD.
Methods
We conducted a multicenter, retrospective, longitudinal study including 239 sera from 62 MOGAD patients assessed from 1995 to 2023 in a discovery and validation setup. Sera were measured for sNfL and sGFAP with a single-molecule array assay and for MOG-IgG with a live cell-based assay. sNfL and sGFAP Z scores and percentiles adjusted for age, body mass index, and sex (sGFAP) were calculated from a healthy control normative database. Mixed-effects regression models were used to characterize biomarkers' dynamics and to investigate associations between serum biomarkers, clinical variables, and disease activity status.
Results
Among the 62 study participants, 29 (46.8%) were female, with a median age at baseline of 40.0 years (interquartile range [IQR] 29.5–49.8) and a median duration of follow-up of 20.0 months (IQR 3.0–62.8). sNfL and sGFAP Z scores were nonlinearly associated with time from attack onset (p < 0.001 and = 0.002, respectively). During attacks, both biomarkers presented higher median values (sNfL Z score 2.9 [IQR 1.4–3.5], 99.8th; sGFAP Z score 0.4 [IQR −0.5 to 1.5], 65.5th) compared with remission (sNfL Z score 0.9 [IQR −0.1 to 1.6], 81.6th, p < 0.001; sGFAP Z score −0.2 [IQR −0.8 to 0.5], 42.1th; p < 0.001) across all clinical phenotypes. sNfL values consistently discriminated disease activity status in the discovery and validation cohorts, showing a 3.5-fold increase in the odds of attacks per Z score unit (odds ratio 3.5, 95% confidence interval 2.3–5.1; p < 0.001). Logistic models incorporating sNfL Z scores demonstrated favorable performance in discriminating disease activity status across both cohorts.
Discussion
sNfL Z scores may serve as a biomarker for monitoring disease activity in MOGAD.
Introduction
Disability in patients with myelin oligodendrocyte glycoprotein (MOG) antibody-associated disease (MOGAD) tends to occur incrementally, as a consequence of attacks, and is generally not attributed to progression independent from relapses.1 Subclinical disease activity is infrequent, generally not requiring surveillance.2-4 As a result, acute disease activity is presently detected through neurologic examination and imaging in case of new symptoms or worsening of disability.1 However, this approach may be insufficient in the presence of subtle symptoms or during symptom presentation without deterioration of clinical deficits or MRI abnormalities.
In this context, although structural biomarkers are validated and used for identifying disease activity in patients with multiple sclerosis (MS), previous studies have provided conflicting evidence regarding the utility of serum neurofilament light chain (sNfL) and serum glial fibrillary acidic protein (sGFAP) as monitoring biomarkers in MOGAD.5-8 The limited generalizability of previous studies may result from the small sized cohorts, the low incidence of attacks during follow-up, and the lack of biomarker value adjustments for clinical and demographic features.
To address the above-mentioned challenges, we conducted a multicenter study aiming to characterize the dynamics of sNfL and sGFAP following attacks, and to investigate whether sNfL and sGFAP Z scores can be used to discriminate the status of disease activity (remission vs attacks) in patients with MOGAD.
Methods
Study Design and Participants
We retrospectively investigated 239 sera of 62 adult patients fulfilling the proposed criteria for the diagnosis of MOGAD1 in a discovery and validation setup (Figure 1). Patients were longitudinally assessed at 3 centers (University Hospital Basel, Switzerland, n = 27 patients; University of São Paulo, Brazil, n = 22 patients; and National Cancer Center, South Korea, n = 13 patients) between October 1995 and December 2023. Analyses were conducted in the discovery cohort, comprising participants from the University Hospital Basel. Subsequently, samples from the University of São Paulo and the National Cancer Center were procured for validation of the findings. Serum samples from healthy controls from the National Cancer Center (n = 199) and the University of São Paulo (n = 44) were investigated for comparability with previously described healthy control normative databases.9-11
Figure 1. Study Design.
Flow diagram of study design. MOGAD = MOG antibody-associated disease; sGFAP = serum glial fibrillary acidic protein; sNfL = serum neurofilament light chain. Numbers in parenthesis indicate numbers of serum samples.
Standard Protocol Approvals, Registrations, and Patient Consents
The study was approved by each center's institutional review board. All patients provided written informed consents.
Clinical Data
Relevant clinical data, including participants' age, sex, body mass index (BMI), and clinical history (e.g., attack history, treatment status, and Expanded Disability Status Scale [EDSS] at sampling) were obtained from medical records. Treatment status was classified as “treated” in patients who were under the effect of any acute (e.g., methylprednisolone, plasmapheresis, or intravenous immunoglobulin) or preventive therapy (e.g., azathioprine, mycophenolate mofetil, anti-CD20 monoclonal antibodies, or other immunosuppressive therapies) at the time of sampling. Conversely, patients sampled without the therapeutic influence of the aforementioned therapies were classified as “untreated.” Attacks were defined as the occurrence of new symptoms or worsening of previous symptoms not attributed to other causes, occurring more than 30 days following the onset of a previous attack.1 Attacks were confirmed by neurologic examination and MRI scans revealing new or contrast-enhancing lesions. Clinical phenotypes were defined as the clinical syndrome presented during the attack preceding sampling. Disease course was defined as monophasic in patients with a history of a single attack during a minimum follow-up duration of 3 years and as relapsing in patients with a history of more than 1 attack, regardless of duration of follow-up. Patients with single attacks and follow-up duration <3 years (n = 3) were not included in analyses involving disease course.
Assays
MOG-IgG seropositivity was assessed for confirmation of patient eligibility with the previously described live cell-based assays of each center.12,13 In addition, MOG-IgG measurements were conducted for all samples using the assay from the University Hospital Basel, with results expressed as a geometric mean fluorescence intensity ratio (MFI).12 sNfL and sGFAP concentrations (pg/mL) were measured in duplicates with an ultrasensitive single-molecule array (Simoa, Quanterix) duplex assay according to the manufacturer's instructions. Samples that were not collected according to standard operating procedures14 (n = 22 for sGFAP) or that did not produce signals above the analytical sensitivity of the assay (n = 4 for sGFAP) were excluded from the study. Samples with intra-assay coefficients of variation (CV) higher than 20% were remeasured. Two batches of samples along with 3 quality control (QC) serum samples (one spiked with human CSF) were included in each batch. The interassay CV from the QCs between the first and the second batch of samples for sNfL was 5.6% (9.2 pg/mL, sample 1), 4.1% (15.1 pg/mL, sample 2), and 5.4% (121.6 pg/mL, sample 3); for sGFAP were 8.3% (68.8 pg/mL, sample 1), 10.3% (129.0 pg/mL, sample 2), and 4.2% (347.4 pg/mL, sample 3).
Statistical Analysis
Participant characteristics are described using medians and interquartile ranges (IQR). The significance cutoff of the statistical analyses was set at p < 0.05. Percentiles and Z scores for sNfL adjusted for age and BMI were calculated based on a healthy control normative database.9 Likewise, sex-, age-, and BMI-adjusted sGFAP percentiles and Z scores were calculated using a healthy control normative database.10,11 All analyses including sNfL and sGFAP as variables used Z scores for the calculations.
The dynamics of sNfL and sGFAP within the first 24 weeks after the onset of clinical attacks were modeled using linear mixed-effects multivariable spline regression models, with the biomarkers as dependent variables, time from attack, sex, and age as independent variables, and patients as random intercepts. To address the potential bias arising from individuals with a high sampling frequency over a short timeframe, we assigned weights to repeated measures. Specifically, we calculated the weights as the inverse number of measurements within intervals smaller than 30 days.15 The number of degrees of freedom of the splines (3 in the final model) was selected based on the Akaike Information Criterium. Samples collected within 90 days after the onset of attacks were classified as active, whereas samples collected more than 90 days after the onset of an attack were classified as remission.
Exploratory analyses of sNfL and sGFAP (dependent variables) across different demographic (age and sex), clinical and paraclinical variables (disease activity status, EDSS, MOG-IgG, clinical phenotypes, disease courses, type of attack, and treatment status) were performed using linear mixed-effects models with patients as random intercepts and inverse probability weighting as argument to adjust for repeated measures in condensed timeframes.15 For comparisons between sNfL and sGFAP Z scores from active samples from different clinical phenotypes, we performed pairwise ANOVA with patient as a blocking factor and adjusted p values according to Bonferroni correction.
Forward stepwise nested generalized mixed-effects logistic modeling was performed to investigate the incremental discriminatory value of additional independent variables (EDSS, treatment status, age, sex, MOG-IgG, sGFAP, and sNfL) beyond the baseline discriminator (patient) in explaining the variability in the outcome variable (disease activity status). Inverse probability weighting was used to adjust for repeated measures in condensed timeframes in these models.15 Likelihood ratio tests compared different models, with multiple testing-adjusted p values (Holm method) indicating the significance of incremental performance improvements resulting from specific discriminatory variable inclusion. Receiver operating characteristic (ROC) curves were used to visualize the effect of incorporating different variables in generalized logistic models. To avoid repeated measures, logistic models for ROC curves were calculated using the average value of the samples per patient, grouped by disease activity status (active vs remission). For the analysis, only 1 averaged sample per patient was included. The corresponding areas under the curve were calculated to summarize the performance of the models.
The estimates from the linear models represent additive changes in the dependent variables (biomarkers) for each one-unit change in the independent variables. By contrast, the odds ratios (OR) from the logistic models represent the multiplicative effect of the independent variables (EDSS, treatment status, age, sex, MOG-IgG, sGFAP, and sNfL) on the likelihood of attacks.
The logistic models were derived in the discovery cohort and subsequently validated in the validation cohort. All analyses were performed using Prism 9 version 9.4.1 or R version 4.3.0. EQUATOR (STROBE) reporting guidelines were followed.
Data Availability
Anonymized data will be made available by the corresponding authors on reasonable request.
Results
General Clinical and Demographic Features of Participants
Among the 62 participants included in the study (Figure 1), 29 (46.8%) were female, with a median age at baseline of 40.0 years (IQR 29.5–49.8), a median EDSS at baseline of 2.0 (IQR 1.0–4.0), and a median follow-up duration of 20.0 months (IQR 3.0–62.8), during which a median of 2.0 sera per patient (IQR 1.0–6.0) were included. Individual longitudinal plots and detailed clinical and demographic features are presented in eFigure 1 and Table 1, respectively.
Table 1.
Clinical and Demographic Features of Study Cohorts
| All patients (n = 62) | Discovery cohort (n = 27) | Validation cohort (n = 35) | |
| Sample per participant, median (IQR) | 2.0 (1.0 to 6.0) | 3.0 (2.0 to 6.5) | 1.0 (1.0 to 6.0) |
| Age at baseline, y, median (IQR) | 40.0 (29.5 to 49.8) | 40.0 (28.0 to 48.5) | 40.0 (33.0 to 50.0) |
| Female: male, (% female) | 29:33 (46.8) | 7:20 (25.9) | 22:13 (62.8) |
| Duration of disease at baseline, mo, median (IQR) | 23.5 (4.3 to 85.5) | 9.0 (1.0 to 38.0) | 41.0 (9.5 to 93.0) |
| Duration of follow-up, mo, median (IQR) | 20.0 (3.0 to 62.8) | 18.0 (5.0 to 65.0) | 22.0 (2.5 to 58.5) |
| EDSS at baseline, median (IQR) | 2.0 (1.0 to 4.0) | 2.3 (0.8 to 4.0) | 2.0 (1.5 to 3.8) |
| EDSS at last follow-up, median (IQR) | 1.5 (0 to 3.1) | 1.0 (1.0 to 2.5) | 2.0 (0.0 to 3.8) |
| sNfL (Z scorea) median, (IQR) | |||
| Attacks | 2.9 (1.4 to 3.5) | 2.5 (1.1 to 3.1) | 3.2 (2.3 to 3.6) |
| Remission | 0.9 (−0.1 to 1.6) | 0.6 (−0.1 to 1.3) | 1.3 (0 to 1.8) |
| sGFAP (Z scorea) median, (IQR) | |||
| Attacks | 0.4 (−0.5 to 1.5) | 0.2 (−0.5 to 1.4) | 0.6 (−0.4 to 1.6) |
| Remission | −0.2 (−0.8 to 0.5) | −0.5 (-1 to -0.1) | 0.2 (−0.1 to 0.7) |
| MOG-IgG (MFI)b median, (IQR) | |||
| Attacks | 12.3 (4.0 to 43.2) | 14.4 (7.8 to 66.8) | 10.6 (2.0 to 31.0) |
| Remission | 6.2 (2.9 to 19.9) | 11.0 (4.9 to 21.2) | 2.9 (1.8 to 10.6) |
| Number of attacks, median (IQR) | |||
| Baseline | 2.0 (1.0 to 4.0) | 1.0 (1.0 to 2.0) | 3.0 (2.0 to 4.0) |
| Last follow-up | 2.0 (1.0 to 4.0) | 1.0 (1.0 to 2.0) | 3.0 (2.0 to 4.0) |
| Treatment at baseline, patient number (%) | |||
| Naive | 29.0 (46.8) | 22.0 (81.5) | 7.0 (20.0) |
| Treated | 33.0 (53.2) | 5 0 (18.5) | 28.0 (80.0) |
| Treatment at last follow-up, patient number (%) | |||
| Naive | 39.0 (62.9) | 19.0 (70.4) | 4.0 (11.4) |
| Treated | 23.0 (37.1) | 8.0 (29.6) | 31.0 (88.6) |
Abbreviations: EDSS = Expanded Disability Status Scale; IgG = immunoglobulin G; IQR = interquartile range; MFI = geometric mean fluorescence intensity ratio; MOG = myelin oligodendrocyte protein; sGFAP = serum glial fibrillary acidic protein; sNfL = serum neurofilament light chain.
Z scores of 0 reflect the mean biomarker concentration in the healthy control population.
2.4 ≤ MFI <3.0, low-positive; MFI 3.0 ≥, clear-positive.
Dynamics of sNfL and sGFAP After Attacks
sNfL and sGFAP Z scores were nonlinearly associated with the time from attack onset (p < 0.001 and = 0.002, respectively): sNfL Z scores of 1.6 (95% confidence interval [95% CI] 1.0–2.3, 94.5th percentile) were observed at attack onset, subsequently increasing to a peak Z score of 2.5 (95% confidence interval [95% CI] 1.8–3.0; 99.4th) at 6.0 weeks (95% CI 5.4–6.7), decreasing to a steady Z score of 1.3 (95% CI 0.6–2.0; 90.3th) at 4.7 months (95% CI 4.5–4.8) (Figure 2A). By contrast, sGFAP Z scores reached a peak of 0.9 (95% CI 0.4–1.5; 81.6th) at 5.3 weeks (95% CI 4.8–5.9), with subsequent decline to a steady level of −0.5 (95% CI −1.2 to −0.1; 30.9th) at 4.2 months (95% CI 4.1–4.4) (Figure 2B).
Figure 2. sNfL and sGFAP Dynamics.
Mixed-effects multivariable linear spline regression models of (A) serum neurofilament light chain (sNfL) Z scores and (B) serum glial fibrillary acidic protein (sGFAP) Z scores during the first 24 weeks following onset of clinical attacks. Marginal effects of time (black lines), 95% confidence intervals (bands), individual biomarker values (black dots), and mean biomarker concentrations in the healthy control reference population (dark blue lines). aZ scores of 0 reflect the mean biomarker concentration in the healthy control population.
Notably, sNfL and sGFAP median Z scores were higher for samples collected within 90 days after the onset of attacks (sNfL Z score 2.9 [IQR 1.4–3.5], 99.8th; sGFAP Z score 0.4 [IQR -0.5 to 1.5], 65.6th) compared with samples collected during remission (sNfL Z score 0.9 [IQR −0.1 to 1.6], 81.6th; p < 0.001; sGFAP Z score −0.2 [IQR −0.8 to 0.5], 42.1th; p < 0.001) (Table 1). Despite variability across different clinical phenotypes (median sNfL Z score optic neuritis 2.1 [IQR 0.6–3.2; 98.2th]; myelitis with/without optic neuritis, cerebral, or cerebellar 3.0 [IQR 2.6–3.4; 99.9th]; cerebral and cerebellar 3.5 [IQR 3.2–3.9; 99.9th]; other 1.0 [IQR 1.0–1.0; 84.1th]), sNfL Z scores were consistently higher during attacks vs remission in all clinical phenotypes (Figure 3, A and C, eTable 1). This observation was supported by a model indicating an additive elevation of 1.1 units in sNfL Z scores per attack (estimate 1.1 [95% CI 0.9–1.4]; p < 0.001), even after adjusting for clinical phenotypes, age, and sex (eTable 2). By contrast, sGFAP Z scores exhibited minimal variation across different clinical phenotypes, consistently presenting higher levels during attacks than in remission, independently of phenotype (estimate 0.6 [95% CI 0.3–0.9]; p < 0.001) (Figure 3, B and D, eTable 3).
Figure 3. sNfL and sGFAP According to Clinical Phenotype and Disease Activity Status.
Z scores of (A) serum neurofilament light chain (sNfL) and (B) serum glial fibrillary acidic protein (sGFAP) according to clinical phenotype (orange, cerebral and/or cerebellar; navy-blue, myelitis with/without optic neuritis, cerebral, or cerebellar; mauve, optic neuritis; light-blue, other) and status of disease activity (red dots, attacks; black dots, remission). Adjusted p values from pairwise ANOVA, ns = not significant, *p < 0.05, **p < 0.01, ***p < 0.001. Mixed-effects multivariable linear spline regression models of (C) sNfL and (D) sGFAP Z scores following the onset of clinical attacks. Marginal effects of time (black lines; 95% CIs [bands]), individual biomarker values according to clinical phenotype (orange, cerebral and/or cerebellar; navy-blue, myelitis with/ without optic neuritis, cerebral, or cerebellar; mauve, optic neuritis; light-blue, other), and mean biomarker concentrations in the normative healthy control reference population (dark blue lines). a Z scores of 0 reflect the mean biomarker concentration in the healthy control population.
Associations of sNfL and sGFAP With Clinical Features
No associations were found between sNfL and sGFAP Z scores with MOG-IgG MFI (p = 0.19 and = 0.42, respectively). However, sNfL and sGFAP were strongly associated with EDSS (estimate 0.19, p = 0.003 and estimate 0.18, p = 0.001, respectively) and with each other (estimate 0.70, p < 0.001). The significant association with sNfL and EDSS did not persist when testing active and remission samples separately. Furthermore, no differences were present between the acute sNfL Z scores (p = 0.74) and sGFAP Z scores (p = 0.83) of patients with monophasic compared with relapsing disease courses, nor between sNfL Z scores (p = 0.17) and sGFAP Z scores (p = 0.78) of initial attacks compared with relapses (eTable 1). Similarly, no differences were detected between sNfL Z scores (p = 0.63) and sGFAP Z scores (p = 0.26) of untreated compared with treated patients, regardless of disease activity status, sex, or age (covariate in the model).
Associations Between Biomarkers, Clinical Features, and Status of Disease Activity
Increasing sNfL levels were consistently associated with attacks in both the discovery and validation cohorts (Figure 4; eTable 4). Specifically, patients exhibited a 3.5-fold increase in the odds of attacks per unit rise in sNfL Z scores (OR 3.5 [95% CI 2.3–5.1]; p < 0.001) (eTable 4). Conversely, EDSS, age, and sGFAP showed inconsistent associations with disease activity status across both the discovery and validation cohorts (Figure 4; eTable 4). The models including treatment status, sex, and MOG-IgG as separate independent variables did not discriminate disease activity status in any of the cohorts (Figure 4; eTable 4).
Figure 4. Associations Between Biomarkers, Clinical Features, and Disease Activity Status.
Estimates from univariable mixed-effects logistic regression models for discrimination of status of disease activity (dependent variable) across the discovery (black) and validation (gray) cohorts. Odds ratios (dots), 95% confidence intervals (95% CI; lines), and p values are indicated. The reference levels are untreated for treatment status and female for sex, with age scaled and MOG-IgG displayed in the log2 scale. EDSS = Expanded Disability Status Scale; IgG = immunoglobulin G; MOG = myelin oligodendrocyte protein; sGFAP = serum glial fibrillary acidic protein; sNfL = serum neurofilament light chain.
For models incorporating treatment status, sex, age, MOG-IgG or sGFAP, the likelihood ratio tests did not indicate improvements in model fit compared with baseline, suggesting that these variables may not substantially enhance the performance of discriminatory models (eTable 5). However, models including sNfL showed consistent enhancements in model fit across all cohorts (p < 0.001), indicating its relevance in discriminating the status of disease activity at sampling (eTable 5). Further comparisons of models, including the combination of EDSS, and sNfL as independent variables in one model, demonstrated enhanced model fit for discriminating status of disease activity, in contrast to models with single independent variables (Figure 5; eTable 6). Overall, our data demonstrate that including sNfL Z scores improves the identification of disease activity compared with relying solely on neurologic examination.
Figure 5. Discrimination of Disease Activity Status.
ROC curves of logistic regression models for discrimination of status of disease activity (dependent variable) across all patients (n = 61, 1 patient excluded because of missing EDSS value). For each patient, samples were averaged according to disease activity status. Green, EDSS and sNfL as independent variables; ochre, sNfL as an independent variable; blue, EDSS as an independent variable. AUC = area under the curve; EDSS = Expanded Disability Status Scale; ROC = receiver operating characteristic; sNfL = serum neurofilament light chain.
Discussion
The precise and reliable detection of disease activity in MOGAD is crucial because it allows for prompt initiation of acute treatment and escalation of immunosuppressive therapy, which are associated with better clinical outcomes.16 In this study, we investigated the dynamics of sNfL and sGFAP following the onset of attacks and characterized their performance in discriminating disease activity status in patients with MOGAD. Notably, our findings demonstrate a consistently enhanced performance of models including sNfL, compared with those without, in distinguishing attacks from remission. These findings suggest that sNfL may serve as a biomarker for monitoring disease activity in addition to neurologic examination in patients with MOGAD.
The dynamics of sNfL, with a peak approximately 2 months after the onset of attacks, likely reflects the temporal evolution of acute inflammation and neuronal injury that occur in MOGAD pathology.17 Conversely, the detection of sNfL Z scores exceeding the normal range of healthy controls and persisting for more than 4 months following attacks in certain patients, suggests prolonged neuroaxonal damage in these individuals. This challenges prevailing notions on the frequency of subclinical disease activity in MOGAD. Further investigation in larger cohorts with paired serum, EDSS, and MRI data and longer follow-up is required to address whether subclinical disease activity exists in a subgroup of patients with MOGAD. The mean sNfL Z score of 1.6 observed in patients with MOGAD at attack onset suggests that axonal damage may precede clinical symptoms in some cases. This finding warrants further investigation through prospective studies.
The variation of sNfL levels according to clinical phenotypes potentially reflects the size of the neuronal structures affected. However, sNfL levels consistently rose during attacks regardless of clinical phenotypes, emphasizing its utility as a biomarker for monitoring disease activity. Furthermore, acute sNfL levels did not differ between patients with monophasic compared with relapsing disease course, nor between initial attacks compared with subsequent relapses, in contrast to previous reports.18,19 These differences likely result from previous studies' smaller sample sizes, shorter follow-up duration, timing of sampling, and the lack of sNfL adjustment for demographic features, as addressed in our study design. Overall, the lack of correlation between sNfL and sGFAP with specific clinical variables in this study may suggest limitations in the use of sNfL and sGFAP Z scores as a predictive biomarker for clinical outcomes such as disease course.
In contrast to sNfL, the dynamics we observed for sGFAP were not influenced by clinical phenotypes. Moreover, although sGFAP levels were elevated during periods of attacks compared with remission, the fluctuations in biomarker levels were mild, resulting in an inconsistent ability to differentiate the status of disease activity. These findings likely reflect the reactive gliosis that is associated with acute inflammation and absent during disease remission.17 This is consistent with observations from other demyelinating diseases such as MS, which similarly exhibit minimal astrogliosis during attacks.11,17
Our study was limited by its retrospective nature, the absence of samples during the weeks preceding attacks, the lack of MRI data paired to each of the sampling timepoints, and the irregular sampling intervals. These limitations precluded us from investigating the time of initiation of biomarker elevation, frequency of subclinical disease activity, and additive value of sNfL compared with combined EDSS and MRI assessment. Consequently, determining the optimal sampling timepoint for distinguishing disease status was not feasible. Owing to the limited number of relapses in our cohort, we could not conduct predictive analyses on future disease activity. Yet, this is the largest MOGAD population collectively investigated for sNfL and sGFAP attack dynamics to date. In addition, analyses were performed with demographic features-adjusted Z scores and percentiles expressing biomarker levels' deviation from a large healthy control population through a rigorous approach with independent discovery and validation cohorts.
Our study reports the dynamics of sNfL and sGFAP after attacks and indicates that elevations in sNfL Z scores can support the identification of relapses in patients with MOGAD. Further studies should prospectively investigate larger cohorts to corroborate these findings, validate optimal timing and cutoffs, and investigate the performance of these biomarkers in predicting future disease activity in patients with MOGAD.
Glossary
- BMI
body mass index
- CV
coefficients of variation
- EDSS
Expanded Disability Status Scale
- IgG
immunoglobulin G
- MOG
myelin oligodendrocyte glycoprotein
- OR
odds ratios
- QC
quality control
- ROC
receiver operating characteristic
- sGFAP
serum glial fibrillary acidic protein
- sNfL
serum neurofilament light chain
Author Contributions
A.B.A.G.R. Gomes: drafting/revision of the manuscript for content, including medical writing for content. Major role in the acquisition of data; analysis and interpretation of data. S.-H. Kim: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. R. Pretzsch: drafting/revision of the manuscript for content, including medical writing for content; analysis and interpretation of data. L. Kulsvehagen: major role in the acquisition of data. S. Schaedelin: analysis or interpretation of data. J. Lerner: major role in the acquisition of data. N.S. Wetzel: major role in the acquisition of data. P. Benkert: analysis or interpretation of data. A. Maleska Maceski: major role in the acquisition of data. J.-W. Hyun: major role in the acquisition of data. A.-C. Lecourt: major role in the acquisition of data. P. Lipps: major role in the acquisition of data. V.A. Schoeps: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. A.D.M.B. Matos: major role in the acquisition of data. N.T. Mendes: major role in the acquisition of data. S.L. Apóstolos-Pereira: major role in the acquisition of data. M. Mehling: drafting/revision of the manuscript for content, including medical writing for content. T. Derfuss: drafting/revision of the manuscript for content, including medical writing for content. L. Kappos: drafting/revision of the manuscript for content, including medical writing for content. D. Callegaro: drafting/revision of the manuscript for content, including medical writing for content. J. Kuhle: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. H.J. Kim: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. A.-K. Pröbstel: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data.
Study Funding
The study was funded by an ECTRIMS Clinical Fellowship and a Swiss Government Excellence Scholarship (to A.B.A.G.R.G.), a grant from the National Research Foundation of Korea (to S.H.K.), a research grant (“Young talents in clinical research”) by the Swiss Academy of Medical Sciences (SAMS) and the Gottfried and Julia Bangerter-Rhyner Foundation (to R.P.), and a research grant (medMS) from the Hertie foundation (to R.P.), a doctoral fellowship from the Goldschmidt Jacobson Foundation (to P.L.), and grants from the Swiss National Science Foundation (SNF Eccellenza Professorship: PCEFP3_194609; SNF Starting Grant: TMSGI3_211318), the National MS Society (FG-1708-28871), the Fondation Pierre Mercier pour la Science, the Propatient Foundation, the Goldschmidt Jacobson Foundation, and the Gottfried and Julia Bangerter-Rhyner Foundation (all to A.-K.P.).
Disclosure
A.B.A.G.R. Gomes has received a research grant from Roche, paid to the University of São Paulo; S.-H. Kim has lectured, consulted, and received honoraria from Bayer Schering Pharma, Biogen, Genzyme, Merck Serono, and UCB; R. Pretzsch has received travel funds from Teva; A.-C. Lecourt was supported by the Horizon 2020 Eurostar program (grant E!113682); A. Maleska Maceski is an investigator in a phase III clinical trial for the treatment of MOGAD sponsored by Hoffmann-LaRoche; S.L. Apóstolos-Pereira has received research grants and honoraria as a speaker and member of advisory boards by AMGEN/Horizon, Alexion, Biogen, Genzyme, Merck, Novartis, Roche; M. Mehling has received research grants from the Swiss National Science Foundation, Roche, and Merck. His institution (University Hospital Basel) has received fees from his participation on the advisory boards of Merck, Roche, Novartis, and Biogen outside the submitted work; T. Derfuss received speaker fees, research support, travel support, and/or served on Advisory Boards or Steering Committees of Actelion, Alexion, Biogen, Celgene, GeNeuro, MedDay, Merck, Mitsubishi Pharma, Novartis, Roche, and Sanofi-Genzyme; he received research support from Alexion, Biogen, Novartis, Roche, Swiss National Research Foundation, University of Basel, and Swiss MS Society; L. Kappos reported having a patent for Neurostatus UHB-AG with royalties paid Payments made to institution (University Hospital Basel); being CEO of RC2NB (employment by University Hospital Basel), part of the MAGNIMS Steering Committee and a board member of the European Charcot Foundation. J. Kuhle has received speaker fees, research support, travel support, and/or served on advisory boards by Swiss MS Society, Swiss National Research Foundation (320030_189140/1), University of Basel, Progressive MS Alliance, Bayer, Biogen, Bristol Myers Squibb, Celgene, Merck, Novartis, Octave Bioscience, Roche, and Sanofi; H.J. Kim received a grant from the National Research Foundation of Korea and research support from Aprilbio and Eisai; received consultancy/speaker fees from Alexion, Aprilbio, Altos Biologics, Biogen, Celltrion, Daewoong, Eisai, GC Pharma, Handok, Horizon Therapeutics, Kaigene, Kolon Life Science, MDimune, Mitsubishi Tanabe Pharma, Merck Serono, Novartis, Roche, Sanofi Genzyme, Teva-Handok, and UCB; is a co-editor for the Multiple Sclerosis Journal and an associated editor for the Journal of Clinical Neurology; A-K. Pröbstel (institution) received financial compensation for participation in advisory boards, and consultations from Biogen, Novartis, Roche, and UCB, all used for research support; All other authors report no potential conflicts of interest related to this study. Go to Neurology.org/NN for full disclosures.
References
- 1.Banwell B, Bennett JL, Marignier R, et al. Diagnosis of myelin oligodendrocyte glycoprotein antibody-associated disease: International MOGAD Panel proposed criteria. Lancet Neurol. 2023;22(3):268-282. doi: 10.1016/S1474-4422(22)00431-8 [DOI] [PubMed] [Google Scholar]
- 2.Jarius S, Ruprecht K, Kleiter I, et al. MOG-IgG in NMO and related disorders: a multicenter study of 50 patients. Part 2: epidemiology, clinical presentation, radiological and laboratory features, treatment responses, and long-term outcome. J Neuroinflammation. 2016;13(1):280. doi: 10.1186/s12974-016-0718-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Syc-Mazurek SB, Chen JJ, Morris P, et al. Frequency of new or enlarging lesions on MRI outside of clinical attacks in patients with MOG-antibody–associated disease. Neurology. 2022;99(18):795-799. doi: 10.1212/WNL.0000000000201263 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Havla J, Kümpfel T, Schinner R, et al. Myelin-oligodendrocyte-glycoprotein (MOG) autoantibodies as potential markers of severe optic neuritis and subclinical retinal axonal degeneration. J Neurol. 2017;264(1):139-151. doi: 10.1007/s00415-016-8333-7 [DOI] [PubMed] [Google Scholar]
- 5.Kim H, Lee EJ, Kim S, et al. Serum biomarkers in myelin oligodendrocyte glycoprotein antibody–associated disease. Neurol Neuroimmunol Neuroinflamm. 2020;7(3):e708. doi: 10.1212/NXI.0000000000000708 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.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 Neuroinflammation. 2021;18(1):105. doi: 10.1186/s12974-021-02138-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.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]
- 8.Hyun JW, Kim SY, Kim Y, et al. Absence of attack-independent neuroaxonal injury in MOG antibody-associated disease: longitudinal assessment of serum neurofilament light chain. Mult Scler. 2022;28(6):993-999. doi: 10.1177/13524585211063756 [DOI] [PubMed] [Google Scholar]
- 9.Benkert P, Meier S, Schaedelin S, et al. 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]
- 10.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]
- 11.Benkert P, Maleska Maceski A, Schaedelin S, et al. Serum glial fibrillary acidic protein and neurofilament light chain levels reflect different mechanisms of disease progression under b-cell depleting treatment in multiple sclerosis. Ann Neurol. Published online October 16, 2024. doi: 10.1002/ana.27096 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Ayroza Galvão Ribeiro Gomes AB, Kulsvehagen L, Lipps P, et al. Immunoglobulin A antibodies against myelin oligodendrocyte glycoprotein in a subgroup of patients with central nervous system demyelination. JAMA Neurol. 2023;80(9):989-995. doi: 10.1001/jamaneurol.2023.2523 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Kim Y, Hyun JW, Woodhall MR, et al. Refining cell-based assay to detect MOG-IgG in patients with central nervous system inflammatory diseases. Mult Scler Relat Disord. 2020;40:101939. doi: 10.1016/j.msard.2020.101939 [DOI] [PubMed] [Google Scholar]
- 14.Teunissen CE, Petzold A, Bennett JL, et al. A consensus protocol for the standardization of cerebrospinal fluid collection and biobanking. Neurology. 2009;73(22):1914-1922. doi: 10.1212/WNL.0b013e3181c47cc2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Hernán MA, Hernández-Díaz S, Robins JM. A structural approach to selection bias. Epidemiology. 2004;15(5):615-625. doi: 10.1097/01.ede.0000135174.63482.43 [DOI] [PubMed] [Google Scholar]
- 16.Stiebel-Kalish H, Hellmann MA, Mimouni M, et al. Does time equal vision in the acute treatment of a cohort of AQP4 and MOG optic neuritis? Neurol Neuroimmunol Neuroinflamm. 2019;6(4):e572. doi: 10.1212/NXI.0000000000000572 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Höftberger R, Guo Y, Flanagan EP, et al. The pathology of central nervous system inflammatory demyelinating disease accompanying myelin oligodendrocyte glycoprotein autoantibody. Acta Neuropathol. 2020;139(5):875-892. doi: 10.1007/s00401-020-02132-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Mariotto S, Gastaldi M, Grazian L, et al. NfL levels predominantly increase at disease onset in MOG-Abs-associated disorders. Mult Scler Relat Disord. 2021;50:102833. doi: 10.1016/j.msard.2021.102833 [DOI] [PubMed] [Google Scholar]
- 19.Wendel EM, Bertolini A, Kousoulos L, et al. Serum neurofilament light-chain levels in children with monophasic myelin oligodendrocyte glycoprotein-associated disease, multiple sclerosis, and other acquired demyelinating syndrome. Mult Scler. 2022;28(10):1553-1561. doi: 10.1177/13524585221081090 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Anonymized data will be made available by the corresponding authors on reasonable request.