Highlights
- Cryptogenic new-onset refractory status epilepticus (cNORSE) was associated with strikingly higher neurofilament light chain (NfL) levels in serum and cerebrospinal fluid than etiology-defined status epilepticus (eSE), chronic epilepsy, and healthy controls.
- S100B did not meaningfully distinguish cNORSE from comparator groups, suggesting that NfL is the more informative marker of acute neuroaxonal injury in this setting.
- NfL rose quickly over the first 3 weeks after onset, supporting the concept that the window to prevent ongoing brain injury may be narrow.
- Higher serum NfL was independently associated with worse functional status at discharge, reinforcing its potential prognostic value.
Background: Why this study matters
New-onset refractory status epilepticus is one of the most urgent syndromes in clinical neurology and neurocritical care. It describes prolonged, treatment-resistant seizures in a patient without a prior epilepsy history or a clear acute structural or metabolic explanation at presentation. When no cause is found despite extensive workup, the syndrome is termed cryptogenic NORSE. In contrast, etiology-defined status epilepticus refers to prolonged seizures in which an underlying cause is identified, such as stroke, infection, autoimmune disease, or a known epilepsy syndrome.
The central clinical question is not simply how to stop the seizures, but how much secondary brain injury is accumulating while treatment is being escalated. In NORSE, that question is especially important because delays in effective therapy may allow ongoing excitotoxic injury, inflammation, and neuronal loss. Biomarkers that reflect real-time or near-real-time tissue injury could therefore help clinicians estimate severity, track response, and define a practical therapeutic window.
This study focused on two biomarkers. Neurofilament light chain is a structural protein abundant in axons; when axons are injured, NfL enters the cerebrospinal fluid and then the bloodstream. S100B is a calcium-binding protein largely associated with astrocytic and glial injury. In theory, both could signal brain injury, but they do not necessarily perform equally well in status epilepticus.
Study design
This was an international cross-sectional study conducted between 2013 and 2025 across 36 hospitals in the US, 2 in Canada, and 1 each in Italy, France, and Belgium. The investigators enrolled 78 patients with cNORSE and 2 independent cohorts of patients with eSE, including 211 patients in one cohort and 73 in the other. Biological samples were obtained during ongoing seizure activity. Comparison groups without status epilepticus included patients with chronic epilepsy and healthy participants. None of the eligible participants were excluded.
The primary exposure variables were serum and CSF concentrations of NfL and S100B. The main outcomes were the degree of neuronal and glial damage, inferred from biomarker concentrations, and their association with short-term functional outcome at discharge. Functional outcome was reported using the Glasgow Outcome Scale extended, with poor outcome defined as scores of 1 to 4.
Key findings
The most important result was the magnitude of NfL elevation in cNORSE. Compared with the eSE cohorts, median CSF NfL was about 10-fold higher and serum NfL about 4-fold higher in cNORSE. Specifically, CSF NfL was 6408 pg/mL in cNORSE versus 694 pg/mL in eSE, and serum NfL was 231 pg/mL versus 55 pg/mL, respectively; both comparisons were highly significant with P values below .001.
When cNORSE was compared with participants without status epilepticus, serum NfL was nearly 20-fold higher. Median serum NfL was 231 pg/mL in cNORSE, compared with 11 pg/mL in the chronic epilepsy cohort and 7 pg/mL in healthy controls. This separation is clinically meaningful because it suggests that cNORSE is not merely a severe seizure disorder, but a syndrome associated with major acute neuroaxonal injury.
Serum and CSF NfL were strongly correlated, with Spearman rho of 0.75 and P below .001. That relationship supports the idea that serum measurement may provide a practical surrogate for CSF-based assessment in many settings, although it does not remove the need for clinical and electroencephalographic evaluation.
Temporal analysis was also notable. NfL increased sharply after onset: serum concentrations had a median of 101 pg/mL in week 1, 197 pg/mL in week 2, and 598 pg/mL in week 3, with a significant upward trend over time. This pattern is important because it suggests that injury may continue to evolve for days to weeks after seizure onset. In practical terms, a single early sample may underestimate the full burden of damage, and serial measurements may be more informative than a one-time test.
By contrast, S100B was disappointing as a discriminator. Concentrations did not differ between groups and did not show a consistent temporal pattern. In this cohort, S100B therefore appeared less useful than NfL for identifying acute injury related to cNORSE or for distinguishing cNORSE from eSE. The result does not mean S100B is biologically irrelevant, but it does suggest that it may be too nonspecific or too insensitive for this particular clinical problem.
NfL also showed discriminative performance. It separated cNORSE from eSE with an area under the receiver operating characteristic curve of 0.79, which indicates good but not perfect discrimination. The biomarker performed even better when cNORSE was compared with non-status controls, with an AUROC of 0.99. In other words, elevated NfL was highly characteristic of cNORSE relative to chronic epilepsy and healthy states, but there was still overlap when cNORSE was compared with other forms of status epilepticus.
Finally, higher serum NfL was independently associated with poor functional outcome at discharge. The reported odds ratio was 1.01, with a 95% CI of 1.00 to 1.03 and a P value of .03. Although the per-unit effect is small, the wide range of NfL values means that clinically relevant differences may accumulate across large biomarker shifts. The finding supports the idea that NfL is not just a marker of seizure presence, but a marker of clinically meaningful brain injury.
Selected numerical results
| Group | Serum NfL, median [IQR], pg/mL | CSF NfL, median [IQR], pg/mL | Main interpretation |
|---|---|---|---|
| cNORSE | 231 [99-855] | 6408 [1503-22 963] | Highest acute neuroaxonal injury burden |
| eSE | 55 [20-135] | 694 [219-2389] | Elevated, but much lower than cNORSE |
| Chronic epilepsy | 11 [7-19] | Not reported | Far below cNORSE |
| Healthy controls | 7 [5-14] | Not reported | Baseline range |
Expert commentary
From a translational perspective, the study strengthens a biologically plausible model of cNORSE as a neurodestructive emergency. NfL reflects axonal injury, which is consistent with the concept that prolonged seizures, inflammatory cascades, metabolic stress, and possible blood-brain barrier dysfunction can damage neurons in a cumulative fashion. The rapid increase over time also fits the clinical impression that cNORSE can evolve relentlessly despite early treatment attempts.
The contrast with S100B is equally informative. Glial markers can be affected by many processes, including systemic illness, barrier leakage, and nonspecific stress responses. In this study, S100B did not track with syndrome severity or timing in a useful way. That does not invalidate glial biology in NORSE, but it does suggest that not all injury biomarkers are equally suited to bedside decision-making.
For clinicians, the most immediate implication is conceptual rather than procedural: cNORSE should be approached as a time-sensitive brain-injury syndrome, not only as a seizure-control problem. NfL could eventually help identify patients who are accumulating damage despite apparent partial seizure suppression, or help stratify risk in biomarker-enriched clinical trials. It may also serve as a pharmacodynamic endpoint for therapies aimed at limiting neuronal injury.
However, this is not yet a bedside practice-changing biomarker. The study is cross-sectional, and that design cannot establish causality or determine whether specific treatments lowered NfL. Sampling times varied, which matters because NfL kinetics are delayed and dynamic. The cohorts were also heterogeneous, especially the eSE groups, which were older and etiologically diverse. Age, comorbid illness, seizure duration, intensive care exposures, and treatment strategies could all influence biomarker levels and outcome. Short-term discharge status may also underestimate later recovery or disability.
Generalizability is another consideration. The work drew on a large international network, which is a major strength, but it also likely reflects referral-center populations enriched for severe disease. That is appropriate for NORSE research, yet it means the findings should be validated prospectively in standardized cohorts before NfL is used to guide treatment thresholds or prognostic counseling.
Conclusion
This study provides strong evidence that cNORSE is associated with substantial acute neuroaxonal injury, captured most clearly by elevated NfL in serum and CSF. The biomarker rose rapidly over the first 3 weeks and was linked to worse short-term functional outcome, while S100B did not offer comparable signal. For neurocritical care teams, the message is clear: cNORSE likely carries a narrow therapeutic window, and early effective treatment remains essential. NfL is promising as a biomarker of injury and prognosis, but it still needs prospective validation before it can be integrated into routine management algorithms.
Funding and clinicaltrials.gov
Funding was not stated in the PubMed abstract provided here. ClinicalTrials.gov registration was not reported, which is consistent with the observational cross-sectional design of the study.
References
- Meletti S, Hanin A, Giovannini G, et al. Acute Brain Injury in New-Onset Refractory Status Epilepticus and Etiology-Defined Status Epilepticus. JAMA Neurol. Published online April 27, 2026. PMID: 42043830.
- Trinka E, Cock H, Hesdorffer D, et al. A definition and classification of status epilepticus: Report of the ILAE Task Force on Classification of Status Epilepticus. Epilepsia. 2015;56(10):1515-1523.
- Khalil M, Teunissen CE, Otto M, et al. Neurofilaments as biomarkers in neurological disorders. Nat Rev Neurol. 2018;14(10):577-589.