Abstract
Background and purpose
The purpose was to explore the prognostic utility of neurofilament light chain (NfL) in patients with immune‐mediated polyradiculoneuropathies (IMPs).
Methods
This retrospective monocentric study analysed serum and cerebrospinal fluid samples from patients diagnosed with IMP collected prior to treatment initiation. NfL concentrations were correlated with clinical outcomes, including F score and hospitalization duration.
Results
Amongst 115 IMP patients tested, baseline cerebrospinal fluid and serum NfL (sNfL) concentrations were higher in acute inflammatory axonal polyradiculoneuropathy (AIAP) than other IMP variants. In the AIAP cohort, a positive correlation was observed between baseline sNfL concentrations, F score and hospitalization duration. Multivariate linear regression analysis further supported the predictive relationship between elevated baseline sNfL concentrations and clinical outcomes. Using receiver operating characteristic analysis, a cut‐off value for sNfL of 351 pg/mL was found to predict an F score >3 in AIAP with a sensitivity of 40% and specificity of 81.8%. AIAP patients with sNfL concentrations above this threshold required longer hospitalization (extended by 15 days).
Discussion
Our findings highlight the potential of baseline sNfL as an effective marker for distinguishing between IMP variants and predicting the prognosis of AIAP. Further validation may facilitate translation of sNfL into clinical practice, potentially identifying high‐risk patients for tailored treatment approaches.
Keywords: acute axonal immune neuropathy, GBS, immune neuropathies, NfL
INTRODUCTION
Immune‐mediated polyradiculoneuropathies (IMPs), such as Guillain–Barré syndrome (GBS) and chronic inflammatory demyelinating polyneuropathy (CIDP), comprise a heterogeneous group of debilitating autoimmune diseases of the peripheral nervous system characterized by diverse clinical presentations and variable disease courses that can rapidly progress to fatal outcomes [1, 2].
Identifying prognostic markers at the disease onset remains a critical unmet need in clinical practice. Despite numerous studies, predicting disease progression is challenging [3, 4, 5]. The need for biomarkers capable of early risk stratification remains. Neurofilament light chain (NfL), a structural component of the axonal cytoskeleton, has emerged as a promising biomarker for neuroaxonal damage, yet its role in IMP has not been sufficiently explored. Axonal damage, a prominent feature in axonal GBS variants, could lead to elevated serum NfL (sNfL) concentrations, reflecting disease severity [6, 7, 8, 9]. Despite promising findings, prior studies were often limited by small cohort sizes and lacked subgroup classification. Therefore, the aim was to explore the utility of sNfL as a predictor of disease severity.
METHODS
Patient cohort
This retrospective monocentric study was conducted at the Department of Neurology of the Technical University of Munich, a tertiary centre for neurological diseases, and analysed the recorded clinical data of patients with IMP from whom cerebrospinal fluid (CSF) and serum samples were available for analysis at initial presentation.
All patients with acute and chronic IMP according to approved diagnostic criteria for GBS [2] and CIDP [1] were identified. For subgroup analysis, acute inflammatory demyelinating polyneuropathy (AIDP), acute inflammatory axonal polyneuropathy (AIAP), CIDP and Miller–Fisher syndrome (MFS) were differentiated according to the modified electrodiagnostic criteria proposed by Rajabally and colleagues [10]. Suspected axonal variants on initial examination were confirmed in a follow‐up nerve conduction study.
Neurofilament light chain analysis
Neurofilament light chain concentrations were measured using the Quanterix Simoa platform with the NF‐light Advantage kit on the HD‐X analyser according to the manufacturer's protocol. Two independent experiments were conducted to measure samples, with an inter‐assay coefficient of variation of 8.17%.
Ethics
This study was approved by the local ethics committee (approval number 2022‐204‐S). All procedures were performed in accordance with relevant guidelines and regulations. Written informed consent was obtained from every participant.
Statistical analysis
The Mann–Whitney U test and the Kruskal–Wallis test were used to compare nonparametric values; Spearman's correlation for nonparametric correlations and multivariate linear regression were used to explore relationships with disease outcome measures adjusted for age, sex, the duration of symptoms, CSF/serum albumin quotient (Q Alb), sNfL and plasmapheresis. The Shapiro–Wilk test confirmed no significant deviations in the residuals; receiver operating characteristic (ROC) curve analysis with Mantel–Cox log‐rank and Gehan–Breslow–Wilcoxon test for event‐time analysis were used to predict disease outcomes. All statistical analyses were performed with GraphPad Prism (v10.2.2).
RESULTS
Baseline characteristics
In this retrospective monocentric study, NfL concentrations in serum and CSF samples were measured from patients with IMP before treatment initiation. Serum from individuals with AIDP, AIAP, MFS and CIDP as well as CSF samples from individuals with AIDP and AIAP were analysed (Figure 1a). Demographic parameters, comorbidities and CSF profiles were consistent between groups and in line with published data (Table 1) [11, 12]. MFS patients, by definition, had a higher frequency of cranial nerve palsy and ataxia, clearly distinguishing them from other IMP variants. AIAP patients had lower rates of sensory involvement, yet displayed a more severe disease course, more frequently requiring plasma exchange as a rescue therapy. Substantially elevated sNfL concentrations were observed in individuals with AIAP compared to the other groups, with similar trends and unusually high outliers in CSF, reaching upper detection limits in two independent experiments (Table 1, Figure 1b). Of note, in comparison to our AIAP cohort, AIAP patients without available biospecimens exhibited a tendency towards a more severe disease course, with a longer median hospitalization duration and higher rates of both invasive mechanical ventilation and rescue therapy with plasmapheresis (data not shown).
FIGURE 1.
Serum neurofilament light chain (sNfL) concentrations correlate with disease severity in acute inflammatory axonal polyneuropathy (AIAP). (a) Serum and (b) CSF samples from patients with immune‐mediated neuropathies obtained during initial presentation were preserved in our local biobank and tested for neurofilament light chain (NfL) with the Quanterix Simoa kit according to the manufacturer's protocol. Data are shown as box and whisker plots, following the Tukey method tested with the Kruskal–Wallis test and corrected for multiple testing with (a) Dunn's test and (b) the Mann–Whitney U test. *p < 0.05, ****p < 0.0001. sNfL concentration was found to correlate with disease severity (F score) (c) and hospitalization duration (d) in patients with AIAP. The data were computed using the Spearman correlation coefficient. sNfL concentrations were associated with F score and hospitalization duration in a multivariate linear regression model adjusted for age, sex, the duration of symptoms, CSF/serum albumin quotient (Q Alb), sNfL and plasmapheresis (e). The prognostic value of sNfL for an F score greater than 3 was analysed using a ROC curve to determine cut‐off values and their corresponding sensitivity and specificity (f). sNfL concentrations above 351 pg/mL were associated with longer hospitalization durations, as tested with the Mantel–Cox log‐rank and Gehan–Breslow–Wilcoxon tests (g). AIAP, acute immune‐mediated axonal polyneuropathy; AIDP, acute immune‐mediated demyelinating polyneuropathy; AUC, area under the curve; CIDP, chronic immune‐mediated demyelinating polyneuropathy; MFS, Miller–Fisher syndrome; Q Alb, albumin quotient CSF and serum; SE, standard error; ROC, receiver operating characteristic.
TABLE 1.
Baseline parameters of immune‐mediated polyradiculoneuropathies.
| AIDP (n = 66) | AIAP (n = 16) | MFS (n = 15) | CIDP (n = 18) | |
|---|---|---|---|---|
| Age (years), median (IQR) | 61 (42.5–67) | 58 (51–67) | 53 (40–58) | 53 (45–61) |
| Sex (females), n (%) | 33/66 (50) | 6/16 (37.5) | 5/15 (33.33) | 3/18 (16.67) |
| Duration of symptoms (days). median (IQR) a | 7 (4–14) | 7 (6.25–17.25) | 2 (1–6) | 14 (8.5–52.5) |
| Time to spinal tap (days), median (IQR) | 1 (1–2) | 1 (1–2.5) | 1 (1–2) | 2 (1–3) |
| Hospitalization (days), median (IQR) b | 13 (7.25–19.75) | 22 (18–33.25) | 12 (6.5–19.5) | 10.5 (8–12.75) |
| Arterial hypertension, n (%) | 27/66 (40.91) | 6/16 (37.5) | 5/15 (33.33) | 6/18 (33.33) |
| Diabetes, n (%) | 5/66 (7.58) | 1/16 (6.25) | 0/15 | 4/18 (22.22) |
| Smoking, n (%) | 15/43 (34.88) | 4/9 (44.44) | 3/10 (30) | 3/13 (23.08) |
| History of cancer, n (%) | 8/66 (12.12) | 2/16 (12.5) | 0/15 | 3/18 (16.67) |
| MGUS/MM, n (%) | 0/66 | 0/16 | 0/15 | 2/18 (11.11) |
| Autoimmune disease, n (%) | 5/66 (7.58) | 1/16 (6.25) | 0/15 | 1/18 (5.56) |
| Infectious disease, n (%) | 4/62 (6.45) | 0/16 | 1/15 (6.67) | 0/12 |
| Prior vaccination, n (%) c | 6/53 (11.32) | 0/12 | 0/13 | 4/8 (50) |
| Prior infectious disease, n (%) c | 41/64 (64.06) | 11/16 (68.75) | 12/15 (80) | 6/13 (46.15) |
| Symptoms | ||||
| Sensory, n (%) | 61/66 (92.42) | 12/16 (75) | 12/15 (80) | 17/18 (94.44) |
| Motor, n (%) | 56/66 (84.85) | 14/16 (87.5) | 7/15 (46.67) | 17/18 (94.44) |
| Ataxia, n (%) | 30/63 (47.62) | 8/16 (50) | 9/15 (60) | 8/18 (44.44) |
| Areflexia/hyporeflexia, n (%) | 54/66 (81.82) | 14/15 (93.33) | 12/15 (80) | 16/18 (88.89) |
| Cranial nerve palsy, n (%) | 22/66 (33.33) | 5/16 (31.25) | 15/15 (100) | 7/18 (38.89) |
| Autonomic dysfunction, n (%) | 13/66 (19.70) | 4/16 (25) | 4/15 (26.67) | 4/18 (22.22) |
| Invasive ventilation, n (%) | 6/66 (9.09) | 2/16 (12.5) | 4/15 (26.67) | 0/18 |
| Peak F score, median (IQR) | 3 (2–4) | 4 (3–4) | 3 (2–4.5) | 3 (2–4) |
| CSF cells/μL, median (IQR) | 2 (1–4) | 2.5 (1–6.5) | 1 (1–2) | 2 (0.5–3.5) |
| CSF Q Alb, median (IQR) | 11.6 (7.48–17.95) | 9.15 (6.35–18.9) | 6.25 (4.7–8.8) | 11 (7.03–18.88) |
| Serum NfL pg/mL, median (IQR) | 34.4 (15.9–125.3) | 1755 (345–3768) | 33.2 (6.1–61.8) | 59.9 (33.1–201.8) |
| CSF NfL pg/mL, median (IQR) | 1530 (521.5–4395) | 4710 (0–59,275) | na | na |
| Treatment | ||||
| IVIg, n (%) | 35/66 (53.03) | 9/16 (56.25) | 7/15 (46.67) | 18/18 (100) |
| Plasma exchange, n (%) | 37/66 (56.06) | 11/16 (68.75) | 6/15 (40) | 11/18 (61.11) |
Note: All parameters were obtained during first presentation.
Abbreviations: AIDP, acute inflammatory demyelinating polyneuropathy; AIAP, acute inflammatory axonal polyneuropathy; CIDP, chronic inflammatory demyelinating polyneuropathy; CSF, cerebrospinal fluid; IQR, interquartile range; IVIg, intravenous immunoglobulins; MFS, Miller–Fisher syndrome; MGUS, monoclonal gammopathy of unknown significance; MM, multiple myeloma; na, not available; NfL, neurofilament light chain; Q Alb, albumin quotient CSF and serum.
Duration of symptoms as stated at hospital admission.
Time from hospital admission until discharge.
Prior vaccination or infectious disease (e.g., gastrointestinal, pulmonary etc.) up to 3 months.
Severe disease course associated with baseline sNfL
Further analysis in the AIAP group revealed a positive correlation between baseline sNfL concentrations and disease severity as assessed by the GBS Disability Scale at the peak of the disease (‘peak F score’, Spearman's r 0.6099, 95% confidence interval [CI] 0.148–0.8534, p = 0.013) [4] and hospitalization duration (r = 0.7847, 95% CI 0.4603–0.9242, p = 0.0005) (Figure 1c,d). Notably, no correlation was observed in the AIDP group (data not shown). Baseline sNfL concentrations were independent of symptom duration, Q Alb and CSF NfL indicating a peripheral origin (data not shown).
Multivariate linear regression analyses were performed to investigate the predictors of disease severity (F score) and hospitalization duration in patients with AIAP (Figure 1e). Whilst the model for disease severity was not statistically significant (p = 0.1568), baseline sNfL concentrations emerged as a significant predictor (p = 0.0420). Conversely, the model for hospitalization duration approached statistical significance (p = 0.0554), with baseline sNfL as a significant predictor (p = 0.0060).
Baseline sNfL predicts severe disease outcome
To predict disease outcome using baseline sNfL concentrations, ROC analysis was performed for the clinical endpoint of a peak F score greater than 3, indicating an inability to walk less than 5 m (Figure 1f). Here, the area under the curve of 0.8364 (95% CI 0.6379–1.000, p = 0.0361) showed a discriminatory ability for baseline sNfL concentrations.
Based on the ROC analysis, different cut‐off values for baseline sNfL concentrations were evaluated with different sensitivities and specificities in an event analysis examining the duration of hospitalization (Figure 1f). Of all the concentrations tested, the lowest cut‐off concentration at which the assay was both sensitive (40%) and specific (81.82%) was 351 pg/mL. Patients exceeding this threshold had a prolonged hospital stay (median hospitalization duration in days 30.5 vs. 15.5, chi‐squared 5.55, p = 0.0185) (Figure 1g). Furthermore, patients with baseline sNfL concentrations below this cut‐off had a hazard ratio of 0.3011 (95% CI 0.0549–1.651), whereas those above the cut‐off had a hazard ratio of 3.321 (0.6056–18.21). Of note, when this cut‐off was applied to a cohort consisting of AIAP and AIDP, it did not discriminate between long and short hospitalization duration (data not shown, chi‐squared 0.1378, p = 0.7105).
DISCUSSION
Diagnostic uncertainty in IMP at the time of initial presentation is a substantial issue in daily neurological practice [13]. Identification of disease subsets and prediction of disease outcome is an unmet need for early risk stratification. Here, evidence is reported for the predictive value of baseline sNfL for disease severity of axonal variants of IMP. Elevated sNfL concentrations were associated with increased disease severity and hospitalization duration. The identified cut‐off of 351 pg/mL shows potential for predicting prolonged hospitalization with moderate sensitivity but high specificity.
The detection of the axonal cytoskeletal protein NfL has emerged as a promising biomarker across neurological disorders including IMP. In previous studies [6, 7, 8], mean concentrations in CSF and serum samples exceeded those of healthy controls, and subgroup analysis suggested a discriminative value of sNfL in distinguishing primarily axonal from demyelinating variants, although the diagnostic criteria for primarily axonal neuropathies remain controversial. Nevertheless, the available evidence suggests the potential use as a surrogate for the disease course [7, 9]. Our results confirm these findings and contribute to the translation of pathophysiological correlates into the quest for biomarkers capable of early risk stratification. Remarkably, CSF and serum NfL concentrations in individual AIAP patients in our cohort exceeded those in patients with neurodegenerative diseases such as amyotrophic lateral sclerosis [14], highlighting the importance of contextualizing paraclinical parameters with the clinical syndrome. The use of CSF/serum ratios suggesting a peripheral origin of NfL may be a reasonable approach [9].
Our findings complement existing prognostic models [12] and provide means for early risk stratification. The strong correlation between baseline sNfL levels and disease severity in AIAP supports the use of sNfL as a tool for identifying patients at risk of severe disease progression. Identification of elevated sNfL could facilitate more intensive monitoring and tailored therapeutic interventions, potentially improving patient outcomes. The discrepancy between the non‐significant regression model and the predictive power of baseline sNfL suggests that other unmeasured factors, such as genetic predispositions, environmental influences or additional biomarkers, might influence disease severity. This is similar to the recalibration challenges encountered in clinical prediction models like the modified Erasmus GBS Outcome Score, indicating a necessity for further refinement to enhance predictive accuracy across diverse patient populations [12].
Although the sNfL cut‐off value demonstrates high specificity, its limited sensitivity means that it may miss a substantial number of severe cases. Therefore, sNfL should be used with other clinical assessments to accurately identify high‐risk patients. The limitations of our study, including its retrospective design, small cohort size and lack of external validation, may affect the generalizability of these findings. Critically ill patients might be underrepresented due to challenges in obtaining consent for sampling, as noted in an AIAP cohort without available biospecimens. Moreover, the cross‐sectional measurement of sNfL does not capture temporal changes in its levels.
In conclusion, sNfL is a promising biomarker for predicting disease severity in IMP, particularly in axonal variants. Further validation may help integrate sNfL into clinical practice for early identification of high‐risk patients and personalized treatment strategies. Of particular interest may be its potential as a general biomarker for axonal degeneration in other neuropathies (including traumatic neuropathies, rheumatological disease), although comorbidities such as diabetes may attenuate its significance [15].
AUTHOR CONTRIBUTIONS
Ali Maisam Afzali: Conceptualization; investigation; writing – original draft; methodology; validation; visualization; formal analysis. Milan Vosko: Writing – review and editing; investigation; data curation. Nya Reinhardt: Investigation; data curation. Iaroslav Zinevych: Investigation; data curation. Vincent Gmeiner: Investigation. Thomas Korn: Writing – review and editing. Christiane Gasperi: Writing – review and editing; formal analysis; validation; methodology. Achim Berthele: Writing – review and editing. Emily Feneberg: Writing – review and editing; investigation; validation; methodology; formal analysis. Bernhard Hemmer: Funding acquisition; writing – review and editing; investigation; conceptualization; project administration; supervision.
FUNDING INFORMATION
This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy within the framework of the Munich Cluster for Systems Neurology (EXC 2145 SyNergy – ID 390857198, Macroscale Hub), by the Bundesministerium für Bildung und Forschung within the framework of the Clinspect‐M project and by the Else Kröner‐Fresenius‐Stiftung (2021_EKEA.164).
CONFLICT OF INTEREST STATEMENT
The authors declare no conflict of interest relevant to the study.
ACKNOWLEDGEMENTS
All contributors of the study are thanked, especially the study nurses and laboratory technicians, for their motivated collaboration and recruitment efforts, and all patients and relatives for their participation and support.
Afzali AM, Vosko M, Reinhardt N, et al. Serum neurofilament light chain predicts disease severity in axonal variants of acute immune neuropathies: A retrospective monocentric cohort study. Eur J Neurol. 2025;32:e16539. doi: 10.1111/ene.16539
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.
REFERENCES
- 1. Bunschoten C, Jacobs BC, Van den Bergh PYK, Cornblath DR, van Doorn PA. Progress in diagnosis and treatment of chronic inflammatory demyelinating polyradiculoneuropathy. Lancet Neurol. 2019;18(8):784‐794. doi: 10.1016/s1474-4422(19)30144-9 [DOI] [PubMed] [Google Scholar]
- 2. Shahrizaila N, Lehmann HC, Kuwabara S. Guillain–Barré syndrome. Lancet. 2021;397(10280):1214‐1228. doi: 10.1016/S0140-6736(21)00517-1 [DOI] [PubMed] [Google Scholar]
- 3. Durand MC, Porcher R, Orlikowski D, et al. Clinical and electrophysiological predictors of respiratory failure in Guillain–Barré syndrome: a prospective study. Lancet Neurol. 2006;5(12):1021‐1028. doi: 10.1016/S1474-4422(06)70603-2 [DOI] [PubMed] [Google Scholar]
- 4. van Koningsveld R, Steyerberg EW, Hughes RA, Swan AV, van Doorn PA, Jacobs BC. A clinical prognostic scoring system for Guillain–Barré syndrome. Lancet Neurol. 2007;6(7):589‐594. doi: 10.1016/S1474-4422(07)70130-8 [DOI] [PubMed] [Google Scholar]
- 5. Walgaard C, Lingsma HF, Ruts L, van Doorn PA, Steyerberg EW, Jacobs BC. Early recognition of poor prognosis in Guillain–Barré syndrome. Neurology. 2011;76(11):968‐975. doi: 10.1212/WNL.0b013e3182104407 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Altmann P, De Simoni D, Kaider A, et al. Increased serum neurofilament light chain concentration indicates poor outcome in Guillain–Barré syndrome. J Neuroinflammation. 2020;17(1):86. doi: 10.1186/s12974-020-01737-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Martin‐Aguilar L, Camps‐Renom P, Lleixa C, et al. Serum neurofilament light chain predicts long‐term prognosis in Guillain–Barré syndrome patients. J Neurol Neurosurg Psychiatry. 2020;92:70‐77. doi: 10.1136/jnnp-2020-323899 [DOI] [PubMed] [Google Scholar]
- 8. van Tilburg SJ, Teunissen CE, Maas C, et al. Dynamics and prognostic value of serum neurofilament light chain in Guillain–Barré syndrome. EBioMedicine. 2024;102:105072. doi: 10.1016/j.ebiom.2024.105072 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Körtvelyessy P, Kuhle J, Düzel E, et al. Ratio and index of neurofilament light chain indicate its origin in Guillain–Barré syndrome. Ann Clin Transl Neurol. 2020;7(11):2213‐2220. doi: 10.1002/acn3.51207 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Rajabally YA, Durand MC, Mitchell J, Orlikowski D, Nicolas G. Electrophysiological diagnosis of Guillain–Barré syndrome subtype: could a single study suffice? J Neurol Neurosurg Psychiatry. 2015;86(1):115‐119. doi: 10.1136/jnnp-2014-307815 [DOI] [PubMed] [Google Scholar]
- 11. Doets AY, Verboon C, van den Berg B, et al. Regional variation of Guillain–Barré syndrome. Brain. 2018;141(10):2866‐2877. doi: 10.1093/brain/awy232 [DOI] [PubMed] [Google Scholar]
- 12. Doets AY, Lingsma HF, Walgaard C, et al. Predicting outcome in Guillain–Barré syndrome. Neurology. 2022;98(5):e518‐e532. doi: 10.1212/wnl.0000000000013139 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Fokke C, van den Berg B, Drenthen J, Walgaard C, van Doorn PA, Jacobs BC. Diagnosis of Guillain–Barré syndrome and validation of Brighton criteria. Brain. 2014;137(Pt 1):33‐43. doi: 10.1093/brain/awt285 [DOI] [PubMed] [Google Scholar]
- 14. Feneberg E, Oeckl P, Steinacker P, et al. Multicenter evaluation of neurofilaments in early symptom onset amyotrophic lateral sclerosis. Neurology. 2018;90(1):e22‐e30. doi: 10.1212/WNL.0000000000004761 [DOI] [PubMed] [Google Scholar]
- 15. Määttä LL, Andersen ST, Parkner T, et al. Longitudinal change in serum neurofilament light chain in type 2 diabetes and early diabetic polyneuropathy: ADDITION‐Denmark. Diabetes Care. 2024;47:986‐994. doi: 10.2337/dc23-2208 [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
The data that support the findings of this study are available from the corresponding author upon reasonable request.