Article Text


The prognostic value of CSF neurofilaments in multiple sclerosis at 15-year follow-up
  1. Axel Petzold1,2
  1. 1Department of Molecular Neuroscience, UCL, ION, London, UK
  2. 2Department of Neurology, VU Medical Center, Amsterdam, The Netherlands
  1. Correspondence to Dr Axel Petzold, Department of Molecular Neuroscience, UCL, ION, Queen Square House, 9th floor, Room 914, London WC1N 3BG, UK; a.petzold{at}

Statistics from

Neurofilament proteins have been established as one of the most reliable body fluid biomarkers for neurodegeneration in multiple sclerosis.1 ,2 One key feature is that neurofilaments are specific for neurons and axons.3 Recent literature reviews consistently showed that elevated body fluid neurofilament protein levels are a poor prognostic sign.4 ,5 A common limitation of these studies was the relative short clinical follow-up period.

Therefore, the aim of the present study was to test the long-term prognostic value of cerebrospinal fluid (CSF) neurofilament protein levels in a previously described cohort.6 More specifically, it was tested if high CSF neurofilament heavy chain (NfH) levels predicted a clinically relevant increase of disability as assessed by progression on the Expanded Disability Status Scale (EDSS) and its functional systems.

Setting: a chart and database review of 51 patients from a previously reported cohort6 who were followed up clinically. Patient recruitment started in 1996 and the last clinical follow-up was in 2013. Clinical assessment and assays were as described.7 Of note, the longitudinal CSF NfH data on 29 of the 51 patients who also had undergone a second lumbar puncture were reported before,7 but the baseline CSF NfH data from the entire cohort subject to this study have not been published before.

Descriptive statistics and non-parametric analyses were performed as described earlier, using SAS (V.9.4).7 Owing to non-Gaussian distribution, the median values and the 25–75% IQR were shown.

Analyses of covariance (ANCOVA) were applied to test if elevated CSF NfH levels were of long-term prognostic value. The cut-off value for elevated CSF NfHSMI35 from the 29 previously published patients was 20 pg/mL and for CSF NfHSMI34 it was 11  pg/mL.7 Poor prognosis was defined as progression on the EDSS over time. Individual trajectories were combined with pooled linear trajectories and ANCOVA to track the change over time in clinical groups.

The patients baseline characteristics are summarised in online supplementary table 1. Patients with a progressive disease course (secondary progressive, primary progressive multiple sclerosis, SP/PP MS) were older and more severely disabled than patients with a relapsing–remitting disease course (RR MS).

The trajectories of the EDSS of individual patients are shown in figure 1. The figure shows a significantly steeper slope of the averaged group trajectory, indicating more severe progression on the EDSS, in patients with RRMS with high CSF NfHSMI35 levels compared to those with normal levels (p=0.034, figure 1A). Likewise, there was a steeper slope of the averaged group trajectory for patients with RRMS with high CSF NfHSMI34 levels (p=0.0007, figure 1B).

Figure 1

Trajectories of Expanded Disability Status Scale (EDSS) scores for the 15-year follow-up period for individual patients (dashed lines) and average trajectories (solid lines). The baseline is indicated by the grey vertical reference line. (A) Clinical subtypes (relapsing–remitting, secondary progressive, primary progressive (RR, SP/PP)) were dichotomised according to the CSF NfHSMI35 baseline levels (high, normal). In RR MS (multiple sclerosis) the slope of the averaged trajectory was significantly steeper if baseline cerebrospinal fluid neurofilament heavy chain (CSF NfHSMI35) levels were high compared to normal baseline CSF NfHSMI35 levels (p=0.035). In SP/PP MS there was no difference between the slopes (p=0.9186). (B) Clinical subtypes (RR, SP/PP) were dichotomised according to the CSF NfHSMI34 baseline levels. In RR MS the slope of the averaged trajectory was significantly steeper if baseline CSF NfHSMI34 levels were high compared to normal baseline CSF NfHSMI35 levels (p=0.0007). In SP/PP MS statistical significance was narrowly missed (p=0.0642).

For patients with a progressive disease course a similar pattern was observed, but disease progression on the EDSS was slower, and significance was missed for CSF NfHSMI35 (p=0.9186) and for CSF NfHSMI34 (p=0.0642, figure 1A, B).

Kaplan-Meier analyses did not reveal any relationship for time to progression with either CSF NfHSMI35 levels (RR MS p=0.19, SP MS p=0.29) or CSF NfHSMI35 levels (RR MS p=0.27, SP MS p=0.76).

The functional systems responsible for the relationship with high CSF NfHSMI35 levels were the visual, pyramidal, cerebellar and sensory systems in RR MS (see online supplementary table S2). The group trajectories for these four systems showed a significantly steeper slope in patients with RRMS with high CSF NfHSMI35 levels, with the highest level of significance for the pyramidal system (p=0.005). Progression on the pyramidal system was also significantly related to high CSF NfHSMI34 levels in patients with RR (p=0.0125) and SP/PP disease course (p=0.0132).

The 15-year follow-up data provide evidence that high CSF NfH levels are of clinically relevant prognostic value. The data also show that the prognostic value was most marked in patients with an RR disease course. Across all clinical subgroups and CSF NfH phosphoforms, the strongest association was found with loss of function in the pyramidal system. These longitudinal data are consistent with and extend on previous cross-sectional studies reviewed elsewhere.1 ,4 ,5

The clinical observation that high CSF NfHSMI35 and NfHSMI34 levels were observed with more extensive neurodegeneration of the pyramidal system is consistent with postmortem tissue and in vivo CSF data demonstrating a rostrocaudal gradient of CSF NfH levels. Neurofilaments are most densely packed in axons and degeneration of the long axons comprising the pyramidal system are a likely source for high NfH levels in the lumbar CSF. Likewise, neurodegeneration affecting the posterior columns and spinothalamic tracts may explain the relationship between high CSF NfHSMI35 levels and progression of sensory symptoms. This interpretation is consistent with the extensive literature on imaging of spinal cord atrophy.

The association of high CSF NfHSMI35 levels with progression on the visual system is in line with previous reports on CSF and blood levels of NfHSMI35.3–5 Recent translational data strongly support the association between high body fluid NfH levels, loss of axons in the optic nerve, loss of retinal ganglion cells in the retina and impaired visual functioning.

Interestingly, in progressive disease, only CSF NfHSMI34 levels were associated with accumulation of disability in the pyramidal system. The 3-year follow-up report of a subgroup of only 29 patients from the here reported 51 patients described an increase of CSF NfHSMI35 levels in patients with SP/PP.7 A review on postmortem data suggested a heterogeneous distribution of NfH phosphoforms in the central nervous system, with a higher degree of phosphorylation, indicating more chronic damage due to neurodegeneration.3 This argument is supported by translational data suggesting phosphorylation of NfH to be an early pathological sign of neurodegeneration.8 The data on protein phosphorylation and neurodegeneration has not yet been investigated extensively in demyelination but there is a large body of literature from the dementia literature and amyotrophic lateral sclerosis.3

Limitations of the study include the non-systematic clinical follow-up interval and time. The EDSS and its subsystem are not of linear scale and patient numbers were small.

In conclusion, the 15-year follow-up data suggest that CSF NfH levels are of potential value for clinical trials. On a group level, their strongest prognostic value is for patients with RR MS who suffer damage to their spinal cord axons.


The author thanks all patients and their relatives for participating. The author is grateful for permission to use data and for the many excellent discussions with the team from the MS Centre VUMC and at the UCL Institute of Neurology.


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors AP developed the NfH ELISA and analysed the samples. AP had the idea of presenting this long-term follow-up study, reviewed the clinical data, performed the statistical analysis and wrote the manuscript.

  • Funding This project was funded by the Dutch MS Research Foundation (grant number 09-538d).

  • Competing interests None.

  • Ethics approval VUmc METc, Amsterdam, The Netherlands.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement This is proprietary data that could potentially permit identification of patients if shared on the internet.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.