765666

research-article2018
MSJ0010.1177/1352458518765666Multiple Sclerosis JournalN Siller, J Kuhle

MULTIPLE
SCLEROSIS MSJ
JOURNAL

Original Research Paper

Serum neurofilament light chain is a Multiple Sclerosis Journal

1­–9

biomarker of acute and chronic neuronal DOI: 10.1177/

https://doi.org/10.1177/1352458518765666
1352458518765666
https://doi.org/10.1177/1352458518765666

damage in early multiple sclerosis © The Author(s), 2018.

Reprints and permissions:
http://www.sagepub.co.uk/
journalsPermissions.nav

Nelly Siller, Jens Kuhle, Muthuraman Muthuraman, Christian Barro, Timo Uphaus,
Sergiu Groppa, Ludwig Kappos, Frauke Zipp and Stefan Bittner

Abstract
Background: Monitoring neuronal injury remains one key challenge in early relapsing-remitting mul- Correspondence to:
S Bittner
tiple sclerosis (RRMS) patients. Upon axonal damage, neurofilament – a major component of the neuro- Department of Neurology,
axonal cytoskeleton – is released into the cerebrospinal fluid (CSF) and subsequently peripheral blood. University Medical Center
of the Johannes Gutenberg
Objective: To investigate the relevance of serum neurofilament light chain (sNfL) for acute and chronic University Mainz, 55131
Mainz, Germany.
axonal damage in early RRMS. stefan.bittner@unimedizin-
Methods: sNfL levels were determined in 74 patients (63 therapy-naive) with recently diagnosed clini- mainz.de
Nelly Siller
cally isolated syndrome (CIS) or RRMS using Single Molecule Array technology. Standardized 3 T mag- Muthuraman Muthuraman
netic resonance imaging (MRI) was performed at baseline and 1–3 consecutive follow-ups (42 patients; Timo Uphaus
Sergiu Groppa
range: 6–37 months). Frauke Zipp
Results: Baseline sNfL correlated significantly with T2 lesion volume (r = 0.555, p < 0.0001). There was Stefan Bittner
Department of Neurology
no correlation between baseline sNfL and age, Expanded Disability Status Scale (EDSS) score or other and Focus Program
Translational Neuroscience
calculated MRI measures. However, T2 lesion volume increased (r = 0.67, p < 0.0001) and brain paren- (FTN), Research Center for
chymal volume decreased more rapidly in patients with higher baseline sNfL (r = −0.623, p = 0.0004). Immunotherapy (FZI), Rhine-
Main Neuroscience Network
Gd-enhancing lesions correlated positively with sNfL levels. Initiation of disease-modifying treatment (rmn2), University Medical
Center of the Johannes
led to a significant decrease in sNfL levels. Gutenberg University Mainz,
Conclusion: sNfL indicates acute inflammation as demonstrated by correlation with Gd+ lesions. It is Mainz, Germany

a promising biomarker for neuro-axonal damage in early multiple sclerosis (MS) patients, since higher Jens Kuhle
Christian Barro
baseline sNfL levels predicted future brain atrophy within 2 years. Ludwig Kappos
Neurologic Clinic and
Policlinic and Departments of
Medicine, Biomedicine and
Keywords: Multiple sclerosis, neurodegeneration, neurofilament, biomarker, clinical progression, MRI, Clinical Research, University
disease activity Hospital Basel, University of
Basel, Basel, Switzerland

Date received: 29 August 2017; revised: 19 January 2018; accepted: 6 February 2018

Introduction chain (200–220 kDa). Neurofilament light chain

Neurofilament as a major component of the axonal (NfL) has been studied most extensively and an asso-
cytoskeleton is released into the interstitial fluid, ciation with acute neuro-axonal damage was demon-
cerebrospinal fluid (CSF) and subsequently serum in strated for NfL which is elevated in CSF during all
low levels upon neuro-axonal damage occurring in stages of multiple sclerosis (MS) with a heightened
neurological disorders.1 Modern magnetic resonance (10 times) increase in times of acute relapses.5
imaging (MRI) techniques and immunohistochemical
studies have revealed neuronal damage in early stages Advances in analytic sensitivity using Single
of the disease course of patients with clinically Molecule Array (Simoa) technology provides the
isolated syndrome (CIS) or relapsing-remitting multi- basis for assessment of NfL levels in serum samples
ple sclerosis (RRMS).2–4 Based on the apparent (sNfL) as an important step towards its use as a bio-
molecular mass of the mammalian subunits, three marker in clinical routine.6 Indeed, NfL levels in
major subunits of neurofilament have been defined: serum and CSF are highly correlated and sNfL in
light (65−70 kDa), medium (140–160 kDa) and heavy RRMS patients is increased compared to healthy

journals.sagepub.com/home/msj 1
Multiple Sclerosis Journal 00(0)

Figure 1. Study design. sNfL was determined in serum of 63 therapy-naive patients; standardized MRI was performed at
baseline. 1−3 consecutive MRI follow-ups (FU) in 42 patients after 637 months were conducted. Follow-up MRIs were
classified into two groups, one with MRI after less than 18 months (n = 38, median = 11 months) and one with MRI at least
18 months later (n = 33, median = 24 months). Correlation between sNfL levels and gadolinium (Gd)-enhancing lesions
was evaluated in 74 patients, 62 of the naive cohort and an additional 12 patients (one patient in the therapy-naive cohort
refused gadolinium administration). Follow-up MRIs with consecutive serum samples were included leading to 107 sNfL
values and MRI scans.

controls.7,8 A recent study using this technology in Center Mainz and University Hospital Basel) were
MS has proposed sNfL as a biomarker to monitor or involved in the study. Serum samples were obtained
potentially even predict neuronal damage in MS in Mainz, pseudonymized, and sent and analysed
patients,8 while it is still largely unclear, whether sNfL without clinical data or other information.
could predict neuronal damage in very early MS or Standardized 3 T MRI protocol was performed at
CIS. Therefore, we investigated the association baseline; 42 patients underwent 1−3 consecutive
between sNfL and clinical and MRI measurements in follow-ups after 6–37 months. Follow-up MRIs
a therapy-naive cohort of CIS and RRMS patients in were classified into two groups, one with MRI
the very early stage of disease. after less than 18 months (n = 38, median: 11 months)
and one with MRI at least 18 months later (n = 33,
median: 24 months). Imaging was performed using
Methods sagittal three-dimensional (3D) T1-weighted mag-
netization-prepared rapid gradient-echo (MP-
Study design and population RAGE) sequence (TE/TI/TR = 2.52/900/1900 ms, flip
In all, 74 patients (63 therapy naive) with recently angle = 9°, field of view (FOV) = 256 × 256 mm2,
diagnosed CIS or RRMS were included in the study matrix size = 256 × 256, slab thickness = 192 mm,
(see also Figure 1 for study design), which was voxel size = 1 × 1 × 1 mm3), sagittal 3D T2-weighted
approved by the local ethics committees; all patients fluid-attenuated inversion recovery (FLAIR) sequence
provided written informed consent. sNfL was deter- (TE/TI/TR = 388/1800/5000 ms, echo-train length = 848,
mined by a previously described Simoa NfL assay.8 FOV = 256 × 256 mm2, matrix size = 256 × 256, slab
Measurements were performed in a completely blinded thickness = 192 mm, voxel size = 1 × 1 × 1 mm3). MRI
fashion, since two centres (University Medical data were obtained and analysed by blinded raters

2 journals.sagepub.com/home/msj
 N Siller, J Kuhle et al.

who had no further information about the sNfL levels from UmanDiagnostics. Calibrators ranged from 0 to
or clinical data of the investigated patients. 2000 pg/mL. Batch prepared calibrators were stored
at −80°C. Inter-assay coefficients of variation (CV)
Correlation between sNfL levels and gadolinium were 4%, 2%, and 3% for control samples with mean
(Gd)-enhancing lesions was evaluated in all 74 concentrations of 13, 33 and 283 pg/mL, respectively.
patients: 62 of the naive cohort and an additional 12 The mean intra-assay CV of duplicate determinations
patients with therapy already at baseline. One patient for concentration was 4%. Repeated measuring was
in the treated cohort did not receive gadolinium and performed for few samples with intra-assay CV above
could therefore not be included in the correlation 20%.
analysis between NfL and Gd-enhancing lesions.
Altogether, 107 sNfL values and corresponding MRI
scans were available for analysis. Statistical analysis
Summary statistics are presented as mean ± SD,
Consecutive serum samples after a median 13 months median (range) and percentage, where applicable. For
were available in 21 patients, 14 patients had started univariate analysis, the Mann–Whitney U test and
an immunomodulatory therapy and 6 had not. One chi-square test were used as appropriate. For correla-
patient quit immunomodulatory therapy after 4 weeks tion, non-parametric Spearman analysis was used.
and was added to the latter group, assuming no long- Multivariate regression analysis was performed with
term therapeutic effect of dimethyl fumarate after this Gd lesion number as dependent variable and sNfL,
short period of time. Informed consent was obtained age, sex and so on as independent variables. Receiver
from all patients and the study was approved by local operating characteristic (ROC) analysis was used to
Medical Ethics Committee. estimate the predictive discriminating values. For
every individual cut-off point, there is a pair of diag-
nostic sensitivity and specificity values. To construct
MRI analysis a ROC graph, we plotted these pairs of values with
Initially, lesion maps were drawn on T2-weighted 3D 1 – specificity on the x-axis and sensitivity on the
FLAIR images using the MRIcron software (http:// y-axis. From the ROC curves, we estimated Youden’s
www.mccauslandcenter.sc.edu/mricro/mricron/). index, which is defined for all points of an ROC
Using the lesion segmentation toolbox (LST) which is curve, and the maximum value of the index may be
part of the statistical parameter mapping (SPM8) soft- used as a criterion for selecting the optimum cut-off
ware, 3D FLAIR images were co-registered to 3D T1 point between two variables. The index is represented
images and bias corrected. After partial volume esti- graphically as the height above the 50% chance line,
mation (PVE), lesion segmentation was performed and it is also equivalent to the area under the curve
with 20 different initial threshold values for the lesion (AUC) represented by a single operating point. p-val-
growth algorithm. By comparing automatically and ues less than <0.05 were considered statistically sig-
manually estimated lesion maps, the optimal thresh- nificant. All statistical analyses were performed by
old (κ value) for lesion detection was determined for means of SPSS 23.0 package software (IBM, Armonk,
each patient and used for automatic lesion volume NY, USA). Prediction accuracy of brain atrophy by
estimation and lesion filling. Subsequently, the filled NfL was calculated using support vector machine
3D T1 images as well as the native 3D T1 images (SVM). SVM is a powerful tool for nonlinear predic-
were segmented into grey matter, white matter and tion between two variables.9 In short, the algorithm
CSF and normalized to Montreal Neurological looks for an optimal threshold between the two vari-
Institute (MNI) space. Finally, the quality of the seg- ables by maximizing the correlation between classes’
mentations was visually inspected. closest points. In most cases, the linear prediction is
not ideal so a projection into a higher-dimensional
space is performed where the data points effectively
Serum NfL measurements become linearly correlated. Here, we have used the
Serum samples were collected by treating physicians radial basis function kernel for this projection due to
during the past few years at the University Medical its good performance as discussed previously9 and
Center Mainz. Samples were processed at room tem- used the grid search (min = 1; max = 10) to find the
perature within 2 hours. Serum samples were spun at few optimal input parameters namely C (Type of
2000g at room temperature for 10 minutes, aliquoted regression algorithm; 1 to 1000) and gamma (0.25).
in polypropylene tubes and stored at −80°C. Serum The selection was checked by 10-fold cross validation
NfL was measured by Simoa NfL assay as described by taking 70% of the data for training and 30% for
previously.8 Bovine lyophilized NfL was obtained testing.

journals.sagepub.com/home/msj 3
Multiple Sclerosis Journal 00(0)

Table 1. Detailed characteristics of patient data and MRI parameters for treatment-naive cohort.

Variable Naive cohorta p valueb (rho value)

n 63
sNfL (pg/mL) 36.26 (24.1–63.2)
Age 35 (27–46) 0.405 (0.107)
Female 44 (69.8%)
DD (months) 1 (0–2) 0.350 (0.120)
EDSS 1 (0–2) 0.568 (0.073)
Gap time (blood 53 (6–137) 0.137 (−0.189)
withdrawal-MRI; days)
LV (mL) 1.76 (0.8–4.39) <0.0001* (0.555)
GMV (mL) 620.6 (569.7–673.5) 0.757 (−0.040)
WMV (mL) 568.8 (534.3–623.3) 0.500 (0.087)
BP (mL) 1187.2 (1124.5–1281.2) 0.857 (0.023)
TV (mL) 1412.5 (1336.4–1523.9) 0.488 (0.089)
sNfL: serum neurofilament light chain; MRI: magnetic resonance imaging; EDSS: Expanded Disability Status Scale; DD: disease
duration at time point of serum sampling; LV: lesion volume; GMV: grey matter volume; WMV: white matter volume; BP: brain
parenchyma; TV: total volume.
aUnless noted otherwise, data reported as median (range).
bDerived from non-parametric correlation (Spearman-rho) with sNfL.
*Statistically significant.

Table 2. Detailed characteristics of patient data and MRI parameters for patients included in the gadolinium-enhancing
lesion analysis.

Variable Gadolinium cohorta p value

n 25 (+) 82 (–)
NfL (pg/mL) 63.2 (37.6−159) 28.1 (18.8−37.6) <0.0001b,*
Age 36 (28−46) 40 (29−47) 0.614b
Female 20 (80%) 56 (68%) 0.320c
DD (months) 1 (1−4.5) 8 (1−20.5) 0.045b,*
EDSS 1 (0−2) 1 (1−2) 0.576b
Gap time (blood 8 (−20 to 54) 0 (−3.2 to 23) 0.449b
withdrawal-MRI; days)
NfL: neurofilament light chain; MRI: magnetic resonance imaging; EDSS Expanded Disability Status Scale; DD: disease duration at
time point of serum sampling
aUnless noted otherwise, data reported as median (range).
bDerived from Mann–Whitney U test.
cDerived from chi-square test.
*Statistically significant.

Results in our naive cohort at baseline was 36.26

(24.1−63.2) pg/mL. In our therapy naive group, 43
Baseline sNfL levels significantly correlate with patients had been diagnosed with RRMS and 20
baseline T2 lesion volume and T2 lesion volume patients reached criteria for CIS. RRMS patients had
change in MS patients significantly higher median (45.33 pg/mL) sNfL lev-
To assess the clinical value of measuring sNfL levels els in comparison to CIS patients (median = 28.39 pg/
in early MS patients, we focused on therapy-naive mL, p < 0.05; Supplementary Figure 1). There were
patients (n = 63) with newly diagnosed CIS/RRMS. no significant correlations between baseline NfL and
The median disease duration from first symptoms disease duration (r = 0.120, p = 0.350), Expanded
onset was 1 month. See Tables 1 and 2 for detailed Disability Status Scale (EDSS) score (r = 0.073,
patient characteristics. The median NfL serum level p = 0.568) or age (r = 0.107, p = 0.405) in our cohort.

4 journals.sagepub.com/home/msj
 N Siller, J Kuhle et al.

Figure 2. Correlation of baseline sNfL levels with T2 lesion volume at baseline and brain atrophy rates after 1 and
2 years of follow-up. (a) Baseline sNfL correlated significantly with T2 lesion volume (r = 0.55, p < 0.0001). One outlier
was excluded due to technical difficulties: movement artefacts during acquisition prevented further analysis. The
regression analysis refers to all data points; however, for improved visualization the x/y axes were adjusted leading to
9 data points out of range. (b) Baseline sNfL correlated significantly with T2 lesion volume change over time (r = 0.67,
p < 0.0001), comparing T2 lesion volume at baseline with MRI follow-up after 6−37 months. (c) Brain parenchyma
decreased more rapidly in patients with higher baseline sNfL. Analysis was performed after a median MRI follow-up of
11 months (Group 1, 6–17 months, n = 38) and after 24 months (Group 2, >18 months, n = 33). (d) Non-linear prediction
(SVM) showed 69.2% prediction accuracy (70% training and 30% testing) of brain atrophy after median 24 months with
baseline sNfL.

Baseline sNfL correlated significantly with T2 lesion subsequent brain atrophy (Figure 2(c)). In our cohort,
volume (Figure 2(a), r = 0.555, p < 0.0001) and change baseline sNfL levels showed only a low prediction
in T2 lesion volume over time (Figure 2(b), r = 0.67, accuracy of brain atrophy after 1 year (group 1,
p < 0.0001). Additional MRI measures did not show 29.4%), compared to relatively high prediction accu-
any significant correlation with baseline sNfL (white racy (69.2%) after 2 years (group 2). These findings
matter volume (r = 0.087, p = 0.5), grey matter volume might reflect the temporal latency of structural MRI
(r = −0.040, p = 0.757), brain parenchyma (r = 0.023, quantification towards neurodegenerative processes.
p = 0.857) or total volume (r = 0.089, p = 0.488)).

Association between sNfL and gadolinium-

Baseline sNfL levels predict brain atrophy enhancing lesions
progression Elevations of sNfL levels reflect neuro-axonal dam-
As neurofilament levels are thought to reflect ongoing age, mediated both by acute inflammatory damage in
neuro-axonal damage, we assessed whether patients an acute relapse and ‘chronic’ ongoing neurodegen-
with higher sNfL levels would suffer from increased erative processes. We first addressed the relation
brain atrophy at later time points. Indeed, over the between Gd+ MRI lesions and sNfL levels. For this
next 6–37 months, brain parenchyma decreased more analysis, all available MRIs (both baseline and fol-
rapidly in patients with higher baseline sNfL (all low-up data) with corresponding serum samples were
patients: r = −0.623, p = 0.0004; MRI group 1: r = −0.652, pooled. Median NfL serum levels in patients with
p = 0.0003; MRI group 2: r = −0.415, p = 0.002, Gd+ lesions were 63.2 (37.6–159) pg/mL and 28.1
Figure 2(b)). Predictive analysis using SVM was per- (18.8−37.6) pg/mL in patients without acute inflam-
formed to assess prediction accuracy of sNfL for mation (p < 0.0001, Figure 3(a)). There was a positive

journals.sagepub.com/home/msj 5
Multiple Sclerosis Journal 00(0)

Figure 3. Association between sNfL levels and gadolinium-enhancing lesions. (a) Median sNfL in group with Gd-
enhancing lesion was 45.3 pg/mL, n = 19 (−6 outliers), median NfL in group without Gd-enhancing lesion was 27.9 pg/
mL, n = 79 (−3 outliers), p < 0.0001. Boxes include median, 25th and 75th percentiles. (b) Positive correlation, using non-
parametric Spearman correlation, between number of Gd-enhancing lesions and baseline serum NfL levels in 25 patients
(r = 0.484, p = 0.014).

correlation between the number of Gd-enhancing β-interferons, two with glatiramer acetate and one
lesions and serum NfL levels (r = 0.484, p = 0.014, patient with teriflunomide. Median baseline sNfL was
Figure 3(b)). We found no significant differences in not significantly different between groups (61.57 pg/
age, sex, EDSS score or gap time between blood with- mL in the therapy group, 34.94 pg/mL in the naive
drawal and MRI assessment between the two groups. group, p = 0.488). Median change in NfL in the ther-
Disease duration was median 8 months in patients apy group was −34.2 pg/mL, whereas the therapy
without Gd-enhancing lesions and only 1 month in the naive group had a median reduction of only −3.7 pg/
group with Gd-enhancing lesions (p = 0.045). mL (p = 0.0074, Figure 4(a) and (b)) in the second
Multivariate regression analysis confirmed sNfL as sNfL analysis. Of note, while a correlation between
the only parameter influencing Gd lesion numbers treatment initiation and reduced sNfL levels is highly
(p < 0.001) while age, sex, disease duration or EDSS probable, one has to point out significant differences
score were not significant. ROC analysis was per- concerning age, sex and disease duration in the groups
formed to determine the predictive discriminating (age: 40 vs 50 years, p = 0.020; sex: 8 females and 6
value of sNfL. sNfL below or above 34 pg/mL pre- males vs 7 females, p = 0.05; disease duration: 1
dicted gadolinium enhancement in MRI with a sensi- month vs 8 months, p = 0.04).
tivity of 84% and specificity of 66% (AUC = 0.845);
the Youden index was 0.53. In clinical practice, where
repetitive MRI assessments are common in young MS Discussion
patients, sNfL measurements might therefore be a The search for valid biomarkers for patients with MS
supportive tool to decide which patient needs gado- has been unsatisfying throughout many years.
linium application and when MRI without gadolinium Particularly at the beginning of the disease, prospec-
is supported. tive biomarkers could facilitate therapy decisions,
permit the detection of neuroimmunological activity
before structural damage becomes prominent10 and
therefore lead to better patient outcome and decreased
Effect of immunomodulatory therapy on sNfL incidence of transition to progressive disease course
levels: longitudinal changes of sNfL over time in the long term.11 So far, female sex, pure sensory
Longitudinal serum samples were available for a lim- onset, lower age at onset and other prognostic factors
ited number of patients (n = 21) within our naive are associated with a benign course in early RRMS
cohort (follow-up serum sample after median patients.12 However, availability of prognostic param-
13 months), allowing us to address the impact of ther- eters in addition to MRI measurements in clinical
apy initiation on sNfL levels. In total, 14 out of 21 practice is limited. Easily accessible, reliable and sen-
newly diagnosed patients started an immunomodula- sitive biomarkers are therefore needed.
tory therapy after baseline sNfL measurement,
whereas 7 patients remained therapy naive. Five In our study, we were able to show in a cohort of
patients started with dimethyl fumarate, six with newly diagnosed CIS/RRMS patients that sNfL levels

6 journals.sagepub.com/home/msj
 N Siller, J Kuhle et al.

Figure 4. Effect of immunomodulatory therapy on sNfL levels. (a) 14 therapy-naive patients who started an
immunomodulatory therapy in comparison to 7 naive patients staying without therapy, each with a consecutive serum
sample (both groups median 13 months after baseline; t0 = baseline, t1 = follow-up). Immunomodulatory therapy
was started median 12 months before second serum sample. (b) sNfL decreases significantly after initiation of any
immunomodulatory therapy (** p = 0.0074, obtained by Mann–Whitney U test).

at the time point of diagnosis correlate significantly this study addressed mostly patients with established,
with T2 lesion volume at baseline and at any time ongoing disease (disease duration ~7 years and EDSS
point with the presence and number of Gd-enhancing ~3.0), in contrast to our investigation focusing on
lesions. Longitudinal investigations showed a very early MS patients (disease duration ~1 month
decrease in sNfL after initiating disease-modifying and EDSS ~1.0).
therapies. Patients with higher baseline sNfL showed
a more rapid decrease in brain volume and a more So far, MRI-derived metrics are most commonly used
rapid increase in T2 lesion volume after 2 years. Acute in clinical practice to monitor subclinical disease activ-
inflammatory neuronal damage, illustrated by ity and – so far only in study cohorts – progressive
Gd-enhancing lesions, was clearly associated with brain atrophy. Other parameters such as NEDA (no evi-
sNfL levels. Strong correlation of T2 lesion volume dence of disease activity, that is, no relapse, EDSS
and of brain atrophy rates with baseline sNfL lead to worsening or new MRI lesions) have been suggested as
the question whether treatment decisions in individual novel monitoring compound parameters for treating
patient care may be adjusted to neuronal damage MS patients.14,15 However, a recent study showed that
already in the beginning of the disease. clinical and MRI data included in NEDA after 2 years
did not correlate with the long-term outcome (>90% of
Recently, a pilot study in 22 versus 20 patients from the subjects had a 10-year follow-up) in a cohort of
a phase II trial of riluzole as add-on treatment to actively treated MS and CIS patients (n = 517).16 The
weekly interferon-β (IFN-β)-1a addressed, for the authors questioned both annual MRI assessments for
first time, the potential of using sNfL as a biomarker monitoring in individual MS care and commonly used
for neuronal damage in MS. Interestingly, sNfL lev- clinical and MRI features over 2 years in most rand-
els correlated with EDSS score, neuropsychological omized clinical trials. Disability accumulation in
outcome and Gd-enhancing lesions and predicted patients despite meeting NEDA criteria was possibly
whole brain atrophy at 1 and 2 years, possibly reflect- overlooked due to a lack of spinal cord MRI and brain
ing progressive neuronal loss and ongoing disease atrophy evaluation. While it is important to point out
activity.13 In a larger study assessing two independent that in contrast to MRI, NfL levels cannot provide
Swiss cohorts, these initial findings were largely con- localization information about the specific site of
firmed and expanded as sNfL levels correlated with axonal damage, a contribution of spinal cord damage to
EDSS assessments, presence of relapses and sNfL levels was recently depicted by Disanto et al.8
Gd-enhancing lesions. Disease-modifying treatments
reduced sNfL levels and patients with higher sNfL We found a significant reduction of sNfL in patients
levels were at risk of future relapses and EDSS pro- that started immunomodulatory treatment in compari-
gression.8 While these findings are indeed promising, son to patients that stayed naive. Reduction of NfL in

journals.sagepub.com/home/msj 7
Multiple Sclerosis Journal 00(0)

CSF of RRMS patients under second line/escalation in the multiple sclerosis brain. Brain 2016; 139:
therapies natalizumab,17 fingolimod18 and rituximab19 807–815.
and recently also in plasma for fingolimod20 has been 4. Trapp BD, Peterson J, Ransohoff RM, et al. Axonal
described before. Our cohort was too small to detect transection in the lesions of multiple sclerosis. N Engl
differences in sNfL decrease between different drugs J Med 1998; 338: 278–285.
in the therapy group or to address NfL levels of
5. Malmestrom C, Haghighi S, Rosengren L, et al.
patients with insufficient treatment responses. Neurofilament light protein and glial fibrillary acidic
protein as biological markers in MS. Neurology 2003;
Several questions remain that need to be addressed 61: 1720–1725.
in larger cohorts of early RRMS patients with longi-
tudinal sNfL and MRI measurements. Is it possible 6. Kuhle J, Barro C, Andreasson U, et al. Comparison
of three analytical platforms for quantification of the
to monitor individual treatment responses via sNfL?
neurofilament light chain in blood samples: ELISA,
Can we predict neuronal loss or even transition into electrochemiluminescence immunoassay and Simoa.
long-term progressive disease course in the very Clin Chem Lab Med 2016; 54: 1655–1661.
early stage of disease and match immunomodula-
tory therapy regimes? Our results underline the 7. Kuhle J, Barro C, Disanto G, et al. Serum
potential value of NfL as a biomarker for neuronal neurofilament light chain in early relapsing remitting
MS is increased and correlates with CSF levels and
damage in MS patients and extend these findings to
with MRI measures of disease severity. Mult Scler
a group of patients with very early disease course of 2016; 22: 1550–1559.
RRMS and CIS. Serial measurements of serum NfL
levels might be used to monitor the extent of inflam- 8. Disanto G, Barro C, Benkert P, et al. Serum
matory-driven neuronal damage and could be useful neurofilament light: A biomarker of neuronal
for treatment decisions. Larger, prospective studies damage in multiple sclerosis. Ann Neurol 2017; 81:
857–870.
are warranted.
9. Cortes C and Vapnik V. Support-vector networks.
Acknowledgements Mach Learn 1995; 20: 273–297.
The authors thank Cheryl Ernest for proofreading and 10. Londono AC and Mora CA. Evidence of disease
editing the manuscript. N.S. and J.K. contributed control: A realistic concept beyond NEDA in the
equally as first authors. F.Z. and S.B. contributed treatment of multiple sclerosis. F1000Res 2017; 6:
equally as senior authors. 566.
11. Larochelle C, Uphaus T, Prat A, et al. Secondary
Declaration of Conflicting Interests progression in multiple sclerosis: Neuronal
The author(s) declared no potential conflicts of inter- exhaustion or distinct pathology? Trends Neurosci
est with respect to the research, authorship and/or 2016; 39: 325–339.
publication of this article.
12. Sartori A, Abdoli M and Freedman MS. Can we
predict benign multiple sclerosis? Results of a 20-year
Funding
long-term follow-up study. J Neurol 2017; 264:
The author(s) disclosed receipt of the following finan- 1068–1075.
cial support for the research, authorship and/or publi-
cation of this article: This work was supported by the 13. Kuhle J, Nourbakhsh B, Grant D, et al. Serum
German Research Council (DFG, CRC-TR-128 to neurofilament is associated with progression of brain
S.G., F.Z. and S.B.). atrophy and disability in early MS. Neurology 2017;
88: 826–831.
14. Bevan CJ and Cree BA. Disease activity free
status: A new end point for a new era in multiple
References sclerosis clinical research? JAMA Neurol 2014; 71:
1. Teunissen CE and Khalil M. Neurofilaments as biomarkers 269–270.
in multiple sclerosis. Mult Scler 2012; 18: 552–556.
15. Giovannoni G, Tomic D, Bright JR, et al. ‘No evident
2. Gracien RM, Reitz SC, Hof SM, et al. Longitudinal disease activity’: The use of combined assessments in
quantitative MRI assessment of cortical damage the management of patients with multiple sclerosis.
in multiple sclerosis: A pilot study. J Magn Reson Mult Scler 2017; 23: 1179–1187.
Imaging 2017; 46: 1485–1490.
16. Cree BA, Gourraud PA, Oksenberg JR, et al. Long-
3. Haider L, Zrzavy T, Hametner S, et al. The term evolution of multiple sclerosis disability in the
topograpy of demyelination and neurodegeneration treatment era. Ann Neurol 2016; 80: 499–510.

8 journals.sagepub.com/home/msj
 N Siller, J Kuhle et al.

17. Gunnarsson M, Malmestrom C, Axelsson M, et al. 19. de Flon P, Gunnarsson M, Laurell K, et al. Reduced
Axonal damage in relapsing multiple sclerosis is inflammation in relapsing-remitting multiple sclerosis
markedly reduced by natalizumab. Ann Neurol 2011; after therapy switch to rituximab. Neurology 2016;
87: 141–147.
69: 83–89.
20. Piehl F, Kockum I, Khademi M, et al. Plasma
18. Kuhle J, Disanto G, Lorscheider J, et al. Fingolimod neurofilament light chain levels in patients with MS Visit SAGE journals online
and CSF neurofilament light chain levels in relapsing- switching from injectable therapies to fingolimod. journals.sagepub.com/
home/msj
remitting multiple sclerosis. Neurology 2015; 84: Mult Scler. Epub ahead of print 1 June 2017. DOI:
1639–1643. 10.1177/1352458517715132. SAGE journals

journals.sagepub.com/home/msj 9