European Journal of Neurology 2006, 13: 135–140

Prediction of response to IVIg treatment in patients with lower motor neurone

disorders
N. Strigl-Pilla, A. Königb, M. Schrödera, H. Beranekb, B.G.H. Schosera, M. Spaetha, D. Pongratza
and W. Müller-Felbera
a
Friedrich-Baur-Institute and bDepartment of Internal Medicine, Ludwig-Maximilians-University Munich, Munich, Germany

Keywords:
amyotrophic lateral
sclerosis, IVIg, multifocal
motor neuropathy,
treatment
Received 5 August 2004
Accepted 19 January 2005
The distinction between multifocal motor neuropathy, treatable by intravenous im-
munoglobulins (IVIg), and degenerative motor neurone disorders is often difficult. To
find predictive factors for the response to IVIg treatment, 40 consecutive patients with
pure lower motor neurone disorders (LMND) were prospectively examined. They all
received at least two times IVIg (2 g/kg bodyweight). Prior to the first and before all
the following treatments a standardized evaluation was performed including clinical
examination, neurophysiological and laboratory evaluation. According to changes in
the neurological examination and the Neuromuscular Symptom Score, the patients
were divided into responders and non-responders after the second course of treatment.
In our study, no single clinical, neurophysiological, or laboratory parameter was
sensitive enough to predict response. The only single parameter that highly correlated
with a positive response to treatment was an elevated GM1 antibody titre. Lack of
response to IVIg treatment is likely in patients with generalization of electromyo-
graphic signs of denervation beyond the clinically involved site, proximal localization
of the weakness, and an elevated level of the creatinekinase. Conduction blocks do not
distinguish between both groups. We propose a scoring system combining clinical,
serological and neurophysiological data in order to decide which patients with LMND
may receive IVIg.

no improvement or even worsening of MMN with

Introduction
Multifocal motor neuropathy (MMN) was first des-
cribed in 1985 in four patients who suffered from
progressive weakness similar to motor neurone dis-
orders (MND) [1]. The neurophysiological hallmarks
were motor conduction blocks. In 1988 IgM anti-
bodies to the GM1 ganglioside directed against gly-
colipids in the region of the nodes of Ranvier were
found, suggesting an autoimmune process [2]. A re-
cent report showed, that the increase of GM1 anti-
body titre in MMN is usually followed by an increase
of the light chain (k ¼ lambda and j ¼ kappa) and
the light chain ratio (k/j) of the antibody whereas in
classical amyotrophic lateral sclerosis (ALS) the light
chain ratio is normal [3].
MMN, originally related to chronic inflammatory
demyelinating polyneuropathy or MND is now con-
sidered as a separate entity. Typical findings that con-
firm the diagnosis of MMN are conduction blocks. In
contrast to other demyelinating neuropathies, there is
Correspondence: Prof. W. Müller-Felber, Friedrich-Baur-Institute,
Ziemssenstr. 1 a, 80336 Munich, Germany (Tel.: ++49 89 51607470;
fax: ++49 89 51602203; e-mail: wolfgang.mueller-felber@fbs.med.
uni-muenchen.de).
steroids or plasmapheresis. In 1993 improvement of
MMN following treatment with intravenous immuno-
globulins (IVIg) was reported for the first time [4]. In
addition, an axonal motor neuropathy similar to MMN
without conduction blocks was successfully treated by
IVIg [5]. A study using a threshold tracking technique
showed, that besides focal demyelination, axonal exci-
tability changed significantly and contributed to con-
duction failure. Furthermore, the axo-glial interaction
was disturbed [6]. Pathologic alterations at the site of
conduction block showed focal regions of fibre loss,
clusters of regenerating fibres, and a low frequency of
fibre degeneration without onion bulbs or segmental
demyelination. These changes were seen only in the
motor fibres of mixed nerves [7].
ALS starts with pure lower motor neurone disorder
(LMND) in up to 20% of all cases. In a prospective,
population-based study of ALS in Ireland, 52% of
patients showed pure lower motor neurone signs at
onset [8].
For that reason the major clinical problem is to find
those patients with LMND suitable for the expensive
IVIg treatment. All criteria that have been proposed so
far lack sufficient sensitivity and specificity to predict
response to treatment.

Ó 2006 EFNS 135

136 N. Strigl-Pill et al.
The goal of this examination was to analyse clinical,
serological and neurophysiological parameters that
allow the prediction of successful IVIg treatment in
patients with pure LMND.
Patients and methods
Between 1998 and 2002, 6495 patients were seen in our
neuromuscular centre. Of these 40 patients were inclu-
ded in the study. The elected patients suffered from pure
LMND (nine females, 31 males) that is asymmetric
limb weakness in the territory of more than one motor
nerve without respiratory failure or cranial nerve
affection. In all patients the duration of the disease
before treatment was more than 1 year.
All patients received IVIg treatment. The average age
was 58 years (33–75 years).
Inclusion criteria were: isolated lesion of the motor
neurone (i.e. clinical and neurophysiological signs of
denervation without affection of the sensory nerves or
upper motor neurone); progressive asymmetric weak-
ness; electromyographical proof of a neurogenic lesion
which was not related to a root compression, poly-
neuropathy or entrapment syndrome.
Exclusion criteria were: respiratory involvement;
lesion of the cranial-nerves; hyper-reflexia or other
signs of an involvement of the upper motor neurone;
sensory impairment; neoplastic disorder; positive
family history as to neuromuscular diseases; poly-
neuropathy.
Positive GM-1 antibodies or the presence of a con-
duction block were not an imperative condition.
Study design
Intravenous immunoglobulins were given at 2 g/kg/
bodyweight over 5 days every 4–6 weeks. Every patient
received at least two courses of IVIg. The average
number of treatment courses was eight (range 2–43).
Examination before and during IVIg treatment
Before treatment was started, the examination of the
cerebrospinal fluid was performed in all patients.
Magnetic resonance imaging (MRI) was performed to
exclude compression of the spinal cord and of the nerve
roots. MRI of the lumbar or cervical plexus was not
performed.
Previous to IVIg and before each treatment period a
detailed clinical, laboratory and neurophysiological
analysis by experienced neurologists was performed
using a standardized protocol. Muscle strength was
examined bilaterally according to the Medical Research
Council (MRC) Score [9]. Activities of daily living were
assessed by the Neuromuscular Symptom Score (NSS)
[10].
Patients were divided into a responder and non-
responder group after the second course of treatment:
Positive response was defined as a lasting improvement
in the NSS and/or in the neurological testing by
means of the MRC Score at the third examination
when compared with the first one. A non-response
was defined as a worsening of the NSS and/or of the
MRC-Score. In the NSS 14 items which describe
activities of daily living were taken into account (i.e. for
example opening of bottles or jam glasses, buttoning
up, binding shoelaces). Each item can be answered
in four ways: very restricted, moderately restricted,
little restricted or normal (points 0, 1, 2, 3). Besides a
worsening of the neurological status or appearance of
new symptoms such as upper motor signs, involvement
of cranial nerves, or of the respiratory system was
defined as a non-response.
For determining the patient’s global assessment of
the treatment, patients were asked about subjective
improvement after each treatment period, using a
semi-quantitative 3-point scale (worse, unchanged,
improved).
The electromyographic examination was performed
with concentric needle electrodes (Medtronic Keypoint,
Copenhagen, Denmark). Proximal as well as distal
muscles of each limb were examined. Pathological
spontaneous activity (PSA) was described semi-
quantitatively (0 ¼ absent to +++ ¼ profuse PSA).
Duration, amplitude and recruitment pattern of the
motor unit action potentials were recorded.
Motor neurography was performed with surface
stimulation and registration electrodes. The motor
nerve conduction velocities (NCV) and distal motor
latencies of the ulnar, median, radial, tibial and
peroneal nerve were bilaterally recorded in the proxi-
mal, middle, and distal part. NCV and compound
muscle action potentials (CMAP) were measured
(baseline to peak). Care was taken to perform all
measurements at 32°C. A conduction block was
defined as a reduction of the area under the CMAP
for more than 50% or a reduction of the amplitude
>50%. F-wave latencies and persistence were recor-
ded at the abductor pollicis brevis, abductor digiti
minimi, and extensor digitorum brevis muscle. To
exclude entrapment syndromes inching of the nerve
was performed at common entrapment sites (elbow,
head of the fibula).
Sensory NCV and sensory nerve action potentials
(SNAP) of the median, ulnar, superficial radial, and
sural nerves were registered on both sides.
Prior to treatment creatinekinase (CK) levels, GM1-
and asialo-GM1 antibody titres were examined.
Ó 2006 EFNS European Journal of Neurology 13, 135–140
 Prediction of response to IVIg treatment in patients with LMND 137

Measurements of GM1- and asialo-GM1 antibody titres Table 1 Characteristics of the patients before treatment

Antibody titres were measured by a commercial Number of patients

laboratory (Institute of Clinical Chemistry, Hospital Responders Non-responders
Lippe-Lemgo, Germany, K.H. Müller and Dr W. (%) (%)
Parameter P-value
Siede). A commercially available enzyme immunoassay
(Buhlmann Laboratories, Switzerland) was used. Sex
Male 13 (42) 18 (58) 0.4697
Female 5 (56) 4 (44)
Statistical analysis Weakness
Prox. and distal 6 (24) 19 (76) 0.0006
Statistical analysis was performed by a statistician distal 12 (80) 3 (20)
(H.B., co-author) who used the BioMedical Computer Creatinekinase (U/l)
Program (BMDP, Department of Biomathematics, >80 5 (21) 19 (79) 0.0002
<80 13 (81) 3 (19)
School of Medicine, University of California, Los GM1 antibody titre
Angeles, CA, USA). To compare clinical, laboratory Positive (IgM > 5950 BTU) 3 (100) 0 (0) 0.0424
and electrophysiological findings between the respond- Negative 7 (33) 14 (67)
ers and non-responders, the Kruskal–Wallis H-test for Low titre 8 (57) 6 (43) 0.0630
continuous variables (rank variance analysis) and the Acute or chronic denervation
Generalized denervation 0 (0) 10 (100) 0.0010
chi-squared test (multifactorial analysis of frequency (a non-paretic limb affected)
tables) for categorical variables were used. A P < 0.05 Local 11 (52) 10 (48)
was considered to be significant. No 7 (78) 2 (22) 0.0019
Conduction block
Yes 11 (65) 6 (35) 0.0313
Results No 7 (30) 16 (70)
Upper limb 9 (75) 3 (25)
The cerebrospinal fluid (CSF) was examined in all pa- Lower limb 1 (25) 3 (75)
tients. 33 patients showed a normal cerebrospinal fluid Upper and lower limb 2 (100) 0 (0)
(15 responders, 18 non-responders); six showed a pro-
tein elevation <100 mg/dl (three responders, three non-
responders); one non-responder showed a protein ele- distal and proximal weakness 76% did not profit. Of
vation of 126 mg/dl. Immunoglobulin synthesis in the the patients who suffered only from distal weakness
central nervous system or an increase of cell account of 20% did not respond (P ¼ 0.0006).
the CSF was not found. The mean levels of CK were significantly higher in
Sensory NCV and sensory nerve action potentials the non-responders (230 U/l) when compared with the
were normal in all patients. responders (86 U/l). Of the patients with an abnormal
Characteristics of the patients before treatment are CK-level (normal range <80 U/l) 21% responded to
shown in Table 1. In total 18 patients responded to treatment, whereas 81% of patients with normal CK-
IVIg treatment (42% responder) and 22 did not (58% levels responded to treatment (P ¼ 0.0002).
non-responder). Nineteen non-responders received Of all patients 7.5% had a high titre of GM1 anti-
more than two courses of IVIg. bodies (IgM > 5950 BTU). All these patients respon-
The age of onset was not different between both groups ded to therapy. Of the patients with negative antibody
(responder: 56.9 years, range 34–76, non-responder: titre 33% responded and 57% with low titre (P ¼
58.5 years, range 33–75). In our patients, the interval 0,0630). Asialo-GM1 antibodies were screened in 83%
between the onset of the clinical symptoms and the of the patients. Fifty-four per cent with positive asialo-
beginning of the treatment was at mean 4.8 years (range GM1 antibody titre and 56% with negative titre were
1–13 years). The duration of symptoms before the onset responders (P ¼ 0.9122).
of treatment was longer in responders (6.0 years, range Of all patients 25% showed electromyographic signs
1–11) than in non-responders (3.7 years, range 1–13 ). of denervation exceeding the clinical affected muscles.
Gender did not influence the response to treatment: All these patients were non-responders to IVIg. In the
58% of the men and 44% of the women did not profit group without generalization of electromyographic
(P ¼ 0.4697). changes (no PSA or local PSA only), 12 patients were
At the beginning of IVIg treatment, weakness was non-responders (40%). Of the patients without PSA
localized distally in 15 patients (two lower limbs, eight 78% responded to IVIg (P ¼ 0.0019).
upper limb, five upper and lower limb) and distally as A conduction block was found in 43% of all patients.
well as proximally in 25 patients. Of the patients with Sixty-five per cent of the patients with conduction

Ó 2006 EFNS European Journal of Neurology 13, 135–140

138 N. Strigl-Pill et al.

10 Responders
to the course of the disease. Five of them received
additional immunosuppressive therapy.
9 Non-responders
8
Number of patients

7 Discussion
6 In contrast to disorders such as ALS and spinal mus-
5 cular atrophy, MMN responds to immunomodulating
4 treatment with IVIg. Unfortunately, there are no widely
3
accepted criteria which permit a clear distinction be-
tween these disorders, especially in the early course of
2
the disease. Several groups have tried to define clinical,
1 laboratory, and neurophysiological criteria [11–13]. The
0 aim is to give the very expensive IVIg treatment to all
0 1 2 3 4 5
who need it, without wasting medical resources on
Score
those patients who will not respond to them.
Figure 1 Scoring of the responders and non-responders. Though most authors suggest that the presence of a
conduction block is mandatory for the diagnosis of
blocks and 30% of the patients without a conduction MMN, recent studies have raised concerns that by a
block responded to treatment (P ¼ 0,0313). Seventy- strict application of this criterion some treatable pa-
five per cent of the patients with a conduction block tients would be excluded from treatment [5,14]. High
exclusively of the upper limbs responded to therapy and titre ganglioside antibodies are rather specific but lack
only 25% of the patients with a conduction block of the sensitivity. It is currently estimated that only 50% of
lower limbs. Both patients with conduction blocks of patients with treatable motor neuropathy have these
upper and lower limbs profited. antibodies [15].
We computed a scoring system with the five param- Consequently, we performed a prospective study with
eters which differed significantly between responders 40 consecutive patients suffering from pure lower motor
and non-responders (CK level, localization of weak- neurone disease that received repeated courses of IVIg
ness, acute or chronic denervation of a limb without treatment. Inclusion criteria were based on clinical
clinical signs, GM1 antibody titre, conduction block). findings, only. The repeated response to treatment was
We gave one point for each of the following condi- used as a gold standard to distinguish between MMN
tions: normal CK level; distal weakness; electromyo- and other forms of MND. The purpose of the study was
graphic signs of denervation only in clinical affected to define a better set of criteria to predict the response
limbs; IgM-GM1 antibody titre >5950 BTU; existence to IVIg treatment.
of a conduction block. Therefore a maximal score of 5 In our study there was no single clinical, neuro-
could be achieved. physiological or laboratory parameter which allows to
No non-responder obtained more than two points select only those patients who might respond to treat-
and no responder less than two points. Seven non- ment. Nevertheless, there is one parameter which pre-
responders and five responders got two points (30% of dicts a non-response to treatment: patients with a
all patients) (Figure 1). generalization of electromyographic signs of denerva-
tion beyond the clinical involved site did not respond to
treatment. This indicates that in treatable multifocal
Follow-up of the non-responders
motor neurone disease, in contrast to spinal muscular
Five patients were lost to follow-up. One patient atrophies or ALS there is no subclinical denervation
showed progression of weakness and signs for the first preceding the onset of symptoms. To our knowledge
motor neurone. Two patients showed a progression of there are no data in the literature concerning the spatial
weakness without signs for the first motor neurone. distribution of electromyographic findings in this group
Seven patients died of respiratory failure. Seven pa- of patients. Most reports describe neurophysiological
tients showed neither progression nor improvement of changes in MMN, including an exhaustive discussion
the paresis. The diagnosis of them is still unclear. about the definition and relevance of conduction blocks
[13,16,17]. Certainly the existence of a definite con-
duction block is a major criterion for MMN; in our
Follow-up of the responders
study a significant correlation could be demonstrated
Four were lost to follow-up. In the remaining 14 between response to treatment and existence of a con-
responders the dosage of IVIg was adapted according duction block which was defined by restrictive criteria.

Ó 2006 EFNS European Journal of Neurology 13, 135–140

 Prediction of response to IVIg treatment in patients with LMND 139

Nevertheless, if we had used the presence of a con- Table 2 Conduction block or score: Comparison of hypothetical
duction block as main inclusion criterion for IVIg treatment decisions

treatment, 30% of all treatable patients (responders Hypothesis Hypothesis correct Hypothesis wrong
without conduction block) would have been remained
CB ¼ responder 27 (67.5%) Patients 13 (32.5%) Patients
untreated. On the other hand 35% of patients with a
Absent CB ¼ 11 Responders 6 Non-responders with CB
conduction block did not respond to treatment. non-responder 16 Non-responders 7 Responders without CB
In patients without conduction blocks, Ellis et al. [14]
found a response to treatment in four of 10 patients and Score 0–1 ¼ 28 (70%) Patients 12 (30%) Patients
non-responder 15 Non-responders 5 Responders: Score 2
Katz et al. [5] in three of six treated patients. Divergent
Score 3–5 ¼ 13 Responders 7 Non-responders: Score 2
to this, in studies with patients who have conduction responder
blocks about 30% of patients [12,18] did not respond to
treatment. Similar to our findings, this indicates that in
quite a considerable number of patients a conduction
potential responder remained untreated, if a score of
block is of limited value for the decision to treat with
above two points was used as criterion.
IVIg.
If we had used the criteria suggested by van den Berg-
Elevated CK levels make a success doubtful, but do
Vos (inclusion criterion: conduction block) 30% of
not exclude it. Of the patients with elevated CK levels
treatable patients would have been missed (seven
79% did not respond to treatment. The mean CK val-
responders of the 23 patients without conduction block)
ues differed significantly between responders and non-
and 35% non-responders with conduction block would
responders. Similar findings have been reported by van
have been treated (six non-responders of the 17 patients
den Berg-Vos et al. [12]. In this study, 30% of
with conduction block) (Table 2).
responders and 83% of non-responders had elevated
This indicates that the use of conduction block cri-
CK levels. Earlier studies have shown that especially
teria results in a considerable overlap between
patients with ALS show elevated CK levels [19].
responders and non-responders. Our score helps to re-
The only single parameter that highly correlated with
duce this grey zone. The number of responders, which
a positive response to treatment in our study was an
would not get treatment, because they do not have a
elevated GM1 antibody titre. All patients which had a
conduction block, can be reduced from 30%
high titre of GM1 antibodies responded to treatment.
(responders without conduction block) to zero (all
In clinical practice, the major problem is, that only a
responders got 2 or more points). The number of non-
minority of treatable patients show highly elevated
responders can be reduced from 35% (non-responders
antibody titres, whereas the majority has normal titres
with conduction block) to 28% (seven non-responders
or unspecific low titre. In our study only three patients
with 2 points in the group of patients with 2 or more
had high titres of GM1 antibodies, another 14 patients
points).
had unspecific low to moderate titres. Previous studies
This indicates, that our scoring is able to increase the
have shown a wide range in the number of patients with
number of treatable patients and to reduce the number
GM1 antibodies. Between 22% [20] and 79% of pa-
of non-responders without the risk of excessive costs by
tients [21] had elevated titres. According to Pestronk
unjustified treatment.
and Choksi [2], the rate of patients with positive titres
can be increased up to 85% by using a modified assay.
This finding was not confirmed by Carpo et al. [22]. Acknowledgements
In order to improve the detection rate of treatable
The study was supported by the Fordan Stiftung and by
patients without an excessive increase in costs due to
the Friedrich-Baur-Stiftung.
large number of non-responders, we propose a scoring
system that combines clinical, laboratory and neuro-
physiological data. By the use of this system all References
responders could be identified. Every patient which got 1. Parry GJ, Clarke S. Multifocal acquired demyelinating
2 or more points should be treated with IVIg. neuropathy masquerading as motor neuron disease.
Nevertheless a grey zone exists in our scoring system Muscle Nerve 1988; 11: 103–107.
at a score of 2 points. Twelve patients got 2 points. Five 2. Pestronk A, Choksi R. Multifocal motor neuropathy.
Serum IgM anti-GM1 ganglioside antibodies in most
of them were responders (28% of all responders), seven patients detected using covalent linkage of GM1 to ELI-
were non-responders (32% of all non-responders). SA plates. Neurology 1997; 49: 1289–1292.
However, all 15 patients with a score of 0 or 1 were 3. Niebroj-Dobosz I, Janik P, Kwiecinski H. Serum IgM
identified as non-responders and 72% of patients who anti-GM1 ganglioside antibodies in lower motor neuron
got 2 or more points responded. For that reason no syndromes. Eur J Neurol 2004; 11: 13–16.

Ó 2006 EFNS European Journal of Neurology 13, 135–140

140 N. Strigl-Pill et al.
4. Nobile-Orazio E, Meucci N, Barbieri S, et al. High-dose
intravenous immunoglobulin therapy in multifocal motor
neuropathy. Neurology 1993; 43: 537–544.
5. Katz JS, Barohn RJ, Kojan S, et al. Axonal multifocal
motor neuropathy without conduction block or other
features of demyelination. Neurology 2002; 58: 615–620.
6. Kaji R. Physiology of conduction block in multifocal
motor neuropathy and other demyelinating neuropathies.
Muscle Nerve 2003; 27: 285–296.
7. Taylor BV, Dyck PJB, Engelstad J, et al. Multifocal
motor neuropathy: pathologic alterations at the site of
conduction block. J Neuropathol Exp Neurol 2004; 63:
129–137.
8. Traynor JB, Codd MB, Corr B, et al. Clinical features of
amyotrophic lateral sclerosis according to the El Escorial
and Airline House diagnostic criteria. Arch Neurol 2000;
57: 1171–1176.
9. Medical Research Council. Aids to the Examination of the
Peripheral Nervous System. London, UK: Her Majesty’s
Stationery Office, 1976: 1–2.
10. Soueidan SA, Dalakas MC. Treatment of inclusion body
myositis with high-dose intravenous immunoglobulin.
Neurology 1993; 43: 876–879.
11. Taylor BV, Gross L, Windebank AJ. The sensitivity and
specificity of anti-GM1 antibody testing. Neurology 1996;
47: 951–955.
12. Berg-Vos RM, Franssen H, Wokke JH, et al. Multifocal
motor neuropathy: diagnostic criteria that predict the
response to immunoglobulin treatment. Ann Neurol 2000;
48: 919–926.
13. Olney RK, Lewis RA, Putnam TD, et al. Consensus cri-
teria for the diagnosis of multifocal motor neuropathy.
Muscle Nerve 2003; 27: 117–121.
14. Ellis CM, Leary S, Payan J, et al. Use of human intra-
venous immunoglobulin in lower motor neuron syn-
dromes. J Neurol Neurosurg Psychiatry 1999; 67: 15–19.
15. Bouche P, Moulonguet A, Younes-Chennoufi AB, et al.
Multifocal motor neuropathy with conduction block: a
study of 24 patients. J Neurol Neurosurg Psychiatry 1995;
59: 38–44.
16. Van den Berg LH, Franssen H, Wokke JH. The long-term
effect of intravenous immunoglobulin treatment in
multifocal motor neuropathy. Brain 1998; 121: 421–428.
17. Taylor BV, Wright RA, Harper CM, et al. Natural his-
tory of 46 patients with multifocal motor neuropathy with
conduction block. Muscle Nerve 2000; 23: 900–908.
18. Federico P, Zochodne DW, Hahn AF, et al. Multifocal
motor neuropathy improved by IVIg: randomized, dou-
ble-blind, placebo-controlled study. Neurology 2000; 55:
1256–1262.
19. Edmonds PJ, Ziegler DK. Diagnostic value of serum
creatine phosphokinase in motor neuron disease. South
Med J 1975; 68: 1388–1390.
20. Le Forestier N, Chassande B, Moulonguet A, et al.
Multifocal motor neuropathies with conduction blocks.
39 cases. Rev Neurol (Paris) 1997; 153: 579–586.
21. Kornberg AJ, Pestronk A. Chronic motor neuropathies:
diagnosis, therapy, and pathogenesis. Ann Neurol 1995;
37(Suppl. 1): S43–S50.
22. Carpo M, Allaria S, Scarlato G, et al. Marginally im-
proved detection of GM1 antibodies by Covalink ELISA
in multifocal motor neuropathy. Neurology 1999; 53:
2206–2207.
Ó 2006 EFNS European Journal of Neurology 13, 135–140