Gait & Posture 33 (2011) 556561

Contents lists available at ScienceDirect

Gait & Posture

journal homepage: www.elsevier.com/locate/gaitpost

Gait analysis may help to distinguish hereditary spastic paraplegia from

cerebral palsy
Sebastian I. Wolf a,*, Frank Braatz a, Dimitrios Metaxiotis a,b, Petra Armbrust a, Thomas Dreher a,
Leonhard Doderlein a,c, Ralf Mikut d
a

Department of Orthopaedic Surgery, University of Heidelberg, Schlierbacher Landstr. 200a, 69118 Heidelberg, Germany
Orthopaedic Department, Papageorgiou Hospital, Thessaloniki, Greece
Behandlungszentrum Aschau, Germany
d
Institute for Applied Computer Science, Karlsruhe Institute of Technology (KIT), Germany
b
c

A R T I C L E I N F O

A B S T R A C T

Article history:
Received 27 July 2010
Received in revised form 13 January 2011
Accepted 15 January 2011

Hereditary spastic paraplegia (HSP) designates a group of genetic disorders typically leading to spasticity
in the lower limbs and consequently to gait disorders. Although the symptoms are similar to those of
cerebral palsy (CP), the correct diagnosis is important for treatment recommendations as one condition
is progressive in nature whereas the other is not. Due to the heterogeneity of HSP, genetic testing is
complex and in some genetic forms still not possible. The aim of this study was, therefore, to investigate
if instrumented 3D-gait analysis could help distinguish between these two conditions.
The gait pattern of 29 patients with HSP was compared with that of 29 patients with CP who were
matched in age, sex, and the extent of gait disturbance and also to 29 typically developing subjects for
reference. More than 3000 gait parameters were evaluated for their relevance to classify patients into
diagnostic groups. Cluster analysis revealed that these gait features may classify only subgroups of
symptoms as the gait pattern is very heterogeneous within each diagnosis group. However, prolonged
hip extension, knee extension, and ankle plantar exion were identied as indicators for HSP. In addition,
large trunk tilt velocities appear unique in some cases of HSP. These indicators in gait pattern may
contribute in establishing the diagnosis of HSP, which is important in predicting outcome when planning
surgical treatment for functional improvements in these patients.
2011 Elsevier B.V. All rights reserved.

Keywords:
Hereditary spastic paraplegia
Gait analysis
Automated methods
Classication
Feature extraction
Pattern recognition

1. Introduction
Hereditary spastic paraplegia (HSP) designates a heterogeneous
group of genetic disorders of which not all can yet be diagnosed by
routine genetic testing [1]. In its pure form, HSP leads to
progressive spasticity in the lower limbs due to pyramidal tract
degeneration [2]. HSP is classied as complex when complicated
by other neurological signs such as ataxia, mental retardation,
dementia, extrapyramidal signs, visual dysfunction, or epilepsy [3].
Pure HSP represents around 70% of the cases and is typically
inherited in an autosomal dominant manner [3,4]. In most cases,
the phenotype includes slowly progressive spasticity in the lower
limbs, resulting in gait deciencies and loss of mobility starting in
the second decade after the onset of symptoms. Clinically, the
onset of this type of HSP can be from childhood through to late
adult life, whereas more than half of mutation carriers do not
develop symptoms until after the age of 30 years [5]. Although the

* Corresponding author. Tel.: +49 6221 966724.

E-mail address: sebastian.wolf@med.uni-heidelberg.de (S.I. Wolf).
0966-6362/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.gaitpost.2011.01.009

clinical ndings may vary, two groups of clinical types have been
described based on the age at presentation: type I before and type II
after 35 years of age [6]. With an onset in the rst few years of life,
combined with delayed motor milestones, the diagnosis is more
suggestive of cerebral palsy (CP), particularly if the clinical picture
is relatively static [5]. However, due to the progressive nature of
the clinical symptoms in HSP, these patients are typically treated
differently to patients with CP. Since genetic testing is not always
possible and both the clinical status and the patients history may
not always be sufcient to establish the diagnosis, a detailed
analysis of the gait pattern may help in the decision for an
appropriate treatment.
Unlike CP, where gait has been studied in detail using 3D
motion capture, only few studies have objectively described gait
function in HSP. Braschinsky et al. [7] reported on reduced walking
speed in 46 patients set in the context of reduced active hip exion
and abduction ROM as well as reduced ankle dorsiexion ROM,
which were determined clinically. Earlier, the group of Klebe found
a reduced ROM in knee exion and reduced walking speed due to
reduced stride length and cadence along with an increase in step
width in 22 adult patients (age 3561 years) when compared to

S.I. Wolf et al. / Gait & Posture 33 (2011) 556561

healthy controls [8]. The most detailed ndings so far were

reported in 15 children with HSP (aged 615 years) by comparing
their gait to that of 40 patients with CP and 20 healthy control
subjects with the aim to quantitatively assess differences in gait
between the two conditions [9]. Temporospatial and kinematic
parameters were found to be similar in the two patient groups but
knee exion at initial contact was particularly increased when
compared to the CP group. Both groups tended to hyperextend the
knee in mid-stance, however, a longer duration of hyperextension
was observed in patients with HSP.
These rst objective ndings on gait characteristics in HSP and
the clinical observation that some patients show a gait pattern
atypical for CP suggest that objective gait observation may help in
distinguishing the two conditions. Therefore, the aim of this
retrospective study was to identify pathology-related gait patterns
or characteristic features irrespective of the patient history. For
this, a systematic approach based on both gait feature computation
as described in [10] and on the use cluster analysis was employed.
Instrumented gait analysis could then be used both for a better
diagnostic work-up and for determining the best treatment
options for the patient.
2. Patients
Thirty-ve patients with clinical symptoms of HSP were seen
between 1996 and 2008 at the outpatient clinics of our hospital.
Instrumented 3D-clinical gait analysis (CGA) based on a conventional gait model [11] was conducted to establish further
treatment recommendations. Of these patients, 29 (aged 563
years, 8f/21m) were included in the study since (a) gait data were
reasonably reliable, i.e., data from ve or more trials were
measured and averaged, (b) video documentation was available,
and (c) the diagnosis of HSP was conrmed by either a documented
family history or by late age of onset (for HSP type II cases). The
gross motor function score (GMFCS, [12]) in eight cases was

557

classied as GMFCS 1, in another 18 patients as GMFCS 2, and in

three patients as GMFCS 3. On their rst visit, 24 subjects had not
previously had any operations whereas ve had received softtissue operations (calf muscles), among them two who also
received proximal procedures (adductors, hamstrings, rectus
femoris). More details about the subjects are given in Table 1.
Further, a group of 385 patients diagnosed with diplegic type CP
who presented for their rst visit in our hospital and who had also
undergone CGA were scanned for the same inclusion criteria as in
the HSP group. Sorted by age and sex, best matches to the HSP
group were manually picked in a blinded manner only by the
information on age, sex, and Gillette Gait Index (GGI) [13]. In this
group seven cases were classied as GMFCS 1, in another 16
patients as GMFCS 2, and in six patients as GMFCS 3. It turned out
(by accident) that, similarly to the HSP group, ve of these 29
patients had also undergone soft-tissue operations (calf muscles),
among them three who had also had proximal procedures,
including one case of bony correction.
Hence the two groups showed very similar clinical backgrounds
but with different diagnoses. For reference and testing reasons, a
third group of typically developing subjects (NORM) was selected
retrospectively out of the gait data base, aiming for best matches in
age and sex.
3. Methods
For the instrumented gait analyses a Vicon 370 motion capture system was used
during the years 19962001 and was then replaced by a Vicon 612 system applying
a conventional gait model [11]. For each patient, the data of at least ve strides of
different trials were averaged. To assess gait patterns, a previously developed
methodological modular framework was applied [10]. This framework formalizes
the processing steps of data selection, gait parameter calculation, and evaluation, as
well as classication according to the clinical problem. For these steps, several
mathematical methods were selected and the validity of the approach was tested by
applying it to the clinical problem of Botulinum Toxin-A treatment for spastic
equinus [10]. The methodology is only briey described here. An extended
summary can be found in the electronic appendix. Original time series, i.e., 3D

Table 1
Subject characteristics.
Triple

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

HSP group

CP group

Reference

Sex

Age

GGI

Walking aid

Previous surgeries

HSP type

Sex

Age

GGI

Walking aid

Previous surgeries

Sex

Age

m
f
m
m
m
f
m
m
f
f
m
m
m
m
m
m
m
m
m
f
f
m
m
m
m
m
m
f
f

5.6
6.2
8.2
8.7
9.0
9.1
9.5
9.9
11.9
12.1
14.0
14.2
14.5
15.3
15.7
16.4
17.6
19.9
21.2
29.0
34.9
36.4
36.6
37.5
37.7
38.7
46.7
52.3
63.5

72
132
131
360
305
277
1511
45
606
271
100
118
149
545
994
398
299
1135
235
182
148
408
220
169
260
326
505
990
180

No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Yes
No
No
No
No
No
No
No
No
Yes
Yes
No

No
No
No
No
No
No
No
No
No
Yes
No
No
No
No
No
Yes
No
No
No
Yes
No
No
No
Yes
No
Yes
No
No
No

I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
II
I
II
I
II
II
II

m
f
m
m
m
f
m
m
f
f
m
m
m
m
m
m
m
m
m
f
f
m
m
m
m
m
m
f
f

4.8
6.0
7.7
8.7
9.3
8.8
9.6
10.3
11.7
11.7
13.3
14.4
14.4
15.5
16.0
15.8
17.7
22.6
21.5
29.6
32.1
35.3
35.1
38.0
38.3
46.1
55.3
49.1
46.2

89
111
98
303
304
266
1584
137
581
279
128
189
235
526
1101
419
233
853
240
252
228
528
244
122
543
423
415
626
186

No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Yes
No
No
No
No
No
No
Yes
No
No
Yes
Yes

No
No
No
No
No
No
No
No
No
No
No
Yes
No
No
No
No
No
No
No
No
No
No
Yes
Yes
Yes
No
No
Yes
No

m
f
m
m
m
f
m
m
f
f
m
m
m
m
m
m
m
m
m
f
f
m
m
m
m
m
m
f
f

5.0
5.9
7.9
8.7
9.1
9.4
9.1
10.2
11.4
11.7
11.7
12.7
13.5
15.5
15.7
23.9
24.7
25.8
27.2
29.3
33.0
35.6
35.7
37.8
41.4
43.2
51.0
50.4
51.7

S.I. Wolf et al. / Gait & Posture 33 (2011) 556561

558

Table 2
List of the 10 most relevant gait parameters to distinguish the gait pattern of patients with HSP from the NORM using mutual information. TDP, time distance parameter; Str,
stride; St, stance; Sw, swing; MSw, mid swing.
Rank

Value

Time
series

Gait
phase

NORM

HSP

CP

Relevance
NORMHSP

p-Value
NORMHSP

Relevance
NORMCP

p-Value
NORMCP

1
2
3
4
5
6
7
8
9
10

ROM
ROM
TDP
ROM
ROM
ROM
MIN
ROM
TDP
ROM

Knee FlexExt
PelvicTilt
Double Support
Knee FlexExt
PelvicTilt Vel
Knee FlexExt
Knee FlexExt
PelvicTilt
Walking Speed
Knee FlexExt Vel

Sw
Str

54.2 (4.5)
2.8 (0.8)
20.7 (2.8)
29.4 (3.8)
0.4 (0.1)
55.5 (4.2)
3.9 (3.4)
2.5 (0.8)
1.3 (0.2)
6.6 (0.7)

26.1 (11.0)
8.9 (3.8)
31.6 (12.1)
10.0 (5.5)
1.2 (0.6)
38.4 (13.7)
21.4 (13.5)
8.1 (3.7)
0.9 (0.3)
4.1 (1.4)

23.4 (10.1)
9.3 (3.4)
33.5 (10.8)
8.0 (5.0)
1.1 (0.4)
34.9 (13.2)
27.7 (11.6)
8.5 (3.2)
0.8 (0.3)
3.8 (1.5)

0.587
0.580
0.519
0.485
0.476
0.456
0.454
0.454
0.453
0.445

1.89E
7.70E
1.19E
2.58E
3.49E
3.65E
2.36E
5.46E
4.78E
3.90E

0.628
0.745
0.561
0.540
0.524
0.500
0.603
0.606
0.560
0.476

2.72E
1.33E
1.98E
1.71E
7.64E
1.84E
8.39E
6.18E
7.91E
3.33E

MSw
Str
Str
Sw
St
Str

joint angles of pelvis, hip, knee, and ankle obtained by conventional, instrumented
3D gait analysis [11] and three derived time series, i.e., joint angle velocities, norm
distances, and norm distances for joint angle velocities to calculate single
features, i.e., scalar quantities of gait parameters. These parameters included range
of motion, mean, maximum, and minimum values of time series as well as the
temporal position of maximum and minimum values in the gait cycle or in
functional sub phases as dened by Perry [14]. In total, 3187 gait parameters,
including 19 additional scalar time distance parameters such as cadence and
walking speed, are generated. The feature set is very similar to [10], but without the
features from the stride-to-stride kinematic variance and with nine additional
time-distance parameters. To cover prolonged knee extension, the most successful
feature found in [9], the duration in the gait cycle with values smaller than zero (in
this case knee hyperextension) was systematically added as an additional
parameter for all joint angle and joint angle velocity time series. Driven by the
subjective clinical impression that trunk movement in some HSP patients appears
unique and distinct, 3D-data on trunk movements available in 18 (out of 29)
patients with HSP and 25 (out of 29) patients with CP were also analyzed via singlefeature computation in the same manner as described above.
The features were then ranked for relevance between zero (unable to separate
the different groups) and one (unique assignment) of the patient groups in question
using mutual information and a priori relevance. In other words, we assessed the
extent to which each parameter could suitably distinguish between gait patterns of
HSP and CP (and physiologic gait). Finally, when also taking cross correlations
between parameters into account by subsequently excluding variables when they
are less relevant than another highly correlated one, a ranking list with a set of
largely independent gait parameters is produced, characterizing the given clinical
(classication) problem. In this contribution, these lists were produced pairwise for
the groups HSP, CP, and NORM. Statistical signicance with a signicance level of
p < 0.05 was determined for each parameter using the non-parametric Wilcoxon
rank-sum test and Bonferroni correction for multiple testing.
Furthermore, unsupervised cluster analysis methods were applied to account for
heterogeneous gait patterns within each patient group. Here, a standard Fuzzy CMeans algorithm with raw data-based time series, a normalization of time series to
mean values of zero and a standard deviation of one, a fuzzier m = 1, and a xed
number of clusters C = 4, 5, and 6 were chosen [15]. Sagittal plane kinematics
including joint angles and joint angle velocities of pelvis, hip, knee, and ankle was
used to classify the sets of 4, 5, and 6 clusters across patient groups. The analysis was
done with the open-source Matlab toolbox Gait-CAD [19,20].

4. Results
In Table 2, a ranking list is given for the 10 most relevant gait
parameters classifying the group of patients with HSP from agematched reference normal subjects (NORM), corresponding values

19
16
13
20
14
11
15
15
13
16

20
19
17
20
19
14
20
20
18
17

for the classication HSPNORM, and p-values for statistical

differences between groups (Wilcoxon rank test as many
parameters are non-normally distributed). Furthermore, mean
group averages and standard deviations are given for all three
groups. In Table 3 these gait data are ranked for the direct
classication between pathological conditions (HSPCP).
The relevance of the gait parameters in Table 2 is around 0.5
both for the HSP and the CP group when classifying gait with
respect to the NORM group, with slightly higher values for CP, i.e.,
this group can be slightly better distinguished from the NORM
group. All of these parameters were highly signicant in the
statistical comparison. In fact, 1257 parameters out of the total
number of 3187 parameters investigated reached statistical
signicance with a signicance level a of 0.05 (p < a = 0.05) using
the non-parametric Wilcoxon rank-sum test (equivalent to the
MannWhitney U-test) and Bonferroni correction with a/
3187 = 0.000016 for the comparison HSP vs. NORM and 1379/
3187 parameters for CP vs. NORM, respectively. Interestingly, the
relevance ranking list for the classication HSPNORM is almost
identical to that of CPNORM. Eight out of the top 10 parameters
coincide, indicating already that the gait patterns of HSP and CP are
very similar. In contrast, the relevance of the top 10 parameters
classifying HSPCP (Table 3) is very small (0.1), and many of
these parameters were not statistically signicant upon comparison. With values p  0.003, the most promising candidates for
classication are the durations of hip and ankle (dorsi) exion < 0
(i.e., hip extension and ankle plantar exion) in the complete stride
and in stance, respectively (ranked 46 in Table 3). However, none
of the 3187 parameters investigated reached statistical signicance after Bonferroni correction for the comparison HSP vs. CP.
Even when combining several gait parameters to generate a rule
for classication as described in [10], the classication was very
weak (not shown).
When using cluster methods on sagittal gait kinematics, the
best results were obtained when allowing a classication of ve
clusters (Fig. 1). This classication closely resembled those
patterns described by Sutherland and Davids [16] as typical gait

Table 3
List of the 10 most relevant gait parameters distinguishing the gait patterns of patients with HSP from those with CP. Values: TDP, time distance parameter; T < 0 = time
duration in which the parameter is smaller then 0; ROM, range of motion. Gait phases: Str, stride; St, stance; Sw, swing; TSt, terminal stance.
Rank

Value

Time series

1
2
3
4
5
6
7
8
9
10

TDP
TDP
Mean
T<0
T<0
T<0
Mean
ROM
MIN
Mean

Stride Time
Step Time
PelvicTilt
Hip FlexExt
Hip FlexExt
DorsiPlanFlex
PelvicTilt
Knee Rotation
Pelvic Rotation
Hip FlexExt

Gait phase

NORM

HSP

CP

Relevance HCPCP

p-Value HSPCP

Str
St
Str
Str
St
Str
Sw
TSt

1.0 (0.1)
0.6 (0.1)
11.1 (4.3)
20.0 (8.6)
22.6 (10.6)
39.9 (11.9)
11.2 (4.3)
17.2 (4.6)
4.5 (2.5)
0.7 (5.6)

1.3 (0.7)
0.7 (0.3)
17.5 (6.7)
12.7 (11.9)
13.7 (13.3)
55.5 (36.3)
17.4 (6.6)
14.5 (5.2)
7.3 (5.9)
7.2 (11.2)

1.2 (0.4)
0.6 (0.2)
16.8 (7.3)
6.4 (10.7)
7.1 (12.5)
34.9 (32.7)
16.7 (7.4)
11.8 (5.5)
7.4 (9.6)
11.8 (9.9)

0.105
0.104
0.082
0.080
0.077
0.073
0.067
0.061
0.060
0.060

0.793
0.529
0.722
0.003
0.003
0.002
0.657
0.002
0.976
0.004

PelvicTilt

S.I. Wolf et al. / Gait & Posture 33 (2011) 556561

Crouch

40
20

20

HipFlexExt
DorsiPlanFlex KneeFlexExt

80
60
40
20
0

50

50

100

100

50

50

100

100

80
60
40
20
0
0

50

100

20
0
-20
-40
-60
-80

50

100

100

50

100

50

50

100

100

80
60
40
20
0
0

50

100

20
0
-20
-40
-60
-80
0
50
100
Gait cycle [%]

0
0

60
40
20
0
-20

60
40
20
0
-20

50

100

50

100

50

100

80
60
40
20
0
0

50

100

20
0
-20
-40
-60
-80
0
50
100
Gait cycle [%]

Norm like

40
20

0
50

80
60
40
20
0

20
0
-20
-40
-60
-80
0
50
100
Gait cycle [%]

20

0
60
40
20
0
-20

Jump Knee

40

0
0

60
40
20
0
-20

Stiff Knee

40
20

0
0

60
40
20
0
-20

Recurvatum

40

559

20
0
-20
-40
-60
-80
0
50
100
Gait cycle [%]

0
50
100
Gait cycle [%]

Fig. 1. Clusters of sagittal gait kinematics found across all subjects (HSP, CP, and NORM).

abnormalities in CP and which were therefore used for naming the

clusters. Type 1 is characterized by excessive knee exion
throughout the gait cycle, known as crouch gait. Type 2 shows
prominent knee extension or hyperextension in mid-stance
(recurvatum gait) whereas type 3 shows insufcient knee
extension with limited knee range of motion (stiff knee gait).
Finally, type 4 (jump knee gait) is characterized by increased knee
exion at loading response and almost normal knee function later
in the gait cycle as in type 5 (norm like gait). It should be stressed
that the results shown in Fig. 1 were received independently of any
clinical knowledge by taking all sagittal plane kinematics into
account. Interestingly, the classication did not change/improve
signicantly when using only knee exion or ankle kinematics for
classication (not shown).
The relative distribution of each diagnosis group into each gait
pattern cluster (Fig. 2) was unique only for the reference group
with a perfect classication for normal (58/58). In pathological
60
55
50

Jump Knee

Jump Knee

number of legs

45
40
35

Stiff Knee

30
Stiff Knee

25
20

Norm like

Recurvatum

15
10
5
0

Crouch

Crouch

Norm like

Norm like
CP

HSP

Norm

Fig. 2. Relative distribution of each patient group into gait pattern clusters.

conditions, cases contributed to any of the ve gait patterns

regardless of the type of pathology (HSP or CP). However, the
contribution of each pathological condition to the norm-like
cluster was very small (3 and 4 of 58, respectively) and comprised
the mildest cases. Hence, HSP showed in principle the same
phenotypes as CP, the only noticeable difference being that only
one leg of the CP group shows a recurvatum knee as compared to
14 legs in the HSP group.
The analysis of 3D-data of the trunk revealed that higher
maximum trunk tilt velocities found in patients with HSP most
signicantly (p = 2.3E 005) differed between diagnosis groups
HSP and CP.
5. Discussion
HSP is associated with progressive impairment of gait;
however, information on functional ability and especially the gait
pattern in these patients is still limited and mainly derived from
subjective observational gait analysis. Nevertheless, a number of
researchers have attempted to describe the gait problems in
patients with HSP [7,17]. The largest cross-sectional study was
performed by Erichsen et al. [18], who reported on 194 subjects
with HSP in southeast Norway. Among these, 31% were classied
as walkers with mild symptoms, 32% as walkers but unable to run,
25% as walkers restricted to walking aids, and 11% as wheel-chairdependent. Unfortunately, none of the studies provide any
information about specic gait deviations.
To our knowledge the only two studies that examined gait
patterns objectively are the ones by Klebe et al. [8], comparing 22
adult patients with healthy controls, and Cimolin et al. [9], who
compared 15 children with HSP to a group of 40 patients with CP
and to 20 healthy controls. The latter study reports on prolonged
knee hyperextension in stance, a gait feature also found to be
signicantly different between diagnosis groups in our patients
(p < 0.002 without Bonferroni correction; feature not shown in
Table 3 since it is ranked 32 with relevance 0.047). More prominent

560

S.I. Wolf et al. / Gait & Posture 33 (2011) 556561

and relevant, however, are the differences with respect to

prolonged times of hip extension and ankle plantar exion (comp.
Table 3 features ranked 46). Cimolin and co-workers speculated
that a motor strategy different to that in CP, a compensatory knee
stabilization strategy, was responsible for prolonged knee
hyperextension in HSP. Knee hyperextension would not be
accompanied by ankle plantar exion and hence without the
presence of a plantar exion/knee extension couple. In contrast, we
found knee hyperextension linked to plantar exion in HSP (comp.
Fig. 1; second column recurvatum). We suspect that this link was
overlooked when only group mean averages were examined and
thus the presence of functional phenotypes within the pathological
condition was neglected. Instead, the relative distribution of each
patient group into gait pattern clusters with only one out of 58 legs
classied as recurvatum (comp. Fig. 2) reects the clinical nding
that patients with CP who have not previously received surgical
treatment are unlikely to show knee hyperextension. It has to be
noted that some HSP patients in this recurvatum cluster did not
show true knee hyperextension but a rather increased knee
extension to be described by almost full extension in stance in
combination with reduced knee exion in swing. Hence, the mere
nding of knee hyperextension or increased knee extension as
such in patients with no history of surgical treatment would
indicate a high probability of HSP. The cause for this nding in HSP
may primarily represent weakness rather than spasticity. This is in
contrast to (typically post-surgical) patients with CP who show
dominance of quadriceps spasticity rather than weakness. This
reasoning is corroborated by the nding of extended time of hip
extension (comp. Table 3, feature ranked 5). The predominance of
either weakness or spasticity may indeed relate to the different
cause of the two pathologies, i.e., the lesion of the immature central
nervous system in CP and the group of genetic conditions leading
to problems with axonal transport (HSP). However, since any
rigorous evidence about the interdependency of spasticity and
weakness is still missing this reasoning remains speculative. More
detailed investigations of other rare diseases with similar
functional ndings such as the Segawa-Syndrome (idiopathic
dystony) or the Curschmann-Steinert-Syndrome (myotonic dystrophy) may help in elucidating the relation of weakness and
spasticity with the primary cause.
Furthermore, we tried to assess the subjective impression of a
unique pelvistrunk coordination pattern in HSP observed on
video. By adding trunk motion to the formalized algorithm
developed for classifying lower limb kinematics [10], we found
signicantly increased peak trunk tilt velocities in HSP as
compared to CP. Four patients with HSP showed a noticeable,
quick forward and backward movement of the trunk at the end of
loading response and at stance-swing transition which was not
present to this extent in any of our patients with CP. There might
also be a link between this nding in trunk motion and prolonged
knee extension in stance. However, this was not examined in
further statistical detail because it only applied to a minority of
subjects within this group and because trunk data were not
available for all of them.
The cluster analysis revealed that the gait pattern alone could
not fully classify the patients into diagnosis groups as the gait
pattern was heterogeneous within each diagnosis group.
However, for subgroups of symptoms, namely for recurvatum
gait, specic characteristics can be found which indicate the
presence of HSP. These are prolonged hip extension in stance,
knee hyperextension, and ankle plantar exion in the lower limb
followed by large trunk tilt velocities in loading response and
stance/swing transition. Clinically these may be important
ndings when other diagnostic factors, i.e., familial history
and genetic testing are not reassuring enough for the diagnosis
of HSP.

As HSP is a rare group of diseases, sample sizes are typically

small. With a sample size of 29 subjects, we are far from claiming
to present a complete clinical picture but with an age range from
5 to 63 years a large variety of subject characteristics was denitely
covered. Effort has been spent in generating adequate samples for
comparing HSP to CP and to physiologic gait. By scanning the gait
database for diplegic patients with CP with the same inclusion
criteria except for diagnosis, we expect to have an adequate
comparison. We tried to minimize the number of limiting factors in
this comparison, such as inhomogeneity between groups due to
different functional levels, age, sex, and previous surgeries, by
picking matched pairs. None of these factors differed signicantly
between the groups. However, the inclusion of a wide age range
and a wide spectrum of involvement in each group in what are
already heterogeneous conditions may have introduced an
additional and possibly even articial challenge in differentiating
between these groups. The comparison of more homogenous
subgroups would be possible but with limited statistical power
and the risk of additional bias effects. Namely, age and body height
might be inuencing factors as with increasing age and body
height weakness becomes an increasing disadvantage for the
biomechanics of walking.
6. Conclusion
The clinical resemblance of HSP and CP is reected on the strong
similarity of gait patterns found by using a rigorous CGA approach
in the largest sample of HSP in the literature. The gait pattern alone
cannot fully classify the patients into diagnosis groups as each
group is heterogeneous. However, prolonged hip extension in
stance, knee hyperextension, and ankle plantar exion as well as
large trunk tilt velocities in loading response and stance/swing
transition are indicative of HSP. These diagnostic indicators may
well be clinically relevant when surgical treatment is in question as
HSP is progressive in nature whereas CP is not. A further study
monitoring the factor of time both with and without orthopaedic
intervention in this patient group is underway.
Conict of interest statement
None of the authors had any nancial or personal relationships
with other people or organizations that could inappropriately bias
their work.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.gaitpost.2011.01.009.
References
[1] Fink JK. Hereditary spastic paraplegia. Curr Neurol Neurosci Rep
2006;6(1):6576.
[2] Depienne C, Stevanin G, Brice A, Durr A. Hereditary spastic paraplegias: an
update. Curr Opin Neurol 2007;20(6):67480.
[3] Harding AE. Classication of the hereditary ataxias and paraplegias. Lancet
1983;1(8334):11515.
[4] Fink JK. Advances in the hereditary spastic paraplegias. Exp Neurol
2003;184(Suppl. 1):S10610.
[5] Salinas S, Proukakis C, Crosby A, Warner TT. Hereditary spastic paraplegia:
clinical features
and
pathogenetic
mechanisms.
Lancet
Neurol
2008;7(12):112738.
[6] Harding AE. Hereditary pure spastic paraplegia: a clinical and genetic study
of 22 families. J Neurol Neurosurg Psychiatry 1981;44(10):87183.
[7] Braschinsky M, Parts K, Maamagi H, Gross-Paju K, Haldre S. Functional
assessment of lower extremities in hereditary spastic paraplegia. Arch Phys
Med Rehabil 2009;90(11):188790.
[8] Klebe S, Stolze H, Kopper F, Lorenz D, Wenzelburger R, Volkmann J, et al. Gait
analysis of sporadic and hereditary spastic paraplegia. J Neurol
2004;251(5):5718.

S.I. Wolf et al. / Gait & Posture 33 (2011) 556561

[9] Cimolin V, Piccinini L, DAngelo MG, Turconi AC, Berti M, Crivellini M, et al. Are
patients with hereditary spastic paraplegia different from patients with spastic diplegia during walking? Gait evaluation using 3D gait analysis. Funct
Neurol 2007;22(1):238.
[10] Wolf S, Loose T, Schablowski M, Doderlein L, Rupp R, Gerner HJ, et al.
Automated feature assessment in instrumented gait analysis. Gait Posture
2006;23(3):3318.
[11] Kadaba MP, Ramakrishnan HK, Wootten ME. Measurement of lower extremity
kinematics during level walking. J Orthop Res 1990;8(3):38392.
[12] Palisano RJ, Hanna SE, Rosenbaum PL, Russell DJ, Walter SD, Wood EP, et al.
Validation of a model of gross motor function for children with cerebral palsy.
Phys Ther 2000;80(10):97485.
[13] Schutte LM, Narayanan U, Stout JL, Selber P, Gage JR, Schwartz MH. An index
for quantifying deviations from normal gait. Gait Posture 2000;11(1):2531.
[14] Perry JP. Gait analysis: normal and pathological function. Thorofare, NJ:
Slack; 1992 .

561

[15] Rozumalski A, Schwartz MH. Crouch gait patterns dened using k-means
cluster analysis are related to underlying clinical pathology. Gait Posture
2009;30(2):15560.
[16] Sutherland DH, Davids JR. Common gait abnormalities of the knee in cerebral
palsy. Clin Orthop Relat Res 1993;288:13947.
[17] von Fellenberg J, Paternotte C, Prudhomme JF, Weissenbach J, Hazan J,
Burgunder JM. Klinische und molekulargenetische Analyse bei vier schweitzerischen Familien mit der reinen Form der hereditaren spastischen Spinalparalyse. Schweiz Med Wochenschr 1998;128(26):104350.
[18] Erichsen AK, Koht J, Stray-Pedersen A, Abdelnoor M, Tallaksen CM. Prevalence
of hereditary ataxia and spastic paraplegia in southeast Norway: a populationbased study. Brain 2009;132(Pt 6):157788.
[19] Mikut R, Burmeister O, Braun S, Reischl M. The open source Matlab toolbox
Gait-CAD and its application to bioelectric signal processing. In: Proceedings of
DGBMT-Workshop Biosignal Analysis, Potsdam; 2008. p. 10911.
[20] http://sourceforge.net/projects/gait-cad/.