Research in Developmental Disabilities 32 (2011) 1957–1964

Contents lists available at ScienceDirect

Research in Developmental Disabilities

Arm swing during walking at different speeds in children withCerebral Palsy and typically developing children

Pieter Meyns a,*, Leen Van Gestel b, Firas Massaad a, Kaat Desloovere b,c,Guy Molenaers c, Jacques Duysens a,d

a Department of Biomedical Kinesiology, Faculty of Kinesiology and Rehabilitation Sciences, K.U.Leuven, Heverlee, Belgiumb Department of Rehabilitation Sciences, Faculty of Kinesiology and Rehabilitation Sciences, K.U.Leuven, Heverlee, Belgiumc Clinical Motion Analysis Laboratory, CERM, University Hospital Leuven, Leuven, Belgiumd Department of Research, Development and Education, Sint Maartenskliniek, Nijmegen, The Netherlands

A R T I C L E I N F O

Article history:

Received 29 March 2011

Accepted 30 March 2011

Available online 4 May 2011

Keywords:

CP

Gait

Arm movement

Rehabilitation

Walking speed

Speed increment

A B S T R A C T

Children with Cerebral Palsy (CP) have difficulties walking at a normal or high speed. It is

known that arm movements play an important role to achieve higher walking speeds in

healthy subjects. However, the role played by arm movements while walking at different

speeds has received no attention in children with CP. Therefore we investigated the use of

arm movements at two walking speeds for children with diplegia (DI) and hemiplegia (HE)

as compared to typically developing (TD) children. Arm and leg swing lengths were

determined in 11 HE children and 15 DI children and compared to 24 TD children using 3D

gait analysis at their preferred and ‘‘as fast as possible’’ walking speeds. We found that TD

children increased walking speed more than both CP groups. HE children showed larger

arm swings on the non-hemiplegic compared to the hemiplegic side for both walking

speeds. In contrast to TD or DI children, the HE group did not show an increase in arm

swing length with increasing walking speed. Their leg swing length was larger on the non-

hemiplegic than on the hemiplegic side but only at the preferred walking speed. The DI

children exhibited smaller leg swings at both walking speeds. Since arm swing is used both

by DI (to increase speed) and by HE children (to compensate for the reduced movement on

the affected side) it is argued that these movements are important and should be allowed

(or even encouraged) in gait training procedures (such as treadmill training).

� 2011 Elsevier Ltd. All rights reserved.

1. Introduction

An increase in gait speed can be achieved through several strategies. The legs can produce more propulsion either byusing stronger activation of ankle plantar flexors or by increasing the hip extensor moment (Neptune, Sasaki, & Kautz, 2008).After brain lesions, such as in patients with Cerebral Palsy (CP) and Traumatic Brain Injury (TBI), walking speed is oftenreduced compared to unimpaired controls (Abel & Damiano, 1996; McFadyen, Swaine, Dumas, & Durand, 2003). In TBI thisreduced speed is mainly attributed to a reduced ankle power generation, since patients preferentially increase hip jointpower generation to walk faster (Williams, Morris, Schache, & McCrory, 2010). The role played by arm movements, however,has received less attention. In typical subjects there is a transition from a passive to a more active arm swing as walking

* Corresponding author at: Department of Biomedical Kinesiology, K.U.Leuven, Tervuursevest 101 Box 1501, B-3001 Heverlee, Belgium.

Tel.: +32 16 32 90 65; fax: +32 16 32 91 97.

E-mail address: Pieter.meyns@faber.kuleuven (P. Meyns).

0891-4222/$ – see front matter � 2011 Elsevier Ltd. All rights reserved.

doi:10.1016/j.ridd.2011.03.029

P. Meyns et al. / Research in Developmental Disabilities 32 (2011) 1957–19641958

speed increases from slow to normal velocities (Wagenaar & Van Emmerik, 1994). With further increasing speeds, arm swingamplitudes increase (Perry, 1992). This progressive recruitment of arm movements indicates that they are likely to play animportant role to achieve higher gait speeds.

In patients with CP, however, arm swing can be affected by paralysis or spasticity, associated with deficient equilibrium. Itcan therefore be questioned how such deficits would affect the ability of patients to increase walking speed. In strokepatients, previous studies showed that the hemiplegic arm swings less than the non-hemiplegic arm during treadmillwalking at several walking velocities (Ford, Wagenaar, & Newell, 2007b) and with cadence increments (Ford, Wagenaar, &Newell, 2007c). Furthermore, studies on adults with hemiplegia revealed that an impaired arm swing can be a limiting factorfor the walking speed. Inversely, treatment of the spastic arm with Botulinum toxin improved walking characteristics inthese patients (Hirsch et al., 2005; Esquenazi, Mayer, & Garreta, 2008). Closely related is the question of whether armmovements should be included in gait rehabilitation with robotic devices (DGO’s; Driven Gait Orthosis). It is currentlyunknown to what extent arm movements should be allowed, or even encouraged, in this task specific training especiallysince some of these devices do not allow arm movements (Meyer-Heim et al., 2007). Moreover, the use of arm movementshas been found to improve clinical mobility test scores in healthy participants compared to when arm movements were notallowed (Milosevic, McConville, & Masani, 2011). This emphasizes the potential for the use of arm movements in gaitrehabilitation.

In children with CP, a variety of topographical lesions and locomotor patterns may emerge depending on brain lesionlocation and secondary musculoskeletal deformities (Forssberg, 1992). Studies on the effect of walking speed on armmovements in children with CP are lacking, since the focus of gait studies has been mainly on leg movements. In childrenwith diplegia the legs are more involved than the arms (Dabney, Lipton, & Miller, 1997). Hence it can be hypothesized thatthey should have a lower preferred speed. However, it is expected that they can fully use their arm movements when askedto walk faster. In contrast, in children with hemiplegia one side of the body is more involved than the other (Dabney et al.,1997). Therefore, we expect a deficit in arm movement on the affected side. It is hypothesized that such deficits can becompensated for by the use of the least affected side but that these compensations may significantly limit the amount ofspeed increase.

In the current study we aimed to investigate the use of arm movements during walking at two walking speeds (preferredand ‘‘as fast as possible’’) for children with diplegia and hemiplegia in terms of the amplitude of the swing movement. Theresults can have important implications for gait rehabilitation of CP children.

2. Methods

2.1. Participants

Twenty-six children with CP (age range 4–12 years) and 24 Typically Developing (TD) children (age range 5–12 years)participated. The CP group included eleven children with hemiplegia and fifteen children with diplegia. The children with CPincluded in the study were ambulant (without walking aids), diagnosed with predominantly spastic type of CP, had sufficientcooperation to follow verbal instructions, did not undergo Botulinum Toxin treatment within the past six months, and didnot receive lower limb surgery. Children with CP were recruited from the Laboratory of Clinical Motion Analysis of theUniversity Hospital Leuven (U.Z.Leuven). All experiments were approved by the local ethical committee and informed,written consent was obtained from the subjects’ parents. Characteristics of the three groups (TD, diplegia, hemiplegia) arepresented in Table 1.

The three groups did not differ with respect to age or weight. However, children with hemiplegia were slightly smallerthan TD children and children with diplegia.

2.2. Protocol

All children underwent a clinical investigation by an experienced physical therapist. Subjects walked at a self-selectedspeed along the 10 m walkway in a straight line. After three successful trials, the subjects walked as fast as possible along the

Table 1

Characteristics of TD children and children with CP.

TD Diplegia Hemiplegia

N 24 15 11

Gender (M/F) 12/12 11/4 8/3

GMFCS (I/II/III) – 8/6/1a 7/4/0

Age (years) 9.40� 2.16 9.96� 2.47 7.83� 2.98

Weight (kg) 31.72� 8.64 31.54� 13.36 23.87� 7.57

Height (cm) 137.83� 14.24 133.61� 18.69 122.1� 15.4*

Note that age, weight and height are presented as follows: mean� standard deviation. N = number of subjects, M/F = male/female, GMFCS = Gross Motor

Function Classification System.a 1 subject with a GMFCS level III was included because this subject was able to complete the walking trials of the experiment without walking aids.* Significantly different from TD (one way analysis of variance: F = 3.69; p = 0.03).

P. Meyns et al. / Research in Developmental Disabilities 32 (2011) 1957–1964 1959

walkway (without running). A successful trial included four consecutive foot strikes with full marker-visibility. Trials wereexcluded when excessive movements were made unrelated to the walking task. The subject was granted some practice trialsfor each condition. Three successful trials for each condition were used for further analysis.

Three-dimensional full-body kinematic data were recorded at 100 Hz with an eight camera Vicon system (OxfordMetrics, Oxford, UK) with the total body PlugInGait marker set. The marker coordinates were filtered and smoothed usingWoltring’s quintic spline routine with a predicted mean-squared error of 15 (Woltring, 1986). Workstation (ViconWorkstation 5.2beta 20, Oxford Metrics) and Polygon software (Version 3.1, Oxford Metrics) were used to define the gaitcycles, to estimate the internal joint centers, and to determine the spatio-temporal parameters. The most affected side inchildren with CP was determined as the side on which the leg had the highest spasticity scores (Modified AshworthScale).

2.3. Measures

The time-courses of the displacement of the hand and toe marker, projected on the sagittal (Fig. 1A left), transverseand frontal plane, were recorded. These measures were used to analyze the arm and leg swing length, determined as thedifference of maximum and minimum position of the respective marker along the x-axis (anterior–posterior direction;Fig. 1B), the y-axis (medio-lateral) and the z-axis (top-bottom). To correct for ongoing body motions, displacement of themarker on the spinous process of the fifth lumbar vertebra was subtracted from the displacements of the hand and toemarker. Arm and leg swing length were normalized by dividing them by the participants height (Hof, 1996). Additionally,the within-subject variability of the swing length was expressed by the coefficient of variation (CV). Since arm swinglength can be affected by trunk displacement, we also compared the trunk rotations between the groups. The trunkrotation to the most affected/non dominant and least affected/dominant side was determined as the angle between theline connecting the two shoulders and the line connecting the two superior anterior iliac spines of the pelvis in thetransverse plane.

To check whether arm swing length was a representative measure of the degree of swinging motion of the arm(especially in HE children), we analyzed the amplitude of the angle between the upper arm segment and the vertical[()TD$FIG]

Fig. 1. Schematic presentation of the displacement of the hand marker along the anterior–posterior direction (A left) and angular displacement of the upper

arm segment with respect to the vertical (A right). Typical examples are presented of the displacement of the hand marker along the anterior–posterior

direction normalized to body size in a child with diplegia (DI) and a typically developing child (TD) during a gait cycle (B). Note that the arm swing length is

depicted as the arrows between the horizontal lines.

P. Meyns et al. / Research in Developmental Disabilities 32 (2011) 1957–19641960

axis (Fig. 1A right). Comparison of the two methods showed that the same HE children who had increased arm swinglengths on the least affected side also presented with increased upper arm elevation angles on this side compared to themost affected side. This was considered a confirmation of the validity of the swing length measure to describe armswing amplitude.

2.4. Statistics

For the comparison of age, weight and height between TD children and children with CP a one way analysis of variancewas used. A repeated measures ANOVA with group and walking condition as factors was used to compare walking speedsbetween the three groups.

To compare the three groups for each variable, we used a general linear model with group as a factor, and two repeatedmeasures factors (walking condition, side of body). Since the walking speeds differed between the groups, we includedactual walking speed as a covariate in our analysis. The walking velocities of both speed conditions were highly correlated(r = 0.778, p< 0.001). Therefore only the walking speeds of preferred walking were used as a covariate. For the comparison ofthe swing lengths along the three different axes an additional repeated measure factor (direction) was used. Tukey’s post hoccomparisons were systematically applied. Statistical significance was set at a = 0.05.

3. Results

3.1. Walking speed

In the self-selected speed (FW) condition, DI children walked significantly slower than TD children (Fig. 2; mean� SD;0.94 m/s� 0.24 versus 1.20 m/s� 0.16, respectively, p = 0.013, 27.8% decrease). No significant differences in walking speedbetween the other groups were found.

In the ‘‘walking as fast as possible’’ (FFW) condition, TD children showed a significantly higher walking speed compared tothe DI and HE children (1.95 m/s� 0.17 versus 1.41 m/s� 0.41, p< 0.001 and 1.67 m/s� 0.18, respectively, p = 0.007). Nosignificant differences between the CP groups were found. All groups were able to increase their walking speed significantly forthe FFW condition, but the TD children showed a much larger increase than the other groups (increase in speed; TD: 64.2%; DI:50.0%; HE: 51.8%; p< 0.001).

3.2. Swing lengths

The comparison of the arm swing lengths along the three different axes yielded a significant higher swing length in theanterior–posterior direction compared to the medio-lateral and the top-bottom directions for the three groups (Fig. 3; all:p< 0.001). Since we have found substantially larger movements in the anterior–posterior direction, only the swing lengthsin that direction will be considered in the rest of this section.

The first analysis was based on the mean swing length for both legs and both arms in the three groups (Fig. 4A).The leg swing length of the DI group was significantly smaller compared to TD and HE children (DI: 0.40� 0.07 versus TD:

0.49� 0.04, p< 0.001; and HE: 0.47� 0.04, p< 0.001). This clear difference between groups was not found for arm swing length(DI: 0.19� 0.07, TD: 0.22� 0.09, HE: 0.20� 0.09; group: p = 0.87).

Secondly, the three groups were compared for the two arms and legs separately (Fig. 4B). For the legs, HE children showeda significant larger swing length on the least affected side compared to the most affected side in the FW condition (Fig. 4B;5.5% increase; 0.47� 0.03 versus 0.44� 0.03, respectively, p< 0.001). For the arms, the HE group also showed a significantlylarger swing length on the least affected side compared to the most affected side (Fig. 4B; 104.0% increase; 0.27� 0.11 versus0.13� 0.07, respectively, p< 0.001). For the other groups the arm and leg swing length did not show significant side differences.

[()TD$FIG]

Fig. 2. Comparison of walking speeds. Walking speed of TD children (empty bars), DI children (gray bars) and HE children (black bars) at a self-selected

speed condition (FW) and a ‘‘walking as fast as possible’’ condition (FFW). Bars: Mean. Whiskers: SD. Horizontal lines: statistically significant differences

(Tukey’s post hoc).

[()TD$FIG]

Fig. 3. Comparison of arm swing lengths (normalized to body size) in the different directions. Normalized arm swing length averaged for side of the body

and walking condition of typically developing children (TD: empty bars), children with diplegia (DI: gray bars) and children with hemiplegia (HE: black

bars) in the anterior–posterior direction (A-P), the medio-lateral direction (M-L) and the top-bottom direction (T-B). Bars: Mean. Whiskers: SD. Horizontal

lines: statistically significant differences (Tukey’s post hoc).

P. Meyns et al. / Research in Developmental Disabilities 32 (2011) 1957–1964 1961

Thirdly, the effect of the increase of walking speed on arm and leg swing length was examined. When walking faster, TDchildren increased their leg swing length significantly more than the CP groups (Fig. 4B; TD-FW: 0.45� 0.04 FFW: 0.52� 0.04,p< 0.001, 14.8% increase; DI-FW: 0.39� 0.06 FFW: 0.41� 0.07, p = 0.33, 5.6% increase; HE-FW: 0.46� 0.03 FFW: 0.49� 0.05,p = 0.08, 7.6% increase). The arm swing length also increased with speed but differently for the various groups. It increasedsignificantly more for TD and DI children than for the HE children (Fig. 4B; TD-FW: 0.16� 0.07 FFW: 0.27� 0.10, p> 0.001, 65.1%increase; DI-FW: 0.16� 0.07 FFW: 0.21� 0.08, p = 0.034, 30.0% increase; HE-FW: 0.19� 0.08 FFW: 0.20� 0.09, p = 0.99, 5.7%increase).

Finally, the effect of the increase of walking speed in the three groups was examined for the two legs and arms separately.For the legs, the HE children showed side differences for leg swing length at the preferred speed but not when the walkingspeed was increased (Fig. 4B; FFW: most affected: 0.49� 0.05 versus least affected: 0.49� 0.05, p = 0.71). For the arms, the HEgroup retained a significantly larger swing length on the least affected side compared to the most affected side when walkingfaster (Fig. 4B; 0.27� 0.10 versus 0.13� 0.09, respectively, p< 0.001).

In contrast, the variability (as expressed in CV) of the arm and leg swing length was not significantly altered when walkingspeed increased. No statistically significant differences were found for the CV of either the arm or leg swing length betweenthe three groups.

[()TD$FIG]

Fig. 4. Comparison of swing lengths normalized to body size. (A) Normalized arm and leg swing length averaged for side of the body of typically developing

children (TD: empty bars), children with diplegia (DI: gray bars) and children with hemiplegia (HE: black bars) averaged for walking conditions. (B)

Normalized arm and leg swing length for the most affected (or non dominant) side of the body (Most Aff/Non Dom: empty bars) and the least affected (or

dominant) side of the body (Least Aff/Dom: striped bars) for the three groups (TD, DI and HE) at the preferred (FW: left) and ‘‘as fast as possible’’ (FFW: right)

walking speed. Bars: Mean. Whiskers: SD. Horizontal lines: statistically significant differences (Tukey’s post hoc).

P. Meyns et al. / Research in Developmental Disabilities 32 (2011) 1957–19641962

3.3. Leg length and cadence

To address whether the leg swing asymmetry in children with HE was related to individual leg length differences,correlations were determined between the leg swing difference (most versus least affected) and the percentage of leg lengthdifference (most versus least affected). No significant correlations were found (FW: p = 0.93; FFW: p = 0.45).

For cadence, no significant differences were found between the groups (TD: 151.3 steps/min� 15.8, DI: 146.5 steps/min� 27.2, HE: 158.8 steps/min� 17.9; group: p = 0.31). Only a main effect of speed was found, which indicates that all groupsincreased their cadence for the FFW condition (FW: 129.17 steps/min� 15.52 versus FFW: 173.83 steps/min� 25.99, p< 0.001,increase of 34.6%). Additionally, correlations were made between the percentage of difference in cadence between the FW andFFW condition and percentage of difference in leg swing length between the FW and FFW condition for the most affected/nondominant and least affected/dominant side. Only in TD children significant negative correlations were found for both sides of thebody (non dominant: r =�0.665, p< 0.001; dominant: r =�0.515, p = 0.012). For TD children this indicates that in order to walk asfast as possible, a high increase in cadence coincides with a low increase in leg swing length and vice versa.

3.4. Trunk rotation

We compared trunk rotations since it could be argued that children with CP show larger trunk movements (which mayinfluence amplitude of arm swing) than TD children. Although TD children seemed to show smaller trunk rotations for bothspeed conditions, they were not statistically different from CP children (TD: 11.018� 5.13 versus DI: 13.558� 5.88; HE:13.468� 4.84; group: p = 0.061).

Finally, correlations were made between trunk rotation and arm swing length for both sides of the body in children withhemiplegia. Significant correlations were found between the trunk rotation to the least affected side and the least affectedarm swing for both walking speeds (FW: r = 0.601, p = 0.047; FFW: r = 0.660, p = 0.027), indicating that large trunk rotationsto the least affected side coincide with large arm swings on the least affected side.

4. Discussion

The present study investigated the arm movements in CP gait. Children with hemiplegia are characterized by weakness,spasticity and lack of motor control on the hemiplegic side. If arm movements are important to compensate for the paresis,one would expect the least affected arm to produce larger arms swings. This was indeed observed as the arm swing on theleast affected side was, on average, 53.25% larger than the arm swing of TD children (0.26 versus 0.17, Fig. 4B). In addition, thehemiplegic arm swing was 22.9% smaller than in TD children (0.13 versus 0.16, Fig. 4B). These two effects resulted in an armswing on the least affected side that was about double the amplitude of the involved arm swing. Similarly, previous studies instroke patients also showed that the arm swing amplitude on the non paretic side was also about double the amplitude onthe paretic side (for the ‘‘no instruction’’ conditions) (Ford et al., 2007b, 2007c).

Children with hemiplegia seemed to involve the trunk as well to aid in this compensatory arm swing, because trunkrotations towards the least affected side correlated with arm swing lengths on the least affected side. The resultingasymmetry in the arm movement was substantially larger than the one seen in the legs. The swing of the non hemiplegic legwas 5.5% longer than the one of the hemiplegic leg. This difference in leg swing length, however, does not seem clinicallyrelevant and might be ascribed to the fact that the currently evaluated children with hemiplegia were mildly affected. Theasymmetry in swing length difference between arms and legs could have been the result of the compensation by the armswing (and the effects of load constraint for the legs). However, one should also consider the fact that the legs may be lessaffected than the arms in children with hemiplegia (Dabney et al., 1997).

As compared to TD children, children with hemiplegia in this study used their arms much less while speeding up (only5.7% increase). In contrast, TD children exhibited an increase of 65.1% in arm swing length. The limitation in the maximalwalking speed in the children with CP goes hand in hand with the constraint to increase the arm swing length. Therefore, it ispossible that the children with hemiplegia were already at their full capacity to compensate for their hemiplegic side duringthe self selected speed. This is in contrast to literature on stroke patients, who did show an increase in arm swing on the non-hemiplegic side when asked to walk faster (Ford et al., 2007b; Stephenson, Lamontagne, & De Serres, 2009).

In children with diplegia the legs are more involved than the arms. Therefore we assumed they would walk at a slowerpreferred speed compared to TD children. This was indeed observed since they walked 27.8% slower and their leg swinglength was, on average, 14.9% smaller compared to the control subjects (0.39 versus 0.45). This limitation in walking speedwas the result of smaller leg swing lengths induced by the lower extremity impairment. Despite their bilateral lower limbimpairment, we hypothesized that these children with diplegia were still able to fully use their arms when walking faster.This was confirmed since, in contrast to children with hemiplegia, the children with diplegia were able to significantlyincrease their arm swing length (30% increase) with an increment in walking speed. However, as expected, this increase wasnot sufficient to bring them to speeds comparable to TD children.

When walking as fast as possible, the increments in arm swing (TD: 65.1%, DI: 30% and HE: 5.7%) went in parallel withrelatively small increases in leg swing length (TD: 14.8%, DI: 5.6% and HE: 7.6%). In contrast, the cadence increased for allgroups about 34%. This indicates that children with CP gained walking speed mainly by increasing their cadence. TD children,in contrast, increased their speed either by increasing their cadence or their leg swing length. This pattern was not clear for

P. Meyns et al. / Research in Developmental Disabilities 32 (2011) 1957–1964 1963

CP children, since no clear relationship was found between the difference in cadence and difference in leg swing length withan increased walking speed. The observation that CP children, especially those with diplegia, increased their arm swinglength when walking faster, supports the notion that arm movements might be used to increase walking speed in thesechildren. Children with diplegia and children with hemiplegia are thus able to increase their walking speed to the samedegree by using different strategies. Children with diplegia use their arm swing more and their leg swing less compared tochildren with hemiplegia in order to increase their gait speed. Interestingly, the typical asymmetry in leg swing disappearedin children with hemiplegia when walking faster. This means that they could increase their leg swing length on the mostaffected side to a similar size as the least affected leg swing length. This is clearly different from evidence in stroke (Roerdink,Lamoth, Kwakkel, van Wieringen, & Beek, 2007) and it may have advantageous implications for the gait rehabilitation in CP.

The evaluated group of children with CP were quite mildly affected (mostly GMFCS level I). Hence it may be expected thatthe present results underestimate the effects seen in a more general population of CP. Furthermore, in the current study theamount of arm movement was determined as swing length in the sagittal plane only. This was based on the finding that thelargest arm movements were present in the anterior–posterior direction in the currently evaluated children. Previousstudies investigating the arm movements often documented joint angles (8) (Romkes et al., 2007). This was not possible forthe present population, since we encountered gimbal-lock (Doorenbosch, Harlaar, & Veeger, 2003) problems for several ofthe currently tested children. However, it was possible to determine upper arm elevation angles and these results were in fullagreement with those found for arm swing length. Nevertheless both measures were in the sagittal plane, which wasjustified for the present population but may be a simplification for other groups that are more affected. Romkes andcolleagues, for example, found increased movements of the arm in the frontal plane for children with diplegia (Romkes et al.,2007). Similarly, they observed increased movements of the thorax in the transverse plane as well as in the sagittal andfrontal planes. The difference in results between both studies can be ascribed to the difference in the two patient groups. Thechildren with diplegia in Romkes’ study were older and taller and they might be differently affected (the GMFCS level was notreported in their study) compared to the currently evaluated children. The impact of the trunk movements on the armmovements during gait is relatively unknown. Huang et al. found that patients with low back pain show an alteredcoordination pattern between thorax rotations and arm swing compared to unimpaired controls (Huang et al., 2011).Despite this change in coordination, they did not find differences in coordination between the arms and legs and arm swingamplitude between the two groups. Therefore, further research is required to investigate the coordination between the trunkand the upper limb movements in children with CP.

The current results can yield promising implications for rehabilitation. Stroke patients show a self-selected comfortableand a ‘‘as fast as possible’’ walking speed that is lower compared to healthy adults (Roerdink et al., 2007). In the current studythe comfortable walking speed of children with diplegia was also lower compared to TD children. However children withhemiplegia walked as fast as TD children. Similarly as in studies on stroke, the ‘‘as fast as possible’’ walking speed of controlsubjects was higher compared to both groups of CP children. We suggest that children with CP might be partly hindered fromwalking faster due to the impairment in the coordination between the arms and legs. This hypothesis is supported by a studyon adults with hemiplegia. Esquenazi et al. (2008) demonstrated that the walking speed of these patients increased bytreating the elbow flexor spasticity with botulinum toxin. Additionally, Hirsch et al. showed that injection of botulinum toxinin the spastic upper limb improved the stride time of the paretic leg in all hemiplegic stroke participants tested (Hirsch et al.,2005). This suggests that spasticity in the upper limb may be an important limiting factor to increase the walking speed inpersons with hemiplegia. Unfortunately there was no separate test of arm spasticity in the current study. In the future itwould be of interest to evaluate the contribution of elements such as spasticity and muscle weakness to the observed armmovements. The results can help to determine whether arm swing movements should be included in a gait rehabilitationprogram. In other pathologies, such as spinal cord injury (Behrman & Harkema, 2000) and stroke (Stephenson, De Serres, &Lamontagne, 2010; Stephenson et al., 2009), it has been claimed that the addition of arm movements can improve gaitrehabilitation. Inversely, preventing the arms to move during walking induces negative changes in gait characteristics (Eke-Okoro, Gregoric, & Larsson, 1997) and coordination patterns (Ford, Wagenaar, & Newell, 2007a). Whether these effects arealso of importance in robotic assistance for gait rehabilitation of CP children remains an open question.

Acknowledgements

This project was supported by a grant from ‘bijzonder onderzoeksfonds’ K.U.Leuven [OT/08/034]. LVG received a PhDfellowship of the Research Foundation Flanders (FWO).

References

Abel, M. F., & Damiano, D. L. (1996). Strategies for increasing walking speed in diplegic cerebral palsy. Journal of Pediatric Orthopaedics – Part B, 16, 753–758.Behrman, A. L., & Harkema, S. J. (2000). Locomotor training after human spinal cord injury: A series of case studies. Physical Therapy, 80, 688–700.Dabney, K. W., Lipton, G. E., & Miller, F. (1997). Cerebral palsy. Current Opinion in Pediatrics, 9, 81–88.Doorenbosch, C. A. M., Harlaar, J., & Veeger, D. (2003). The globe system: An unambiguous description of shoulder positions in daily life movements. Journal of

Rehabilitation Research and Development, 40, 147–155.Eke-Okoro, S. T., Gregoric, M., & Larsson, L. E. (1997). Alterations in gait resulting from deliberate changes of arm-swing amplitude and phase. Clinical Biomechanics

(Bristol, Avon), 12, 516–521.Esquenazi, A., Mayer, N., & Garreta, R. (2008). Influence of botulinum toxin type A treatment of elbow flexor spasticity on hemiparetic gait. American Journal of

Physical Medicine and Rehabilitation, 87, 305–310.

P. Meyns et al. / Research in Developmental Disabilities 32 (2011) 1957–19641964

Ford, M. P., Wagenaar, R. C., & Newell, K. M. (2007a). Arm constraint and walking in healthy adults. Gait and Posture, 26, 135–141.Ford, M. P., Wagenaar, R. C., & Newell, K. M. (2007b). Phase manipulation and walking in stroke. Journal of Neurologic Physical Therapy, 31, 85–91.Ford, M. P., Wagenaar, R. C., & Newell, K. M. (2007c). The effects of auditory rhythms and instruction on walking patterns in individuals post stroke. Gait and

Posture, 26, 150–155.Forssberg, H. (1992). Evolution of plantigrade gait – Is there a neuronal correlate. Developmental Medicine and Child Neurology, 34, 920–925.Hirsch, M. A., Westhoff, B., Toole, T., Haupenthal, S., Krauspe, R., & Hefter, H. (2005). Association between botulinum toxin injection into the arm and changes in

gait in adults after stroke. Movement Disorders, 20, 1014–1020.Hof, A. L. (1996). Scaling gait data to body size. Gait & Posture, 4, 222–223.Huang, Y. P., Bruijn, S. M., Lin, J. H., Meijer, O. G., Wu, W. H., Abbasi-Bafghi, H., et al. (2011). Gait adaptations in low back pain patients with lumbar disc herniation:

Trunk coordination and arm swing. European Spine Journal, 20, 491–499.McFadyen, B. J., Swaine, B., Dumas, D., & Durand, A. (2003). Residual effects of a traumatic brain injury on locomotor capacity: A first study of spatiotemporal

patterns during unobstructed and obstructed walking. Journal of Head Trauma Rehabilitation, 18, 512–525.Meyer-Heim, A., Borggraefe, I., Ammann-Reiffer, C., Berweck, S., Sennhauser, F. H., Colombo, G., et al. (2007). Feasibility of robotic-assisted locomotor training in

children with central gait impairment. Developmental Medicine and Child Neurology, 49, 900–906.Milosevic, M., McConville, K. M., & Masani, K. (2011). Arm movement improves performance in clinical balance and mobility tests. Gait and Posture, 33, 507–509.Neptune, R. R., Sasaki, K., & Kautz, S. A. (2008). The effect of walking speed on muscle function and mechanical energetics. Gait & Posture, 28, 135–143.Perry, J. (1992). Gait analysis: Normal and pathological function. Thorofare, NJ: SLACK Incorporated.Roerdink, M., Lamoth, C. J., Kwakkel, G., van Wieringen, P. C., & Beek, P. J. (2007). Gait coordination after stroke: Benefits of acoustically paced treadmill walking.

Physical Therapy, 87, 1009–1022.Romkes, J., Peeters, W., Oosterom, A. M., Molenaar, S., Bakels, I., & Brunner, R. (2007). Evaluating upper body movements during gait in healthy children and

children with diplegic cerebral palsy. Journal of Pediatric Orthopaedics – Part B, 16, 175–180.Stephenson, J. L., De Serres, S. J., & Lamontagne, A. (2010). The effect of arm movements on the lower limb during gait after a stroke. Gait and Posture, 31, 109–115.Stephenson, J. L., Lamontagne, A., & De Serres, S. J. (2009). The coordination of upper and lower limb movements during gait in healthy and stroke individuals. Gait

and Posture, 29, 11–16.Wagenaar, R. C., & Van Emmerik, R. E. A. (1994). Dynamics of pathological gait. Human Movement Science, 13, 441–471.Williams, G., Morris, M. E., Schache, A., & McCrory, P. R. (2010). People preferentially increase hip joint power generation to walk faster following traumatic brain

injury. Neurorehabilitation and Neural Repair, 24, 550–558.Woltring, H. J. (1986). A FORTRAN package for generalized, cross-validatory spline smoothing and differentiation. Advanced Engineering Software, 8, 104–113.