fn

. uv

Gait & Posture 40 (2014) 346351

Contents lists available at ScienceDirect

Gait & P

journa l homepage: www.e l seDepartment of Rehabilitation Medicine, MOVE Research Institute Amsterdam, VU University Medical Center, Amsterdam, the NetherlandsfKU Leuven Department of Mechanical Engineering, Leuven, Belgium

A R T I C L E I N F O

Article history:Received 17 February 2014Received in revised form 8 April 2014Accepted 29 April 2014

Keywords:Cerebral palsySpasticityHypertoniaTorqueMuscle modelingBotulinum toxin-A

A B S T R A C T

Clinical assessment of spasticity is compromised by the difculty to distinguish neural from non-neuralcomponents of increased joint torque. Quantifying the contributions of each of these components iscrucial to optimize the selection of anti-spasticity treatments such as botulinum toxin (BTX). The aim ofthis study was to compare different biomechanical parameters that quantify the neural contribution toankle joint torque measured during manually-applied passive stretches to the gastrocsoleus in childrenwith spastic cerebral palsy (CP). The gastrocsoleus of 53 children with CP (10.9 3.7 y; females n = 14;bilateral/unilateral involvement n = 28/25; Gross Motor Functional Classication Score IIV) and 10 age-matched typically developing (TD) children were assessed using a manually-applied, instrumentedspasticity assessment. Joint angle characteristics, root mean square electromyography and joint torquewere simultaneously recorded during passive stretches at increasing velocities. From the CP cohort, 10muscles were re-assessed for between-session reliability and 19 muscles were re-assessed 6 weeks post-BTX. A parameter related to mechanical work, containing both neural and non-neural components, wascompared to newly developed parameters that were based on the modeling of passive stiffness andviscosity. The difference between modeled and measured response provided a quantication of theneural component. Both types of parameters were reliable (ICC > 0.95) and distinguished TD from spasticmuscles (p < 0.001). However, only the newly developed parameters signicantly decreased post-BTX(p = 0.012). Identifying the neural and non-neural contributions to increased joint torque allows for thedevelopment of individually tailored tone management.

2014 Elsevier B.V. All rights reserved.

1. Introduction

Common clinical assessment of spasticity in children withcerebral palsy (CP) is based on manipulation of the joint to feel theresistance in a passively stretched muscle. In 1954, Tardieu andcolleagues emphasized the importance of differentiating betweendifferent causes of this increased resistance, or hypertonia [1].According to the currently prevailing denition of spasticity, anincrease in resistance during passive muscle stretch is termedspasticity when there is an accompanying velocity-dependentpathological stretch reex activation resulting in muscle activity

that resists or stops the motion [2]. In the absence of muscleactivation, all other excessive increase in resistance during stretchis thought to be caused by passive stiffness and viscosity due toalterations of intra- and extracellular muscle, soft-tissue, and jointstructures. Therefore, different components contribute to thefeeling of increased resistance in passively stretched muscle:neural and non-neural components. Direct quantication of thedifferent components in a clinical setting is highly relevantallowing for comprehensive tone assessment.

Tone-reducing medication, such as botulinum toxin-A (BTX)targets the neural component by blocking the release ofacetylcholine at the cholinergic nerve terminals, which preventsthe muscle from contracting. This treatment does not work whenthe increased resistance is of non-neural origin [3]. Methods totreat passive stiffness include casting, orthotic management, orwhen xed contractures arise, orthopedic surgery. Therefore, to

* Corresponding author at: Clinical Motion Analysis Laboratory, UniversityHospital Leuven, Weligerveld 1, Pellenberg, 3212, Belgium.Tel.: +32 16341295/38024.

http://dx.doi.org/10.1016/j.gaitpost.2014.04.2070966-6362/ 2014 Elsevier B.V. All rights reserved.Identication of the neural component oapplied spasticity assessments in childre

L. Bar-On a,b,*, K. Desloovere a,b, G. Molenaers a,c,d, JaUniversity Hospital Pellenberg, Clinical Motion Analysis Laboratory, University Hospital, LebKU Leuven Department of Rehabilitation Sciences, Leuven, BelgiumcKU Leuven Department of Development and Regeneration, Leuven, BelgiumdUniversity Hospital Pellenberg, Department of Orthopedics, Leuven, Belgium torque during manually- with cerebral palsy

Harlaar e, T. Kindt b, E. Aertbelin f

en, Belgium

osture

evier .com/ loca te /gai tpost

problems that impeded assessment; BTX injections within 6months prior to rst testing; previous lower-limb orthopedic orneuro-surgery. Age-matched TD children acted as a control group.The hospitals ethical committee approved the protocol(B32220072814) and all childrens parents signed an informedconsent.

To assess between-session reliability, a subgroup of childrenwith CP underwent a repeated assessment (including replacementof all sensors) after a two hour rest interval in which no treatmentwas administered. As part of an individually-dened, multileveltreatment, a second subgroup of children with CP were addition-ally measured 48 weeks after intramuscular BTX injections(Allergan, UK) in the gastrocnemius under short anesthesia. Afterinjections, the children underwent lower-leg casting for 10 days,intensive rehabilitation, and orthotic management (day and night).

In children with unilateral CP, only the affected side was tested.In children with bilateral involvement, if time allowed, both sideswere tested. If not, the most involved side was tested as dened bythe clinical spasticity scales [4,15].

2.2. Experimental procedure

L. Bar-On et al. / Gait & Posture 40 (2014) 346351 347provide the appropriate treatment to children with CP, it isimperative to differentiate and quantify the components.

The clinical test proposed by Tardieu, and its modications [4]attempt to differentiate components by comparing the range ofmotion (ROM) during a slow muscle stretch (R1 angle) to the catchangle during a fast stretch (R2). However, muscle activation duringslow stretch has been reported in some spastic muscles [5,6].Without verifying whether the muscle is inactive, the validity of R1is uncertain. Additionally, inaccuracies when manually determin-ing R2 have been reported [5] and its reliability is furthercompromised since the velocity of the fast stretch is not measured.Only an assessment that simultaneously collects electro physio-logical and biomechanical signals during passive muscle stretchhas the potential to differentiate the inuence of differentcomponents in a valid and reliable way. This approach is used inresearch settings using motor-driven devices [79] that permitstandardization of position displacements [7] or of the appliedforce [9]. For example, using a torque-motor that appliedsinusoidal movements at a constant force, Lakie et al. [9] usedthe peak resonance frequency as a measure of linear stiffness andviscosity in the wrist. On the other hand, by controllingdisplacement position, De Vlugt et al. [7] modeled the contributionof different non-linear components to measured ankle torque.However, to what extent clinically assessed spasticity can bereplicated by a motor-driven device is questionable. Rabita et al.[10] have shown that fewer stretch-reexes are elicited whenspastic muscles are stretched by a robot, than by an examiner.

A different group of evaluation techniques make use of thestraight forward clinical application combined with a quantitativeapproach, and are referred to as instrumented manual techniques[11,12]. These methods replicate a clinical spasticity test by havingan examiner apply muscle stretches while simultaneouslycollecting synchronized electromyography (EMG), kinematics,and/or joint torque. By examining the change in signals withincreasing muscle lengthening velocity, parameters that quantifyspasticity have been developed and have been shown to beapplicable and valid in clinical settings [6,11]. However, in thesestudies, the velocity-dependent effects from both neural and non-neural components are reected in the parameters that quantifyjoint torque. To differentiate between the components, furthermuscle modeling is required.

The non-linear behavior of passive stiffness and viscosity havebeen well described in healthy and hemiplegic subjects [7,13].However, to the best of our knowledge, these models have rarelybeen applied to data from an instrumented manual spasticityassessment [14]. The aim of this study is to quantify the amount ofneural contribution to the joint torque measured during stretchesof the gastrocsoleus in children with spastic CP. To achieve this, wemodel the non-neural components of passive stiffness andviscosity on data collected during stretches at increasing velocities.We then assume that the difference between modeled andmeasured response represents the neural component. Specically,we hypothesized that: (1) the neural component will have goodbetween-session reliability; (2) all components will be higher inchildren with CP than in typically developing (TD) children; and (3)in comparison to a previously-described parameter [11], the neuralcomponent will be more sensitive to the effect of BTX treatment.

2. Method

2.1. Participants

Children with spastic CP aged 518 years were recruited fromthe University Hospital Leuven. Exclusion criteria were: presenceof ataxia or dystonia; ankles with xed varus or valgus deformitieshindering pure sagittal plane passive ankle motion; cognitiveAll assessments were performed by the same trained assessoras detailed in [11]. Joint motion was tracked using two inertialmeasurement units, joint forces and torques were measured usinga 6 dof load-cell, and surface EMG (sEMG) was collected from thelateral gastrocnemius and tibialis anterior (Fig. 1). The subjectswere asked to remain relaxed throughout the measurement. Theankle joint was moved through the full ROM, at low velocity during5 s, at a medium velocity (1 s), and nally at high velocity,performed as fast as possible. At each velocity, four repetitionswere carried out with an interval of 7 s.

2.3. Data analysis

2.3.1. Data processingData visualization and analyses were performed in MATLAB.

To estimate joint angles, a Kalman smoother [16] was applied onthe inertial measurement unit data. Average maximum angularvelocity (Vmax) was calculated per velocity trial. Using measuredsegment-lengths and moment-arms, the net internal ankle jointtorque was calculated from the measured external forces andmoments as outlined in [11]. The EMG onset in the lateral

Fig. 1. Test starting position and direction of stretch (white arrow). Muscle activitywas measured with surface electromyography (1); joint-angle characteristics withinertial measurement units (2); and torque using a force-sensor (3) attached to afoot orthosis. The mass of the foot plus orthosis is considered negligible. Theresulting ankle joint torque was calculated after compensation for movement innon-perpendicular directions and the moments exerted on the handle (see [11] formore detail).

gastrocnemius was dened according to [17] and used to conrmthe presence or absence of muscle activation during stretches.Stretch repetitions were excluded when performed out-of-plane(see Supplement 1 in [11]), at inconsistent velocities betweenrepetitions within a velocity trial (difference >20/s), and in caseof poor quality sEMG (low signal-to-noise ratio or obviousartefacts). In children with CP, stretches were additionallyexcluded when there was no EMG onset detected during highvelocity stretches.

2.3.2. Outcome parametersThe following model to describe healthy muscle passive

stiffness and viscosity was approximated from de Vlugt et al. [7]:

T 1 bdudt

ekuu0 T0 (1)

where T is the total predicted internal joint torque; b, a viscositycoefcient; k, a passive stiffness coefcient; u, angular position; u0,angular position offset; and T0, offset torque. In contrast to theoriginal model, the model is represented in angular positions andtorques rather than muscle lengths and forces, i.e., assuming thateffects of a changing lever arm are negligible. The model was ttedto a region starting at the time of Vmax to the time at 90% ROM.Based on data visualization, in this region, only the agonist, ratherthan antagonist muscle was elongated, while the inuence of endROM discomfort was excluded. An optimal t of the model to themeasurement was computed using the LevenbergMarquardtalgorithm [18]. Fig. 2 shows the measured and modeled torque

ROM) of the squared deviation between T and Tm divided by the timeinterval between tVmax and t90%ROM:

Average model deviation

1t90%ROM tVmax

Z t90%ROMtVmax

T Tm2dts

(2)

Average work-deviation was dened as the integral from theangular position at Vmax (uVmax ) to the angular position at 90% ROM(u90%ROM) of the absolute deviation between T and Tm divided bythe position interval between uVmax and u90%ROM.

Average work deviation 1u90%ROM uVmax

Z u90%ROMuVmax

jT Tmjdu (3)

This parameter corresponded to the amount of work exchangedbetween assessor and subject during the motion.

Additionally, work was calculated as the integral of Tm overangular position from uVmax to u90%ROM to compare the modelparameters to previous literature [11]:

Work Z u90%ROMuVmax

Tmudu (4)

Average work over the four repetitions at both high and lowvelocity was calculated. The difference of these two was dened aswork-change, see also [11].

2.4. Statistics

-

tc

348 L. Bar-On et al. / Gait & Posture 40 (2014) 346351during one low- and one high-velocity stretch in a CP and a TDchild. However, rather than modeling individual stretches [7], datafrom multiple stretches at all velocities were used to t a model perassessed muscle providing more information on which to estimatethe distinct model components (Fig. 3).

To capture how the measured torque Tm deviated from themodeled torque T, the following two parameters were developed:average model deviation, was dened as the square root of theintegral from the time at Vmax (tVmax ) to the time at 90% ROM (t90%

-30 0

5

10

15

-15 0

5

10

15

-40 -2 0 0 20 400

5

10

15

-10 0 10 20 30 400

5

10

15

Torq

ue (N

m)

Torq

ue (N

m)

Torq

ue (N

m)

Torq

ue (N

m)

Joint ang le(deg)

Joint ang le(deg)

I II

III IV

Fig. 2. Measured and predicted torque during low (I, II) and high (III, IV) velocity strestretch trial were used to create the modelled torque.Model-deviation, work-deviation and work-change were cal-culated for each assessed muscle. Intra-class correlation coef-cients (ICC1,1) [19] were calculated and interpreted according to[20]. Standard error of measurement (SEM) values were derivedfrom the square root of the mean square error from one-wayANOVA [21]. Parameters were compared using t-tests (CP vs. TD) orpaired-sample t-tests (pre- vs. post-BTX). The average differencebetween groups and between pre- and post-BTX was evaluated in

-20 -1 0 0 10 20

10 -5 0 5 10 15

Joint ang le(deg)

Joint ang le(deg)

Measured to rque

Mode l pred icon

hes from a healthy (I, III) subject, and from a subject with CP (II, IV). Data from one

L. Bar-On et al. / Gait & Posture 40 (2014) 346351 349view of the minimal detectable change (MDC) calculated fromSEM-values [22]. The relation between the developed parametersand subject age, diagnosis, clinical spasticity scores [4,15], and theGross Motor Functional Classication Scale were investigatedusing either Pearson correlation coefcients (continuous param-eters), or ANOVA and post-hoc Tukey tests (categorical param-eters). Analyses were performed using SPSS. Signicance was setat p < 0.05.

3. Results

Fifty-three children with CP and 10 TD children were included(Table 1). Three children with CP were assessed bilaterally, 10 werere-assessed for between-session reliability, and 16 (19 muscles)underwent re-assessment on average 53.3 11.6 days post-BTX(average dosage injected into the gastrocnemius was 4.52 0.79 U/kg).

Study results can be found in Table 2. During high velocity trials,average Vmax was higher in the TD group than in the CP and in theCP group post-BTX than pre-BTX. ICC values were high for all

Fig. 3. Measured, modeled torque, and the components of the modeled torque during lmuscle of a subject with CP pre- (II, V), and post- (III, VI) treatment with botulinum toxinparameters except for Vmax at medium velocity. The differencesbetween the mean values for CP and TD children for work-change(1.78 J), model-deviation (1.80 Nm), and work-deviation (1.54 J)were higher than their corresponding MDC values (0.82 J, 0.84 Nm,and 0.72 J, respectively). Post-BTX, only model-deviation, andwork-deviation signicantly decreased. The mean pre-post de-crease in model-deviation (0.93 Nm) was larger than its MDC value(0.84 Nm). Model-deviation, work-deviation, and work-changehad fair and good correlations with age (r = 0.32; r = 0.31; r = 0.61,p < 0.05, respectively).

4. Discussion

The aim of this study was to compare different biomechanicalparameters that estimate the amount of neural contribution toincreased joint torque during manually-applied, passive anklemovements at different velocities. The developed parameters wereassessed for their between-session reliability and validated bycomparing children with CP to TD children and by the response ofmuscles with spasticity to BTX-treatment.

ow (IIII) and high (IVVI) velocity stretches of a healthy muscle (I, IV), and from a-A. Data from stretch trials at all velocities were used to create the modeled torque.

Muscle models have mostly been applied to data collected withmotor-driven devices [7,13,23]. However, instrumented manualassessments have the advantage of being portable, easy toadminister and to tolerate. Their disadvantage is decreased controlcreating the risk that extracted data are inuenced by measure-ment performance. Nevertheless, in the current study, a modelapproximated from [7], was successfully applied to data collectedduring manually-applied movements. The model accurately

component cannot be tted properly by the applied model. Thecurrent approach thus holds a middle between data-drivenparameters (such as in [25]) and a model-driven approach, suchas in [7].

Several experimental and modeling assumptions in the currentstudy need to be recognized. Firstly, it was assumed that the anklemotion during stretch occurred in the sagittal plane. However, aspreviously reported [11], small out-of-plane movements (

L. Bar-On et al. / Gait & Posture 40 (2014) 346351 351suggest that the measurement error is still too high to detectimprovement in all subjects. Therefore, more effort is required toincrease reliability and improve performance standardization.

The effects of BTX were not captured by work-change, whichwas not surprising as work-change was also inuenced by stiffnessand viscosity. This nding conrms that by removing the non-neural components, the deviation parameters better captured theneural inuence. This is in line with previous ndings, that reportlittle effect of BTX on passive myotendinous stiffness in the calfmuscles of children with CP [3].

Muscle contractures already start developing at an early age[26] and can affect gait in children with CP [27]. This emphasizesthe need for proper torque differentiation and more dedication indeveloping treatments that target passive stiffness. Casting,combined with BTX, has been found to be more effective thanBTX alone in improving gait [28]. Despite the short casting periodpost-BTX, work-change in the current study, did not signicantlydecrease post-treatment. Recent results suggest that growthvelocity, rather than spasticity, plays a crucial role in contracturedevelopment [29]. In line with these ndings, muscle stiffnessincreases, whereas spasticity decreases, with age [30]. In thecurrent study, age had a better correlation to work-change than tothe deviation parameters.

There was also a large response variability to BTX-treatment.This variation highlights the importance of searching for predictorsof treatment response in order to better the tailor individualizedtreatment. Apart from age, no other relations were found betweenthe biomechanical parameters and subject characteristics, butlarger studies are required. Future studies should investigatewhether muscles with a low model-deviation value are poorerresponders to BTX. Accordingly, the current method offers a way inwhich to ne-tune treatment to the individual muscles' underlyingetiology. However, more explorations are required to understandthe relation between spasticity as assessed at rest in a non-weightbearing position and functional movements, such as gait.

In the current study, EMG was not analyzed and was onlyused to conrm the presence/absence of muscle activation. Theamount of EMG was not used as it has previously been shownthat the relative contribution of the lateral gastrocnemius to thetotal neural torque component in the ankle is only 3% [7]. Afuture study should collect EMG from more muscles acting onthe ankle to validate the relation between the identied neuralcomponent and the amount and timing of pathological muscleactivation.

In summary, the neural contribution to increased ankletorque during stretches of spastic gastrocsoelus muscles wascaptured in two biomechanical parameters (model-deviationand work-deviation) that also decreased post-BTX. Work-change, that includes passive stiffness and viscosity, was higherin spastic than in TD muscles, but was not affected by BTX.Therefore, while there was a large response variability,treatment with BTX combined with casting, predominantlytreated the neural component. This highlights the need tobetter ne-tune treatment modalities for the gastrocsoleus inchildren with CP.

Acknowledgements

This work was made possible by a grant from the DoctoralScholarships Committee for International Collaboration with nonEER-countries (DBOF) of the KU Leuven, Belgium. This work wasfurther supported by a grant from Applied Biomedical Researchfrom the Flemish agency for Innovation by Science and Technology(IWT-TBM: grant number 060799); and by an unrestrictededucational grant from Allergan, Inc. (USA).References

[1] Tardieu G, Shentoub S, Delarue R. A la recherche dune technique de mesure dela spasticite imprime avec le periodique. Neurologique 1954;91:1434.

[2] Lance JW. The control of muscle tone, reexes, and movement: RobertWartenbeg lecture. Neurology 1980;30:1303.

[3] Alhusaini AAA, Crosbie J, Shepherd RB, Dean CM, Scheinberg A. No change incalf muscle passive stiffness after botulinum toxin injection in children withcerebral palsy. Dev Med Child Neurol 2011;53:5538.

[4] Boyd RN, Graham HK. Objective measurement of clinical ndings in the use ofbotulinum toxin type A for the management of children with cerebral palsy.Eur J Neurol 1999;6:2335.

[5] Van den Noort JC, Scholtes VA, Becher JG, Harlaar J. Evaluation of the catch inspasticity assessment in children with cerebral palsy. Arch Phys Med Rehabil2010;91:61523.

[6] Bar-On L, Aertbelin E, Molenaers G, Desloovere K. Muscle activation patternswhen passively stretching spastic lower limb muscles of children with cerebralpalsy. PLoS One 2014;9:e91759.

[7] De Vlugt E, de Groot JH, Schenkeveld KE, Arendzen JH, van der Helm FCT,Meskers CGM. The relation between neuromechanical parameters andAshworth score in stroke patients. J Neuroeng Rehabil 2010;7:35.

[8] Alibiglou L, Rymer WZ, Harvey RL, Mirbagheri MM. The relation betweenAshworth scores and neuromechanical measurements of spasticity followingstroke. J Neuroeng Rehabil 2008;5:18.

[9] Lakie M, Walsh EG, Wright GW. Assessment of human hemiplegic spasticity bya resonant frequency method. Clin Biomech (Bristol Avon) 1988;3:1738.

[10] Rabita G, Dupont L, Thevenon A, Lensel-Corbeil G, Prot C, Vanvelcenaher J.Differences in kinematic parameters and plantarexor reex responsesbetween manual (Ashworth) and isokinetic mobilisations in spasticityassessment. Clin Neurophysiol 2005;116:93100.

[11] Bar-On, Aertbelin E, Wambacq H, Severijns D, Lambrecht K, Dan B, et al. Aclinical measurement to quantify spasticity in children with cerebral palsy byintegration of multidimensional signals. Gait Posture 2012;38:1417.

[12] Van den Noort JC, Scholtes VA, Harlaar J. Evaluation of clinical spasticityassessment in cerebral palsy using inertial sensors. Gait Posture 2009;30:13843.

[13] Harlaar J, Becher JG, Snijders C, Lankhorst GJ. Passive stiffness characteristics ofankle plantar exors in hemiplegia. Clin Biomech 2000;15:26170.

[14] Chen J-JJ, Wu Y-N, Huang S-C, Lee H-M, Wang Y-L. The use of a portable muscletone measurement device to measure the effects of botulinum toxin type a onelbow exor spasticity. Arch Phys Med Rehabil 2005;86:165560.

[15] Bohannon RW, Smith MB. Interrater reliability of a modied Ashworth scale ofmuscle spasticity. Phys Ther 1987;67:2067.

[16] Rauch HE, Tung F, Striebel C. Maximum likelihood estimates of linear dynamicsystems. Am Inst Aeronaut Astronaut J 1965;3:144550.

[17] Staude G, Wolf W. Objective motor response onset detection in surfacemyoelectric signals. Med Eng Phys 1999;21:44967.

[18] Nocedal J, Wright S. Numerical optimization (Springer series in operationsresearch and nancial engineering). 2nd ed. Springer; 2006.

[19] McGraw KO, Wong SP. Forming inferences about some intraclass correlationcoefcients. Psychol Methods 1996;1:3046.

[20] Katz JN, Larson MG, Phillips CB, Fossel AH, Liang MH. Comparativemeasurement sensitivity of short and longer health status instruments.Med Care 1992;30:91725.

[21] Weir JP. Quantifying test-retest reliability using the intraclass correlationcoefcient and the SEM. J Strength Cond Res 2005;19:23140.

[22] De Vet HC, Terwee CB, Ostelo RW, Beckerman H, Knol DL, Bouter LM. Minimalchanges in health status questionnaires: distinction between minimallydetectable change and minimally important change. Health Qual LifeOutcomes 2006;4:54.

[23] De Gooijer-van de Groep KL, de Vlugt E, de Groot JH, van der Heijden-MaessenHCM, Wielheesen DHM, van Wijlen-Hempel RMS, et al. Differentiationbetween non-neural and neural contributors to ankle joint stiffness in cerebralpalsy. J Neuroeng Rehabil 2013;10:81.

[24] Bishop CM. Pattern recognition and machine learning. Springer; 2006.[25] Pandyan AD, Van Wijck FMJ, Stark S, Vuadens P, Johnson GR, Barnes MP. The

construct validity of a spasticity measurement device for clinical practice: analternative to the Ashworth scales. Disabil Rehabil 2006;28:57985.

[26] Willerslev-Olsen M, Lorentzen, Nielsen JB, Sinkjaer T. Passive muscleproperties are altered in children with cerebral palsy before the age of 3years and are difcult to distinguish clinically from spasticity. Dev Med ChildNeurol 2013;55:61723.

[27] Willerslev-Olsen M, Andersen JB, Sinkjaer T, Nielsen JB. Sensory feedback toankle plantar exors is not exaggerated during gait in spastic children withcerebral palsy. J Neurophysiol 2014;111:74654.

[28] Desloovere K, Molenaers G, Jonkers I, De Cat J, De Borre L, Nijs J, et al. Arandomized study of combined botulinum toxin type A and casting in theambulant child with cerebral palsy using objective outcome measures. Eur JNeurol 2001;8:7587.

[29] vehlk M, Leistritz L, Kraus T, Zwick EB, Steinwender G, Linhart WE. Thegrowth and the development of gastro-soleus contracture in cerebral palsy.Gait Posture 2013;38:S12.

[30] Hgglund G, Wagner P. Development of spasticity with age in a totalpopulation of children with cerebral palsy. BMC Musculoskelet Disord2008;9:150.

Identification of the neural component of torque during manually-applied spasticity assessments in children with cerebral palsy1 Introduction2 Method2.1 Participants2.2 Experimental procedure2.3 Data analysis2.3.1 Data processing2.3.2 Outcome parameters

2.4 Statistics

3 Results4 DiscussionAcknowledgementsReferences