Muscle strength training to improve gait function in children with cerebral palsy


* Correspondence to first author at Regionala barn-och ungdomshabiliteringen, Box 21 062, SE 418 04 Göteborg, Sweden


The aim of the study was to investigate the influence of muscle strength training on gait outcomes in children with cerebral palsy. Sixteen children (two females, 14 males, Gross Motor Function Classification System levels I–II, mean age 12y 6mo, range 9y 4mo–15y 4mo) underwent muscle strength measurement using a handheld device, Gross Motor Function Measure (GMFM) assessment, three-dimensional gait analysis, joint range of motion assessment, and grading of spasticity before and after 8 weeks of training. All participants had a diagnosis of spastic diplegia and could walk without aids. Training consisted of exercises for lower extremity muscles with free weights, rubber bands, and body weight for resistance, three times a week. Values for muscle strength below normal were identified in all children; this was most pronounced at the ankle, followed by the hip muscles. After training, muscle strength and GMFM scores increased, velocity was unchanged, stride length increased, and cadence was reduced. There was an increase in hip extensor moment and power generated at push off. Eight weeks of muscle strength training can increase muscle strength and improve gait function.

Parents of children with cerebral palsy (CP) often ask about and focus on their child’s ability to walk. Population-based studies show that about 69% of children with CP are classified as walking with or without aids1,2 according to the Gross Motor Function Classification System (GMFCS), and that age at first walking is often delayed.3 A Canadian study of gross motor development showed improvement of motor function at early ages until around 5 years, when an average of 90% of motor development potential has been reached, and that there is a plateau in gross motor development at about 7 years of age.4 Bell et al. and Johnson et al. demonstrated a decrease in gait function and pattern through adolescence and adulthood expressed as a decrease in gait velocity, stride length, and sagittal joint excursions over time.5,6 Surveys of adults with CP show decreased walking ability with age.3,7

Possible factors interfering with normal gait pattern in CP include spasticity, muscle contractures, bony deformities, loss of selective motor control, and muscle weakness.8 Recent research has focused on muscle weakness. Wiley and Damiano,9 and Ross and Engsberg10 described muscle weakness as more pronounced distally and found an imbalance across joints. Several studies of different methods for muscle strength training show that it is possible to increase muscle strength in children with CP.11–15 The influence of muscle strength training on walking ability has been reported as an increase in the Gross Motor Function Measure scores (GMFM).12–17

Gait can be measured with three-dimensional (3D) gait analysis, which provides a description of gait pattern geometry (kinematics) and forces (kinetics). Variables such as gait velocity, cadence, and stride length have been reported to increase with training11,14 as well as improved kinematics in the sagittal plane.17 Only one report on kinetics has been identified, which showed no change in ankle plantar moment after training of dorsiflexors and plantarflexors.17 Gait can also be measured in terms of efficiency, energy expenditure index, or physical cost index. One study reported better efficiency after training15 and two others saw no change.11,13

Muscle strength in children with CP can be measured reliably with handheld devices.18,19 Damiano et al. demonstrated that data should be reported as torque (force × lever arm) to enable comparisons between individuals and over time.20 A recent study reported normative data obtained with a handheld device and calculations of torque, and based on this presented equations for predicted values for each muscle group based on age, body weight, and sex.21 This makes comparison possible both over time in one individual and between individuals of different age and weight.

The aim of this study was to investigate whether muscle strength training could influence gait function (as measured by the GMFM) and gait pattern (kinematics and kinetics) in children with CP between the ages of 9 and 15 years. The aim was also to use torque and normative values in the comparison.



Potential participants were identified from the medical records of habilitation centres around Göteborg. Inclusion criteria were bilateral spastic CP, age 10 to 15 years, and GMFCS levels I and II (walking without aids). Children had to be able to follow instructions and participate in a group. Exclusion criteria were orthopaedic surgery or botulinum toxin injections in the past 12 months. The Ethics Committee at the Medical Faculty at the University of Göteborg approved this study. Written informed consent was obtained from parents of each participant.


Examination before and after a period of muscle strength training consisted of: muscle strength measurement, GMFM, 3D gait analysis, joint range of motion (ROM) measurement, and grading of spasticity. In terms of the International Classification of Functioning, Disability and Health,22 these methods measure ‘body function’ (muscle strength, 3D gait analysis, ROM, and spasticity) and ‘activity’ (GMFM).

Muscle strength was measured with a handheld device: a myometer (adapted Chatillon dynamometer; Axel Ericson Medical AB, Göteborg, Sweden) using the ‘make’ technique, where the child gradually builds up force against the myometer for about 5 seconds. Three attempts were made for each muscle group and the maximum-recorded force value was used for data analysis. Lever arm was measured with a tape measure, and torque (Nm) was calculated by multiplying force by the length of each lever arm. Data were also compared to a normative predicted value based on an equation with parameters for age, body weight, and sex.21 Measurements were divided by the predicted value giving a percentage for every muscle group in each child. This made it possible to compare muscle groups and to compare children with different ages and weights. Eight muscle groups were tested (hip extensors, flexors, abductors and adductors; knee extensors and flexors; and ankle dorsiflexors and plantarflexors) with the same testing positions as in the normative study.21 For ankle plantarflexors no normative predicted value was available for children over the age of 9 years.

Gross motor function was tested with GMFM domains D: standing, and E: walking, running, and jumping. The Gross Motor Ability Estimator software which was included in the GMFM-66 version23 was used to calculate a score.

3D gait analysis was carried out with a motion capture system consisting of six infrared cameras (ProReflex Qualisys AB, Göteborg, Sweden) and two Kistler force plates (Kistler 9281C, Kistler Instruments AG, Winterthur, Switzerland) working synchronically at 240Hz. Recordings of motion and calculations were made with the software QtracC version 2.51, QtracV version 2.60, and QGait 2.0 (Qualisys Medical AB, Göteborg, Sweden).24 At least three acceptable trials for each child were collected. Parents confirmed that the performance was representative of their children’s regular gait pattern. 3D gait data were compared with the laboratory reference database for children between 10 and 15 years of age (27 children, mean age 12y 11mo). Gait velocity, stride length, and cadence were compared with age norms.25

ROM was measured in the lower extremities using a regular plastic goniometer (hip extension, hip abduction, knee flexion, popliteal angle, and ankle dorsiflexion). Deviations outside 2 standard deviations from normal values were noted.26,27

Spasticity was tested in four muscle groups with the Ashworth scale:28 0=hypotonus; 1=normal; 2=resistance through less than half ROM; 3=resistance through most of ROM; and 4=difficult to move passively. Examination was done with children lying in a relaxed supine position for hip adductors, hamstrings, and plantarflexors, and in a prone position for rectus femoris. Scores for each child were totalled and divided by number of muscles tested, a method described by Ostensjø et al.29 The summary index was then categorized: no increase (≤1), mild (>1 to ≤2), or moderate (>2).

Muscle strength measurements, GMFM, ROM, and spasticity were tested by two of the authors (MNE, KA who are physiotherapists and gait analysis was carried out by the gait laboratory staff). Testing was performed the week before training started and a week after training. Muscle strength was also tested 2 weeks before training and 2 weeks after training. Mean muscle strength measurements in Nm before, were compared with mean measurements after training.

Before training started there was an individual analysis of each child based on muscle strength measurements and 3D gait analysis, identifying muscles with the most pronounced muscle weakness and most important gait pattern abnormalities (like a small abducting moment in the hip, too much knee flexion during stance, or no power generation in the ankle at push off). This resulted in an individual training programme for each child, containing specific training instructions for four different muscle groups.

The training period lasted for 8 weeks, three times a week: twice a week at home with parental assistance and once a week in a small group with a physiotherapist at the physiotherapy department after school. Children were divided into small training groups based on age. At home they carried out the individual programme with three sets of 10 repetitions for each muscle group: first set easy, second medium, and third with a heavy load – 10 Repetition Maximum (10RM). Resistance was provided by adjustable weight cuffs for the medium and heavy sets. Rubber bands and body weight were used for the easy set and when it was not possible to use weight cuffs. Weight resistance was increased during the training period when the children could do more than 10 repetitions with the 10RM weight.

The group session consisted of a low intensity short warm-up session with a cycle ergometer, rowing machine, or step-up. After initial warm-up, children carried out their individual programmes with strength training exercises. Stretching of hamstrings, rectus femoris, and plantarflexor muscles followed the training session. All meetings ended with a group activity chosen by the children (different games). The sessions lasted for one hour and a half in total. Each child had a training diary, indicating dates and resistance. Group training sessions and decisions on increasing resistance were carried out by two physiotherapists.

Statistical analyses

Non parametric tests were used as the group was small and data could not be proved to be normally distributed. Gait data for the CP group and laboratory reference database were analyzed with the Mann–Whitney U test. All differences before and after training were analyzed with the Wilcoxon signed rank test except for spasticity grading where the paired sign test was used. Significance level was set at p<0.05. Statistical tests were performed using StatView software (SAS Institute Inc).


Twenty-four children met inclusion criteria, and of these eight chose not to participate. Of the 16 participating children there were 14 males and two females with a mean age of 12 years and 6 months (range 9y 4mo–15y 4mo). According to the GMFCS, 10 children were classified at Level I and six at Level II (Table I). Two training diaries were not available. Of a possible 24 training sessions, 14 children performed a mean of 18 sessions (range 11–23y; 76%). There was no report of negative events caused by the training during the period. There was no change in children’s regular therapy and no one was using orthoses.

Table I.   Mean (SD) age, weight, and height of participants according to Gross Motor Function Classification System (GMFCS) level
GMFCS levelAge, yWeight, kgHeight, cm
I, n=1012.2 (1.8)36 (7.1)146 (12)
II, n=613 (2)43.9 (8.9)150 (10)

Muscle strength

All children could lift their legs against gravity with resistance. Values below normative predicted were identified in all children (Table II). Weakness was most pronounced at the ankle, followed by hip muscles. Muscles around the knee were strongest, with values within normative predicted ranges for knee extensors in 12 children. Torque values for plantarflexors were below normative values (for 9-year-old children 40.2 Nm).21 Most of the children undertook training exercises for ankle dorsiflexors (n=12) and plantarflexors (n=12), hip abductors (n=12), and hip extensors (n=10), based on the pre-testing examination.

Table II.   Muscle strength before and after traininga
Muscle groupBefore training period, median (range) and % of predicted normal valueAfter training period, median (range)Comparison before and after training, p-value
All n=32Targeted muscle groupNon-targeted muscle group
  1. aResults are presented as torque (Nm) and as a percentage of predicted normal value21 (except for ankle plantarflexors which is only torque). n=number of legs. Wilcoxon signed rank p-values based on Nm data.

Hip extensors, Nm
49.8 (22.2–106.5)
61.6 (34.9–104.6)
55.2 (26.7–161.5)<0.0010.001, n=200.049, n=12
Hip flexors, Nm44.5 (16.9–97)
67.1 (30–110.1)
49 (20.9–121.3) <0.0010.025, n=8<0.001, n=24
Hip abductors, Nm38.8 (19.3–89.7)
62.8 (37–113.4)
42.5 (23–109.3)<0.001<0.001, n=240.263, n=8
Hip adductors, Nm39.7 (21.8–94.3)
66.4 (41.9–109.4)
45.4 (24.7–105.5) 0.0010.002, n=120.117, n=20
Knee extensors, Nm51.1 (24.2–106.4)
84.6 (52.1–113.5)
57.7 (22.1–98.6)0.9550.917, n=60.869, n=26
Knee flexors, Nm44 (23.2–97.1)
83.8 (51.3–124.9)
46.3 (27.6–107.9) 0.0010.075, n=60.006, n=26
Ankle dorsiflexors, Nm10.9 (0.6–20.5)
46 (2–68.8)
11.5 (0–25.7) 0.0570.339, n=250.018, n=7
Ankle plantarflexors, Nm30.4 (13–65.3)33 (14.1–73.8)0.1320.014, n=250.311, n=7

After the training period there was a significant increase in muscle strength in all hip muscle groups and in knee flexors (Table II). When results were divided into groups of targeted or non-targeted muscle groups, there was a visible increase in hip extensors and flexors in all children. Stronger hip abductors, adductors, and ankle plantarflexors were found in those who had targeted these muscle groups. Knee flexors and ankle dorsiflexors were stronger in those who did not target these muscle groups.


Three of the children scored 100% before training. For the other children there was a statistically significant increase after training (Table III). This increase was mostly attributable to the ability to stand on one leg (seven children had a higher item score than before training) and to hop on one foot (four had a higher item score).

Table III.   Results from Gross Motor Function Measure (GMFM) and time–distance gait parameters (n= 16)
 Before training, median (range)After training, median (range)Wilcoxon signed rank p
GMFM84.8 (66.7–100)90 (67.4–100)0.003
Velocity, m/sec1.2 (1–1.5)1.25 (0.9–1.6)0.859
Stride, m1.1 (0.9–1.4)1.15 (0.9–1.5)0.059
Cadence, steps/min132 (108–151)130.5 (104–149)0.016


Gait velocity, stride length, and cadence were within age norms except in two cases where stride length was short for age. Velocity did not change after the training period, cadence was reduced, and there was a tendency to increased stride length (Table III).

Kinematic and kinetic gait pattern parameters were tested, and children with CP differed significantly from normal for sixteen of these variables which were chosen for statistical testing (Table IV). There was a significant increase in hip extensor moment and plantarflexor generating power at push-off after the training period (see example Fig. 1).

Table IV.   Gait analysis data for 16 children (both legs)
  Gait variableBefore training, median (range)After training, median (range)pc
  1. aNm/kg body weight for moment, and W/kg for power; bKnee flexion moments reported as negative values (in italic) reflects an extension moment of the knee; cWilcoxon signed-ranks test p<0.05.

Kinematics,°HipExtension in stance−2.9 (−19.8 to 12.5)−3.8 (−19.5 to 18.4)0.135
Flexion in stance45.4 (17.4 to 71.4)46.7 (16.9 to 76.1)0.155
KneeInitial contact17.9 (5.7 to 43.7)19.5 (3 to 46.9)0.421
Extension in stance6.1 (−18.5 to 28.1)6 (−20 to 25.6)0.231
AnkleInitial contact−4.3 (−22.1 to 5.8)−5.9 (−17.5 to 3.9)0.940
Dorsiflexion in stance8.9 (−8.6 to 17.5)7.1 (−4.1 to 19.5)0.903
Dorsiflexion in swing−0.4 (−17 to 30)−1.1 (−17.1 to 14.8)0.184
KineticsaHipExtension moment0.66 (−0.1 to 1.38)0.78 (0.11 to 1.30)0.015
Flexion moment0.59 (0.33 to 1.30)0.71 (0.37 to 1.30)0.067
Abduction moment 10.52 (0.12 to 1.25)0.51 (0.23 to 1.28)0.860
Abduction moment 20.44 (0.12 to 0.86)0.48 (0.16 to 0.91)0.088
KneeExtension moment0.76 (0.19 to 1.69)0.67 (0.26 to 2.32)0.899
Flexion momentb−0.41 (−0.93 to 0.10)−0.38 (−0.99 to 0.14)0.318
AnklePlantarflexion moment1.08 (0.70 to 1.60)1.11 (0.75 to 1.71)0.193
Generating power1.59 (0.64 to 3.17)1.91 (0.35 to 3.75)0.046
Absorbing power1.22 (0.44 to 7.52)1.18 (0.52 to 4.34)0.772
Figure 1.

Ankle generating power. Grey band, lab reference, dotted line, mean before, solid line, mean after.

Contractures and spasticity

Decreased ROM outside 2SDs from normal was present in all children. After the training period there was a statistically significant increase in hamstring length, with a median (range) of 135° (120–165°) before, and 136° (125–155°) after (p=0.002) but no change in other muscle groups.

Increased muscle tone was graded as no increase in two of the children, mild in 12, and moderate in two. There was no statistical difference in spasticity grading after the training period.


This study shows an increase in muscle strength, and measurable positive changes in gait function and pattern with training. This is comparable with the findings from earlier studies, which also showed some significant changes and trends in both muscle strength and gait function.12,14

GMFM scores increased at group level, which was slightly unexpected as three children had already scored 100 (full) before training. Items that showed changes were most often standing on one leg and hopping on one foot, which require stability and strength at hip and ankle. The ability to balance on one leg is very important for many tasks in everyday life, such as negotiating obstacles and climbing stairs. Several studies on the natural history of gross motor development in children with CP shows that there is not much increase in the GMFM after 7 years of age.4,30,31

This study shows that an intervention can make a change in the GMFM even after these ages. Wang and Yang report a change of 3.71 on the GMFM-66 to be a clinically visible improvement.32

Gait analysis showed increased stride length and plantarflexor generating power at push off after training. This could be explained by better stability around both hip and knee that increases stability in stance and which makes it easier for the ankle plantarflexors to push off actively. The increased stride length and push off corresponds well with the increase in muscle strength around the hips and with the increase in balance on one leg visible in the GMFM.

There is a need for further analysis of the relationship between muscle strength and gait pattern to augment understanding of gait deviations in CP, and for planning effective treatment and training. Some earlier studies of the effects of strength training on gait have shown increased velocity with training.11,14 This was not found in the current study. Some children varied in velocity during the same session but there was no pattern of systematic change after the training period. Children in the current study all walked without assistive devices and their gait velocity was already within normal age variation. Studies with increased velocity after training included children walking with assistive devices, who probably had a low walking velocity before training.

Is increased muscle strength the factor that determines better gait? Many of the children in the current study had reduced ROM before the training started, and the programme included a stretching session. However, the only change in ROM that could be observed was a small increase in the hamstrings (which was clinically not remarkable and within measurement error) and there was no change in spasticity grading. Studies on the relationship between clinical measurements and gait parameters found that muscle strength and selective motor control correlated better with gait data than ROM and spasticity.33,34

Gormley stated that CP is a disorder that manifests differently in each child, and for optimal results of interventions an individual analysis is needed together with an individual plan for treatment.8 This is the main reason this study developed individual programmes for the children, but it led to problems when comparing data, as the group was small which made the use of subgroups difficult. The result that there was increased strength in muscle groups that were not targeted in the training was unexpected. It is likely that this effect can be explained in terms of the human body being a system where activation of one muscle group also requires activation of other muscle groups needed for stabilization of adjacent joints. This stabilizing mechanism may give a training effect and can explain why, for example, muscle strength in the hip flexors increased although they were only included in the programme for some of the children. Almost all the children in this study had weakness in the dorsiflexors, and these muscles were included in the programmes for most of them; however, there was no increase of strength in the dorsiflexors. This is a muscle group that is often difficult to activate in isolation and, therefore, may need special attention. This has been carried out in a previous study focusing on training dorsiflexors with the help of computer-assisted biofeedback, and increased strength was reported.35


Eight weeks of individually-designed training with a focus on muscle strength not only increased muscle strength, but also improved gait function in children with CP. Increase in muscle strength is one of the major factors for the improvement in gait function so the current results support muscle strength training as a means of improving gait in children with CP.


We thank the Norrbacka-Eugenia foundation, the Research and Development Foundation of Göteborg and Bohuslän, Linnéa och Josef Carlsson foundation, and Petter Silfverskiöld foundation for funding this research.