Mechanically assisted walking training for walking, participation, and quality of life in children with cerebral palsy

Abstract

Background

Cerebral palsy is the most common physical disability in childhood. Mechanically assisted walking training can be provided with or without body weight support to enable children with cerebral palsy to perform repetitive practice of complex gait cycles. It is important to examine the effects of mechanically assisted walking training to identify evidence‐based treatments to improve walking performance.

Objectives

To assess the effects of mechanically assisted walking training compared to control for walking, participation, and quality of life in children with cerebral palsy 3 to 18 years of age.

Search methods

In January 2020, we searched CENTRAL, MEDLINE, Embase, six other databases, and two trials registers. We handsearched conference abstracts and checked reference lists of included studies.

Selection criteria

Randomized controlled trials (RCTs) or quasi‐RCTs, including cross‐over trials, comparing any type of mechanically assisted walking training (with or without body weight support) with no walking training or the same dose of overground walking training in children with cerebral palsy (classified as Gross Motor Function Classification System [GMFCS] Levels I to IV) 3 to 18 years of age.

Data collection and analysis

We used standard methodological procedures expected by Cochrane.

Main results

This review includes 17 studies with 451 participants (GMFCS Levels I to IV; mean age range 4 to 14 years) from outpatient settings. The duration of the intervention period (4 to 12 weeks) ranged widely, as did intensity of training in terms of both length (15 minutes to 40 minutes) and frequency (two to five times a week) of sessions. Six studies were funded by grants, three had no funding support, and eight did not report information on funding. Due to the nature of the intervention, all studies were at high risk of performance bias.

Mechanically assisted walking training without body weight support versus no walking training

Four studies (100 participants) assessed this comparison. Compared to no walking, mechanically assisted walking training without body weight support increased walking speed (mean difference [MD] 0.05 meter per second [m/s] [change scores], 95% confidence interval [CI] 0.03 to 0.07; 1 study, 10 participants; moderate‐quality evidence) as measured by the Biodex Gait Trainer 2™ (Biodex, Shirley, NY, USA) and improved gross motor function (standardized MD [SMD] 1.30 [postintervention scores], 95% CI 0.49 to 2.11; 2 studies, 60 participants; low‐quality evidence) postintervention. One study (30 participants) reported no adverse events (low‐quality evidence). No study measured participation or quality of life.

Mechanically assisted walking training without body weight support versus the same dose of overground walking training

Two studies (55 participants) assessed this comparison. Compared to the same dose of overground walking, mechanically assisted walking training without body weight support increased walking speed (MD 0.25 m/s [change or postintervention scores], 95% CI 0.13 to 0.37; 2 studies, 55 participants; moderate‐quality evidence) as assessed by the 6‐minute walk test or Vicon gait analysis. It also improved gross motor function (MD 11.90% [change scores], 95% CI 2.98 to 20.82; 1 study, 35 participants; moderate‐quality evidence) as assessed by the Gross Motor Function Measure (GMFM) and participation (MD 8.20 [change scores], 95% CI 5.69 to 10.71; 1 study, 35 participants; moderate‐quality evidence) as assessed by the Pediatric Evaluation of Disability Inventory (scored from 0 to 59), compared to the same dose of overground walking training. No study measured adverse events or quality of life.

Mechanically assisted walking training with body weight support versus no walking training

Eight studies (210 participants) assessed this comparison. Compared to no walking training, mechanically assisted walking training with body weight support increased walking speed (MD 0.07 m/s [change and postintervention scores], 95% CI 0.06 to 0.08; 7 studies, 161 participants; moderate‐quality evidence) as assessed by the 10‐meter or 8‐meter walk test. There were no differences between groups in gross motor function (MD 1.09% [change and postintervention scores], 95% CI ‐0.57 to 2.75; 3 studies, 58 participants; low‐quality evidence) as assessed by the GMFM; participation (SMD 0.33 [change scores], 95% CI ‐0.27 to 0.93; 2 studies, 44 participants; low‐quality evidence); and quality of life (MD 9.50% [change scores], 95% CI ‐4.03 to 23.03; 1 study, 26 participants; low‐quality evidence) as assessed by the Pediatric Quality of Life Cerebral Palsy Module (scored 0 [bad] to 100 [good]). Three studies (56 participants) reported no adverse events (low‐quality evidence).

Mechanically assisted walking training with body weight support versus the same dose of overground walking training

Three studies (86 participants) assessed this comparison. There were no differences between groups in walking speed (MD ‐0.02 m/s [change and postintervention scores], 95% CI ‐0.08 to 0.04; 3 studies, 78 participants; low‐quality evidence) as assessed by the 10‐meter or 5‐minute walk test; gross motor function (MD ‐0.73% [postintervention scores], 95% CI ‐14.38 to 12.92; 2 studies, 52 participants; low‐quality evidence) as assessed by the GMFM; and participation (MD ‐4.74 [change scores], 95% CI ‐11.89 to 2.41; 1 study, 26 participants; moderate‐quality evidence) as assessed by the School Function Assessment (scored from 19 to 76). No study measured adverse events or quality of life.

Authors' conclusions

Compared with no walking, mechanically assisted walking training probably results in small increases in walking speed (with or without body weight support) and may improve gross motor function (with body weight support). Compared with the same dose of overground walking, mechanically assisted walking training with body weight support may result in little to no difference in walking speed and gross motor function, although two studies found that mechanically assisted walking training without body weight support is probably more effective than the same dose of overground walking training for walking speed and gross motor function. Not many studies reported adverse events, although those that did appeared to show no differences between groups. The results are largely not clinically significant, sample sizes are small, and risk of bias and intensity of intervention vary across studies, making it hard to draw robust conclusions. Mechanically assisted walking training is a means to undertake high‐intensity, repetitive, task‐specific training and may be useful for children with poor concentration.

Plain language summary

Mechanically assisted walking training for children with cerebral palsy

Background

Children with cerebral palsy have difficulty walking independently. It is thought that they might benefit from mechanically assisted walking training compared with no walking or overground walking (i.e. without mechanical support). Mechanically assisted walking training includes using motorized devices such as a treadmill, a gait trainer (a wheeled walking aid), or a robotic training device (such as a robotic knee brace) to help children with cerebral palsy to walk further. This training can be provided either with or without body weight support (such as a harness, a handrail, or manual physical support).

Review question

What is the effect of mechanically assisted walking compared to no walking or to the same amount of overground walking on walking, participation, and quality of life in children with cerebral palsy 3 to 18 years of age.

Study characteristics

This review includes 17 studies involving a total of 451 children with a mean age range between 4 and 14 years. All children had cerebral palsy. We found four studies comparing mechanically assisted walking without body weight support to no walking; two studies comparing mechanically assisted walking without body weight support to the same amount of overground walking; eight studies comparing mechanically assisted walking plus body weight support to no walking; and three studies comparing mechanically assisted walking with body weight support to the same amount of overground walking. Mechanically assisted training was provided for 15 to 40 minutes a session, two to five times a week, for 4 to 12 weeks. Five studies were funded by a grant, and one study was funded by two different grants. Eight studies did not report funding information, and three studies received no funding support.

The evidence is current to January 2020.

Key results

Mechanically assisted walking without body weight support

1. Compared with no walking, there was a small benefit in terms of walking speed and gross motor function (skills needed to control the large muscles of the body used in walking). In one study, there was no difference between groups in terms of adverse events (undesirable outcomes).

2. Compared with the same amount of overground walking, there was a small benefit in terms of walking speed, gross motor function, and participation. No study reported adverse event rates.

Mechanically assisted walking with body weight support

1. Compared with no walking, there was a small benefit in terms of walking speed but no clear difference in terms of gross motor function, participation, or adverse events.

2. Compared with the same amount of overground walking, there was no benefit in terms of walking speed, gross motor function, or participation. No study reported adverse events.

Conclusions

Moderate‐ and low‐quality evidence suggests that the use of mechanically assisted walking without body weight support may result in small improvements in walking speed and gross motor function, compared to both no walking and the same amount of overground walking. For mechanically assisted walking with body weight support, benefits were seen in walking speed and gross motor function compared to no walking, but not the same amount of overground walking. Not many studies reported adverse events, although those that did appear to show no differences between groups. Mechanically assisted walking can provide high‐dose, repetitive training. It may be a useful way to provide practice for younger children with poor concentration when it is hard to apply the same dose of overground walking.

Summary of findings

Summary of findings 1. Mechanically assisted walking training without body weight support versus no walking training for cerebral palsy.

Mechanically assisted walking training without body weight support versus no walking training for cerebral palsy
Patient or population: children with cerebral palsy classified within GMFCS Levels I to III between 5 and 19 years of age Settings: outpatient setting Intervention: mechanically assisted walking training without body weight support using treadmill and gait trainer Comparison: no walking training
Outcomes	Assumed risk	Corresponding risk	Relative effect (95% CI)	Number of participants (studies)	Quality of evidence (GRADE)	Comments
	No walking training	Mechanically assisted walking training without body weight support
Mobility Measured by: walking speed on a Biodex Gait Trainer 2™ (average speed over 3 × 3 minutes, reported in m/s) Timing of outcome assessment: change scores from baseline to 12 weeks	Mean change score for walking speed in the control group was 0.09 m/s (0.07 m/s to 0.11 m/s)	Mean change score for walking speed in the intervention group was, on average, 0.05 m/s faster (0.03 m/s faster to 0.07 m/s faster)	‐	10 (1 RCT)	⨁⨁⨁◯ Moderatea	‐
Gross motor function Measured by: PDMS‐2‐locomotion (higher score = greater performance, reported as score) or mTUG (higher score = greater performance, reported in ‐seconds) Timing of outcome assessment: postintervention scores at 12 weeks	‐	Mean gross motor function score in the intervention groups was, on average, 1.30 standard deviations better (0.49 to 2.11 better)	‐	60 (2 RCTs)	⨁⨁◯◯ Lowa,b	An effect size of 1.4 is considered to be a large effect size
Adverse events Measured by: no official measure; instead events are reported from observation Timing of outcome assessment: postintervention reports at 12 weeks	‐	‐	‐	30 (1 RCT)	⨁⨁◯◯ Lowa,b	Reported no adverse events
Participation (not measured)	‐	‐	‐	‐	‐	Not measured
Quality of life (not measured)	‐	‐	‐	‐	‐	Not measured
The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; GMFCS: Gross Motor Function Classification System; PDMS‐2: Peabody Developmental Motor Scales, Second Edition; mTUG: modified Timed UP and Go; RCT: randomized controlled trial.
GRADE Working Group grades of evidence. High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

^aWe downgraded the quality of the evidence by one level for imprecision due to insufficient studies.
^bWe downgraded the quality of the evidence by one level for limitations in study design due to several ratings of high or unclear risk of bias.

Summary of findings 2. Mechanically assisted walking training without body weight support versus the same dose of overground walking training for cerebral palsy.

Mechanically assisted walking training without body weight support versus the same dose of overground walking training for cerebral palsy
Patient or population: children with cerebral palsy classified within GMFCS Levels I to III and between 3 and 12 years of age Settings: outpatient setting Intervention: mechanically assisted walking training without body weight support using treadmill and gait trainer Comparison: same dose of overground walking training
Outcomes	Assumed risk	Corresponding risk	Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
	Same dose of overground walking	Mechanically assisted walking training without body weight support using treadmill and gait trainer
Mobility Measured by: 6‐minute walk test or gait analysis via Vicon, reported in m/s Timing of outcome assessment: change scores from baseline to 6 weeks or postintervention scores at 12 weeks	Mean change score for walking speed in control groups was 0.31 m/s (0.20 m/s to 0.42 m/s)	Mean change score for walking speed in intervention groups was, on average, 0.25 m/s faster (0.13 m/s faster to 0.37 m/s faster)	‐	55 (2 RCTs)	⨁⨁⨁◯ Moderatea	‐
Gross motor function Measured by: GMFM‐E (higher score = greater performance), reported as % Timing of outcome assessment: change from baseline to 6 weeks	Mean change score for gross motor function in control group was 8.20% (2.10% to 14.30%)	Mean change score for gross motor function in intervention group was, on average, 11.90% better (2.98% better to 20.82% better)	‐	35 (1 RCT)	⨁⨁⨁◯ Moderatea	‐
Adverse events (not measured)	‐	‐	‐	‐	‐	Not measured
Participation Measured by: PEDI‐mobility score (total scores range from 0 to 59; higher scores = greater participation) Timing of outcome assessment: change scores from baseline to 6 weeks	Mean change score for participation in control group was 3.70 points (1.60 points to 5.80 points)	Mean change score for participation in intervention group was, on average, 8.20 points higher (5.69 points higher to 10.71 points higher)	‐	35 (1 RCT)	⨁⨁⨁◯ Moderatea	‐
Quality of life (not measured)	‐	‐	‐	‐	‐	Not measured
The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; GMFCS: Gross Motor Function Classification System; GMFM: Gross Motor Function Measure—Dimension E; MD: mean difference; PEDI: Pediatric Evaluation of Disability Inventory; RCT: randomized controlled trial.
GRADE Working Group grades of evidence. High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

^aWe downgraded the quality of evidence by one level for imprecision due to insufficient studies.

Summary of findings 3. Mechanically assisted walking training with body weight support versus no walking training for cerebral palsy.

Mechanically assisted walking training with body weight support versus no walking training for cerebral palsy
Patient or population: children with cerebral palsy classified within GMFCS Levels I to V and between 3 and 18 years of age Settings: outpatient setting Intervention: mechanically assisted walking training with body weight support using treadmill and gait trainer Comparison: no walking training
Outcomes	Assumed risk	Corresponding risk	Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
	No walking training	Mechanically assisted walking training with body weight support using treadmill and gait trainer
Mobility Measured by: 10‐meter walk test or 8‐meter walk test (walking speed over a short distance; greater speed = greater functional mobility), reported in m/s Timing of outcome assessment: change scores from baseline to 2 to 12 weeks or postintervention scores at 2 to 12 weeks	Mean score for walking speed in control groups was 0.45 m/s (0.35 m/s to 0.54 m/s)	Mean score for walking speed in intervention groups was, on average, 0.07 m/s faster (0.05 m/s faster to 0.09 m/s faster)	‐	161 (7 RCTs)	⨁⨁⨁◯ Moderatea	‐
Gross motor function Measured by: GMFM‐E (higher score = greater performance), reported as % Timing of outcome assessment: change scores from baseline to 4 to 12 weeks or postintervention scores at 4 to 12 weeks	Mean gross motor function score in control groups was 25.22% (16.61% to 33.84%)	Mean gross motor function score in intervention groups was, on average, 1.09% better (0.57% lower to 2.75% better)	‐	58 (3 RCTs)	⨁⨁◯◯ Lowa,b	‐
Adverse events Measured by: no official measure; instead events are reported from observation Timing of outcome assessment: postintervention reports at 2 to 6 weeks	‐	‐	‐	56 (3 RCT)	⨁⨁◯◯ Lowa,b	Reported no adverse events
Participation Measured by: CAPE‐intensity score or WeeFIM score; higher scores = greater participation Timing of outcome assessment: change scores from baseline to 2 to 12 weeks	Mean change score for participation in control groups was 40.77 points (19.09 points to 62.45 points higher)	Mean change score for participation in intervention groups was, on average, 0.33 standard deviations higher (0.27 lower to 0.93 higher)	‐	44 (2 RCTs)	⨁⨁◯◯ Lowb
Quality of life Measured by: PedsQOL‐CP score (range from 0 to 100; higher scores = better quality of life) Timing of outcome assessment: change scores from baseline to 12 weeks	Mean quality of life score in control group was 62.10 points (56.13 points to 68.07 points)	Mean quality of life score in intervention groups was, on average, 9.50 points higher (4.03 points lower to 23.03 points higher)	‐	26 (1 RCT)	⨁⨁◯◯ Lowb	‐
The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CAPE: Children’s Assessment of Participation and Enjoyment; CI: confidence interval; GMFCS: Gross Motor Function Classification System; GMFM‐E: Gross Motor Function Measure—Dimension E; MD: mean difference; PedsQOL‐CP: Pediatric Quality of Life Cerebral Palsy Module; RCT: randomized controlled trial; WeeFIM: Functional Independence Measure for Children.
GRADE Working Group grades of evidence. High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

^aWe downgraded the quality of evidence by one level for imprecision due to insufficient studies.
^bWe downgraded the quality of evidence by one level for study design limitations due to several ratings of high or unclear risk of bias.

Summary of findings 4. Mechanically assisted walking training with body weight support versus the same dose of overground walking training for cerebral palsy.

Mechanically assisted walking training with body weight support versus the same dose of overground walking training for cerebral palsy
Patient or population: children with cerebral palsy classified within GMFCS Levels I to IV and between 4 and 18 years of age Settings: outpatient setting Intervention: mechanically assisted walking training with body weight support using treadmill and gait trainer Comparison: same dose of overground walking training
Outcomes	Assumed risk	Corresponding risk	Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
	Same dose of overground walking training	Mechanically assisted walking training with body weight support using treadmill and gait trainer
Mobility Measured by: 10‐meter walk test (walking speed over a short distance; greater speed = greater functional mobility) or 6‐minute walk test (distance in meters walked in 6 minutes; greater distance = greater endurance), reported in m/s Timing of outcome assessment: change scores from baseline to 8 to 10 weeks or postintervention scores at 8 to 10 weeks	Mean score for walking speed in control groups was 0.35 m/s (0.19 m/s to 0.51 m/s)	Mean score for walking speed in intervention groups was, on average, 0.02 m/s slower (0.08 m/s slower to 0.04 m/s faster)	‐	78 (3 RCTs)	⨁⨁◯◯ Lowa,b	‐
Gross motor function Measured by: GMFM‐E (higher score = greater performance), reported as % Timing of outcome assessment: postintervention scores at 8 to 10 weeks	Mean gross motor function score in control groups was 53.84% (41.33% to 66.35%)	Mean gross motor function score in intervention groups was, on average, 0.73% lower (14.38% lower to 12.92% higher)	‐	52 (2 RCTs)	⨁⨁◯◯ Lowa,b	‐
Adverse events (not measured)	‐	‐	‐	‐	‐	Not measured
Participation Measured by: SFA score (total scores range from 19 to 76; higher scores = greater participation) Timing of outcome assessment: change scores from baseline to 10 weeks	Mean change score for participation in control group was 5.57 points (4.69 points to 6.45 points)	Mean change score for participation in intervention group was, on average, 4.74 points lower (11.89 points lower to 2.41 points higher)	‐	26 (1 RCT)	⨁◯◯◯ Moderatea	‐
Quality of life (not measured)	‐	‐	‐	‐	‐	Not measured
The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; GMFCS: Gross Motor Function Classification System; GMFM‐E: Gross Motor Function Measure—Dimension E; MD: mean difference; RCT: randomized controlled trial; SFA: School Function Assessment.
GRADE Working Group grades of evidence. High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

^aWe downgraded the quality of evidence for imprecision by one level due to insufficient studies.
^bWe downgraded the quality of evidence by one level for study design limitations due to several ratings of high or unclear risk of bias.

Background

Description of the condition

Cerebral palsy is defined as “a group of permanent disorders of the development of movement and posture, causing activity limitation, that are attributed to non‐progressive disturbances that occurred in the developing fetal or infant brain. The motor disorders of cerebral palsy are often accompanied by disturbances of sensation, perception, cognition, communication, and behavior, by epilepsy, and by secondary musculoskeletal problems" (Rosenbaum 2007). The potential causes of cerebral palsy are associated with certain prenatal, perinatal, and postnatal events such as hypoxic, ischemic, infectious, congenital, or traumatic insults affecting the developing brain (Wright 2012). Cerebral palsy continues to be the most common physical disability seen in childhood, with a worldwide incidence of 2 to 4 per 1000 live births (Oskoui 2013; Stavsky 2017; Wright 2012; Yeargin‐Allsopp 2008). The incidence of cerebral palsy has remained the same worldwide due to increased survival rates of preterm and very preterm infants who have resultant comorbidities (Durkin 2016; Hafström 2018; Stavsky 2017; Thygesen 2016; Van Naarden Braun 2016). Although cerebral palsy is a non‐progressive neurological lesion resulting in motor impairment, the condition is not unchanging, and various impairments of the neuromuscular, musculoskeletal, and sensory systems become features of infancy and early childhood (Wright 2012). The most common type of cerebral palsy is hypertonia/spasticity (85%), followed by dyskinesia (7%), ataxia (4%), and hypotonia (3%) (ACPR 2018). The distribution of motor impairment may be bilateral (as in diplegia, quadriplegia, or triplegia) or unilateral (as in hemiplegia). Within the hypertonia/spasticity type, the most common motor distribution is hemiplegia (40%), followed by diplegia (36%) and quadriplegia (24%) (ACPR 2018). Children with a dyskinetic or ataxic motor type are generally classified as having a quadriplegic distribution. The severity of the impairment can interfere with mobility, such as performance of walking, over the lifespan. Most children (70%) develop sufficient motor control to walk (ACPR 2018; Koop 2009), but their ability to walk can gradually decline over time, especially during adolescence (Hanna 2009), and into adulthood (Opheim 2012).

The Gross Motor Function Classification System (GMFCS) is a five‐level classification system developed by Palisano 1997 to differentiate mobility function in children with cerebral palsy. Children within GMFCS Level I or II (62%) can walk without devices, and children within Level III (11%) may use any type of mobility device for walking and wheeled devices for longer distances (ACPR 2018). Even though children with cerebral palsy who have a GMFCS rating of Level I or II and who can walk independently without devices may retain walking (Hanna 2009; McCormick 2007), gait performance may worsen (Johnson 1997; Maanum 2010), and walking speed may decrease (Gannotti 2008; Maanum 2010). Within GMFCS Level III, walking ability may even be lost during adolescence or young adulthood (Bottos 2001; Day 2007), although children within GMFCS Level IV have very limited walking ability, and those within GMFCS Level V have no walking ability (Palisano 1997).

Altogether, these studies suggest that further research is required to investigate the effectiveness of mobility with walking‐based training using mechanically assisted or overground walking to improve and maintain both of the following.

1. Walking performance for ambulant (GMFCS Level I or II) or semi‐ambulant (GMFCS Level III) children with cerebral palsy.

2. Other functional outcomes at the International Classification of Functioning, Disability and Health (ICF) Activity and Participation levels, or quality of life, or both (Majnemer 2012), for children with restricted walking ability (GMFCS Level I, II, or III) or no walking ability (GMFCS Level IV or V).

Description of the intervention

One intervention that may provide the opportunity for more focused walking training is mechanically assisted training on a treadmill or robotic device. Mechanically assisted training can be provided with or without body weight support to enable children with cerebral palsy to perform repetitive practice of complex gait cycles in the clinical setting (Bryant 2013; Gates 2012; Grecco 2016; Sherief 2015; Swe 2015). The premise of this approach is that children are helped to walk at increasing speed by first providing and then reducing weight‐bearing support for the lower limbs when walking on the device. Improvement in walking may, in turn, be carried over to improved walking overground in the community. Any increase in independent mobility may improve the child’s participation (attendance and involvement) in everyday activities (Palisano 2009; Palisano 2011), which may also contribute to improved quality of life (Whittingham 2010).

Mechanically assisted walking training consists of using a treadmill (with or without body weight support and the assistance of one or more therapists), an end‐effector system (such as a gait trainer, with or without body weight support), or a robotic training device (such as Lokomat® 2018). Treadmill or gait trainers offer body weight support through a solid or fabric seat to support the pelvis and stabilize the trunk during walking (Damiano 2009; Mattern‐Baxter 2009; Mutlu 2009; Paleg 2015; Willoughby 2009). Use of a robotic training device requires the child’s legs to be placed into supports attached to a machine, which then assist walking at a set tempo or step length (Arellano‐Martínez 2013; Drużbicki 2013; Wiart 2016); set up may be time consuming. Robotic training devices offer the advantage of being able to assist with repetitive limb movement without occupational risk to therapists; however, two systematic reviews found that robotic assisted training is not effective for improving activities (e.g. walking speed) in children with cerebral palsy (Carvalho 2017; Lefmann 2017). It is still not clear whether the walking pattern produced through mechanically assisted walking is sufficient to improve overground walking.

Overground walking training is the most common alternative to mechanically assisted walking training. Overground walking is an important activity of daily life that enhances independence across a wide range of participation domains (Palisano 2009; Palisano 2011). However, assisting the practice of walking can be physically demanding for therapists (e.g. sustained bending or holding a leg to guide a proper gait pattern), causing muscle pain in the therapist's back or upper limbs and, as result, may not produce enough repetitions to benefit the patient's performance in everyday life. This provides motivation for therapists to seek an alternative method of walking training for children with cerebral palsy.

How the intervention might work

Similar to learning or adapting any motor skill, the actual practice of walking is necessary to improve walking (Shummway‐Cook 2016). Mechanically assisted walking is a type of locomotor intervention for improving walking ability that is based on the principles of task‐specific training and involves intensive, repetitive practice. Two systematic reviews of more than 20 randomized controlled trials (RCTs) found constraint‐induced movement therapy (CIMT) to be effective in improving upper limb function in children with hemiplegic cerebral palsy, albeit no more effective than the same dose of upper limb therapy without restraint (Chen 2014; Chiu 2016). CIMT involves intensive and targeted practice of the more affected limb during restraint of the less affected limb (i.e. forcing children to find solutions to their movement problems by using only the more affected upper limb); this implies that the mechanism of the effect is the dose of practice rather than the type of practice. Mechanically assisted waking training has the potential to provide extensive task‐specific walking practice, similar to the CIMT approach; thus it is possible that treadmill training may benefit overground walking.

Why it is important to do this review

Retrospective studies have indicated that many children with cerebral palsy are at risk for deteriorating walking status during adolescence and young adulthood (Bottos 2001; Day 2007; Gannotti 2008; Hanna 2009; Jahnsen 2004; Johnson 1997; Voorman 2007). This deterioration can even result in loss of walking, particularly among children whose walking is already poor or borderline (Bottos 2001; Day 2007; Jahnsen 2004). Therefore, implementing interventions targeting walking ability once children with cerebral palsy have developed the ability to walk should be an ongoing priority.

Seven previous systematic reviews have examined the effects of mechanically assisted walking in cerebral palsy: four reviews on treadmill training (Damiano 2009; Mattern‐Baxter 2009; Mutlu 2009; Willoughby 2009); two reviews on robotic assisted training (Carvalho 2017; Lefmann 2017); and one review on gait trainer training (Paleg 2015). Although these reviews have provided preliminary evidence on the effects of mechanically assisted walking training, five of them included all published studies, regardless of design and level of evidence (Carvalho 2017; Damiano 2009; Mattern‐Baxter 2009; Mutlu 2009; Paleg 2015), and one was a review without a meta‐analysis, which was not restricted to participants with cerebral palsy (Paleg 2015). Two systematic reviews with meta‐analyses reported that robotic assisted training confers no beneficial effect on walking speed for children with cerebral palsy (Carvalho 2017; Lefmann 2017), but one of these included only two RCTs (Lefmann 2017), and the other analyzed the results from all published studies without conducting a subgroup analysis for RCTs (Carvalho 2017). Even though two reviews published in 2009 conducted a meta‐analysis (Damiano 2009; Willoughby 2009), neither review reported an effect size; therefore it is not possible to judge the clinical significance of the results. Furthermore, since the time of publication of these reviews, a significant number of clinical trials have examined mechanically assisted walking training for children with cerebral palsy (Drużbicki 2013; Gates 2012; Gharib 2011; Grecco 2013a; Sherief 2015; Smania 2011; Swe 2015; Willoughby 2009).

To improve clinical practice, it is therefore important to examine the effects of mechanically assisted walking training to provide scientific guidance for selecting treatments to improve walking performance, and possibly to prevent walking deterioration with age. In doing so, it is critical to examine the effectiveness of all types of mechanically assisted walking training, including training using (1) a treadmill with or without body weight support; (2) gait training with and without body weight support; and (3) a robotically assisted device.

Objectives

To assess the effects of mechanically assisted walking training compared to control for walking, participation, and quality of life in children with cerebral palsy 3 to 18 years of age.

Methods

Criteria for considering studies for this review

Types of studies

To enable recommendations based on the highest level of evidence, this review included only randomized controlled trials (RCTs), quasi‐RCTs (i.e. where participants are allocated in a way that is not strictly random, such as by alternation or date of birth), and randomized cross‐over trials.

Types of participants

Boys and girls with cerebral palsy, classified within GMFCS Levels I to IV, who were able to understand simple instructions and were 3 to 18 years of age.

Types of interventions

Any type of mechanically assisted walking training (e.g. treadmill, Lokomat, gait trainer) in any direction (forward or backward) with any support (with or without body weight support), irrespective of duration per session.

Comparisons were no or sham walking training or the same dose of overground walking training (with or without the help of therapists or walking aids, or both). No or sham walking training was defined as walking training for less than 30% of the time that the intervention group spent walking. Participants could have received usual therapy as long as both groups received it.

Types of outcome measures

We employed commonly used measures that have been designed specifically for use with children, or measures designed for use with adults and subsequently validated for use in children. We performed separate analyses for our primary and secondary outcomes post intervention. We decided on which primary and secondary outcomes should be included in the analysis through discussion until agreement was reached.

Primary outcomes

1. Mobility, measured directly by tests that produce continuous outcome data, such as the 10‐meter walk test (Watson 2002), or the six‐minute walk test (Maher 2008).

2. Gross motor function, measured through gross motor assessment measures that include walking item(s) and scored on categorical scales (e.g. Gross Motor Function Measure [GMFM] dimensions D and E; Russell 2010).

3. Adverse events, measured using the incidence of adverse outcomes, such as injury and pain, or any other reported adverse events. Pain may have been measured on a binary scale (e.g. pain, no pain), a categorical scale (e.g. the Faces Pain Scale; Bieri 1990), or a continuous scale (e.g. a visual analogue scale). We anticipated that two conditions might have occurred—one was that pain from the intervention did not interfere with training and so training could continue; the other was that pain from the intervention resulted in a need to discontinue training.

Secondary outcomes

1. Participation, measured via ordinal data, for example, the Capacity Profile (CAP; Meester‐Delver 2007), or the Goal Attainment Scale (GAS; Kiresuk 2014).

2. Quality of life, measured via ordinal data, for example, the Pediatric Quality of Life Inventory (PedsQL; Varni 1999).

Search methods for identification of studies

Electronic searches

We searched the following electronic databases and trials registers up to January 2020.

1. Cochrane Central Register of Controlled Trials (CENTRAL; 2020, Issue 1), in the Cochrane Library, which includes the Developmental, Psychosocial and Learning Problems Specialized Register (searched January 28, 2020).

2. MEDLINE Ovid (1946 to January 28, 2020).

3. MEDLINE In‐Progress and Other Non‐Indexed Citations Ovid (searched January 28, 2020).

4. MEDLINE EPub ahead of Print Ovid (searched January 28, 2020).

5. Embase Ovid (1980 to January 28, 2020).

6. Cumulative Index to Nursing and Allied Health Literature (CINAHL) EBSCOhost (1982 to January 28, 2020).

7. Cochrane Database of Systematic Reviews (CDSR; 2020, Issue 1), part of the Cochrane Library (searched January 28, 2020).

8. Epistemonikos (www.epistemonikos.org; searched January 28, 2020).

9. ProQuest Dissertations & Theses (1990 to January 28, 2020).

10. Science Citation Index‐Expanded Web of Science (1991 to January 28, 2020).

11. Physiotherapy Evidence Database (PEDro; www.pedro.org.au; searched January 28, 2020).

12. ClinicalTrials.gov (clinicaltrials.gov; searched January 28, 2020).

13. World Health Organization International Clinical Trials Registry Platform (WHO ICTRP; who.int/ictrp/en; searched January 28, 2020).

We did not limit the search by year, language, or type of publication. The search strategies for each database are reported in Appendix 1. We ran pre‐publication searches in MEDLINE and Embase on June 11, 2020, to check that none of our included studies had been retracted.

Searching other resources

We reviewed the reference lists of included studies to identify additional studies. We handsearched Developmental Medicine and Child Neurology (last searched March 15, 2020) to identify relevant studies from conference reports of organizations such as the American Academy for Cerebral Palsy and Developmental Medicine, the European Academy of Childhood Disability, and the Australasian Academy of Cerebral Palsy and Developmental Medicine. We contacted the authors of any potentially eligible study to request the necessary data to enable us to assess the study for inclusion (Criteria for considering studies for this review).

Data collection and analysis

In successive sections, we report only the methods we used in this review. Please see our protocol—Chiu 2018—and Table 5 for methods that we will use in future updates of this review.

1. Unused methods.

Unused methods	Description	Reasons for non‐use
Measures of treatment effects	Dichotomous data For dichotomous data (e.g. adverse events), we had planned to report the pooled estimate of the intervention effect as a risk difference in occurrence of events, along with 95% confidence intervals	No such data
Unit of analysis issues	Cluster‐RCTs We had not expected to encounter any cluster‐RCTs. However, had we found any, we may have included these studies in the same analysis (i.e. combined single RCTs with cluster‐RCTs), provided the designs and interventions of the cluster‐RCTs had been similar to those of the single RCTs. When appropriate, we would have included cluster‐RCTs in subgroup and sensitivity analyses	We did not find any cluster‐RCTs
Assessment of reporting biases	We had planned to use funnel plots to investigate similarity among studies and Egger's test to investigate the relationship between study effect sizes and their standard error (Egger 1997). Had we found funnel plots with significant asymmetry, we would have examined variations between studies for publication bias and other small study effects. Common tests of publication bias lack sensitivity, so we would have considered whether the effects of smaller trials differed systematically from those of larger trials, as smaller trials tend to have lower methodological quality and tend to produce exaggerated effect sizes. Also, we would have looked at the quality of the studies, sampling of participants, and interventions and outcome measures reported	Too few studies (fewer than 10)
Subgroup analysis and investigation of heterogeneity	We had planned to conduct the following subgroup analyses 1. Ambulatory status of participants (GMFCS Levels I to III vs GMFCS Levels II to IV) 2. Type of mechanical assistance (mechanically assisted [and type] vs overground walking) 3. Dose of intervention (session duration and number of sessions) 4. Intensity (frequency of sessions: intensive vs distributed) 5. Difficulty of the intervention (e.g. treadmill speed) 6. Motor type and distribution (e.g. spastic diplegia vs spastic hemiplegia vs spastic diplegia plus additional dystonia)	Overlap with GMFCS Levels or insufficient data
Sensitivity analysis	We had planned to conduct sensitivity analyses to determine the impact of study quality on the robustness of findings. We had intended to conduct analyses of all included studies, including cluster‐RCTs, using a fixed‐effect model to inform us about the effects of smaller studies or of statistical heterogeneity. If we had found evidence of statistical heterogeneity, we would have conducted further analyses to assess methodological heterogeneity. Specifically, we would have repeated the analyses using the random‐effects model, limited only to those studies that had the following 1. Low risk of selection bias (those with random sequence generation and allocation concealment) 2. Low risk of detection bias (those with blinded measurement) 3. Low risk of attrition bias (those with complete outcome data) We may have conducted additional analyses for any issues due to clinical heterogeneity, such as the amount of usual care between studies, that may have impacted the robustness of findings	Very low heterogeneity, I² close to 0%
GMFCS: Gross Motor Functional Classification System; RCT: randomized controlled trial.

Selection of studies

Two review authors (HC, TB) independently screened all titles and abstracts identified by the search strategy and discarded clearly irrelevant records. They next obtained the full‐text reports of all potentially eligible studies, and one review author (HC) screened the reference lists for additional studies. Two review authors (HC,TB) independently assessed all retrieved papers for eligibility against the inclusion criteria (Criteria for considering studies for this review). They resolved disagreements or ambiguous issues by discussion with a third review author (LA). We recorded key studies excluded after full‐text assessment in a Characteristics of excluded studies table and listed the reasons for exclusion. We illustrated the selection process in a PRISMA diagram (Moher 2009).

Data extraction and management

Two review authors (HC, TB) independently extracted data on each of the following topics, entering them onto a spreadsheet specifically designed for the purpose and piloted. They resolved disagreements or ambiguous issues by discussion with a third review author (LA).

1. Study design.

2. Participants (type of cerebral palsy, age, GMFCS Level).

3. Intervention (type and parameters of mechanically assisted walking, duration of intervention, and comparators).

4. Outcomes (e.g. mobility, participation, quality of life, timing of outcome measures).

We treated multiple reports of the same trial as one study. We contacted the study authors of papers with missing information and data and asked them to provide the additional information (see Dealing with missing data).

Assessment of risk of bias in included studies

Two review authors (HC, TB) independently assessed the risk of bias in each included study using Cochrane's "Risk of bias" tool (Higgins 2017). For each included study, both review authors independently assessed the risk of bias within each of the following seven domains as low, high, or unclear: sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessment; incomplete outcome data (including data on attrition and exclusions, differentiating intention‐to‐treat analyses from per‐protocol (‘as treated') analyses); selective outcome reporting; and other sources of bias. We made judgements according to the guidelines reported in Table 8.5a in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2017); the criteria for these judgements are presented in Appendix 2. Both review authors resolved disagreements or ambiguous issues by discussion with the other review author (LA). We entered our judgements into a "Risk of bias" table in Review Manager 5 (RevMan 5) (Review Manager 2014), along with a brief rationale for the judgements.

Overall risk of bias

We considered a study at low risk of bias overall if it was at low risk of bias for all key domains; at high risk of bias overall if it was at high risk of bias for more than one key domain; and at unclear risk of bias overall if it was at unclear risk of bias for more than one key domain (no study was rated at both high and unclear risk of bias for more than one domain). We define a "key domain" as a domain that we considered important.

We used the results of the "Risk of bias" assessment to conduct a sensitivity analysis (Sensitivity analysis), and to inform GRADE analysis of the quality of the body of evidence.

Measures of treatment effect

Continuous data

For continuous outcomes, when studies used the same measures, we reported the pooled estimate of the intervention effect as a mean difference (MD) and presented this with 95% confidence intervals (CIs). When measures were different, we presented the standardized MD (SMD) with 95% CI.

Dichotomous data

No dichotomous data (e.g. adverse events) were provided in the included studies.

Ordinal data

We converted results from ordinal data (e.g. GMFM dimensions D and E) to a percentage (%).

Unit of analysis issues

Cluster‐randomized trials

We did not find any cluster‐randomized trials.

Cross‐over trials

We combined results from the first phase of cross‐over trials with those from parallel‐group trials. We used only results from the first phase of cross‐over trials because in a cross‐over design, participants are exposed to multiple treatment conditions and repeated measures, which might result in bias due to practice, learning, or carry‐over effects (Portney 2009).

Multiple outcomes

When multiple measures of the same outcome were available, we chose the most direct measure that yielded continuous data, in preference to scales returning ordinal data. For example, when a study included several direct measures of mobility, we selected the 10‐meter walk test (if available), as this best reflects walking performance. If this was not available, we opted for the six‐minute walk test as our second choice. If the study did not report these data on mobility, we selected laboratory walking test to include mobility in the analysis. Once the outcome measure was decided, if both change and postintervention scores were available, we selected change scores for the meta‐analysis.

Multiple time points

When outcomes were measured at multiple time points, we selected outcomes measured immediately post intervention. If these were unavailable, we used the last time point of measurement post intervention.

Dealing with missing data

We identified missing data and dropouts for each included study and reported this information in the "Risk of bias" tables (beneath the Characteristics of included studies tables). We contacted study authors up to three times to obtain missing data. If we were unable to obtain missing data (e.g. missing standard deviations [SDs]) from the study authors, we calculated them using t values, P values, CIs, or standard errors, when reported. If these statistics were not available for calculating and imputing missing data, we did not include the study in the comparison, and we discussed in the Discussion section the extent to which the missing data were likely to influence results.

Assessment of heterogeneity

We assessed clinical heterogeneity by considering diversity in participant characteristics (such as age of participants) and interventions. We also assessed methodology heterogeneity such as the method of group allocation (see Subgroup analysis and investigation of heterogeneity and Sensitivity analysis).

We assessed statistical heterogeneity between studies by visually inspecting the forest plot for overlapping CIs. In addition, we used the Chi² test for homogeneity with a significant level of α (alpha) = 0.1, and we calculated the I² statistic for estimating the percentage of variation in effect estimates due to heterogeneity rather than sample error. We interpreted I² values using the thresholds listed below. These thresholds were suggested by the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2017).

1. 0% and 40% may represent little heterogeneity.

2. 30% and 60% may represent moderate heterogeneity.

3. 50% and 90% may represent substantial heterogeneity.

4. 75% and 100% may represent considerate heterogeneity.

When interpreting the I² statistic, we also took into account the size and direction of effects and the strength of evidence for heterogeneity based on the P value of the Chi² test. Because we used the random‐effects model, we also estimated and presented Tau², along with its CI, to provide an estimate of the extent of between‐study variability.

Assessment of reporting biases

We were unable to assess reporting bias as planned because too few studies (fewer than 10) were included in a meta‐analysis.

Data synthesis

We synthesized the data using RevMan 5 (Review Manager 2014). We performed meta‐analyses when studies had sufficiently similar interventions (e.g. in terms of type of intervention; intervention frequency, intensity, and duration) to ensure that the result was clinically meaningful. This was decided by clinical experts on the review author team (HC, TB). As we still expected some heterogeneity in intervention parameters, we pooled the available data using a random‐effects model. When we could not conduct a meta‐analysis, we provided a narrative description of the results.

Specific comparisons assessed effects of the following.

1. Mechanically assisted walking training without body weight support compared with no walking training.

2. Mechanically assisted walking training without body weight support compared with the same dose of overground walking training.

3. Mechanically assisted walking training with body weight support compared with no walking training.

4. Mechanically assisted walking training with body weight support compared with the same dose of overground walking training.

"Summary of findings" table

We summarized the results per comparison in a "Summary of findings" table, which we constructed using the GRADEpro Guideline Development Tool (GRADEpro GDT). We included all of our primary and secondary outcomes in each table (see Types of outcome measures). We reported outcomes assessed immediately post intervention, unless this information was not available, in which case we reported the last time point of measurement post intervention.

We used the five GRADE considerations (risk of bias, consistency of effect, imprecision, indirectness, and publication bias) to assess the quality of the body of evidence for each outcome, and to draw conclusions about the quality of evidence within the text of the review. Using the GRADE approach (Schünemann 2017), two review authors (HC, TB) independently determined the quality of the evidence for each outcome as high, moderate, low, or very low, and resolved discrepancies through discussion until they reached agreement.

Subgroup analysis and investigation of heterogeneity

We conducted the following subgroup analysis by using a random‐effects model to explore the results and assess heterogeneity.

1. Age of participants.

Sensitivity analysis

We did not conduct sensitivity analyses because of very low heterogeneity; I² was close to 0%.

Results

Description of studies

Results of the search

We identified a total of 5513 records of potentially relevant trials, 1698 of which were duplicates and were subsequently discarded. We screened the titles and abstracts of the 3815 remaining records and excluded 3742 of them. We considered the eligibility of 73 full‐text reports, from which we included 23 reports of 17 studies and excluded 43 studies.

Seven studies are ongoing. See Figure 1 for flow of studies through the review.

1.

Flow of studies through the review.

Included studies

We included 23 reports of 17 studies (Ameer 2019; Bryant 2013; Cherng 2007; Dodd 2007; Druzbicki 2013; El‐Shamy 2017; El‐Shemy 2018; Gates 2012; Gharib 2011; Grecco 2013a; Hösl 2018; Peri 2017; Sherief 2015; Smania 2011; Swe 2015; Wallard 2017; Willoughby 2010). The detailed characteristics of the included studies are described in the Characteristics of included studies tables.

Study design

All studies were RCTs. Two studies used a cross‐over design with random allocation to the order of intervention (Cherng 2007; Hösl 2018). All other studies used a parallel‐group design with true randomization or quasi‐randomization (Peri 2017), or were cluster‐controlled (Dodd 2007).

Location/Setting

All 17 studies took place in an outpatient setting and in several geographical locations: Australia (Dodd 2007; Swe 2015; Willoughby 2010); Belgium (Wallard 2017); Brazil (Grecco 2013a); Egypt (Ameer 2019; El‐Shamy 2017; El‐Shemy 2018; Gharib 2011; Sherief 2015); Germany (Hösl 2018); Italy (Peri 2017); Poland (Druzbicki 2013); Taiwan (Cherng 2007); the UK (Bryant 2013; Smania 2011); and the USA (Gates 2012).

Participants

The 17 studies included a total of 451 participants with cerebral palsy, with mean age across studies ranging from 4 to 14 years. Seven studies (41%) involved participants younger than nine years of age (Ameer 2019; Cherng 2007; Dodd 2007; Grecco 2013a; Peri 2017; Sherief 2015; Wallard 2017).

The GMFCS Level of participants ranged from I to IV. Twelve studies (71%) involved participants within GMFCS Level I, II, or III (Ameer 2019; Cherng 2007; Druzbicki 2013; El‐Shamy 2017; El‐Shemy 2018; Gharib 2011; Grecco 2013a; Hösl 2018; Peri 2017; Sherief 2015; Swe 2015; Wallard 2017).

The number of participants per trial ranged from 10 in Hösl 2018 to 35 in Druzbicki 2013 and Grecco 2013a. Ten studies (59%) involved more than 30 participants. Two studies (13%) involved a small number of participants (eight participants in Cherng 2007 and 10 participants in Hösl 2018).

Interventions

Four studies compared mechanically assisted walking without body weight support to no walking training (El‐Shemy 2018; Gharib 2011; Hösl 2018; Sherief 2015). Two studies compared mechanically assisted walking without body weight support to the same dose of overground walking training (Ameer 2019; Grecco 2013a). Eight studies compared mechanically assisted walking with body weight support to no walking training (Bryant 2013; Cherng 2007; Dodd 2007; Druzbicki 2013; El‐Shamy 2017; Gates 2012; Smania 2011; Wallard 2017). Three studies compared mechanically assisted walking with body weight support to the same dose of overground walking training (Peri 2017; Swe 2015; Willoughby 2010).

Mechanically assisted walking training involved using a treadmill without body weight support in three studies (Grecco 2013a; Hösl 2018; Sherief 2015), a treadmill with body weight support in six studies (Bryant 2013; Cherng 2007; Dodd 2007; Gates 2012; Swe 2015; Willoughby 2010), a gait trainer without body weight support in one study (Gharib 2011), a gait trainer with body weight support in two studies (El‐Shamy 2017; Smania 2011), and a robotic training device in three studies (Druzbicki 2013; Peri 2017; Wallard 2017).

Intensity of mechanically assisted walking without body weight support ranged in length of sessions (15 to 30 minutes) and frequency of sessions (two to three times a week) for an intervention period of 11 weeks (7 to 12 weeks). Intensity of mechanically assisted walking with body weight support ranged in length of sessions (20 to 40 minutes) and frequency of sessions (two to five times a week) for an intervention period of eight weeks (4 to 12 weeks). Nine studies (53%) trained participants for 30 minutes a session (Bryant 2013; Dodd 2007; El‐Shemy 2018; Gates 2012; Grecco 2013a; Peri 2017; Smania 2011; Swe 2015; Willoughby 2010), eight studies (47%) trained participants three times a week (Ameer 2019; Bryant 2013; Cherng 2007; El‐Shamy 2017; El‐Shemy 2018; Gharib 2011; Hösl 2018; Sherief 2015), and seven studies (41%) trained participants for 12 weeks (Ameer 2019; Cherng 2007; El‐Shamy 2017; El‐Shemy 2018; Gates 2012; Gharib 2011; Sherief 2015), using a treadmill (with or without body weight support).

Comparators

Seven studies reported that the control group received no walking training (Cherng 2007; Dodd 2007; Druzbicki 2013; El‐Shamy 2017; El‐Shemy 2018; Gharib 2011; Sherief 2015); four reported that the control group received usual therapy, such as stretching or strengthening exercises, but no walking or sham walking (Bryant 2013; Gates 2012; Hösl 2018; Wallard 2017); and one reported that the control group received overground walking for 25% of the time that the intervention group spent on walking (Smania 2011).

Six of the 12 studies reported that both experimental and control groups received usual therapy (Cherng 2007; Druzbicki 2013; El‐Shamy 2017; El‐Shemy 2018; Gharib 2011; Sherief 2015).

Five studies reported that the control group received the same dose of overground walking training (Ameer 2019; Grecco 2013a; Peri 2017; Swe 2015; Willoughby 2010). Of these five studies, two studies reported that both experimental and control groups received usual therapy (Ameer 2019; Willoughby 2010).

Outcomes

Thirteen studies reported on measures of mobility (Ameer 2019; Cherng 2007; Dodd 2007; Druzbicki 2013; El‐Shamy 2017; Gates 2012; Gharib 2011; Grecco 2013a; Peri 2017; Smania 2011; Swe 2015; Wallard 2017; Willoughby 2010).

Nine studies reported on measures of gross motor function (Bryant 2013; Cherng 2007; El‐Shemy 2018; Grecco 2013a; Hösl 2018; Peri 2017; Sherief 2015Swe 2015; Wallard 2017); four studies reported on measures of participation (Gates 2012; Grecco 2013a; Smania 2011; Willoughby 2010); one study reported on measures of quality of life (Gates 2012); and four studies reported no adverse events at the end of the intervention (Bryant 2013; Dodd 2007; Sherief 2015; Smania 2011).

Missing data

When data were not explicitly reported, we attempted to contact study authors to obtain the missing information, wherever possible. We contacted six study authors to request clarification of some design features or provision of missing information, to complete the quality ratings (correspondence was via email or letter, with a reminder sent after three weeks, then every month up to three times if we did not get a response) (Ameer 2019; Bryant 2013; El‐Shamy 2017; Gates 2012; Hösl 2018; Wallard 2017). We obtained data for gross motor function from Bryant 2013 and clarification on dose of intervention for experimental and control groups from Ameer 2019. No data were provided by, or no contact achieved with, the authors of the other four studies. For El‐Shamy 2017, we used published data for mobility only. No published data were available in Hösl 2018 for inclusion in any analyses. Gates 2012, Johnston 2009, and Johnston 2011 reported different outcomes of the same study. From Gates 2012, we used the Children’s Assessment of Participation and Enjoyment (CAPE) as the outcome for participation, and the Pediatric Quality of Life Cerebral Palsy Module (PedsQLCP) as the outcome for quality of life. From Johnston 2011, we used the 10‐minute walk test as a measure of walking speed (m/s) for the outcome of mobility. Wallard 2017 and Wallard 2018 also reported different outcomes of the same study. From Wallard 2017, we used the Gross Motor Function Measure (GMFM)—dimension E (GMFM‐E)—as the outcome for gross motor function. From Wallard 2018, we used the 10‐minute walk test as a measure of walking speed (m/s) for the outcome of mobility.

Funding sources

Of 17 studies, six studies reported grant support (Bryant 2013; Cherng 2007; Gates 2012; Grecco 2013a; Smania 2011; Willoughby 2010), three reported no funding (Gharib 2011; Sherief 2015; Swe 2015), and eight did not report funding resources (Ameer 2019; Dodd 2007; Druzbicki 2013; El‐Shamy 2017; El‐Shemy 2018; Hösl 2018; Peri 2017; Wallard 2017).

Excluded studies

We excluded 43 studies for the following reasons. Sixteen studies were not RCTs or quasi‐RCTs. In six studies, participants did not have cerebral palsy; in two studies, they were not 3 to 18 years old; and in one study, they were not classified within GMFCS Levels I to IV. The intervention in four studies was not "mechanically assisted walking training with or without body weight support"; and the comparator was not "no walking" or "same dose of overground walking" in 14 studies. See Characteristics of excluded studies tables.

Ongoing studies

We identified seven ongoing studies. All were RCTs that were conducted in Australia (ACTRN12617001410347), Switzerland (NCT00887848), Canada (NCT02196298), the USA and Canada (NCT02391324), The Netherlands (NL8154), Egypt (PACTR201901582864286), and an unknown location (Yang 2019). Five studies used a cross‐over design, and two studies used a factorial design (NCT02391324; PACTR201901582864286).

The number of participants that the studies recruited ranged from 10 in ACTRN12617001410347 to 160 in NCT02391324. Participants ranged in age from 5 to 14 years in ACTRN12617001410347, from 6 to 18 years in NCT00887848 and PACTR201901582864286, from 5 to 12 years in NCT02196298, from 5 to 18 years in NCT02391324, from 6 to 17 years in NL8154, and from 3 to 12 years in Yang 2019.

Three studies involved children with cerebral palsy within GMFCS Levels II to IV (ACTRN12617001410347; NCT00887848; Yang 2019); two studies involved children with cerebral palsy within GMFCS Levels II to III (NCT02196298; NCT02391324); one study involved children with cerebral palsy within GMFCS Levels I to II (NL8154); and the other study had no restriction on GMFCS Levels (PACTR201901582864286).

Seven studies investigated the effects of robot‐assisted gait training compared to wait‐list control (ACTRN12617001410347; NCT00887848; NL8154), to the same dose of usual therapy (NCT02196298; PACTR201901582864286), to the same dose of overground walking (NCT02391324), or to usual care (Yang 2019).

See Characteristics of ongoing studies tables.

Risk of bias in included studies

Two review authors (HC and TB) independently assessed the methodological quality of the included studies using Cochrane's "Risk of bias" tool (across the categories of random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and other sources of bias; see "Risk of bias" tables beneath the Characteristics of included studies tables, as well as Figure 2 and Figure 3).

2.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

3.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

Random sequence generation

Ten of the 17 included studies described the method of random sequence generation and were rated as being at low risk of bias in this domain (Ameer 2019; Bryant 2013; El‐Shamy 2017; El‐Shemy 2018; Gates 2012; Grecco 2013a; Sherief 2015; Smania 2011; Swe 2015; Willoughby 2010). We rated two studies as potentially at high risk of bias because they used quasi‐randomization methods (Dodd 2007; Peri 2017). We rated the five remaining studies as having unclear risk of bias due to lack of information (Cherng 2007; Druzbicki 2013; Gharib 2011; Hösl 2018; Wallard 2017).

Allocation concealment

Nine of the 17 included studies described the method of concealing allocation and were rated as being at low risk of bias in this domain (Ameer 2019; Bryant 2013; El‐Shamy 2017; Gharib 2011; Grecco 2013a; Sherief 2015; Smania 2011; Swe 2015; Willoughby 2010). We rated two studies as potentially at high risk of bias because of no allocation concealment (Dodd 2007; Peri 2017). We rated the six remaining studies as having unclear risk of bias due to lack of information (Cherng 2007; Druzbicki 2013; El‐Shemy 2018; Gates 2012; Hösl 2018; Wallard 2017).

Blinding

Performance bias

None of the 17 included studies described participants or therapists as being blinded to group allocation. As a result, we judged all 17 included studies to be at high risk of performance bias.

Detection bias

We judged 12 of the 17 included studies to be at low risk of detection bias because they described the outcome assessors as blinded to group allocation. We rated the other five studies at high risk of detection bias because they had no blinded assessors (Ameer 2019; Dodd 2007; Hösl 2018; Peri 2017; Sherief 2015).

Incomplete outcome data

We judged three studies to be at high risk of attrition bias because they had dropout rates greater than 15% and did not report an intention‐to‐treat analysis (Bryant 2013; Druzbicki 2013; Gates 2012). We judged one study to be at unclear risk of attrition bias because this study had a dropout rate greater than 15% (Willoughby 2010). We judged 13 studies to be at low risk of bias because they had dropout rates less than 15% and had conducted an intention‐to‐treat analysis (Dodd 2007; Gharib 2011; Grecco 2013a; Swe 2015), or they had no dropouts (Ameer 2019; Cherng 2007; El‐Shamy 2017; El‐Shemy 2018; Hösl 2018; Peri 2017; Sherief 2015; Smania 2011; Wallard 2017).

Selective reporting

We judged all 17 included studies to be at low risk of reporting bias because they reported between‐group differences, along with point estimates and variability in either change scores or preintervention and postintervention scores for each group (Figure 2; Figure 3). We found only one study protocol (Swe 2015), and we noted no evidence of selective reporting of outcomes relevant to this review.

Other potential sources of bias

All 17 included studies had no other obvious issues (Figure 2; Figure 3), but all reported small sample sizes, with 25 participants per trial on average; this could result in vulnerability to small trial bias for this review. Only Swe 2015 had a protocol registered (ACTRN12613001025729), and the methods reported in Swe 2015 were consistent with the published protocol. No other obvious issues were found in all 17 included studies; therefore, they were rated having low risk for other potential sources of bias.

Effects of interventions

See: Table 1; Table 2; Table 3; Table 4

Comparison 1. Mechanically assisted walking training without body weight support versus no walking training

Primary outcomes

Mobility

One study with 10 participants provided change scores for walking speed (meters per second [m/s]) immediately after the intervention (12 weeks); walking speed = average speed over three episodes of three minutes of walking on the Biodex Gait Trainer 2™ (Biodex, Shirley, NY, USA) (Gharib 2011). The use of mechanically assisted walking training without body weight support for children with cerebral palsy increased walking speed compared with no walking training. The pooled mean difference (MD) for walking speed was 0.05 m/s (95% confidence interval [CI] 0.03 to 0.07; P < 0.001; random‐effects model; moderate‐quality evidence; see the illustrative forest plot in Analysis 1.1; Table 1).

1.1. Analysis.

Comparison 1: Mechanically assisted walking training without body weight support vs no walking training, Outcome 1: Mobilty (walking speed): Biodex Gait Trainer 2™ (m/s) change from baseline to end of treatment (12 weeks)

Gross motor function

At 12 weeks post intervention, two studies with a total of 60 participants provided postintervention scores for gross motor function, measured on the locomotive subtest of the Peabody Developmental Motor Scale and reported as a percentage (Sherief 2015), or using the modified Timed Up and Go (mTUG) test and reported in seconds (El‐Shemy 2018). We converted the unit of mTUG from second to ‐second in El‐Shemy 2018 to have the same direction of score for gross motor function. The use of mechanically assisted walking training without body weight support for children with cerebral palsy improved gross motor function compared with no walking training. The pooled standardized mean difference (SMD) was 1.30 (95% CI 0.49 to 2.11; P = 0.002; random‐effects model; Tau² = 0.17; Chi² = 2.00, df = 1 [P = 0.16]; I² = 50%; Analysis 1.2; low‐quality evidence; Table 1).

1.2. Analysis.

Comparison 1: Mechanically assisted walking training without body weight support vs no walking training, Outcome 2: Gross motor function: PDMS‐2, locomotive subtest (%) or mTUG (s) at end of treatment (12 weeks)

Adverse events

One study with a total of 30 participants reported that there were no adverse events (Sherief 2015). We rated the quality of this evidence as low.

No study measured the secondary outcomes of participation and quality of life.

Comparison 2. Mechanically assisted walking training without body weight support versus the same dose of overground walking training

Primary outcomes

Mobility

Two studies (55 participants) provided change scores—Grecco 2013a—or postintervention scores—Ameer 2019—for walking speed (m/s), measured using the 6‐minute walk test and Vicon gait analysis, respectively, from baseline to post intervention or at post intervention (7 to 12 weeks). We converted the unit of walking speed in Grecco 2013a from m to m/s and 95% CI to standard deviation (SD). The use of mechanically assisted walking training without body weight support for children with cerebral palsy resulted in a greater increase in walking speed compared with the same dose of overground walking training. The pooled MD for walking speed was 0.25 m/s (95% CI 0.13 to 0.37; P < 0.001; random‐effects model; Tau² = 0.00; Chi² = 2.42, df = 1 [P = 0.12]; I² = 59%; Analysis 2.1; moderate‐quality evidence; Table 2).

2.1. Analysis.

Comparison 2: Mechanically assisted walking training without body weight support vs the same dose of overground walking training, Outcome 1: Mobility: 10‐meter walk test or gait analysis (m/s) change from baseline to end of treatment or at end of treatment (7 to 12 weeks)

Gross motor function

One study with a total of 35 participants provided change scores (Grecco 2013a) for gross motor function (%), measured using the E subscale—walking, running and jumping—of the GMFM‐88, from baseline to postintervention (seven weeks). The use of mechanically‐assisted walking training without body weight support for children with cerebral palsy improved gross motor function compared with the same dose of overground walking training. The pooled MD for GMFM was 11.90% (95% CI 2.98 to 20.82; P = 0.009; random‐effects model; see the illustrative forest plot in Analysis 2.2; moderate‐quality evidence; Table 2).

2.2. Analysis.

Comparison 2: Mechanically assisted walking training without body weight support vs the same dose of overground walking training, Outcome 2: Gross motor function: GMFM‐E (%) change from baseline to end of treatment (7 weeks)

Secondary outcomes

Participation

One study (35 participants) provided change scores from baseline to post intervention (seven weeks) (Grecco 2013a). The use of mechanically assisted walking training without body weight support for children with cerebral palsy improved participation (measured by the Pediatric Evaluation Disability Inventory (PEDI) mobility subscale; scores from 0 to 59) compared to the same dose of overground walking. The pooled MD was 8.20 (95% CI 5.69 to 10.71; P < 0.001; random‐effects model; see the illustrative forest plot in Analysis 2.3; moderate‐quality evidence; Table 2).

2.3. Analysis.

Comparison 2: Mechanically assisted walking training without body weight support vs the same dose of overground walking training, Outcome 3: Participation: PEDI‐mobility (score) change from baseline to end of treatment (7 weeks)

No study measured the primary outcome of adverse events or the secondary outcome of quality of life.

Comparison 3. Mechanically assisted walking training with body weight support versus no walking training

Primary outcomes

Mobility

Seven studies (161 participants) provided change scores measured by the 10‐meter walk test—Dodd 2007; Druzbicki 2013; and Smania 2011—or postintervention scores measured by the 10‐meter walk test—Cherng 2007; Gates 2012; and Wallard 2017—or the 8‐meter walk test—El‐Shamy 2017—for walking speed (m/s) from baseline to end of treatment or at end of treatment (2 to 12 weeks). We converted the unit of walking speed in Cherng 2007 from cm/s to m/s, and in Dodd 2007 from m/min to m/s. The use of mechanically assisted walking training with body weight support for children with cerebral palsy increased walking speed compared with no walking training. The pooled MD was 0.07 m/s (95% CI 0.06 to 0.08; P < 0.001; random‐effects model; Tau² = 0.00; Chi² = 5.58, df = 6 [P = 0.47]; I² = 0%; Analysis 3.1; moderate‐quality evidence; Table 3).

3.1. Analysis.

Comparison 3: Mechanically assisted walking training with body weight support vs no walking training, Outcome 1: Mobility: 10‐Meter Walk Test or 6‐Minute Walk Test (m/s) change from baseline to end of treatment or at end of treatment (2 to 12 weeks)

Gross motor function

Three studies (58 participants) provided change scores—Bryant 2013—or postintervention scores—Cherng 2007 and Wallard 2017—for gross motor function (%), measured on the GMFM‐E immediately after the intervention (4 to 12 weeks). There was little or no difference in gross motor function between the use of mechanically assisted walking training with body weight support and no walking training. The pooled MD was 1.09% (95% CI ‐0.57 to 2.75; P = 0.20; random‐effects model; Tau² = 0.00; Chi² = 1.69, df = 2 [P = 0.43]; I² = 0%; Analysis 3.2; low‐quality evidence; Table 3).

3.2. Analysis.

Comparison 3: Mechanically assisted walking training with body weight support vs no walking training, Outcome 2: Gross motor function: GMFM‐E (%) change from baseline to end of treatment or at end of treatment (4 to 12 weeks)

Adverse events

Three studies (56 participants) reported that there were no adverse events (Bryant 2013; Dodd 2007; Smania 2011). We rated the quality of this evidence as low.

Secondary outcomes

Participation

Two studies with a total of 44 participants provided change scores for participation immediately after the intervention (2 to 12 weeks) (Gates 2012; Smania 2011). The pooled SMD for the intensity subscale of the CAPE (scored 1 to 7) and the WeeFIM (pediatric version of the Functional Independence Measure; scored 18 to 126) was 0.33 SD (95% CI ‐0.27 to 0.93; P = 0.28; random‐effects model; Tau² = 0.00; Chi² = 0.72, df = 1 [P = 0.40]; I² = 0%; Analysis 3.3; low‐quality evidence; Table 3), indicating greater participation in the intervention group than in the control group.

3.3. Analysis.

Comparison 3: Mechanically assisted walking training with body weight support vs no walking training, Outcome 3: Participation: CAPE‐Intensity or WeeFIM (score) change from baseline to end of treatment (2 to 12 weeks)

Quality of life

One study (26 participants) provided change scores for quality of life immediately after the intervention (12 weeks) (Gates 2012). The pooled MD for PedsQLCP (scored 0 to 100, from bad to good) was 9.50 points (95% CI ‐4.03 to 23.03; P = 0.18; random‐effects model; see the illustrative forest plot in Analysis 3.4; low‐quality evidence; Table 3), indicating the possibility of greater quality of life in the intervention group than in the control group.

3.4. Analysis.

Comparison 3: Mechanically assisted walking training with body weight support vs no walking training, Outcome 4: Quality of life: PedsQOL‐CP (score) change from baseline to end of treatment (12 weeks)

Comparison 4. Mechanically assisted walking training with body weight support versus the same dose of overground walking training

Primary outcomes

Mobility

Three studies (78 participants) provided change scores—Swe 2015 and Willoughby 2010—or postintervention scores—Peri 2017) for walking speed (m/s), measured using the 10‐meter and 6‐minute walk tests, respectively, from baseline to end of treatment or at end of treatment (8 to 10 weeks). We converted the unit of walking speed in Peri 2017 from m to m/s. The use of mechanically assisted walking training with body weight support for children with cerebral palsy did not increase walking speed compared with the same dose of overground walking training. The pooled MD was ‐0.02 m/s (95% CI ‐0.08 to 0.04; P = 0.52; random‐effects model; Tau² = 0.00; Chi² = 0.32, df = 2 [P = 0.85]; I² = 0%; Analysis 4.1; low‐quality evidence; Table 4).

4.1. Analysis.

Comparison 4: Mechanically assisted walking training with body weight support vs the same dose of overground walking training, Outcome 1: Mobility: 10‐Meter Walk Test or 6‐Minute Walk Test (%) change from baseline to end of treatment or at end of treatment (8 to 10 weeks)

Swe 2015 provided different scores for change scores in their abstract (0.016 m/s) and text (0.16 m/s). We found the result in the text to be more accurate. We report the data from walking speed at baseline and post intervention in Table 6.

2. Swe 2015 data for the intervention group.

Study	Walking speed (m/s) at baseline	Walking speed (m/s) at post intervention (8 weeks)	Change score in text	Change score in abstract
Swe 2015	0.92	1.08	0.16	0.016

Gross motor function

Two studies (52 participants) provided postintervention scores for gross motor function (%), measured by the GMFM‐E, immediately after the intervention (8 to 10 weeks) (Peri 2017; Swe 2015). There was little or no difference between gross motor function scores between children treated with mechanically assisted walking training without body weight support and the same dose of overground walking training. The pooled MD was ‐0.73% (95% CI ‐14.38 to 12.92, P = 0.92; random‐effects model; Tau² = 0.00; Chi² = 0.13, df = 1 [P = 0.72]; I² = 0%; Analysis 4.2; low‐quality evidence; Table 4).

4.2. Analysis.

Comparison 4: Mechanically assisted walking training with body weight support vs the same dose of overground walking training, Outcome 2: Gross motor function: GMFM‐E (%) at end of treatment (8 to 10 weeks)

Secondary outcomes

Participation

One study (26 participants) provided change scores for participation immediately after the intervention (10 weeks) (Willoughby 2010). No clear evidence showed a difference in participation, measured by the School Function Assessment (SFA; possible scores range from 19 to 76), between mechanically assisted walking training without body weight support and the same dose of overground walking training. The pooled MD was ‐4.74 (95% CI ‐11.89 to 2.41; P = 0.19; random‐effects model; see the illustrative forest plot in Analysis 4.3; moderate‐quality evidence; Table 4).

4.3. Analysis.

Comparison 4: Mechanically assisted walking training with body weight support vs the same dose of overground walking training, Outcome 3: Participation: School Function Assessment (score) change from baseline to end of treatment (10 weeks)

No study in this comparison measured the primary outcome of adverse events or the secondary outcome of quality of life.

Subgroup analyses

Age of participants

Six studies with a total of 129 participants (mean age between four and nine years old) provided change scores—Dodd 2007 and Grecco 2013a—or postintervention scores—Ameer 2019; Cherng 2007; Peri 2017; Wallard 2017—for walking speed (m/s) immediately after the intervention. The pooled MD for walking speed was 0.13 m/s (95% CI 0.02 to 0.24; P = 0.02; random‐effects model; Tau² = 0.01; Chi² = 23.16, df = 5 [P < 0.001]; I² = 78%; Analysis 5.1; moderate‐quality evidence). Thus, mechanically assisted walking training probably resulted in a slight increase in walking speed in younger participants (nine years old or younger).

5.1. Analysis.

Comparison 5: Subgroup analysis, Outcome 1: Age

Seven studies with a total of 175 participants (mean age between 10 and 13 years old) provided change scores—Druzbicki 2013; Gharib 2011; Smania 2011; Swe 2015; Willoughby 2010—or postintervention scores—El‐Shamy 2017; Gates 2012—for walking speed (m/s) immediately after the intervention. The pooled MD for walking speed was 0.04 m/s (95% CI ‐0.00 to 0.07; P = 0.06; random‐effects model; Tau² = 0.00; Chi² = 10.43, df = 6 [P = 0.11]; I² = 42%; Analysis 5.1; moderate‐quality evidence). This result suggests that the intervention probably results in little or no difference in walking speed in this age group.

Discussion

Summary of main results

This review sought to assess the effects of mechanically assisted walking training for mobility, gross motor function, participation, and quality of life in children with cerebral palsy aged 3 to 18 years. All studies took place in outpatient settings in high‐income countries. Intensity of training also ranged widely in terms of length of sessions (15 minutes to 40 minutes), frequency of sessions (two to five times a week), and duration of intervention periods (4 to 12 weeks). However, most studies trained participants for 30 minutes, three times a week, for 12 weeks using a treadmill (with or without body weight support), which is clinically feasible and should be of sufficient intensity to show an effect.

We included 17 studies with a total of 451 participants. Compared with no walking training, mechanically assisted walking training, with or without body weight support, led to immediate but small increases in walking speed and gross motor function with body weight support. Compared with the same dose of overground walking training, mechanically assisted walking training with body weight support had no immediate effect, but two studies found that mechanically assisted walking training without body weight support had an immediate effect. Of the four studies that collected data on adverse events, all reported finding no adverse events. Of the four studies that measured participation, three studies reported that mechanically assisted walking training had an immediate effect compared with the same dose of overground walking, but one study found that mechanically assisted walking training with body weight support had no immediate effect. We found that younger children (under nine years old) showed an increase in walking speed immediately after mechanically assisted walking training.

Overall completeness and applicability of evidence

We conducted a comprehensive search in all languages for relevant studies, including handsearching for conference abstracts and searching for unpublished studies in trial registries. Although we are confident that our approach revealed all relevant trials, it is possible that we did not identify studies published in the grey literature, such as theses and unpublished abstracts from conference proceedings. However, it is unlikely that this would have had a significant impact on our results because the grey literature tends to report preliminary results of studies with relatively small numbers of participants.

We included data from 16 studies in a meta‐analysis of one of the four research questions. Even though statistical heterogeneity was limited in any meta‐analysis of the four comparisons, only a few studies had small numbers of participants in each (range of sample size = 8 to 35). Hence, the results of this review highlight the need for further clinical trials with greater numbers of participants.

More studies supported the results of the use of mechanically assisted walking training with body weight support than the use of mechanically assisted walking training without body weight support. This implies that evidence on the effect of the use of mechanically assisted walking training without body weight support is limited. Furthermore, few children were classified within Gross Motor Function Classification System (GMFCS) Level IV, and this reduces the applicability of findings to these children. Additionally, trials on the use of mechanically assisted walking without body weight support compared with the same dose of overground walking included only children younger than nine years of age, which implies that the evidence can be generalized only to younger children. Last, given that this intervention requires resources, the cost of the device may restrict its applicability in low‐income countries.

Quality of the evidence

Overall, there is risk of bias in these studies because nearly half of all included studies had or were at unclear risk of allocation bias (eight studies), none of the studies could blind participants or therapists, a third did not blind outcome assessors (five studies), and one‐fifth had or were at unclear risk of attrition bias (four studies). See Characteristics of included studies tables.

Using the GRADE criteria, we rated the quality of evidence for all comparisons as moderate or low. We judged all studies as being at high risk of bias for blinding of participants and personnel. Not blinding participants and therapists may increase the risk of introducing performance bias, but unfortunately this is not feasible for studies implementing an intervention such as mechanically assisted walking training. We judged less than 30% of studies as being at high risk of bias for blinding of outcome assessment. We also judged 41% of the studies as being at high or unclear risk of bias for random sequence generation and 47% of studies as being at high or unclear risk of bias for allocation concealment. Intervention effect estimates may be exaggerated in up to 15% of trials with unclear or inadequate sequence allocation and allocation concealment, but this is mainly true for self‐reported outcomes rather than for scales completed by personnel or objective outcomes (Savović 2012). We judged four trials (23.5%) as being at high or unclear risk of bias for incomplete data, and three trials did not report that they performed an intention‐to‐treat analysis. Excluding participants from the analysis can result in biased estimates of treatment effects, although the direction of bias is unpredictable. Finally, we also downgraded the quality of evidence due to imprecision when an insufficient number of studies was included in a comparison (Nüesch 2009).

In summary, moderate‐quality evidence indicates that the use of mechanically assisted walking training without body weight support increased walking speed, and low‐quality evidence suggests that it increased gross motor function, compared with no walking training. Moderate‐quality evidence also shows that the use of mechanically assisted walking training without body weight support improved walking speed, gross motor function, and participation compared with the same dose of overground walking training. Moderate‐quality evidence shows that the use of mechanically assisted walking training with body weight support increased walking speed, and low‐quality evidence suggests that it increased gross motor function, participation, and quality of life, compared with no walking training. Finally, low‐quality evidence suggests that the use of mechanically assisted walking training with body weight support increased walking speed and gross motor function, and moderate‐quality evidence shows that it increased participation, compared with the same dose of overground walking training.

Potential biases in the review process

One potential source of bias in this review is that we were not able to obtain all data (i.e. three randomized controlled trials [RCTs] and one cross‐over trial without data).

We are confident about the detailed search strategy and handsearching efforts used to identify all relevant trials. We may not have identified studies published in the grey literature, but this is unlikely to have had a significant impact on our results because the grey literature tends to have relatively small numbers of participants and inconclusive results. Therefore, inclusion of the grey literature may decrease the size of the effect reported in our review (McAuley 2000).

Agreements and disagreements with other studies or reviews

We included a greater number of trials than two previous reviews that examined the effects of robotic assisted walking in children with cerebral palsy and conducted meta‐analyses (Carvalho 2017; Lefmann 2017). These reviews included all types of studies, including two of the RCTs included in our review (Druzbicki 2013; Smania 2011).

Carvalho 2017 conducted a meta‐analysis combining randomized trials with single‐group studies and did not conduct a subgroup analysis for the two randomized trials. Lefmann 2017 reported that children with cerebral palsy who received robotic assisted walking training did not walk faster than those who received no walking training (mean difference [MD] 0.11 m/s, 95% confidence interval [CI] ‐0.48 to 0.70). In contrast, for our comparison of mechanically assisted walking plus body weight support versus no walking intervention, we included the same two randomized trials (Druzbicki 2013; Smania 2011), and we found six more—two of which used robotic devices (El‐Shamy 2017; Wallard 2017), and four of which used a treadmill (Bryant 2013; Cherng 2007; Dodd 2007; Gates 2012). We performed a meta‐analysis of these six studies and found small, but clinically insignificant increases in walking speed (MD 0.06 m/s, 95% CI 0.02 to 0.10). We also found a small improvement in gross motor function, participation, and quality of life.

Authors' conclusions

Implications for practice.

Mechanically assisted walking without body weight support may result in small improvements in walking speed and gross motor function compared to both no walking and the same amount of overground walking. For mechanically assisted walking with body weight support, we found improvements in walking speed and gross motor function compared to no walking, but not compared to the same amount of overground walking. Not many studies reported adverse events, although those that did appear to show no differences between groups. Mechanically assisted walking training may provide a way to undertake high‐intensity, repetitive, task‐specific walking training. In particular, it may be useful to structure practice for children with poor concentration when it is hard to apply the same dose of overground walking training.

We found a possible meaningful effect of mechanically assisted walking training without body weight support on mobility involving two studies (Ameer 2019; Grecco 2013a), and on gross motor function and participation involving only one study (Grecco 2013a). This may be because the participants in both studies included younger children (six years old) (Ameer 2019; Grecco 2013a). In addition, one of these studies used a supervised, high‐intensity training regimen with children within GMFCS Levels I to III in a clinical setting (Grecco 2013a). The protocol employed was 2 × 30‐minute sessions per week over seven consecutive weeks. Training velocity was based on a cardiopulmonary effort test. Maximal tolerance of 60% to 80% reached on the test was used for the training sessions. The children walked at 60% maximal speed in the first and final five minutes of the session, and at 80% maximal speed for the other 20 minutes. Thus, training was performed at the aerobic threshold, which may have led to better physical conditioning, thereby allowing participants to walk a significantly longer distance in six minutes and to achieve better daily participation with regard to mobility, compared with the same amount of overground walking. These findings suggest that a supervised, high‐intensity program is more likely to yield results in younger children with cerebral palsy who use walking as their main method of ambulation. Furthermore, these findings suggest that the addition of body weight support to mechanically assisted walking training might not be necessary for children with cerebral palsy within GMFCS Levels I to III. Body weight support was designed for children with severe motor impairments when they are beginning to walk. The use of body weight support for children with cerebral palsy who have already achieved independent walking may hinder learning. Perhaps body weight support could be reserved for children within GMFCS Level IV.

Implications for research.

The current review highlights the need for large, fully powered, randomized controlled trials on the effects of mechanically assisted walking for children with cerebral palsy. These trials should be well designed and should include allocation concealment, blinding of outcome assessment, and strategies to overcome attrition bias. In terms of participants, it would be useful to recruit younger participants (i.e. the intervention begins once children with cerebral palsy are just becoming ambulatory). Then participants could be assigned to an appropriate intervention, that is, those within GMFCS Levels I, II, and III could receive no body weight support, and those within GMFCS Levels IV and V could be given body weight support. Additionally, those within GMFCS Levels I to III could be trained at high intensity in terms of treadmill speed and inclination, for example, an aerobic threshold (as in Grecco 2013a) could be used to structure training. Most of the studies in the current review included participants within multiple GMFCS Levels, so it is impossible to tease out the effects of ambulatory status on outcomes. In terms of assessment of mobility, trials could measure all the common tests, that is, walking speed (10‐meter walk test), walking distance (6‐minute walk test), and Gross Motor Function Measure—Dimension E (GMFM‐E). Also, it is important to investigate whether increases in mobility carry over to improvements in participation and quality of life. Last, follow‐up of at least six months beyond the intervention is necessary to assess the sustainability of improvements.

History

Protocol first published: Issue 8, 2018
Review first published: Issue 11, 2020

Appendices

Appendix 1. Search strategies

Cochrane Central Register of Controlled Trials (CENTRAL)

#1MeSH descriptor: [Cerebral Palsy] this term only
#2(cerebral NEXT pals*):ti,ab,kw
#3"Little disease":ti,ab,kw
#4(spastic* near/5 pals*):ti,ab,kw
#5(athetoid near/5 pals*):ti,ab,kw
#6(ataxic near/5 pals*):ti,ab,kw
#7(spastic* near/5 cerebral*):ti,ab,kw
#8MeSH descriptor: [Muscle Hypertonia] this term only
#9MeSH descriptor: [Muscle Spasticity] explode all trees
#10MeSH descriptor: [Hemiplegia] explode all trees
#11(hemipleg* NOT stroke*):ti,ab,kw
#12(unilateral near/3 spastic*):ti,ab,kw
#13(unilateral near/3 CP):ti,ab,kw
#14(hemipleg* near/3 CP):ti,ab,kw
#15(hemipleg* near/3 spastic*):ti,ab,kw
#16(diplegi* near/3 CP):ti,ab,kw
#17(diplegi* near/3 spastic*):ti,ab,kw
#18(monoplegi* near/3 CP):ti,ab,kw
#19(monoplegi* near/3 spastic*):ti,ab,kw
#20(triplegi* near/3 CP):ti,ab,kw
#21(triplegi* near/3 spastic*):ti,ab,kw
#22MeSH descriptor: [Quadriplegia] this term only
#23(quadriplegi* near/3 CP):ti,ab,kw
#24(quadriplegi* near/3 spastic*):ti,ab,kw
#25{OR #1‐#24}
#26MeSH descriptor: [Physical Therapy Modalities] this term only
#27MeSH descriptor: [Motor Skills Disorders] explode all trees and with qualifier(s): [rehabilitation ‐ RH]
#28MeSH descriptor: [Exercise Movement Techniques] explode all trees
#29MeSH descriptor: [Robotics] explode all trees
#30MeSH descriptor: [Exoskeleton Device] explode all trees
#31(Locomot* near/3 (assist* or device* or train or intervention* or therap*)):ti,ab,kw
#32(robotic near/3 (assist* or device* or train* or intervention* or therap*)):ti,ab,kw
#33(mechanical* near/3 (assist* or device* or train* or intervention* or therap*)):ti,ab,kw
#34((electronic* or electromechanical*) near/3 (assist* or device* or train* or intervention* or therap*)):ti,ab,kw
#35(orthotic* or orthosis*):ti,ab,kw
#36Lokomat*:ti,ab,kw
#37(treadmill OR Tread next mill):ti,ab,kw
#38((bodyweight or body NEXT weight) near/3 (relief* or reliev* or support* or suspend* or
unsupport*)):ti,ab,kw
#39(gait near/3 (assist* or devic* or train* or intervention* or therap*)):ti,ab,kw
#40(walk* near/3 (assist* or device* or mechanical*)):ti,ab,kw
#41(ambulat* near/3 (assist* or device* or mechanical*)):ti,ab,kw
#42(end next effector*):ti,ab,kw
#43MeSH descriptor: [Weight‐Bearing] this term only
#44(weight near/1 bearing):ti,ab,kw
#45{OR #26‐#44}
#46[mh "Body Weight"] AND [mh Walking]
#47#45 or #46
#48#25 and #47 in Trials

MEDLINE (Ovid)

1 Cerebral Palsy/
2 cerebral pals$.tw,kf.
3 Little disease.tw,kf.
4 (spastic$ adj5 pals$).tw,kf.
5 (athetoid adj5 palsy$).tw,kf.
6 (ataxic adj5 pals$).tw,kf.
7 (spastic$ adj5 cerebral$).tw,kf.
8 Muscle Hypertonia/
9 Muscle Spasticity/
10 hemiplegia/
11 hemipleg$.tw,kf.
12 (unilateral adj3 spastic$).tw,kf.
13 (unilateral adj3 CP).tw,kf.
14 (hemipleg$ adj3 CP).tw,kf.
15 (hemipleg$ adj3 spastic$).tw,kf.
16 (diplegi$ adj3 cp).tw,kf.
17 (diplegi$ adj3 spastic$).tw,kf.
18 (monoplegi$ adj3 CP).tw,kf.
19 (monoplegi$ adj3 spastic$).tw,kf.
20 (triplegi$ adj3 CP).tw,kf.
21 (triplegi$ adj3 spastic$).tw,kf.
22 quadriplegia/
23 (quadriplegi$ adj3 CP).tw,kf.
24 (quadriplegi$ adj3 spastic$).tw,kf.
25 or/1‐24 (53425)
26 Physical Therapy Modalities/
27 Motor Skills Disorders/rh, th
28 Exercise Movement Techniques/
29 Robotics/
30 Exoskeleton Device/
31 (Locomot$ adj5 (assist$ or device$ or train$ or intervention$ or therap$)).tw,kf.
32 (robotic$ adj5 (assist$ or device$ or train$ or intervention$ or therap$)).tw,kf.
33 (mechanical$ adj5 (assist$ or device$ or train$ or intervention$ or therap$)).tw,kf.
34 ((electronic$ or electromechanical$) adj5 (assist$ or device$ or train$ or intervention$ or therap$)).tw,kf.
35 (orthotic$ or orthosis$).tw,kf.
36 Lokomat$.tw,kf.
37 (treadmill or tread‐mill).tw,kf. (28000)
38 Body Weight/ and Walking/
39 ((bodyweight or body‐weight) adj3 (relief$ or reliev$ or support$ or suspend$ or unsupport$)).tw,kf.
40 (gait adj3 (assist$ or devic$ or train$ or intervention$ or therap$)).tw,kf.
41 (walk$ adj3 (assist$ or device$ or mechanical$)).tw,kf.
42 (ambulat$ adj3 (assist$ or device$ or mechanical$)).tw,kf.
43 end effector$.tw.
44 Weight‐Bearing/
45 (weight adj1 bearing).tw,kf.
46 or/26‐45
47 randomized controlled trial.pt.
48 controlled clinical trial.pt.
49 randomi#ed.ab.
50 placebo$.ab.
51 drug therapy.fs.
52 randomly.ab.
53 trial.ab.
54 groups.ab.
55 or/47‐54
56 exp animals/ not humans.sh.
57 55 not 56
58 25 and 46 and 57

Ovid MEDLINE(R) In‐Process & Other Non‐Indexed Citations AND Ovid MEDLINE(R) Epub Ahead of Print

11 cerebral pals$.tw,kf.
2 Little disease.tw,kf.
3 (spastic$ adj5 pals$).tw,kf.
4 (athetoid adj5 palsy$).tw,kf.
5 (ataxic adj5 pals$).tw,kf.
6 (spastic$ adj5 cerebral$).tw,kf.
7 (Musc$ adj3 hypertonia$).tw,kf.
8 (Musc$ adj3 spasticit$).tw,kf.
9 hemipleg$.tw,kf.
10 (unilateral adj3 spastic$).tw,kf.
11 (unilateral adj3 CP).tw,kf.
12 (hemipleg$ adj3 CP).tw,kf.
13 (hemipleg$ adj3 spastic$).tw,kf.
14 (diplegi$ adj3 cp).tw,kf.
15 (diplegi$ adj3 spastic$).tw,kf.
16 (monoplegi$ adj3 CP).tw,kf.
17 (monoplegi$ adj3 spastic$).tw,kf.
18 (triplegi$ adj3 CP).tw,kf.
19 (triplegi$ adj3 spastic$).tw,kf.
20 (quadriplegi$ adj3 CP).tw,kf.
21 (quadriplegi$ adj3 spastic$).tw,kf.
22 or/1‐21
23 (Exoskeleton adj3 device$).tw,kf.
24 (Locomot$ adj5 (assist$ or device$ or train$ or intervention$ or therap$)).tw,kf.
25 (robotic$ adj5 (assist$ or device$ or train$ or intervention$ or therap$)).tw,kf.
26 (mechanical$ adj5 (assist$ or device$ or train$ or intervention$ or therap$)).tw,kf.
27 ((electronic$ or electromechanical$) adj5 (assist$ or device$ or train$ or intervention$ or therap$)).tw,kf.
28 (orthotic$ or orthosis$).tw,kf.
29 Lokomat$.tw,kf.
30 (treadmill or tread‐mill).tw,kf.
31 ((bodyweight or body‐weight) adj3 (relief$ or reliev$ or support$ or suspend$ or unsupport$)).tw,kf.
32 (gait adj3 (assist$ or devic$ or train$ or intervention$ or therap$)).tw,kf.
33 (walk$ adj3 (assist$ or device$ or mechanical$)).tw,kf.
34 (ambulat$ adj3 (assist$ or device$ or mechanical$)).tw,kf.
35 end effector$.tw.
36 (weight adj1 bearing).tw,kf.
37 or/23‐36
38 22 and 37
39 (random$ or control$ or group$ or cluster$ or placebo$ or trial$ or assign$ or allocat$ or prospectiv$ or meta‐analysis or systematic review or longitudinal$).tw,kf.
40 38 and 39

Embase Ovid

1 cerebral palsy/
2 cerebral pals$.tw,kw.
3 Little disease.tw,kw.
4 (spastic$ adj5 pals$).tw,kw.
5 (athetoid adj5 palsy$).tw,kw.
6 (ataxic adj5 pals$).tw,kw.
7 (spastic$ adj5 cerebral$).tw,kw.
8 Muscle Hypertonia/
9 spasticity/
10 hemiplegia/
11 hemipleg$.tw,kw.
12 (unilateral adj3 spastic$).tw,kw.
13 (unilateral adj3 cp).tw,kw.
14 (hemipleg$ adj3 CP).tw,kw.
15 (hemipleg$ adj3 spastic$).tw,kw.
16 (diplegi$ adj3 cp).tw,kw.
17 (diplegi$ adj3 spastic$).tw,kw.
18 (monoplegi$ adj3 CP).tw,kw.
19 (monoplegi$ adj3 spastic$).tw,kw.
20 (triplegi$ adj3 CP).tw,kw.
21 (triplegi$ adj3 spastic$).tw,kw.
22 quadriplegia/
23 (quadriplegi$ adj3 CP).tw,kw.
24 (quadriplegi$ adj3 spastic$).tw,kw.
25 or/1‐24
26 physiotherapy/
27 motor performance/ and rehabilitation/
28 kinesiotherapy/
29 robotic exoskeleton/
30 (Locomot$ adj5 (assist$ or device$ or train$ or intervention$ or therap$)).tw,kw.
31 (robotic$ adj5 (assist$ or device$ or train$ or intervention$ or therap$)).tw,kw.
32 (mechanical$ adj5 (assist$ or device$ or train$ or intervention$ or therap$)).tw,kw.
33 ((electronic$ or electromechanical$) adj5 (assist$ or device$ or train$ or intervention$ or therap$)).tw,kw.
34 orthotics/
35 (orthotic$ or orthosis$).tw,kw.
36 Lokomat$.tw,kw.
37 Treadmill/
38 Treadmill exercise/
39 (treadmill or tread‐mill).tw,kw.
40 body weight/ and walking/
41 ((bodyweight or body‐weight) adj3 (relief$ or reliev$ or support$ or suspend$ or unsupport$)).tw,kw.
42 walking aid/
43 gait trainer/
44 (gait adj3 (assist$ or devic$ or train$ or intervention$ or therap$)).tw,kw.
45 (walk$ adj3 (assist$ or device$ or mechanical$)).tw,kw.
46 (ambulat$ adj3 (assist$ or device$ or mechanical$)).tw,kw.
47 end effector$.tw,kw.
48 weight bearing/
49 (weight adj1 bearing).tw,kw.
50 or/26‐49
51 Randomized controlled trial/
52 controlled clinical trial/
53 Single blind procedure/
54 Double blind procedure/
55 triple blind procedure/
56 Crossover procedure/
57 (crossover or cross‐over).tw.
58 ((singl$ or doubl$ or tripl$ or trebl$) adj1 (blind$ or mask$)).tw.
59 Placebo/
60 placebo.tw.
61 prospective.tw.
62 factorial$.tw.
63 random$.tw.
64 assign$.ab.
65 allocat$.tw.
66 volunteer$.ab.
67 or/51‐66
68 25 and 50 and 67

CINAHL EBSCOhost (Cumulative Index to Nursing and Allied Health Literature)

S1(MH "Cerebral Palsy")
S2TI(cerebral pals*) OR AB(cerebral pals*)
S3"Little disease"
S4TI(spastic* N5 pals*) OR AB(spastic* N5 pals*)
S5TI(athetoid N5 pals*) OR AB(athetoid N5 pals*)
S6TI(ataxic N5 pals*) OR AB(ataxic N5 pals*)
S7TI(spastic* N5 cerebral*) OR AB(spastic* N5 cerebral*)
S8(MH "Muscle Hypertonia")
S9(MH "Muscle Spasticity")
S10(MH "Hemiplegia")
S11TI(Hemiplegia NOT stroke*) OR AB(Hemiplegia NOT stroke*)
S12TI(unilateral* N3 spastic*) OR AB(unilateral* N3 spastic*)
S13TI(unilateral N3 CP) OR AB(unilateral N3 CP)
S14TI (hemipleg* N3 CP) OR AB(hemipleg* N3 CP)
S15TI(hemipleg* N3 spastic*) OR AB(hemipleg* N3 spastic*)
S16TI(diplegi* N3 CP) OR AB(diplegi* N3 CP)
S17TI(diplegi* N3 spastic*) OR AB(diplegi* N3 spastic*)
S18TI(monoplegi* N3 CP) OR AB(monoplegi* N3 CP)
S19TI(monoplegi* N3 spastic*) OR AB(monoplegi* N3 spastic*)
S20TI(triplegi* N3 CP) OR AB(triplegi* N3 CP)
S21TI(triplegi* N3 spastic*) OR AB(triplegi* N3 spastic*)
S22(MH "Quadriplegia")
S23TI(quadriplegi* N3 CP) OR AB(quadriplegi* N3 CP)
S24TI(quadriplegi* N3 spastic*) OR AB(quadriplegi* N3 spastic*)
S25S1 OR S2 OR S3 OR S4 OR S5 OR S6 OR S7 OR S8 OR S9 OR S10 OR S11 OR S12 OR S13 OR S14 OR S15 OR S16 OR S17 OR S18 OR S19 OR S20 OR S21 OR S22 OR S23 OR S24
S26(MH "Body‐Weight‐Supported Treadmill Training")
S27(MH "Motor Skills Disorders/RH/TH")
S28(MH "Gait Training")
S29(MH "Exoskeleton Devices")
S30(MH "Robotics")
S31TI(Locomot* N3 (assist* or device* or train or intervention* or therap*)) OR AB(Locomot* N3 (assist* or device* or train or intervention* or therap*))
S32TI(robotic N3 (assist* or device* or train* or intervention* or therap*)) OR AB(robotic N3 (assist* or device* or train* or intervention* or therap*))
S33TI(mechanical* N3 (assist* or device* or train* or intervention* or therap*)) OR AB(mechanical* N3 (assist* or device* or train* or intervention* or therap*))
S34TI((electronic* or electromechanical*) N3 (assist* or device* or train* or intervention* or therap*)) OR AB((electronic* or electromechanical*) N3 (assist* or device* or train* or intervention* or therap*))
S35TI(orthotic* or orthosis*) OR AB(orthotic* or orthosis*)
S36Lokomat*
S37(MH "Treadmills")
S38TI(treadmill OR tread mill ) OR AB(treadmill OR tread mill )
S39TI((bodyweight or body weight ) N3 (relief* or reliev* or support* or suspend* or unsupport*))
S40TI(gait N3 (assist* or devic* or train* or intervention* or therap*)) OR AB(gait N3 (assist* or devic* or train* or intervention* or therap*))
S41TI(walk* N3 (assist* or device* or mechanical*)) OR AB(walk* N3 (assist* or device* or mechanical*))
S42TI(ambulat* N3 (assist* or device* or mechanical*)) OR AB(ambulat* N3 (assist* or device* or mechanical*))
S43TI(end effector*) OR AB(end effector*)
S44(MH "Weight‐Bearing")
S45S25 AND S44
S46TI(weight bearing) OR AB(weight bearing)
S47S26 OR S27 OR S28 OR S29 OR S30 OR S31 OR S32 OR S33 OR S34 OR S35 OR S36 OR S37 OR S38 OR S39 OR S40 OR S41 OR S42 OR S43 OR S44 OR S45 OR S46
S48(MH "Body Weight") AND (MH "Walking")
S49S47 OR S48
S50S25 AND S49
S51MH ("Randomized Controlled Trials")
S53(MH "Single‐Blind Studies")
S54(MH "Random Assignment")
S55(MH "Pretest‐Posttest Design")
S56MH ("Cluster Sample")
S58AB (random*)
S59TI (trial)
S60(MH "Sample Size") AND AB (assigned OR allocated OR control)
S61MH (Placebos)
S62PT (Randomized Controlled Trial)
S63AB (control W5 group)
S64MH ("Crossover Design") OR MH ("Comparative Studies")
S65AB (cluster W3 RCT)
S66(MH "Animals+")
S67MH ("Animal Studies")
S68TI (animal model*)
S69S66 OR S67 OR S68
S70MH ("Human")
S71S69 NOT S70
S72S51 OR S52 OR S53 OR S54 OR S55 OR S56 OR S57 OR S58 OR S59 OR S60 OR S61 OR S62 OR S63 OR S64 OR S65
S73S72 NOT S71
S74S50 AND S73

Cochrane Database of Systematic Reviews, part of the Cochrane Library

#1MeSH descriptor: [Cerebral Palsy] this term only
#2(cerebral NEXT pals*):ti,ab,kw
#3"Little disease":ti,ab,kw
#4(spastic* near/5 pals*):ti,ab,kw
#5(athetoid near/5 pals*):ti,ab,kw
#6(ataxic near/5 pals*):ti,ab,kw
#7(spastic* near/5 cerebral*):ti,ab,kw
#8MeSH descriptor: [Muscle Hypertonia] this term only
#9MeSH descriptor: [Muscle Spasticity] explode all trees
#10MeSH descriptor: [Hemiplegia] explode all trees
#11(hemipleg* NOT stroke*):ti,ab,kw
#12(unilateral near/3 spastic*):ti,ab,kw
#13(unilateral near/3 CP):ti,ab,kw
#14(hemipleg* near/3 CP):ti,ab,kw
#15(hemipleg* near/3 spastic*):ti,ab,kw
#16(diplegi* near/3 CP):ti,ab,kw
#17(diplegi* near/3 spastic*):ti,ab,kw
#18(monoplegi* near/3 CP):ti,ab,kw
#19(monoplegi* near/3 spastic*):ti,ab,kw
#20(triplegi* near/3 CP):ti,ab,kw
#21(triplegi* near/3 spastic*):ti,ab,kw
#22MeSH descriptor: [Quadriplegia] this term only
#23(quadriplegi* near/3 CP):ti,ab,kw
#24(quadriplegi* near/3 spastic*):ti,ab,kw
#25{OR #1‐#24}
#26MeSH descriptor: [Physical Therapy Modalities] this term only
#27MeSH descriptor: [Motor Skills Disorders] explode all trees and with qualifier(s): [rehabilitation ‐ RH]
#28MeSH descriptor: [Exercise Movement Techniques] explode all trees
#29MeSH descriptor: [Robotics] explode all trees
#30MeSH descriptor: [Exoskeleton Device] explode all trees
#31(Locomot* near/3 (assist* or device* or train or intervention* or therap*)):ti,ab,kw
#32(robotic near/3 (assist* or device* or train* or intervention* or therap*)):ti,ab,kw
#33(mechanical* near/3 (assist* or device* or train* or intervention* or therap*)):ti,ab,kw
#34((electronic* or electromechanical*) near/3 (assist* or device* or train* or intervention* or therap*)):ti,ab,kw
#35(orthotic* or orthosis*):ti,ab,kw
#36Lokomat*:ti,ab,kw
#37(treadmill OR Tread next mill):ti,ab,kw
#38((bodyweight or body NEXT weight) near/3 (relief* or reliev* or support* or suspend* or unsupport*)):ti,ab,kw
#39(gait near/3 (assist* or devic* or train* or intervention* or therap*)):ti,ab,kw
#40(walk* near/3 (assist* or device* or mechanical*)):ti,ab,kw
#41(ambulat* near/3 (assist* or device* or mechanical*)):ti,ab,kw
#42(end next effector*):ti,ab,kw
#43MeSH descriptor: [Weight‐Bearing] this term only
#44(weight near/1 bearing):ti,ab,kw
#45{OR #26‐#44}
#46[mh "Body Weight"] AND [mh Walking]
#47#45 or #46
#48#25 and #47 in Cochrane Reviews and Cochrane Protocols

Epistemonikos (www.epistemonikos.org)

(title:(title:(cerebral palsy OR spastic* OR hemipleg* OR monoplegi* OR diplegi* OR triplegi* OR quadriplegi*) AND (title:(mechanical* OR ROBOTIC* OR electronic OR electromechanical* OR exoskeleton OR orthotic* OR orthosis OR Lokomat OR treadmill OR tread‐mill) OR abstract:(mechanical* OR robotic* OR electronic OR electromechanical* OR exoskeleton OR orthotic* OR orthosis OR Lokomat OR treadmill OR tread‐mill)) AND (title:(ambulat* OR walk* OR bodyweight OR "body weight" OR gait OR locomotor*) OR abstract:(ambulat* OR walk* OR bodyweight OR "body weight" OR gait OR locomotor*))) OR abstract:(title:(cerebral palsy OR spastic* OR hemipleg* OR monoplegi* OR diplegi* OR triplegi* OR quadriplegi*) AND (title:(mechanical* OR ROBOTIC* OR electronic OR electromechanical* OR exoskeleton OR orthotic* OR orthosis OR Lokomat OR treadmill OR tread‐mill) OR abstract:(mechanical* OR ROBOTIC* OR electronic OR electromechanical* OR exoskeleton OR orthotic* OR orthosis OR Lokomat OR treadmill OR tread‐mill)) AND (title:(ambulat* OR walk* OR bodyweight OR "body weight" OR gait OR locomotor*) OR abstract:(ambulat* OR walk* OR bodyweight OR "body weight" OR gait OR locomotor*))))

ProQuest Dissertations & Theses

ti("cerebral pals*" OR spastic* OR hemipleg* OR monoplegi* OR diplegi* OR triplegi* OR quadriplegi*) AND noft ((mechanical* OR robotic* OR electronic OR electromechanical* OR exoskeleton OR orthotic* OR orthosis OR Lokomat OR treadmill OR "tread mill")) AND noft(ambulat* OR walk* OR bodyweight OR "body weight" OR gait OR locomotor*) AND noft (random* OR allocat* OR assign* OR control* OR placebo OR group OR trial OR TAU OR "treatment as usual" OR "usual treatment" OR "usual care")

Science Citation Index—Expanded (Web of Science)

# 38 #37AND #36
Indexes=SCI‐EXPANDED Timespan=All years
# 37TS=(RANDOM* OR RCT OR TRIAL* OR ALLOCAT* OR ASSIGN* OR GROUP OR CONTROL* OR "USUAL TREATMENT" OR "TREATMENT AS USUAL" OR TAU OR "FOLLOW UP")
Indexes=SCI‐EXPANDED Timespan=All years
# 36#35 AND #22
Indexes=SCI‐EXPANDED Timespan=All years
# 35 #34 OR #33 OR #32 OR #31 OR #30 OR #29 OR #28 OR #27 OR #26 OR #25 OR #24 OR #23
Indexes=SCI‐EXPANDED Timespan=All years
# 34 TS=(weight near/1 bearing)
Indexes=SCI‐EXPANDED Timespan=All years
# 33 TS=("end Effector*")
Indexes=SCI‐EXPANDED Timespan=All years
# 32 TS=(ambulat* near/3 (assist* or device* or mechanical*))
Indexes=SCI‐EXPANDED Timespan=All years
# 31 TS= (walk* near/3 (assist* or device* or mechanical*))
Indexes=SCI‐EXPANDED Timespan=All years
# 30 TS=(gait near/3 (assist* or devic* or train* or intervention* or therap*))
Indexes=SCI‐EXPANDED Timespan=All years
# 29 TS=((bodyweight or "body weight") near/3 (relief* or reliev* or support* or suspend* or unsupport*))
Indexes=SCI‐EXPANDED Timespan=All years
# 28 TS=(treadmill* OR "Tread mill*" OR "tread‐mill*")
Indexes=SCI‐EXPANDED Timespan=All years
# 27 TS=Lokomat*
Indexes=SCI‐EXPANDED Timespan=All years
# 26 TS=(orthotic* or orthosis*)
Indexes=SCI‐EXPANDED Timespan=All years
# 25 TS=((electronic* or electromechanical*) near/3 (assist* or device* or train* or intervention* or therap*))
Indexes=SCI‐EXPANDED Timespan=All years
# 24 TS=(Locomot* near/3 (assist* or device* or train or intervention* or therap*))
Indexes=SCI‐EXPANDED Timespan=All years
# 23 TS=(mechanical* near/3 (assist* or device* or train* or intervention* or therap*))
Indexes=SCI‐EXPANDED Timespan=All years
# 22 #21 OR #20 OR #19 OR #18 OR #17 OR #16 OR #15 OR #14 OR #13 OR #12 OR #11 OR #10 OR #9 OR #8 OR #7 OR #6 OR #5 OR #4 OR #3 OR #2 OR #1
Indexes=SCI‐EXPANDED Timespan=All years
# TS=(quadriplegi* near/3 spastic*)
Indexes=SCI‐EXPANDED Timespan=All years
# 20 TS=(quadriplegi* near/3 CP)
Indexes=SCI‐EXPANDED Timespan=All years
# 19 TS=(triplegi* near/3 CP)
Indexes=SCI‐EXPANDED Timespan=All years
# 18 TS=(triplegi* near/3 spastic*)
Indexes=SCI‐EXPANDED Timespan=All years
# 17 TS=(monoplegi* near/3 CP)
Indexes=SCI‐EXPANDED Timespan=All years
# 16 TS=(monoplegi* near/3 spastic*)
Indexes=SCI‐EXPANDED Timespan=All years
# 15 TS=(diplegi* near/3 spastic*)
Indexes=SCI‐EXPANDED Timespan=All years
# 14 TS=(diplegi* near/3 CP)
Indexes=SCI‐EXPANDED Timespan=All years
# 13 TS=(hemipleg* near/3 CP)
Indexes=SCI‐EXPANDED Timespan=All years
# 12 TS=(hemipleg* near/3 spastic*)
Indexes=SCI‐EXPANDED Timespan=All years
#11 TS=(unilateral* near/3 CP)
Indexes=SCI‐EXPANDED Timespan=All years
# 10 TS=(unilateral* near/3 spastic*)
Indexes=SCI‐EXPANDED Timespan=All years
# 9 TS=(hemipleg* NOT stroke*)
Indexes=SCI‐EXPANDED Timespan=All years
# 8 TS=("Muscle Spasticity")
Indexes=SCI‐EXPANDED Timespan=All years
# 7 TS="Muscle Hypertonia"
Indexes=SCI‐EXPANDED Timespan=All years
# 6 TS=(spastic* near/5 cerebral*)
Indexes=SCI‐EXPANDED Timespan=All years
# 5 48 TS=(ataxic near/5 pals*)
Indexes=SCI‐EXPANDED Timespan=All years
# 4 TS=(athetoid near/5 pals*)
Indexes=SCI‐EXPANDED Timespan=All years
# 3 TS=(spastic* near/5 pals*)
Indexes=SCI‐EXPANDED Timespan=All years
# 2 TS=("Little disease")
Indexes=SCI‐EXPANDED Timespan=All years
# 1 TS=("cerebral pals*")
Indexes=SCI‐EXPANDED Timespan=All years

PEDro (Physiotherapy Evidence Database; www.pedro.org.au)

Abstract & title = Walk* Topic = cerebral palsy Therapy = orthoses, taping, splinting Topic = cerebral palsy
Abstract & title = treadmill* Topic = cerebral palsy
Abstract & title = mechanical* Topic = cerebral palsy
Abstract & title = robotic* Topic = cerebral palsy
Abstract & title = exoskeleton* Topic = cerebral palsy
Abstract & title = weight bearing Topic = cerebral palsy
Abstract & title = body weight Topic = cerebral palsy

ClinicalTrials.gov (clinicaltrials.gov)

body weight OR bodyweight OR weight bearing OR ambulatory OR walk OR gait OR locomotor | Interventional Studies | Cerebral Palsy | mechanical OR robotic OR electronic OR electromechanical OR exoskeleton OR orthotic OR orthosis OR Lokomat OR treadmill OR tread‐mill

WHO Clinical Trials Registry Platform (WHO ICTRP; who.int/ictrp/en)

Cerebral Palsy AND body weight support OR Cerebral Palsy AND bodyweight support

Synonyms automatically included in the search:
Cerebral Palsy, Cerebral palsied, Cerebral Palsy, cerebral palsy (diagnosis), Cerebral palsy (disorder),
Cerebral Palsy [Disease/Finding], Cerebral palsy NOS, Cerebral palsy unspecified, Cerebral paralysis, cerebral;
paralysis, CNS CONGENITAL ANOMALY, CP (Cerebral Palsy), DIABETES MELLITUS NOS, Kaveggia syndrome, Palsy
cerebral, PALSY CEREBRAL, Palsy;cerebral, paralysis; cerebral AND body weight support OR Cerebral Palsy,
Cerebral palsied, Cerebral Palsy, cerebral palsy (diagnosis), Cerebral palsy (disorder), Cerebral Palsy
[Disease/Finding], Cerebral palsy NOS, Cerebral palsy unspecified, Cerebral paralysis, cerebral; paralysis, CNS
CONGENITAL ANOMALY, CP (Cerebral Palsy), DIABETES MELLITUS NOS, Kaveggia syndrome, Palsy cerebral, PALSY
CEREBRAL, Palsy;cerebral, paralysis; cerebral AND bodyweight support

Appendix 2. ‘Risk of bias' criteria: operational definitions

Sequence generation

1. Low risk of bias: based on a random component in the sequence generation that was judged to be both appropriate and sufficiently described

2. High risk of bias: based on any non‐random component

3. Unclear risk of bias: insufficient information regarding the sequence generation process to permit a judgement of low or high risk of bias

Allocation concealment

1. Low risk of bias: method of concealment allocation used did not allow foresight of patient assignment

2. High risk of bias: method of concealment allocation used allowed possible foresight of patient assignment

3. Unclear risk of bias: method of concealment allocation was not described, or was described in insufficient detail to permit a judgement of low or high risk of bias

Blinding of patients and personnel

Blinding of patients

1. Low risk of bias: patients blinded to allocated intervention and unlikely that blinding was broken; or no or incomplete blinding but judged that the given outcome was unlikely to have been influenced by lack of blinding

2. High risk of bias: patients not blinded to allocated intervention; or patients blinded to allocated intervention but likely that blinding was broken (and a given outcome influenced by lack of blinding)

3. Unclear risk of bias: insufficient information to permit a judgement of low or high risk of bias

Blinding of care provider

1. Low risk of bias: care provider blinded to allocated intervention and unlikely that blinding was broken; or no or incomplete blinding but judged that a given outcome was unlikely to have been influenced by lack of blinding

2. High risk of bias: care provider not blinded to allocated intervention; or care provider blinded to allocated intervention but likely that blinding was broken (and a given outcome influenced by lack of blinding)

3. Unclear risk of bias: insufficient information to permit a judgement of low or high risk of bias

Blinding of outcome assessment

1. Low risk of bias: outcome assessor (including "patients" with respect to self‐reported outcomes) blinded to patients’ allocated intervention, and unlikely that blinding was broken; or no or incomplete blinding but judged that a given outcome was unlikely to have been influenced by lack of blinding

2. High risk of bias: outcome assessor (including "patients" with respect to self‐reported outcomes) unblinded to patients’ allocated intervention; or outcome assessor blinded to allocated intervention but likely that blinding was broken (and a given outcome influenced by lack of blinding)

3. Unclear risk of bias: insufficient information to permit a judgement of low or high risk of bias

Incomplete outcome data

1. Low risk of bias: no missing outcome data; reasons for missing data unlikely related to true outcome; missing outcome data balanced across intervention groups with similar reasons for omissions (dichotomous outcomes: proportion of missing outcomes compared with observed event risk not enough to have clinically relevant impact on the intervention effect estimate; continuous outcomes: difference in means or standardized mean difference [SMD] among missing outcomes not enough to have clinically relevant impact on observed effect size, missing data imputed using appropriate methods, intention‐to‐treat analysis undertaken, less than or equal to 10% dropout rate)

2. High risk of bias: reason for missing outcome data likely related to true outcome (dichotomous outcomes: proportion of missing outcomes compared with observed event risk enough to induce a clinically relevant bias in intervention effect estimate; continuous outcomes: difference in mean or SMD among missing outcomes enough to induce a clinically relevant bias on observed effect size, as‐treated analysis undertaken with substantial departure of intervention received from that assigned at randomization, 30% or greater dropout rate)

3. Unclear risk of bias: insufficient reporting of attrition or exclusions to permit a judgement of low or high risk of bias; greater than 10% and less than 30% dropout rate

Selective outcome reporting

1. Low risk of bias: all primary outcomes of interest were reported adequately, with point estimates and measures of variance for all time points

2. High risk of bias: incomplete reporting of pre‐specified outcomes; one or more primary outcomes reported using measurements, analysis methods, or subsets of data not pre‐specified; one or more reported primary outcomes not pre‐specified; one or more outcomes of interest reported incompletely and could not be entered into a meta‐analysis; or results excluded for a key outcome expected to have been reported

3. Unclear risk of bias: insufficient information to permit a judgement of low or high risk of bias

Other sources of bias

1. Low risk of bias: study appeared to be free of other sources of bias

2. High risk of bias: results may have been confounded by at least one important risk of bias (e.g. study‐design specific, fraudulent, other)

3. Unclear risk of bias: other sources of bias may have been present, but there was insufficient information to assess whether an important risk of bias existed, or rationale or evidence regarding whether an identified problem would introduce bias was insufficient

Data and analyses

Comparison 1. Mechanically assisted walking training without body weight support vs no walking training.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Mobilty (walking speed): Biodex Gait Trainer 2™ (m/s) change from baseline to end of treatment (12 weeks)	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
1.2 Gross motor function: PDMS‐2, locomotive subtest (%) or mTUG (s) at end of treatment (12 weeks)	2	60	Std. Mean Difference (IV, Random, 95% CI)	1.30 [0.49, 2.11]

Comparison 2. Mechanically assisted walking training without body weight support vs the same dose of overground walking training.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Mobility: 10‐meter walk test or gait analysis (m/s) change from baseline to end of treatment or at end of treatment (7 to 12 weeks)	2	55	Mean Difference (IV, Random, 95% CI)	0.25 [0.13, 0.37]
2.2 Gross motor function: GMFM‐E (%) change from baseline to end of treatment (7 weeks)	1		Mean Difference (IV, Random, 95% CI)	Totals not selected
2.3 Participation: PEDI‐mobility (score) change from baseline to end of treatment (7 weeks)	1		Mean Difference (IV, Random, 95% CI)	Totals not selected

Comparison 3. Mechanically assisted walking training with body weight support vs no walking training.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Mobility: 10‐Meter Walk Test or 6‐Minute Walk Test (m/s) change from baseline to end of treatment or at end of treatment (2 to 12 weeks)	7	161	Mean Difference (IV, Random, 95% CI)	0.07 [0.06, 0.08]
3.2 Gross motor function: GMFM‐E (%) change from baseline to end of treatment or at end of treatment (4 to 12 weeks)	3	58	Mean Difference (IV, Random, 95% CI)	1.09 [‐0.57, 2.75]
3.3 Participation: CAPE‐Intensity or WeeFIM (score) change from baseline to end of treatment (2 to 12 weeks)	2	44	Std. Mean Difference (IV, Random, 95% CI)	0.33 [‐0.27, 0.93]
3.4 Quality of life: PedsQOL‐CP (score) change from baseline to end of treatment (12 weeks)	1		Mean Difference (IV, Random, 95% CI)	Totals not selected

Comparison 4. Mechanically assisted walking training with body weight support vs the same dose of overground walking training.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
4.1 Mobility: 10‐Meter Walk Test or 6‐Minute Walk Test (%) change from baseline to end of treatment or at end of treatment (8 to 10 weeks)	3	78	Mean Difference (IV, Random, 95% CI)	‐0.02 [‐0.08, 0.04]
4.2 Gross motor function: GMFM‐E (%) at end of treatment (8 to 10 weeks)	2	52	Mean Difference (IV, Random, 95% CI)	‐0.73 [‐14.38, 12.92]
4.3 Participation: School Function Assessment (score) change from baseline to end of treatment (10 weeks)	1		Mean Difference (IV, Random, 95% CI)	Totals not selected

Comparison 5. Subgroup analysis.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
5.1 Age	13		Mean Difference (IV, Random, 95% CI)	Subtotals only
5.1.1 9 years of age or younger	6	129	Mean Difference (IV, Random, 95% CI)	0.13 [0.02, 0.24]
5.1.2 10 years of age or older	7	175	Mean Difference (IV, Random, 95% CI)	0.04 [‐0.00, 0.07]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Ameer 2019.

Study characteristics
Methods	Design: randomized controlled trial
Participants	Country: Egypt Sample size: 20 Dropouts: 0 Mean age: 6 years (range 4 to 8 years) Type of cerebral palsy: spastic, diplegic GMFCS Levels: I and II Inclusion criteria: aged 4 to 8 years; with cerebral palsy within GMFCS Levels I and II Exclusion criteria: not stated
Interventions	Experimental group (n = 10) Intervention: treadmill walking; 20 minutes × 3 sessions per week × 12 weeks (total time: 720 minutes) Body weight support: no Speed: self‐paced speed Progression: no Control group (n = 10) Intervention: overground walking; 20 minutes × 3 sessions per week × 12 weeks (total time: 720 minutes) Both groups (n = 20) Intervention: usual therapy; 25 minutes × 3 sessions per week × 12 weeks (total time: 900 minutes)
Outcomes	Primary outcomes Mobility: gait analysis (Vicon) (m/s) Gross motor function: not measured Adverse events: no adverse events Secondary outcomes Participation: not measured Quality of life: not measured Timing of outcome assessment: 0 and 12 weeks
Notes	Funding source: not reported Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: used prepared cards to randomly allocate participants to groups, indicating the group assigned
Allocation concealment (selection bias)	Low risk	Comment: used independent and concealed allocation. Participant group assignment was concealed in a prepared card
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	High risk	Comment: no assessors were blinded, thus increasing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no dropoutsa
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

Bryant 2013.

Study characteristics
Methods	Design: randomized controlled trial
Participants	Country: UK Sample size: 24 Dropouts: 4 (3 in experimental group, 1 in control group) Mean age: 14 years (range 8 to 17 years) Type of cerebral palsy: spastic, dyskinetic GMFCS Levels: IV and V Inclusion criteria: aged 8 to 17 years; with cerebral palsy within GMFCS Levels IV and V; "able to pedal on an adapted static bicycle; walk with partial body weight support on a treadmill" (quote) Exclusion criteria: "had undergone orthopedic surgery to the spine or lower limbs within a year before the enrolment; cognitive or behavioral impairment preventing understanding or compliance with instructions" (quote)
Interventions	Experimental group (n = 9) Intervention: treadmill walking; 30 minutes × 3 sessions per week × 6 weeks (total time: 540 minutes) Body weight support: amount not reported Speed: 0.38 m/s Progression: 0.52 m/s Control group (n = 11) Intervention: usual therapy (stretching, mat exercise, standing, swimming); 30 minutes × 3 sessions per week × 6 weeks (total time: 540 minutes)
Outcomes	Primary outcomes Mobility: not measured Gross motor function: Gross Motor Function Measure—Dimension E (Walking, Running, and Jumping, %) Adverse events: no adverse events Secondary outcomes Participation: not measured Quality of life: not measured Timing of outcome assessment: 0 and 6 weeks
Notes	Funding source: National Institute for Health Research (grant number PB‐PG‐0807‐14074) Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: randomly placing sealed envelopes containing a piece of paper in participant files indicating the group
Allocation concealment (selection bias)	Low risk	Comment: used independent and concealed allocation. Participant group assignment was concealed in a sealed envelope
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: the assessor was blinded, thus reducing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment: 4 (17%) dropped out during the intervention phase—3 from the experimental group and 1 from the control group.a Most (13%) dropped out because of personal health issues. Intention‐to‐treat analysis not reported
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

Cherng 2007.

Study characteristics
Methods	Design: randomized controlled trial (cross‐over study)
Participants	Country: Taiwan Sample size: 8 Dropouts: 0 Mean age: 4 years (range 3 to 7 years) Type of cerebral palsy: spastic diplegia GMFCS Levels: II and III Inclusion criteria: aged 3 to 7 years; with cerebral palsy within GMFCS Levels I, II, and III; "able to follow instructions; had no surgical treatment during the preceding 6 months before study onset" (quote) Exclusion criteria: not stated
Interventions	Experimental group (n = 4) Intervention: treadmill walking; 20 minutes × 2 to 3 sessions per week × 12 weeks (total time: 600 minutes) Body weight support: avoid knee collapse during stance Speed: individual set for comfortable level Progression: increased according to individual’s comfort level Control group (n = 4) Intervention: none Both groups (n = 8) Intervention: usual therapy (neurodevelopmental treatment); 30 minutes × 2 to 3 sessions per week × 12 weeks (total time: 600 minutes)
Outcomes	Primary outcomes Mobility: 10‐Meter Walk Test (m/s) Gross motor function: Gross Motor Function Measure—Dimension E (Walking, Running, and Jumping, %) Adverse events: not measured Secondary outcomes Participation: not measured Quality of life: not measured Timing of outcome assessment: 0 and 12 weeks
Notes	Funding source: National Science Council, Taiwan (grant number NSC 92‐2218‐E‐006‐003) Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: method not described
Allocation concealment (selection bias)	Unclear risk	Comment: method not described
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: assessor was blinded, thus reducing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no dropoutsa
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

Dodd 2007.

Study characteristics
Methods	Design: cluster‐controlled trial
Participants	Country: Australia Sample size: 14 Dropouts: 1 (in control group) Mean age: 9 years (range 5 to 15 years) Type of cerebral palsy: spastic diplegia, quadriplegia, athetotic quadriplegia GMFCS Levels: III and IV Inclusion criteria: aged 5 to 18 years; cerebral palsy within GMFCS Levels III and IV; able to understand simple instructions Exclusion criteria: "needed physical assistance from another person to walk; had lower limb surgery, botulinum toxin injections or serial casting in the 6 months before the trial started; had a concurrent medical condition, such as uncontrolled epilepsy, that would limit their ability to participate in a physical activity programme" (quote)
Interventions	Experimental group (n = 7) Intervention: treadmill walking; 30 minutes × 2 sessions per week × 6 weeks (total time: 360 minutes) Body weight support: observed hip and knee flexion Speed: 0.33 m/s Progression: gradually increasing at each session according to observed steps comfortably Control group (n = 7) Intervention: none
Outcomes	Primary outcomes Mobility: 10‐Meter Walk Test (m/s) Gross motor function: not measured Adverse events: no adverse events Secondary outcomes Participation: not measured Quality of life: not measured Timing of outcome assessment: 0 and 6 weeks
Notes	Funding source: not reported Conflict of interest: none reported *Although 1 participant dropped out of the control group, the study authors reported change scores for all 7 participants
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Comment: no randomization; cluster‐controlled trial
Allocation concealment (selection bias)	High risk	Comment: no concealment; experimental group in only 1 school
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	High risk	Comment: no assessors were blinded, thus increasing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: 1 (7%) in control group dropped out because of unexpected surgery before the second assessmenta; intention‐to‐treat analysis reported
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

Druzbicki 2013.

Study characteristics
Methods	Design: randomized controlled trial
Participants	Country: Poland Sample size: 52 Dropouts: 17 (all in control group) Mean age: 10 years (range 6 to 13 years) Type of cerebral palsy: spastic diplegia GMFCS Levels: II and III Inclusion criteria: children with cerebral palsy; spastic diplegia within GMFCS Levels II and III; "able to independently stand and walk or walk with assistance; no disorders of higher mental functions" (quote) Exclusion criteria: had "botulinum toxin during the last 6 months; surgery within a 1‐year period before the date of the examination; active drug‐resistant epilepsy; anatomical leg length discrepancy larger than 2 cm (due to the Lokomat system limitations); fixed contractures; bone and joint deformities; bone‐articular instability (joint dislocation); baclofen therapy using an implanted infusion pump; inhibiting casts during the last 6 months; significant amblyopia and hearing loss; inflammation of the skin and open skin lesions around the trunk or limb; contraindications for training on a treadmill; lack of patient cooperation" (quote)
Interventions	Experimental group (n = 26) Intervention: robotic walking; minutesb (not reported) × 5 sessions per week × 4 weeks (total time: unable to calculate as minutes not reportedb) Body weight support: individual set Speed: individual set Progression: not reported Control group (n = 9) Intervention: none Both groups (n = 35) Intervention: usual therapy (not reported); minutes × (not reportedb) × 5 sessions per week × 4 weeks (total time: unable to calculate as minutes not reportedb)
Outcomes	Primary outcomes Mobility: 10‐Meter Walk Test (m/s) Gross motor function: not measured Adverse events: not measured Secondary outcomes Participation: not measured Quality of life: not measured Timing of outcome assessment: 0 and 4 weeks
Notes	Funding source: not reported Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: method not described
Allocation concealment (selection bias)	Unclear risk	Comment: method not described
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: assessor was blinded, thus reducing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment: process of recruitment not stated; 17 (33%) in control group dropped out for various unstated reasonsa; intention‐to‐treat analysis not reported
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

El‐Shamy 2017.

Study characteristics
Methods	Design: randomized controlled trial
Participants	Country: Egypt Sample size: 30 Dropouts: 0 Mean age: 10 years (range 8 to 12 years) Type of cerebral palsy: spastic diplegia GMFCS Levels: I and II Inclusion criteria: aged 8 to 12 years; cerebral palsy; "able to walk with or without walking aids; had spasticity degree ranging between grades 1, 1+ and 2, according to the Modified Ashworth Scale; able to understand and follow simple verbal instructions" (quote) Exclusion criteria: had "any fixed deformities that interfere with lower limb functions; cardiac or respiratory conditions that are affected by exercise; presence of seizures or lower limb orthopedic surgery in the preceding 12 months; botulinum toxin injections in the previous 6 months" (quote)
Interventions	Experimental group (n = 15) Intervention: gait trainer walking; 20 minutes × 3 sessions per week × 12 weeks (total time: 720 minutes) Body weight support: observed hip and knee flexion Speed: 75% of overground walking Progression: not reported Control group (n = 15) Intervention: none Both groups (n = 30) Intervention: usual therapy (reflex‐inhibiting patterns, stretching exercises, muscle strengthening, proprioception, balance and gait training for both legs); 60 minutes × 3 sessions per week × 12 weeks (total time: 2160 minutes)
Outcomes	Primary outcomes Mobility: 8‐Meter Walk Test (m/s) Gross motor function: not measured Adverse events: not measured Secondary outcomes Participation: not measured Quality of life: not measured Timing of outcome assessment: 0 and 12 weeks
Notes	Funding source: not reported Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: to randomly allocate participants to groups, used sealed envelopes containing a piece of paper indicating experimental or control group
Allocation concealment (selection bias)	Low risk	Comment: used independent and concealed allocation. Participant group assignment was concealed in a sealed envelope
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: assessor was blinded, thus reducing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no dropoutsa
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

El‐Shemy 2018.

Study characteristics
Methods	Design: randomized controlled trial
Participants	Country: Egypt Sample size: 30 Dropouts: 0 Mean age: 12 years (range 11 to 13 years) Type of cerebral palsy: spastic diplegia GMFCS Levels: I and II Inclusion criteria: aged 11 to 13 years; cerebral palsy within GMFCS Level I to II; "had spasticity degree ranging between grades < 2, according to the Modified Ashworth Scale; had normal vision and hearing; able to follow instructions and understand testing and treatment" (quote) Exclusion criteria: had "any fixed contractures or deformities of the spine or lower extremities within 1 year, auditory or visual agnosia or marked language deficits, and cognitive impairment affecting their ability to follow instructions during assessment and training; botulinum toxin injections within the last 6 months" (quote)
Interventions	Experimental group (n = 15) Intervention: gait trainer walking; 30 minutes × 3 sessions per week × 12 weeks (total time: 1080 minutes) Body weight support: amount not reported Speed: 80% of maximal speed on the gait trainer (determined before training) Progression: 60% to 80% of maximal speed on the gait trainer (determined before training) Control group (n = 15) Intervention: none Both groups (n = 30) Intervention: usual therapy (reflex‐inhibiting patterns, stretching exercises, muscle strengthening, proprioception, balance and gait training for both legs); 60 minutes × 3 sessions per week × 12 weeks (total time: 2160 minutes)
Outcomes	Primary outcomes Mobility: not measured Gross motor function: Modified Timed Up and Go Test(s) Adverse events: not measured Secondary outcomes Participation: not measured Quality of life: not measured Timing of outcome assessment: 0 and 12 weeks
Notes	Funding source: not reported Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: used computer‐generated random number list
Allocation concealment (selection bias)	Unclear risk	Comment: does not describe the person who performed the allocation
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: assessor was blinded, thus reducing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no dropoutsa
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

Gates 2012.

Study characteristics
Methods	Design: randomized controlled trial
Participants	Country: USA Sample size: 34 Dropouts: 8 (4 in experimental group, 4 in control group) Mean age: 10 years (range 6 to 13 years) Type of cerebral palsy: spastic diplegia, triplegia, quadriplegia GMFCS Levels: II to IV Inclusion criteria: aged 6 to 13 years; spastic cerebral palsy within GMFCS Levels II to IV; with "gait velocity < 80% of age‐expected values determined by motion analysis; body weight < 150 pounds due to equipment" (quote) restriction; able to take 8 steps independently with or without assistive devices; can follow multiple‐step commands Exclusion criteria: had "medical condition negatively impacted by exercise (e.g. asthma, severe cardiac abnormalities); movement disorders typically associated with cerebral palsy (such as dystonia and athetosis); lower extremity orthopedic surgery in the past year; botulinum toxin in the past 6 months; dorsal rhizotomy in the past 2 years; intrathecal baclofen, flexion contractures > 30 ° at hip, > 20 ° at knee; plantarflexion contractures > 15 ° with knee extended" (quote); unable to follow multiple commands
Interventions	Experimental group (n = 14) Intervention: treadmill walking; 20 to 40 minutes × 5 sessions per week × 12 weeks (total time: 1800 minutes) Body weight support: 50% reduced to 0% Speed: individual set Progression: 50% increase in initial speed at the end Control group (n = 12) Intervention: no walking (strengthening focused on functional tasks, such as standing); 30 to 60 minutes × 5 sessions per week × 12 weeks (total time: 2700 minutes)
Outcomes	Primary outcomes Mobility: 10‐Meter Walk Test (m/s) Gross motor function: not measured Adverse events: not measured Secondary outcomes Participation: Children’s Assessment of Participation and Enjoyment (CAPE)—Intensity (scored 1 to 7) Quality of life: Pediatric Quality of Life Cerebral Palsy Module (PedsQOL‐CP) (scored 0 to 100) Timing of outcome assessment: 0 and 12 weeks
Notes	Funding source: Shriners Hospitals for Children® (grant number 9147) Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: used blocks to randomly allocate participants to groups
Allocation concealment (selection bias)	Unclear risk	Comment: does not describe the person who performed the allocation
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: assessor was blinded, thus reducing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment: 8 (24%) dropped out—4 from the experimental group and 4 from the control group.a All dropouts during the intervention period. Intention‐to‐treat analysis not reported
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

Gharib 2011.

Study characteristics
Methods	Design: randomized controlled trial
Participants	Country: Egypt Sample size: 30 Dropouts: 0 Mean age: 12 years (range 10 to 13 years) Type of cerebral palsy: spastic hemiplegia GMFCS Level: II Inclusion criteria: aged 10 to 13 years; hemiparetic cerebral palsy with mild degree of spasticity in the affected lower limbs (Modified Ashworth Scale score < 2); within GMFCS Level II Exclusion criteria: had "a fracture, sprain or strain injury of the lower extremities in the past six months; neurological or orthopaedic surgery in the last 12 months; botulinum toxin application for at least six months before the study; exercise induced asthma; a congenital heart defect with cardiac compromise; aggressive or self‐harming behaviours; cognitive impairment (not being able to follow simple verbal commands and instructions during tests and training); an uncontrolled seizure disorder" (quote)
Interventions	Experimental group (n = 15) Intervention: gait trainer walking? (Biodex Gait Trainer 2™); 15 minutes × 3 sessions per week × 12 weeks (total time: 540 minutes) Body weight support: no Speed: gradually increased for each child Progression: according to target step length (e.g. the larger the step length, the faster the speed) Control group (n = 15) Intervention: none Both groups (n = 30) Intervention: usual therapy (traditional physical therapy exercise); 30 minutes × 3 sessions per week × 12 weeks (total time: 1080 minutes)
Outcomes	Primary outcomes Mobility: gait analysis (Biodex Gait Trainer 2™) (m/s) Gross motor function: not measured Adverse events: not measured Secondary outcomes Participation: not measured Quality of life: not measured Timing of outcome assessment: 0 and 12 weeks
Notes	Funding source: no funding Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: method not described
Allocation concealment (selection bias)	Low risk	Comment: used independent and concealed allocation
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: assessor was blinded, thus reducing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no dropoutsa; intention‐to‐treat analysis reported
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

Grecco 2013a.

Study characteristics
Methods	Design: randomized controlled trial
Participants	Country: Brazil Sample size: 35 Dropouts: 0 Mean age: 6 years (range 3 to 12 years) Type of cerebral palsy: spastic GMFCS Levels: I to III Inclusion criteria: aged 3 to 12 years; cerebral palsy within GMFCS Levels I, II, and III; "absence of cognitive or visual impairment that could compromise the performance of the tasks; functional ambulation for at least 12 months" (quote) Exclusion criteria: "had orthopedic surgical procedures or neuromuscular block in the 12 months prior to the training sessions; those with orthopedic deformity with indication for surgery" (quote)
Interventions	Experimental group (n = 17) Intervention: treadmill walking; 30 minutes × 2 sessions per week × 7 weeks (total time: 420 minutes) Body weight support: no Speed: 60% increased to 80% of maximum OG speed Progression: no Control group (n = 18) Intervention: overground walking; 30 minutes × 2 sessions per week × 7 weeks (total time: 420 minutes)
Outcomes	Primary outcomes Mobility: 6‐Minute Walk Test (m/s) Gross motor function: Gross Motor Function Measure—Dimension E (Walking, Running, and Jumping, %) Adverse events: not measured Secondary outcomes Participation: Pediatric Evaluation of Disability Inventory (PEDI)—Mobility (scored 0 to 59) Quality of life: not measured Timing of outcome assessment: 0 and 7 weeks
Notes	Funding source: Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: randomly allocated participants to groups using "a set of numbered, sealed, opaque envelopes"
Allocation concealment (selection bias)	Low risk	Comment: used a central office. Participant group assignment was concealed in a sealed envelope
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: assessor was blinded, thus reducing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no dropoutsa; intention‐to‐treat analysis reported
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

Hösl 2018.

Study characteristics
Methods	Design: randomized controlled trial (cross‐over study)
Participants	Country: Germany Sample size: 10 Dropouts: 0 Mean age: 12 years (range 5 to 19 years) Type of cerebral palsy: spastic GMFCS Levels: I and II Inclusion criteria: aged 5 to 19 years; cerebral palsy within GMFCS Levels I and II; not receiving any botulinum toxin within 24 months before participation; not having had any surgery to the lower leg Exclusion criteria: not stated
Interventions	Experimental group (n = 5) Intervention: treadmill backward‐downhill walking; 23 minutes × 3 sessions per week × 9 weeks (total time: 621 minutes) Body weight support: no Speed: 0.47 m/s with ‐11% inclination Progression: 0.64 m/s with 16% inclination Control group (n = 5) Intervention: shame (7 stretch exercises); 23 minutes × 3 sessions per week × 9 weeks (total time: 621 minutes)
Outcomes	Primary outcomes Mobility: gait analysis (Vicon) (m/s) Gross motor function: Gross Motor Function Measure—Dimension E (Walking, Running, and Jumping, %) Adverse events: not measured Secondary outcomes Participation: not measured Quality of life: not measured Timing of outcome assessment: 0 and 9 weeks
Notes	Funding source: not reported Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: method not described
Allocation concealment (selection bias)	Unclear risk	Comment: method not described
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	High risk	Comment: no assessors were blinded, thus increasing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no dropoutsa
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

Peri 2017.

Study characteristics
Methods	Design: quasi‐randomized controlled trial
Participants	Country: Italy Sample size: 22 Dropouts: 0 Mean age: 9 years (range 4 to 17 years) Type of cerebral palsy: spastic diplegia GMFCS Levels: I and III Inclusion criteria: children with cerebral palsy; "able to communicate pain, fear or discomfort; able to walk independently with or without the use of assistive devices or orthoses; able to cooperate for assessment; femur length bigger than 21 cm for an appropriate use of robotic orthoses; having a regular routine in physiotherapy treatment before this study" (quote) Exclusion criteria: had "multi‐level surgery within six months before the onset of the study; botulinum toxin A injections within the previous three months; under baclofen therapy using an implanted infusion pump" (quote)
Interventions	Experimental group (n = 12) Intervention: robotic walking (Lokomat); 30 minutes × 4 sessions per week × 10 weeks (total time: 1200 minutes) Body weight support: 50% Speed: 0.33 m/s Progression: 0.44 to 0.56 m/s Control group (n = 10) Intervention: overground walking; 30 minutes × 4 sessions per week × 10 weeks (total time: 1200 minutes)
Outcomes	Primary outcomes Mobility: 6‐Minute Walk Test (m/s) Gross motor function: Gross Motor Function Measure—Dimension E (Walking, Running, and Jumping, %) Adverse events: not measured Secondary outcomes Participation: not measured Quality of life: not measured Timing of outcome assessment: 0 and 10 weeks
Notes	Funding source: not reported Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Comment: quasi‐randomized by sequential alternative allocation
Allocation concealment (selection bias)	High risk	Comment: no concealment
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	High risk	Comment: no assessors were blinded, thus increasing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no dropoutsa
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

Sherief 2015.

Study characteristics
Methods	Design: randomized controlled trial
Participants	Country: Egypt Sample size: 30 Dropouts: 0 Mean age: 8 years (range 7 to 11 years) Type of cerebral palsy: spastic hemiplegia GMFCS Levels: I to III (GMFCS level estimated by reviewers) Inclusion criteria: children with cerebral palsy were "at spasticity grades ranged from 1 to +1 according to the Modified Ashworth Scale; were able to follow simple verbal commands included in the tests; did not have fixed deformity of both lower limbs; were able to stand with support" (quote) Exclusion criteria: had "shorting or contracture; cardiovascular diseases; surgery within the previous 24 months; sensory defensiveness; inability to follow instructions" (quote)
Interventions	Experimental group (n = 15) Intervention: treadmill walking + Ankle‐Foot Orthosis; 20 minutes × 3 sessions per week × 12 weeks (total time: 720 minutes) Body weight support: no Speed: 75% of comfortable overground speed Progression: no Control group (n = 15) Intervention: none Both groups (n = 30) Intervention: usual therapy; 60 minutes × 3 sessions per week × 12 weeks (total time: 2160 minutes)
Outcomes	Primary outcomes Mobility: not measured Gross motor function: Peabody Developmental Motor Scales, Second Edition – Locomotion (%) Adverse events: no adverse events Secondary outcomes Participation: not measured Quality of ilfe: not measured Timing of outcome assessment: 0 and 12 weeks
Notes	Funding source: no funding Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: randomly allocated participants to groups using sealed envelopes "containing a card labeled with either group A or B"
Allocation concealment (selection bias)	Low risk	Comment: used sealed envelopes opened by participants
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	High risk	Comment: no assessors were blinded, thus increasing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no dropoutsa
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

Smania 2011.

Study characteristics
Methods	Design: randomized controlled trial
Participants	Country: UK Sample size: 18 Dropouts: 0 Mean age: 13 years (range 10 to 18 years) Type of cerebral palsy: spastic diplegia, quadriplegia GMFCS Levels: II to IV Inclusion criteria: aged 10 to 18 years; with diplegic or quadriplegic cerebral palsy within GMFCS Levels II and IV; able to walk independently or with the use of an assistance device for at least 10 minutes; "able to maintain a sitting position without assistance; able to follow instructions; able to participate in the rehabilitative program" (quote) Exclusion criteria: had "lower limb spasticity of > 2 points or higher on the Modified Ashworth Scale; severe lower limb contractures; cardiovascular diseases; orthopedic surgery or neurosurgery in the past 12 months; botulinum toxin injections within 6 months before the beginning of the study" (quote)
Interventions	Experimental group (n = 9) Intervention: gait trainer walking; 30 minutes × 5 sessions per week × 2 weeks (total time: 300 minutes) Body weight support: 30% reduced to 0% Speed: individual set Progression: increasing at each session according to the last session Control group (n = 9) Intervention: overground walking; 7.5 minutes × 5 sessions per week × 2 weeks (total time: 75 minutes)
Outcomes	Primary outcomes Mobility: 10‐Meter Walk Test (m/s) Gross motor function: not measured Adverse events: no adverse events Secondary outcomes Participation: Functional Independence Measure for Children (WeeFIM) (scored 18 to 126) Quality of life: not measured Timing of outcome assessment: 0 and 2 weeks
Notes	Funding source: CariVerona Fondation (PACIS), Verona, Italy Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: used computer‐generated numbers for sequence of randomization and sealed envelopes for assignment of groups
Allocation concealment (selection bias)	Low risk	Comment: used independent and concealed allocation. Participant group assignment was concealed in a sealed envelope
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: assessor was blinded, thus reducing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no dropoutsa
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

Swe 2015.

Study characteristics
Methods	Design: randomized controlled trial
Participants	Country: Australia Sample size: 30 Dropouts: 0 Mean age: 13 years (range 6 to 18 years) Type of cerebral palsy: spastic hemiplegia, diplegia, triplegia, quadriplegia, athetoid GMFCS Levels: II and III Inclusion criteria: children with cerebral palsy within GMFCS Levels II and III Exclusion criteria: had "visual impairment that could compromise the performance of the tasks; concurrent medical condition that posed a risk to their safety during training; lower limb orthopaedic surgery or botulinum toxin injections in the last 6 months" (quote)
Interventions	Experimental group (n = 15) Intervention: treadmill walking + body weight support; 30 minutes × 2 sessions per week × 8 weeks (total time: 480 minutes) Body weight support: individual set, reduced at each session according to observed upright trunk posture Speed: individual set from 0.03 m/s Progression: gradually increasing at each session according to observed steps comfortably Control group (n = 15) Intervention: overground walking; 30 minutes × 2 sessions per week × 8 weeks (total time: 480 minutes)
Outcomes	Primary outcomes Mobility: 10‐Meter Walk Test (m/s) Gross motor function: Gross Motor Function Measure—Dimension E (Walking, Running, and Jumping, %) Adverse events: not measured Secondary outcomes Participation: not measured Quality of life: not measured Timing of outcome assessment: 0 and 8 weeks
Notes	Funding source: no funding Conflict of interest: none reported Trial registration:ACTRN12613001025729
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: used block randomization to allocate equal numbers of participants to groups
Allocation concealment (selection bias)	Low risk	Comment: used sealed envelopes opened by participants
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: assessor was blinded, thus reducing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no dropoutsa; intention‐to‐treat analysis reported
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; outcome measures consistent with those reported in clinical trials registration
Other bias	Low risk	Comment: no other obvious issues

Wallard 2017.

Study characteristics
Methods	Design: randomized controlled trial
Participants	Country: Belgium Sample size: 30 Dropouts: 0 Mean age: 9 years (range 8 to 10 years) Type of cerebral palsy: spastic diplegia GMFCS Level: II Inclusion criteria: children with cerebral palsy, spastic diplegia, and a jump gait within GMFCS Level II Exclusion criteria: not stated
Interventions	Experimental group (n = 14) Intervention: robotic walking (Lokomat); 40 minutes × 5 sessions per week × 4 weeks (total time: 800 minutes) Body weight support: 70% reduced to 40% Speed: 0.19 m/s Progression: 0.39 m/s Control group (n = 16) Interventions: exercises (stretch, balance); 40 minutes × 5 sessions per week × 4 weeks (total time: 800 minutes)
Outcomes	Primary outcomes Mobility: 10‐Meter Walk Test (m/s) Gross motor function: Gross Motor Function Measure—Dimension E (Walking, Running, and Jumping, %) Adverse events: not measured Secondary outcomes Participation: not measured Quality of life: not measured Timing of outcome assessment: 0 and 4 weeks
Notes	Funding source: not reported Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: randomization by drawing lots but the process is not described clearly
Allocation concealment (selection bias)	Unclear risk	Comment: method not described
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: assessor was blinded, thus reducing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no dropoutsa
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

Willoughby 2010.

Study characteristics
Methods	Design: randomized controlled trial
Participants	Country: Australia Sample size: 34 Dropouts: 8 (5 in experimental group, 3 in control group) Mean age: 11 years (range 5 to 18 years) Type of cerebral palsy: spastic GMFCS Levels: III and IV Inclusion criteria: aged 5 to 18 years; cerebral palsy within GMFCS Levels III and IV; "able to understand simple instructions and reliability indicate yes and no" (quote) Exclusion criteria: "if needed physical assistance from another person to walk with their assistive device; had a concurrent medical condition, such as severe cardiorespiratory disease or uncontrolled epilepsy, that posed a risk to their safety during training; had lower limb orthopedic surgery or botulinum toxin injections in the 6 months prior to their participation" (quote)
Interventions	Experimental group (n = 12) Intervention: treadmill walking + body weight support; 30 minutes × 2 sessions per week × 9 weeks (total time: 540 minutes) Body weight support: individual set, reduced when possible according to observed posture Speed: 0.03 to 0.56 m/s Progression: 0.07 to 0.82 m/s Control group (n = 14) Intervention: overground walking; 30 minutes × 2 sessions per week × 9 weeks (total time: 540 minutes) Both groups (n = 26) Intervention: usual therapy (group‐based sessions, number of sessions unavailable)
Outcomes	Primary outcomes Mobility: 10‐Meter Walk Test (m/s) Gross motor function: not measured Adverse events: not measured Secondary outcomes Participation: School Function Assessment—Travel (scored 19 to 76) Quality of life: not measured Timing of outcome assessment: 0 and 10 weeks
Notes	Funding source: Faculty of Health Sciences, La Trobe University, Melbourne, Australia (grant no. 1206/101961) Conflict of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: a set of sequentially numbered, sealed, opaque envelopes used for enrollment, and sealed envelopes for randomized assignment of groups
Allocation concealment (selection bias)	Low risk	Comment: used independent and concealed allocation. Participant group assignment was concealed in a sealed envelope
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: neither participants nor personnel were blinded, thus increasing the risk of performance bias
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: assessor was blinded, thus reducing the risk of detection bias
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Comment: 8 (24%) dropped outa. Most (21%) dropped out because of personal health issues. Intention‐to‐treat analysis reported
Selective reporting (reporting bias)	Low risk	Comment: point estimate and variability reported; no study protocol available
Other bias	Low risk	Comment: no other obvious issues

^aWe judged a study to be at high risk of attrition bias if study authors reported greater than 15% dropout rates and did not report an intention‐to‐treat analysis. We judged a study to be at unclear risk of attrition bias if study authors reported greater than 15% dropout rates or did not report an intention‐to‐treat analysis. We judged a study to be at low risk of bias if study authors reported dropout rates less than 15% and had conducted an intention‐to‐treat analysis, or reported no dropout rates.
^bWrote to study authors for more information but received no response.

GMFCS: Gross Motor Function Classification System.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Angulo‐Barroso 2013	Participants do not have cerebral palsy
Anttila 2008	Not a randomized controlled trial (RCT) or quasi‐RCT
Arellano‐Martinez 2013	Comparator is not "no walking" or "same dose of overground walking"
Bayón 2018	Not a randomized controlled trial (RCT) or quasi‐RCT
Bjornson 2019	Comparator is not "no walking" or "same dose of overground walking"
Booth 2018	Not a randomized controlled trial (RCT) or quasi‐RCT
Borggraefe 2010	Not a randomized controlled trial (RCT) or quasi‐RCT
Brutsch 2011	Not a randomized controlled trial (RCT) or quasi‐RCT
Carvalho 2005	Participants do not have cerebral palsy
Carvalho 2006	Participants do not have cerebral palsy
Carvalho 2017	Not a randomized controlled trial (RCT) or quasi‐RCT
Carvalho de Abreu 2008	Participants do not have cerebral palsy
Chan 2004	Comparator is not "no walking" or "same dose of overground walking"
Chen 2016	Intervention is not "mechanically assisted walking training with/without body weight support"
Cho 2016	Comparator is not "no walking" or "same dose of overground walking"
Chrysagis 2012	Comparator is not "no walking" or "same dose of overground walking"
Damiano 2009	Not a randomized controlled trial (RCT) or quasi‐RCT
Damiano 2017	Comparator is not "no walking" or "same dose of overground walking"
David 2006	Participants do not have cerebral palsy
Druzbicki 2010	Intervention is not "mechanically assisted walking training with/without body weight support"
DuarteNde 2014	Comparator is not "no walking" or "same dose of overground walking"
Elnahhas 2019	Not a randomized controlled trial (RCT) or quasi‐RCT
Emara 2015	Comparator is not "no walking" or "same dose of overground walking"
Emara 2016	Comparator is not "no walking" or "same dose of overground walking"
Franki 2015	Intervention is not "mechanically assisted walking training with/without body weight support"
Grecco 2013b	Not a randomized controlled trial (RCT) or quasi‐RCT
Grecco 2014	Comparator is not "no walking" or "same dose of overground walking"
Krebs 2009	Not a randomized controlled trial (RCT) or quasi‐RCT
Lee 2015	Participants do not have cerebral palsy
Lerner 2017	Not a randomized controlled trial (RCT) or quasi‐RCT
Lorentzen 2017	Participants are not aged 3 to 18 years
Lotfian 2019	Not a randomized controlled trial (RCT) or quasi‐RCT
Mattern‐Baxter 2013	Participants are not aged 3 to 18 years
Meyer‐Heim 2007	Not a randomized controlled trial (RCT) or quasi‐RCT
Olama 2011	Intervention is not "mechanically assisted walking training with/without body weight support"
Parvin 2017	Comparator is not "no walking" or "same dose of overground walking"
Sarhan 2014	Comparator is not "no walking" or "same dose of overground walking"
Su 2013	Participants are not Gross Motor Function Classification System (GMFCS) Levels I to IV
Sukal‐Moulton 2014	Not a randomized controlled trial (RCT) or quasi‐RCT
van Hedel 2016	Not a randomized controlled trial (RCT) or quasi‐RCT
Wu 2017a	Comparator not "no walking" or "same dose of overground walking"
Wu 2017b	Comparator not "no walking" or "same dose of overground walking"
Yazici 2019	Not a randomized controlled trial (RCT) or quasi‐RCT

Characteristics of ongoing studies [ordered by study ID]

ACTRN12617001410347.

Study name	Public title: What are the possible benefits of robotic assisted gait training and how much training is needed: a pilot randomised clinical trial with children with cerebral palsy Scientific title: What is the effect of robotic assisted gait training on self‐perception of performance and participation, using the Canadian Occupational Performance Measure, and how much training is needed to determine change: a pilot randomised clinical trial with children with cerebral palsy
Methods	Design: RCT (cross‐over) Method of randomization: "sequence is computer generated, and stratified to Gross Motor Function Measure to ensure equal numbers of each level in each arm" (quote) Blinding of outcome assessors: yes Intention‐to‐treat analysis: unclear
Participants	Country: Australia Target sample size: 10 Inclusion criteria Children with diagnosis of cerebral palsy GMFCS Levels II to IV Aged 5 to 14 years Able to follow instructions and to participate in a minimum of 45 minutes of active physical therapy Able to communicate pain and discomfort and to report on perceived level of exertion Exclusion criteria Body weight < 15 kg or > 135 kg Femur length < 23 cm or > 47 cm Unable to sustain vertical position for > 20 minutes (orthostatic) Knee flexion contracture > 10 ° Knee valgus > 40 ° Orthopedic surgery < 9 months for soft tissues and < 18 months for lower limb bony surgery Botulinum toxin (type A) injection in the previous 4 months or planned during study period Uncontrolled seizure disorder Weight‐bearing restrictions Pregnancy Behavioral problems/inability to offer consent in addition to guardian
Interventions	Experimental group Intervention: robotic assisted gait training. Arm 1: 30 to 40 minutes × 2 sessions per week × 6 weeks. Arm 2: 30 to 40 minutes × 4 sessions per week × 6 weeks Control group Intervention: 6‐week waiting list, then random allocation to arm 1 or arm 2
Outcomes	Primary outcomes Change in Canadian Occupational Performance Measure Secondary outcomes Gross Motor Function Measure sections D and E Standardized muscle circumference of thigh and calf (centimeters) Borg Rating of Perceived Exertion Scale Australian Spasticity Assessment Scale Range of motion of lower limbs Selective motor control in lower limbs Gross Motor Function Classification Scale Functional Mobility Scale Two‐Minute Walk Test Cerebral Palsy Quality of Life Questionnaire—Child Children’s Assessment of Participation and Enjoyment Preferences for Activities of Children Outcomes will be measured at: 0, 6, and 12 weeks
Starting date	October 6, 2017
Contact information	Name: A/Prof Remo (Ray) Russo, Head of Research Address: Paediatric Rehabilitation Department, Women's and Children's Health Network, 72 King William Road, North Adelaide SA 5006, Australia Email:ray.russo@sa.gov.au
Notes

NCT00887848.

Study name	Public title: Effectiveness of robotic assisted gait training in children with cerebral palsy (PeLoGAIT) Scientific title: Effectiveness of robotic assisted gait training in children with cerebral palsy: a randomized controlled clinical trial including 3D gait analysis
Methods	Design: RCT (cross‐over) Method of randomization: "randomization into the two groups with different intervention sequences is performed using a minimization method with a random factor of 0.9, including the factors severity of impairment (GMFCS‐level II or III/IV), age (6–11 years or 11–18 years) and Botulinum Toxin A‐treatment (present or absent) in the preceding 6 months" (quote) Blinding of outcome assessors: yes Intention‐to‐treat analysis: unclear
Participants	Country: Switzerland Target sample size: 34 Inclusion criteria Bilateral spastic cerebral palsy GMFCS Levels II to IV Exclusion criteria Prior orthopedic surgery on lower extremity or trunk (< 6 months) Prior neurosurgical interventions (< 6 months) Significant mental retardation Severe contractures Prior Lokomat training (< 6 months)
Interventions	Experimental group Intervention: robot‐assisted gait training (Lokomat); 45 minutes × 3 sessions per week × 5 weeks Control group Intervention: 5‐week waiting list
Outcomes	Primary outcome Gross Motor Function Measure—66, section E (GMFM‐66‐E) Secondary outcomes GMFM‐66, section D Gait speed 6‐Minute Walk Test 3D gait analysis Outcomes will be measured at: 0, 6, and 12 weeks
Starting date	October 5, 2019
Contact information	Name: Corinne Ammann‐Reiffer Address: University Children's Hospital Zurich, Rehabilitation Center Affoltern Email:Corinne.ammann@kispi.uzh.ch
Notes

NCT02196298.

Study name	Public title: A randomized trial comparing the Lokomat with a gait‐related physiotherapy program in children with cerebral palsy Scientific title: A randomized cross‐over clinical trial comparing the impact of the Lokomat gait training system with a gait‐related physiotherapy program in children with cerebral palsy
Methods	Design: RCT (cross‐over) Method of randomization: unclear Blinding of outcome assessors: yes Intention‐to‐treat analysis: unclear
Participants	Country: Canada Target sample size: 40 Inclusion criteria Ages 5 to12 years inclusive Assessed as GMFCS Level II or III Able to follow testing instructions and participate in a minimum of 30 minutes of active PT Able to reliably signal pain, fear, and discomfort Passive range of motion (ROM) of hips and knees within minimum range requirement for Lokomat (hip and knee flexion contracture < 10 °, knee valgus < 40 °) Client of Child Development Program at Holland Bloorview Able to commit to attendance twice weekly for 8 weeks (to support the primary efficacy analysis) Exclusion criteria Fixed knee contracture > 10 °, knee valgus > 40 ° such that robotic leg orthosis will not be adaptable to lower limbs Hip instability/subluxation > 45% Orthopedic surgery within the last 9 months (muscle) or 12 months (bone) Botulinum toxin‐A (BTX‐A) injections to lower limb in last 4 months Inability to discontinue BTX‐A for period of 6 months (during trial) due to concerns about ROM or pain Severe spasticity may be a contraindication Any weight‐bearing restrictions Seizure disorder that is not controlled by medication (if on medication, must not have had a seizure in the last 12 months) Open skin lesion or vascular disorder of lower extremities Not able to co‐operate or be positioned adequately within the Lokomat as shown during the 2 Lokomat fitting/acclimatization sessions Not prepared or unable to discontinue a regular therapy intervention during the course of the trial Involved in another intervention study
Interventions	Experimental group Intervention: robot‐assisted gait training (Lokomat); 30 minutes × 1 to 2 sessions per week × 8 to 10 weeks (16 sessions in total) + overground walking; 5 minutes × 1 to 2 sessions per week × 8 to 10 weeks (16 sessions in total) Control group Intervention: usual therapy; 35 minutes × 1 to 2 sessions per week × 8 to 10 weeks (16 sessions in total)
Outcomes	Primary outcome Gross Motor Function Measure—66 (GMFM‐66) Secondary outcomes Six‐minute walk test Advanced motor skills on the Challenge Module (for children in GMFCS Level II) Activities Scale for Kids KidScreen Questionnaire (health‐related quality of life) Children's Assessment of Participation and Enjoyment Gait kinematics, measured with the GaitRite evaluation system (time distance parameters) Gait quality, measured on an observational gait scale Goal Attainment Scaling Canadian Occupational Performance Measure Quality Function Measure Other outcomes Dimensions of Mastery Questionnaire at first baseline (motivation and persistence with difficult tasks) ROM of hip, knee, and ankle Body pain (each treatment session) Outcomes will be measured at: 0 and 8 weeks
Starting date	October 2012
Contact information	Name: Dr Virginia Wright Address: Holland Bloorview Kids Rehabiltation Hospital, Toronto, Ontario, Canada Email:vwright@hollandbloorview.ca
Notes

NCT02391324.

Study name	Public title: Effectiveness of robotic gait training and physical therapy for children and youth with cerebral palsy Scientific title: Evaluation of the effectiveness of robotic gait training and gait‐focused physical therapy programs for children and youth with cerebral palsy: a mixed methods randomized controlled trial
Methods	Design: RCT (factorial) Method of randomization: unclear Blinding of outcome assessors: yes Intention‐to‐treat analysis: unclear
Participants	Country: USA, Canada Target sample size: 160 Inclusion criteria Diagnosis of CP (any type), GMFCS Levels II and III Able to follow GMFM testing instructions and to participate in a minimum of 30 minutes of active PT (as judged by the child's PT or physician if not followed by a PT) Able to reliably signal pain, fear, and discomfort using verbal or non‐verbal signals Passive range of motion (ROM) of hips and knees within minimum range requirement for LOK (hip and knee flexion contracture < 10 °, and knee valgus < 40 °) Parent/child agrees to attend 16 study intervention sessions (given within two 10‐week periods), an LOK fitting/acclimatization session or fPT acclimatization session, and the 3 assessment sessions during the course of the study Parent agrees to contact primary PT (if not already involved by parent in the screening process) and pediatrician or physiatrist to confirm eligibility Parent agrees that regular PT (and other gross motor mobility therapies such as conductive education and medek) will be discontinued from the time of the baseline assessment through the 8 to 10 weeks of active intervention or control. Note that home programs such as stretching and strengthening and treadmill and exercise bike riding (no longer than 10 minutes total per day) will be permitted in all groups. Exclusion criteria Botulinum toxin injection within the past 4 months or planned within the next 6 months Fixed knee contracture > 10 °, knee valgus > 40 ° such that orthosis will not be adaptable to lower limbs Hip instability/subluxation as demonstrated by a migration percentage > 45% Orthopedic surgery (soft tissue releases) within last 9 months, or lower limb bony surgery within last 18 months Severe spasticity may be a contraindication as determined in Lokomat trial session using L‐FORCE assessment Any weight‐bearing restrictions Seizure disorder unless fully controlled by medication and no evidence of seizure in last 12 months and physician provides singed approval to enter the study Open skin lesion or vascular disorder of lower extremities Not able to co‐operate or be positioned adequately within the LOK as shown during the fitting/acclimatization session
Interventions	Experimental group Intervention: arm 1: robot‐assisted gait training (Lokomat); 50 minutes × 2 to 3 sessions per week × 8 to 10 weeks; arm 2: overground walking; 50 minutes × 2 to 3 sessions per week × 8 to 10 weeks; arm 3: robot‐assisted gait training (Lokomat) + overground walking; 50 minutes × 2 to 3 sessions per week × 8 to 10 weeks Control group Intervention: none
Outcomes	Primary outcome Gross Motor Function Measure—66 Secondary outcomes Six‐Minute Walk Test Canadian Occupational Performance Measure Goal Attainment Scaling Adapted Shuttle Run Test Pediatric Berg Balance Scale Quality FM (quality of movement) Activities Balance Confidence Scale Pediatric Evaluation of Disability Inventory—a Computer Adaptive Test Steps/d Physical Activity Self‐Efficacy Scale Participation and Environment Measures for Children and Youth Kidscreen Life Satisfaction Scale Outcomes will be measured at: 0, 4, and 8 weeks
Starting date	January 2016
Contact information	Name: Dr Lesley Wiart Address: University of Alberta, Edmonton, Alberta, Canada Email:lwiart@ualberta.ca
Notes

NL8154.

Study name	Public title: Functional gait training for children and adolescents with cerebral palsy Scientific title: A new functional gait training for children and adolescents with cerebral palsy to improve the gait adaptability: an explorative study
Methods	Design: RCT (cross‐over) Method of randomization: unclear Blinding of outcome assessors: unclear Intention‐to‐treat analysis: unclear
Participants	Country: The Netherlands Target sample size: 30 Inclusion criteria Ages 6 to 17 years Spastic, dyskinetic, or ataxtic cerebral palsy, both unilateral and bilateral GMFCS Level I or II Referral to rehabilitation specialist with a question concerning walking ability Exclusion criteria 1. Surgery in last 2 years such as single‐event multi‐level surgery or selective dorsal rhizotomy or intrathecal baclofen therapy 2. Botulinum toxin injection in lower extremity in the last 6 months 3. Epilepsy, severe vision problems, cognitive problems, or temporary complaints, etc., affecting walking
Interventions	Experimental group Intervention: a treadmill; 45 minutes × 2 sessions per week × 5 weeks Control group Intervention: 5‐week waiting list
Outcomes	Primary outcomes Walking Adaptability Ladder Test for Kids. Participants walk in a ladder of 10 meters, in which the targets decrease by 2 centimeters. The score on the WAL‐K will be determined by means of completion time and failures during the task Secondary outcomes Obstacle avoidance task on the GRAIL Motor plan task on the GRAIL Motor control during walking on the GRAIL: number of synergies during walking based on electromyography signals of the lower extremity 10‐meter walk test: comfortable and maximal walking speeds Functional muscle power of the lower extremity Perceived motor competence Quality of life (KIDSCREEN‐52) GMFCS Level Localization of the CP (unilateral vs bilateral) Outcomes will be measured at: 0, 5, and 10 weeks, and after 3 months
Starting date	September 5, 2019
Contact information	Name: Rosanne Kuijpers Address: Sint Maartenskliniek, Netherlands Email:r.kuijpers@maartenskliniek.nl
Notes

PACTR201901582864286.

Study name	Public title: Effect of locomotor training with a robotic‐gait orthosis (Lokomat) in spasticity modulation of spastic hemiplegic children Scientific title: Effect of locomotor training with a robotic‐gait orthosis (Lokomat) in spasticity modulation of spastic hemiplegic children
Methods	Design: RCT (factorial) Method of randomization: simple randomization using a randomization table created by a computer software program; numbered containers Blinding of outcome assessors: no Intention‐to‐treat analysis: unclear
Participants	Country: Egypt Target sample size: 30 Inclusion criteria Ages 6 to 18 years Grade 1 or 1+ according to modified Ashworth Scale and grade II or III according to Manual Ability Classification Scale Able to understand and follow verbal commands and instructions included in the test No sensory impairment or other neurological or psychological problems other than mild perceptual defects Exclusion criteria Any other neurological deficits such as convulsions Involuntary movements or receiving muscle relaxants Children with impairment of sensation (superficial, deep, and cortical) Children with any other disease such as cardiac disease and with severe mental retardation (intelligence quotient not less than 50) Children with anatomical leg length discrepancy (Lokomat limitation) exceeding 2 cm, fixed contracture, bone and joint deformities, bone articular instability (joint dislocation), inflammation of skin, and open skin lesion around the trunk or limb
Interventions	Experimental group Intervention: Lokomat gait training; 60 minutes × 3 sessions per week × 12 weeks Control group Intervention: traditional exercise program; 60 minutes × 3 sessions per week × 12 weeks
Outcomes	Primary outcomes Gait parameters with 3D motion analysis system for ankle joint angle at initial contact Stride length Cadence Speed Secondary outcomes 1. Modified Ashworth Scale Timing of outcome assessment: 0 and 12 weeks
Starting date	June 5, 2018
Contact information	Name: Mohamed Mostafa Address: 7 Ahmed Elziat St Bein Srayyat Giza, Egypt Email:drsergany_79@hotmail.com
Notes

Yang 2019.

Study name	Public title: Effect of robot‐assisted gait training in children with cerebral palsy Scientific title: Effect of robot‐assisted gait training in children with cerebral palsy
Methods	Design: RCT (cross‐over trial) Method of randomization: not stated Blinding of outcome assessors: yes Intention‐to‐treat analysis: not stated
Participants	Country: unclear Target sample size: 20 Inclusion criteria Children diagnosed with spastic cerebral palsy (ages 3 to 12 years) GMFCS Levels II to IV Height of 98 to 160 centimeters Able to follow instructions and communicate if they feel pain or discomfort (weeFIM score: > 11 points in communication, social cognition domain) Exclusion criteria Children with cognitive impairment History of neurosurgery or orthopedic surgery on limbs Severe joint contracture (knee joint: > 20 ° flexion contracture, hip joint > 40 ° contracture)
Interventions	Experimental group Intervention: robotic training; 3 sessions per week for 6 weeks Control group Intervention: 6‐week waiting list
Outcomes	Primary outcomes Gross Motor Function Measure (GMFM‐88) score Functional Independence Measure (WeeFIM) score Secondary Manual muscle power Range of motion Modified Ashworth Scale Timing of outcome assessment: 0 and 6 weeks
Starting date
Contact information	Name: not stated Address: not stated Email: not stated
Notes

CP: cerebral palsy.
fPT: functional physical therapy program.
GMFCS: Gross Motor Function Classification System.
LOK: Lokomat.
PT: Physical therapist.
RCT: randomized controlled trial.

Differences between protocol and review

1. Authorship. Dr Leanne Jonhston was unable to contribute to the review and hence was removed from the review author line.

2. Data collection and analysis. We removed from this section any methods that we were unable to use in the review and placed them in Table 5.

3. Types of interventions. We added a definition of no/sham walking training for comparators because the control group may have received a small amount of walking training.

4. We ran an additional pre‐publication search to ensure that none of our included studies had been retracted because of fraud, error, or other reasons.

Contributions of authors

Hsiu‐Ching Chiu (HC) contributed to conception and design of the review. HC searched electronic databases, screened tittles and abstracts of records identified by the searches for relevance, assessed and selected trials for inclusion, extracted trial and outcome data, contacted trialists about unpublished data, entered the data into RevMan 5 (Review Manager 2014), assessed the risk of bias for each included study, rated the quality of evidence, carried out the statistical analyses and interpreted the data, and contributed to and approved the final version of the review. HC is the guarantor for this review.

Louise Ada (LA) contributed to conception and design of the review. LA resolved disagreements regarding the selection of trials for inclusion and the "Risk of bias" assessments, evaluated and extracted trial data, assessed the quality of selected trials, guided analyses and interpretation of data, and contributed to and approved the final version of the review.

Theofani A Bania (TB) searched conference proceedings, screened titles and abstracts of records identified by the searches for relevance, assessed and selected trials for inclusion, extracted trial and outcome data, assessed the risk of bias in each included study, rated the quality of evidence, analyzed and interpreted data, and contributed to and approved the final version of the review.

Sources of support

Internal sources

• None, Other

External sources

• None, Other

Declarations of interest

Hsiu‐Ching Chiu—none known.
Louise Ada—none known.
Theofani Bania—none known.

New