The effectiveness of robotic-assisted gait training for paediatric gait disorders: systematic review

Experimental studies were examined to determine the current volume and quality of available literature on RAGT in paediatric participants (aged 5–17 years). The search population focused on children with gait disorders of any aetiology (neurological, orthopaedic or developmental). The intervention included any dedicated period of rehabilitation (e.g. more than a single session) using a specified RAGT device (e.g. Lokomat ®, Gait Trainer GT1 ®), where the effect could be distinguished from concomitant therapy. As discussed, RAGT can provide all or part of a naturalistic gait cycle, contrasting body weight support/treadmill training where the participant or the therapist must provide or augment active motion beyond that provided by the device.

The search strategies were influenced by the PRISMA protocol [8]. A health sciences librarian was consulted for the initial development of the search protocol. Search terms included pediatric OR paediatric; lokomat OR robotic assisted treadmill; child* OR adolescen* OR youth OR young adult* OR teenager*; robotic assisted gait training; gait. Databases searched included CINAHL (1981–2016); EMBASE (1980–2015); Cochrane Database of Systematic Reviews (1993–2016); Physiotherapy Evidence Database (PEDro); MEDLINE (1980–2016); and Scopus (1960–2016). Limiters included age and language. Pearling of the reference lists of identified studies was conducted.

Articles were included if they were:

Articles were excluded if they:

Any level of experimental research design was considered given the need to determine RAGT’s broad clinical effectiveness. Other research manuscripts (e.g. literature reviews) were considered in the initial stages to facilitate pearling of sources, but were not included in the final summary of evidence.

From the original list of studies generated, one reviewer removed studies that were obviously ineligible based on title. After this first cull, abstracts and/or full texts were then reviewed by two reviewers independently and further exclusions made. Decisions for inclusion/exclusion were based on consensus, and facilitated by the third reviewer as necessary. All authors had articles allocated as first, second or third reviewer equally.

Two of the three potential reviewers independently extracted data, including a determination of study type and therefore level of evidence (using the NHMRC levels of evidence tables [9]), population characteristics, practice setting, total number of participants, intervention (and comparison), outcomes and results. The reported outcomes were further categorised into the domains stipulated by the International Classification of Functioning, Disability and Health (ICF).

Included articles were critically appraised by at least two members of the research team using the risk of bias checklist constructed by the Cochrane Collaboration [10]. This list provides the key aspects of trial design that avoid bias in study conduct including randomising, blinded conduct and adequate reporting. The three authors acted independently as primary or secondary reviewers for each study with the third allocated as consensus arbiter as necessary.

All extracted data were reported in table form and as a narrative synthesis. Where appropriate, data were pooled from sufficiently homogenous studies in a meta-analysis [10] to produce summary effects using Revman 5.

Seventeen papers were identified from the search – see Fig. 1 for the flowchart summarising the results of the search. Of the 17 identified, three were RCTs (NHMRC II), one used a non-randomised control group (NHMRC III2), three used an interrupted time-series design (NHMRC III3), and the remaining 10 were pre/post-test design with no control (NHMRC IV). Individual studies and their level of evidence, as described using the NHMRC nomenclature, are summarized in Table 1.

Four hundred and eighty-six participants were recruited across the 17 reports (Table 1). Sample sizes varied from one participant to 89 participants in the included studies. Every study included participants with CP, with the majority rated on the Gross Motor Function Classification System (GMFCS) between level I and IV. Only one study included a participant who had a GMFCS level V [11]. The type and distribution of CP were diverse, with most studies including a variety of presentations including bilateral and unilateral CP of spastic and dyskinetic forms. One study specified its purpose as assessing the RAGT performance of children with CP that had either crouch gait or equinus foot posture (n = 4) [5].

Other clinical conditions besides CP were also reported. Borggraefe and colleagues [12] included participants with paraplegia (cause unspecified), Rett Syndrome, traumatic brain injury, Guillain-Barré syndrome, hip dysplasia and patella ‘luxation’ (dislocation). Meyer-Heim and colleagues [13] also recruited participants with Guillain-Barré syndrome, traumatic brain injury and incomplete paraplegia (cause unspecified), as well as haemorrhagic encephalitis and stroke. Brutsch and colleagues [14] reported participants having myelomeningocele, systemic lupus erythematosus and traumatic brain injury as well as CP.

All studies documented an age range, and the mean age across the studies (including control groups) was 9 years 9 months. The studies by Druzbicki and colleagues [15] and Koenig and colleagues [16] were not included in this calculation as no mean age was reported and could not be independently determined.

The heterogeneous aims and methods across trials meant no consistent RAGT schedule was described – the schedules (dosages) are summarized in Table 1 for both the experimental and control groups. Fifteen studies used the Lokomat ® brand of RAGT, one study used the Gait Trainer GT I ® [4] and one used a novel mobile RAGT device called the CPWalker [17]. Most studies described an intervention schedule of between 2 and 5 training sessions per week, over a 2–6 week testing period. Principal aims of these studies were to investigate the improvement of gross motor performance, or kinematic/tempero-spatial performance in the ICF body structure and function domain, from the use of RAGT. Five studies were found and excluded as they described participants’ involvement in a single RAGT session.

Participants’ total time spent on RAGT devices was inconsistent. Testing schedules reflected a typical therapy session duration, between 25 and 60 min and may have included other warm-ups and stretching components. No clear evidence-based guidelines regarding the development of training schedules were articulated. For example, there was little discussion around the use of exercise parameters (e.g. VO₂ max, heart rate) to determine a personalised training schedule for participants, with most studies appearing to use the RAGT device ‘as tolerated’ by the participant.

A control intervention was only employed in seven of the studies and included variations such as an alternate form of feedback, [5, 14] individual physiotherapy only [15, 18, 19], alternative exercises [4] or combined with functional electrical stimulation [20].

Risks of bias assessments for the studies are detailed in Table 2. Scores were generally indicative of extremely high risk of bias, with the major issues being lack of randomization (both generation and concealment, without a control group), as well as conduct issues of poor blinding of assessors, and poor reporting of data (Table 2). There were general sampling issues – mostly samples of convenience from existing clinics and small numbers with no sample size indications for the power of the study. Clearly no study achieved blinding of participants or personnel administering the intervention.

A variety of outcome measures were reported in the studies, some standardised and some self-developed. The outcome measures used, their ICF domains and a summary of findings are represented in Table 3. The most common measures used were related to mobility, such as the 6 min walk test (6MinWT) and 10 m walk test (10MWT) [4, 5, 13, 19, 21‐24], and the Gross Motor Function Measure-66 (GMFM-66: items D ‘standing’; E ‘walking’) [5, 11, 13, 19, 21‐25]. Other mobility-related outcomes included functional walking capability, and independence or level of assistance for gait (functional ambulation classification (FAC)) [13, 23, 24].

Other outcomes included motivation to participate, balance and gait parameters (from posturography and 3D analyses), range of motion, motivation to participate, general levels of function (WeeFIM), as well as documenting adverse effects from the use of RAGT.

Most outcomes reported in individual studies were in the activity domain; as mentioned these were predominantly for the activity of walking. Eight studies assessed children’s preferred walking speed following RAGT. Considering firstly the meta-analysis, we were able to pool data from two of the three RCTs for the outcome measure of gait speed [4, 18]. We found no significant difference between RAGT and the comparison (exercise or physiotherapy alone) (standardised mean difference of 0.11, 95% confidence intervals −0.48–0.70; p = 0.720. See Fig. 2). For the remaining individual studies all (except Schroeder [19]) showed that participants’ use of RAGT led to improvements in 10mWT performance with the studies by Smania and colleagues [4], Patritti and colleagues [5], Borggraefe and colleagues [21, 22] indicating that the changes were statistically significant at 1, 3, 4 and 6 months follow-up respectively.

Functional gross motor performance was assessed by nine studies using the GMFM-66, items D (standing) and/or E (walking). All studies except van Hedel [24] showed statistically significant improvements in item D (standing) of the GMFM. Six studies showed statistically significant improvements in item E following RAGT [11, 13, 19, 21, 22, 25]. Three of the eight studies evaluating children’s walking endurance (6MinWT) noted statistically significant improvements [4, 21, 22].

The only participation measure used was the Canadian Occupational Performance Measure (COPM) [19] which reported significant differences post intervention.

For the domain of body structure and function, two studies evaluated the changes in children’s balance performance following combined physiotherapy or FES and RAGT, measured using a stabilometric platform [15, 20]. For the Druzbicki study [15], statistically significant improvements were noted in five of 14 balance domains assessed in the experimental group (n = 9), compared with no significant difference in any balance domains in the control group who received physiotherapy intervention only (n = 5). The other study [20] compared RAGT with RAGT plus FES and found a greater percent improvement in balance measures with the combination. Three studies reported the enhanced effect of facilitating active engagement among children using RAGT through ‘augmented feedback’; notably either through the use of virtual reality games (e.g. soccer) or therapist encouragement to help children change their level of response to different training demands during RAGT [5, 14, 16].

No studies reported on any aspects of cost or cost-benefit in relation to alternate forms of therapy.

Table 4 summarises adverse events as reported by one trial. Borggraefe and colleagues [12] conducted the sole study with the primary focus on documenting adverse events among their cohort. They report 47 adverse events among 38 of their 89 participants, with examples including muscle and joint pain, skin erythema and skin lesions, often at the harness site. Five incidents were reported as ‘mild-to-moderate’ [12], and overall very few participants were excluded from RAGT due to an adverse event. No other studies reported this information.

Overall there is weak and inconsistent evidence for the effectiveness of RAGT for children with gait disorders. The majority of the evidence is for children with CP and for the benefit of improving gait attributes such as speed, standing ability and in some instances walking distance. However the studies are generally of low level design, with a small sample size and poor conduct ratings. Therefore a clinical recommendation to adopt RAGT in children and adolescents with gait impairment cannot be made until more consistent findings are reported in larger RCTs.

The disparate aims and training schedules used in the existing literature further complicate the development of appropriate recommendations for the use of RAGT with children. At this stage it is unclear from the scientific evidence what intensity, duration or timing of RAGT is best. Borggraefe and colleagues [22] do provide evidence that a relatively intense program over 12 weeks does seem to produce sustained benefit after 6 months however there was no control group so the effects of time alone are not known. Interestingly, children who participate in RAGT do particularly well when virtual games (e.g. soccer) or therapist encouragement are used to enhance their active participation [5, 11, 14]. While not inferring that the use of virtual games is essential for successful RAGT, it does highlight that active participation may be impeded when children’s focus is on more passive media, such as DVDs that require little engagement [14]. Rehabilitation strategies have been shown to produce measureable neuroplastic changes in the cortex of children with hemiplegic CP associated with improvement in upper limb motor skills [26]. These improvements appeared to be more consistent with active, as opposed to passive interventions. Clinicians should consider the type of feedback and motivational tools required to maintain children’s attention on their walking and body placement during RAGT for maximum exercise benefits and active motor learning.

The available evidence has not yet comprehensively highlighted the effectiveness of RAGT at different life stages. For example, the effect of training on children in their early stages of acquisition of walking compared with teenagers experiencing pubertal growth who already have established gait patterns. Van Hedel and colleagues [24] provide some data that suggest that children with CP who are more severely affected may benefit more than those who are less affected – particularly in the areas of walking-related outcomes. As mentioned, the majority of the studies were either solely children with CP or mixed groups. Further research needs to study well-defined groups across a range of diagnoses. Other determinants of success have been suggested as age and gross motor abilities [11, 19] at baseline. While the evidence suggests RAGT may have significant clinical utility and enjoyment among children, training protocols and baseline standards of its use would be useful information for clinicians so that guidelines are evidence-based and not ad hoc.

It appears that the effectiveness of RAGT for children’s gait can be measured with commonly used, commonly available, activity-based outcome measures such as the GMFM or the 10MWT and that in some cases these can be confirmed as clinically meaningful changes. Based on these findings, it is recommended that clinicians use these available tools to track their clients’ changes, especially in children with CP. Schroeder and colleagues [11] provide some early evidence for benefits in the participation domain (using the COPM) however it was a small sample and a time series control, therefore this needs replication in an RCT. Children with CP who participated in an aerobic exercise fitness program showed improvements in physiological measures of aerobic activity [27], but this does not appear to generalise into daily functional activity [28] or in their level of participation or quality of life [29]. Children’s fitness levels and strengthening would seem to be important targets for an RAGT experience, but how this will impact on their activity and participation levels should be a focus of future research. There is a general paucity of research assessing RAGT outcomes on children’s involvement in the “personal factors”, “participation” and “environmental” domains of the ICF. The translation of gait outcomes into children’s meaningful life roles and environments following RAGT is critical information for clinicians to know, as this can shape the goals developed with families and children.

Finally it is of concern that only one study monitored adverse events. Whilst the majority of those events reported were minor (skin redness etc.) the rate was high at 42%. Further research is needed to ascertain rates and to investigate preventive measures for the 5% reported to have adverse events sufficient to cause them to cease the RAGT session.