Comparison of a robotic-assisted gait training program with a program of functional gait training for children with cerebral palsy: design and methods of a two group randomized controlled cross-over trial

The maintenance and enhancement of walking abilities is an important focus of rehabilitation for children with cerebral palsy (CP) in order to promote the physiological, functional and social benefits associated with ambulation (Stuberg 1992; Eisenberg et al. 2009; McKeever et al. 2013). CP is the most common cause of childhood physical disability (Oskoui et al. 2013) involving heterogeneous motor impairments caused by damage to the fetal or developing brain. Approximately 50% of children with CP require only orthoses or minimal assistive devices for independent mobility (i.e., Levels I and II of the Gross Motor Function Classification System [GMFCS]), while those in GMFCS Levels III and IV need extensive bracing and walkers or wheelchairs to move short and/or long distances (Palisano et al. 2008). Moreover, while the neurological damage is non-progressive in CP, walking skills tend to degrade with growth and age for those in GMFCS levels III and IV (Hanna et al. 2009) leading to increased reliance on non-ambulatory mobility options for efficiency and ease (Bottos and Gericke 2003).

Although wheelchair use does not necessarily lead to physiological or anatomical deterioration (Bottos et al. 2001), the benefits of standing and walking may include enhanced cardiovascular fitness (Park et al. 2001; Bjornson et al. 2007), functional gains (Strifling et al. 2008; Eisenberg et al. 2009) and greater involvement in social roles (Lepage et al. 1998). Furthermore, decreased standing time reinforces sedentary behaviours in children with CP (Verschuren et al. 2014), potentially increasing risk of comorbidities and premature mortality as seen in adults (Peterson et al. 2013). Beyond the health benefits, many families and clinicians emphasize walking because of the dominant societal beliefs that walking holds symbolic value associated with ‘normalcy’ and reduction of the social stigma of disability (Gibson et al. 2012). Combined with the evidence of positive health outcomes, these strong values result in extensive use and development of physiotherapy (PT), orthopaedic and medical interventions that focus on walking (Novak et al. 2013). However, the recent emergence of technologically-based walking interventions has been criticized because of the increased focus on ‘normalcy’ (Phelan et al. 2014), highlighting that persistent efforts towards walking may limit time for other childhood activities and not increase participation (Wiart 2011). In light of these pros and cons, it is important that rehabilitation practitioners seek better understanding of the impact of gait therapies and families’ values related to walking, especially as compelling high technology options such as RAGT become more available (Phelan et al. 2014).

Over the last decade, therapy emphasis for children with CP has shifted from being largely impairment-based (i.e., strength training and range of motion work) to focusing on activity and participation, with incorporation of motor learning principles (Park and Kim 2013). These principles include high repetition and active participation of the learner to promote neuroplasticity and skill acquisition (Levac et al. 2009). In the context of walking-focused therapies, technology can capitalize on motor learning principles (Schindl et al. 2000). For example, partial body-weight supported treadmill training (PBWSTT) systems extend the opportunity for repetitive gait training to children who have lower tolerance for independent ambulation (GMFCS Levels III and IV) (Palisano et al. 2008), supporting a focus on quality of gait and walking experience (motor learning aspects), as well as on walking endurance. Although PBWSTT results in improvements of time-distance aspects of gait and gait-related functional abilities for children in GMFCS II–IV (Willoughby et al. 2009; Zwicker and Mayson 2010; Bar-Haim et al. 2010; Johnston et al. 2011; Chrysagis et al. 2012), the extensive manual assistance required for those in GMFCS III–IV often makes it difficult to use. Functional electrical stimulation (FES) devices can be used to stimulate muscle activation, however these devices typically target one muscle to support movement of one joint (Pool et al. 2014). Many children with CP require multi-joint assistance for repetitive gait training. Robotic-assisted gait training (RAGT) devices, such as the commercially available Lokomat^®Pro (Hocoma AG, Switzerland, www.hocoma.com), were designed to address these physical limitations through use of robotic leg orthoses to guide leg movement, and have been reported to be at least as efficacious as manually assisted PBWSTT (Tefertiller et al. 2011).

There is strong evidence of improved functional outcomes following RAGT in adults with spinal cord injury and stroke, including increased gait speed and endurance in the 6-minute walk test (Tefertiller et al. 2011), but gains are not always superior to traditional PT outcomes (Dobkin and Duncan 2012). In children and youth with CP, knowledge about the impact of RAGT is limited (Castelli 2011). One-group studies have demonstrated an association between RAGT training and moderate to large improvements in gross motor skills, gait velocity and endurance (Meyer-Heim et al. 2007, 2009; Koenig et al. 2008; Brütsch et al. 2010; Drużbicki et al. 2010; Borggraefe et al. 2010a, b; Pajaro-Blazquez et al. 2013; Schroeder et al. 2014b) as well as improvements in performance and satisfaction of activities of daily life/participation (Schroeder et al. 2014a, b). A single published randomized controlled trial (RCT) comparing Lokomat training to a PT-only intervention showed no difference in gait kinematics, yet gross motor function and walking endurance were not evaluated and conclusions are further limited by a greater than 50% drop-out rate in the control group (Druzbicki et al. 2013). At the time of designing this study and receiving funding, only one RCT comparing Lokomat training to a waitlist control group was active (clinicaltrials.gov). A summary of the paediatric studies and key results is provided in Additional file 1.

Stronger evidence is required to provide the necessary understanding of the role and potential of RAGT as a clinical therapy for children with CP (Zwicker and Mayson 2010; Druzbicki et al. 2013; Aurich Schuler et al. 2015). To address this knowledge gap, we developed an RCT with a cross-over design to compare a program of RAGT using the Lokomat to a gait-related PT program in children with CP in GMFCS Levels II and III. These children are expected to have primary goals that are walking-based, and their existing ability to ambulate permits use of gait-related therapeutic content in each intervention arm. The chosen outcome measures extend beyond the GMFM and assessment of gait, and incorporate individualized goals as well as activity and participation outcomes. By controlling for concurrent PT, co-intervention concerns arising from previous studies of RAGT are reduced. Further, an acknowledged limitation of RAGT is that participants can passively move through a standardized gait cycle (Koenig et al. 2011) and if sessions are not designed to be interactive, the benefits may be inferior to an active PT program. Thus both Lokomat and PT intervention protocols used in this study are grounded in motor learning theory, following principles to optimize learning and promote active engagement of the participant by progressively increasing the challenge, introducing variability, promoting high repetition, and practicing activities that are meaningful to the child by specifically targeting the child’s goals (Levac et al. 2009). Intervention protocols are fully operationalized and will be reported in detail (a critical aspect of external validity) to provide clinically applicable information not included within Lokomat study publications thus far.

The primary aim of the study is to determine the impact on gait-related gross motor skills of a gait training program using a robotic-assisted gait trainer (the Lokomat^®Pro) compared with a gait-related physiotherapy (PT) program for children in GMFCS Levels II and III. The primary hypothesis is that children receiving the Lokomat intervention will improve more on the GMFM-66 and 6-minute walk test (6MWT) than those in the gait-related PT program.

The secondary objective is to evaluate the comparative impact of Lokomat training and gait-related PT on: (a) individualized walking/gross motor based goals; (b) the amount/location of walking the child does each day (environmental context and season considered); (c) participation in activities and (d) health-related quality of life.

The study aims to enrol 40 children with CP (ages 5–12 years) as follows: (a) 20 children in GMFCS Level II who have walking limitations but ambulate without a gait device or crutches or canes for most distances; and (b) 20 children in GMFCS Level III who have a greater extent of walking limitation and use walkers or wheelchairs for short to long distances (Russell et al. 2000).

This study uses a cross-over design to compare the two different therapy intervention programs: robotic assisted gait training (Lokomat) versus gait-focused PT (see Fig. 1). The order of treatment (Lokomat or PT first) is designated by random assignment. Children cross-over to the second intervention arm following a six-week non-treatment break between interventions. The second intervention arm begins after a second baseline assessment completed at the end of the break to re-establish the child’s abilities. The use of a cross-over design helps to reduce the impact of confounding variables outside of the control of the study itself, since each cross-over participant acts as his/her own control (i.e., reduces between subject variation) (Senn 2002). Cross-over designs are also more efficient than standard RCTs or repeated measure designs, requiring fewer participants (Louis et al. 1984). While there may be concern that the extended length of involvement in the trial to complete both phases may be an obstacle for enrolment or adherence, cross-over designs have been successfully employed in CP and Lokomat research (McNee et al. 2007; Mayr et al. 2007). Furthermore, the opportunity to receive both intervention arms and be assured access to the Lokomat is thought to outweigh any perceived burden of the time commitment (Law et al. 1997).

Each intervention is given twice weekly for a total of 16 sessions over a targeted treatment period of 8 weeks. This training frequency mirrors a standard clinical PT treatment block. A maximum of 10 weeks is targeted to keep the dose as close to twice a week as possible while taking into account the occasional need to miss a session due to illness, vacation, and other unavoidable interruptions. The total number of sessions is similar to trials in which Lokomat effectiveness was demonstrated (Koenig et al. 2008; Borggraefe et al. 2010a, b; Meyer-Heim and van Hedel 2013; Schroeder et al. 2014a, b) while a chief difference is the total duration of the treatment period, in that other Lokomat trials have used a short intensive burst of therapy (e.g., 15 treatments over three weeks). Our concern from a motor learning and assimilation standpoint is that a short burst may be insufficient time for the child to develop and integrate new skills gained through training into their daily life.

Children are required to discontinue regular PT interventions prior to the first baseline assessment and to abstain from starting any other PT during the trial to reduce the chance of confounding effects associated with any previous or concurrent treatment. The child is allowed to continue any existing program of soft tissue stretches and/or basic walking and standing, but their clinical physiotherapist (PT), or other clinician, is asked not to alter the programs during the course of the study. Activities are monitored at the start of each intervention session (Lokomat or PT), when the child and parent are asked by the study PT about any other gross motor-based treatments and physical activity they did since the last session. These physical activity descriptions are documented in the child’s session log. The study PT is required to alert the research assistant (RA) to any potential co-interventions. Subsequently, the RA calls the parent to ask them to put any activities on hold that are judged to be walking-related co-interventions.

During the six-week break that occurs before crossing to the other intervention arm, the child is allowed to continue their existing regimen of soft tissue stretches and any basic walking/standing home program they had been given prior to the trial, but asked not to embark on any other new therapy. The RA phones the parent mid-way through this phase to check for any treatments that the child may have started/resumed. A request to discontinue any added treatment may be made by the RA after discussion with the co-Principal Investigators (co-PIs) if it appears to be a gait-related co-intervention (i.e., a possible confounder). While the six weeks between the two intervention arms is not a true wash-out period (since effects of rehabilitation are not expected to be reversible when therapy is withheld), it does give a time break for families between the two phases. A new baseline is established before starting the second intervention to ensure intervention effects are isolated for analysis (Mayr et al. 2007).

In both the PT and Lokomat intervention arms, the study PTs have the opportunity to look at the baseline assessment data acquired prior to the first Lokomat or PT session to ensure sufficient clinical understanding of the child to be able to make suitable treatment and goal decisions. Within each intervention arm, a study treatment log of each session is kept for every participant. Each session is fully documented by the study PT on log sheets that require detailed and open disclosure of all activities performed. This log is available to the study PT at each session to permit reference back to previous sessions and support progression. The logs are checked at bi-weekly intervals by the study RA so that any fidelity issues can be addressed in subsequent sessions. Following completion of the intervention arm, completed forms are removed and replaced by a new set of treatment logs. The study RA will also monitor session attendance to problem solve any issues promptly.

The information recorded in these session logs will allow us to fully operationalize all of the aspects of the treatment approaches at the end of the study (an important external validity consideration). The within-session documentation process is also expected to optimize adherence of the study PT to the intervention protocols (treatment fidelity). Furthermore, two of the child’s Lokomat and PT sessions are filmed to permit evaluation at the end of the study of the extent and type of motor learning strategy use within the interventions, using the Motor Learning Strategy Rating Instrument (MLSRI) developed by Levac et al. (2009), and validated by Kamath et al. (2012) at our centre.

The primary outcome measures selected directly reflect those used in previous Lokomat studies (i.e., GMFM-66, 6MWT, Timed Up and Go and Canadian Occupational Performance Measure [COPM]) (Russell et al. 2000; Russell et al. 2013; Thompson et al. 2008; Williams et al. 2005; Law et al. 1990), and expand into other ICF areas including movement quality (Quality FM, basic gait assessment of spatiotemporal variables via electronic walkway and observational assessment) (Wright et al. 2008, 2014; Sorsdahl et al. 2008), ROM and spasticity (Tardieu) (Scholtes et al. 2006), advanced gross motor skills (Challenge Assessment) (Wright et al. 2012; Glazebrook and Wright 2014), functional abilities (i.e., PEDI Caregiver Assistance and ASK-30) (Haley et al. 1992; Young et al. 2000), and participation/QOL (i.e., step activity monitor, CAPE and KIDSCREEN) (King et al. 2007; Ravens-Sieberer et al. 2007; Bjornson et al. 2014). Collectively, these outcome measures provide comprehensive information in areas of activity and participation in alignment with the WHO ICF framework (World Health Organization 2001). Questionnaires are completed by the parent as well as the child if the child is 8 years or older. Each measure is used at each of the four assessment sessions. Personal motivation characteristics of the child (Dimensions of Mastery Questionnaire [DMQ]) (Miller et al. 2014) are assessed at baseline.

The COPM is completed by the child and parent with the assessor at the end of the assessment session, focusing on activity and participation goals. Since the child is not randomized to an intervention group until after the baseline assessment, these are goals that fit a gait-related functional mobility program generally and are not specific to the Lokomat or PT arms. This it is anticipated that these will be higher-level activity and participation related goals that are suitable to either intervention. Goal attainment scaling (GAS) (King et al. 2000) is done during the first two intervention sessions by the study PT with the child/parent. Working from the COPM goals that were created at baseline by the PT assessor, the study PTs re-script these into measurable GAS goals that fit with the intervention (PT or Lokomat) and what they learned about the child from review of their assessment results. This process allows the GAS aspect to be tied in directly to the perceived possibilities of the intervention arm, and fits with the specific outcome scaling process that is required within GAS, i.e., best suited to highly observable goal areas related to body structure and activity. Collectively it is hoped that the combined use of the COPM and GAS will provide a comprehensive evaluation of the extent and nature of goal accomplishments (GAS), as well as perceptions of performance abilities and satisfaction with abilities (COPM).

The study has been designed and powered to ensure adequate sample size for the primary research questions and analysis of children in GMFCS Levels II and III. The starting point for our sample size calculation consideration was a single group Lokomat study with a similar GMFCS sample and Lokomat use protocol in which the GMFM Stand and Walk mean scores changed by approximately 5 points (SD change = 7) (Borggraefe et al. 2010b). Our study is powered to let us detect a difference between the two interventions of three points (SD = 6: small effect size) on the GMFM-66, a magnitude of change that is considered to be a minimally clinically important difference (Oeffinger et al. 2008) when evaluating benefits of treatment. The reason for a difference change target in our study is that we are comparing two active interventions, each of which could result in positive change given the walking focus and boost for many of the children in therapy intensity. This sample size is also sufficient to handle the 6MWT comparisons and the small effect size observed by others in single group Lokomat studies in the 6MWT results (mean gains of 25 m [SD = 50 m], personal communication, Glenrose Hospital program, 2012).

For crossover designs, sample size calculations that link with a matched pairs t test calculation are appropriate (Senn 2002). A paired t test is also the underlying basis for the more sensitive repeated measures ANOVA that we have chosen to use so that we can include a period effect within the analysis. With an alpha of 0.025, beta of 0.20, and a detectable change score difference between intervention of thre points on the GMFM-66 (with an estimated SD of change of 6.0 points), a sample of 32 children is required (Hintze 2001). We increased the sample requirement to 40 to account for a potential drop out/protocol deviation rate of 20%. Sample calculation based on this generous drop-out rate will support the primary efficacy analysis that focuses only on the children who attend more than 70% of the sessions within one or the other intervention phase.

Data entry forms were built for each measure in the study to correspond to the paper forms that the assessors used. REDCap (Harris 2012) requires outer limits for each value, and each paper data sheet is double entered for purposes of data verification process. Collectively these features optimize the data entry accuracy. Study data will be kept for seven years following conclusion of the trial, at which point data will be destroyed following REB requirements.

Descriptive statistics and graphic displays will be presented for all outcomes for the PT and Lokomat interventions (pre, post-test and change scores). Equivalence between baseline_A and baseline_B scores will be evaluated for each outcome; a t-test or Wilcoxon test will be used in this comparison as appropriate. The primary analysis will be for the pre- to post-test differences (change scores) of the GMFM-66 and 6MWT between groups. Assuming normality of data, a repeated measures ANCOVA will be performed on GMFM-66 and 6MWT change scores to compare the effects of the Lokomat and gait-related PT interventions. The child’s relevant baseline score (baseline_A/baseline_B) will be used in the ANCOVA to increase analytic precision. The P value will be adjusted to 0.025 to handle the expected high correlation between the GMFM-66 and timed walk. Evaluation of an intervention order effect (i.e., being the first or second treatment), as well as the interactions with intervention group will be built into this analysis (Senn 2002). Secondary sub-analyses will be done in the same way within GMFCS II and III subgroups to determine any differential response for GMFCS Level.

Due to the lack of expectation of return to baseline in the break period, a second analytic approach will be taken as well using a generalized estimating equation (GEE) adjustment for repeated measures within subjects analysis (Baker et al. 2007). In this approach treatment effects are evaluated via regression analysis and the GEE method will allow the fit of two models. The first is for treatment effect, treatment-period effect, and differential carryover effect adjusting for the baseline scores, and the second is treatment effect and treatment-period effect with exclusion of any differential carryover effect. Factors hypothesized a priori to be associated with GMFM-66 change score magnitude will be entered within this model using a forward stepwise regression procedure, e.g., age, GMFCS Level, motivation (DMQ) score, ASK baseline score.

Finally, a third approach will be taken if there is significant evidence of a carryover effect and significant interaction effects. This will be to focus on the data from the first arm of the study, i.e., an ANCOVA adjusted for baseline GMFM-66 change scores of just the first intervention period (randomisation to either Lokomat or PT with 20 children per group). This comparison in itself would have a power of about 0.60 for the GMFM-66 (with alpha = 0.025 and a medium effect size of 0.50).

A similar set of analytic approaches will be used for the secondary outcome measures (GMFM Stand and Walk, Timed Up and Go, walk velocity/step and stride length [from GAITRite], observational gait scale, average daily step counts [step activity monitor data], Quality FM, Challenge Assessment, ASK-30, PEDI Part II, KIDSCREEN, COPM and GAS scores), using the baseline value of the measure of interest as a covariate within a MANCOVA procedure that will permit simultaneously handling of multiple secondary outcome measure change score evaluations. If both parent and child filled out the questionnaires (i.e., all children 8 years of age and up who were able to do so), the children’s data will be analysed as primary. Otherwise parent questionnaire data will be considered as primary. A comparison of child and parent questionnaire results will be done as a secondary analysis via an ANOVA in which both single point in time (baseline) and change scores will be evaluated. ROM and Tardieu scores will be summarized in a simpler manner, using descriptive statistics and graphical presentations to illustrate scores in each phase. The power of all primary and secondary ‘no difference’ analyses will be determined post hoc.

The primary analysis will include all children who attended ≥70% of sessions for each intervention, and those who discontinued one or both interventions for a reason directly related to the study (e.g., excessive fatigue associated with participation or issues related to dislike of one or both interventions), or dropped out of the trial due to an intervention-associated injury. In the case of dropout due to study fatigue or dislike of the intervention, a strong effort will be made to do a discharge assessment at the time of withdrawal. In the case of injury, it may be possible to follow the child and evaluate some of the outcomes when the child is able (post-minor injury). If the physical assessment is not possible due to compromised physical status, completion of the functional and participation measures may still be appropriate.

A second analysis will include all who adhered to ≥50% of each intervention as well as those who discontinued one or both interventions for reasons noted in the primary efficacy analysis. The third analysis will be an “intent to treat”, including the last available data from all participants regardless of level of adherence or reason for withdrawal. As outlined above, a strong effort will be made to do a discharge assessment at time of withdrawal. Demographic characteristics, baseline scores and change scores of these three adherence groups will be evaluated to determine any systematic differences in their characteristics or outcomes.

Data from the session summary sheets will be compiled. Treatment categories, Lokomat parameter adjustments, and session activities will be summarized through graphic and tabular summaries, and related descriptive information from therapists’ comments will be reported.

At the start, midpoint and end of each session heart rate is taken via pulse, and children are asked about any pain/musculoskeletal issues using a visual analogue scale with pain ratings done as per the FACES^® scale (Wong and Baker 1988) and a generic body diagram. All areas of the child’s skin that underlie the Lokomat straps and cuffs are checked by the study PT before and after every Lokomat session, with any areas of redness or skin breakdown marked on the body diagram. The expectation is that any areas of redness that occur during Lokomat therapy are transient and fully resolved by the next session. Failure to resolve between sessions requires adjustment of the straps or padding, realignment of the Lokomat set-up, use of skin protection tape, or withholding the next session as appropriate. At the end of each session, the study PT enters all AE information noted at that session on an AE summary sheet and alerts the study RA to review the sheet. The RA and PI classify the event in terms of severity and attribution to the Lokomat or PT session, and determine necessary follow-up action according to a pre-set action protocol approved by the REB. This process ensures that a prompt response, precautions and reporting to the REB can be taken as required. All occurrences of open wounds or fracture are reported by the co-PIs within 24 h of the session to the REB with treatment offered through Holland Bloorview. AEs are monitored by a Safety Monitoring Committee (SMC) acting independently of the investigators and study funders. This SMC consists of a pediatrician, methodologist/researcher, nurse, pediatric orthopaedic surgeon, PT, and a lay person. The SMC meets every four months to review any AE forms that have been filed since time of the previous meeting, or convenes a special interim meeting to review any event that required reporting to the REB. Early stopping of the trial will occur if a recommendation to terminate the study is made by the SMC by majority vote. This may occur at any time, with decisions reported to the PI and REB by the SMC Chair immediately by telephone and email along with the reason for their decision. Withdrawals for reasons other than safety concerns will not be cause for study termination.