Locomotor training using an overground robotic exoskeleton in long-term manual wheelchair users with a chronic spinal cord injury living in the community: Lessons learned from a feasibility study in terms of recruitment, attendance, learnability, performance and safety

Design

A single-group longitudinal prospective feasibility study.

Participants

Robotic exoskeleton

The wearable robotic exoskeleton EKSO™ (version 1.1) (Ekso Bionics, Richmond, CA, USA) is a ready-to-wear, battery-powered, motor driven, robotic pair of legs generating motion at the hip and knee joints in a properly sequenced manner. Each joint is independently controlled by different sensors linked to a small, portable, computerized control system attached to the flexible trunk module that also encompass the battery. Information gathered by over 35 different sensors (e.g., accelerometers, speed controllers, gyroscopes, pressure sensors) feed a decisional algorithm loop allowing manual wheelchair users with SCI to perform sit-stand transfers and walk. When walking with the robotic exoskeleton, each step is primarily commanded by combined forward and lateral bodyweight shifts toward the weight-bearing lower extremity before initiating the oscillation with the opposite lower extremity. The certified therapist can control numerous walking features (e.g., speed, step height, step length). The EKSO GT robotic exoskeleton weighs about 28 kg and can technically reach a maximal walking speed of 1.6 m/s. The EKSO GT robotic exoskeleton is approved by Health Canada for clinical use.

Locomotor training program

Initially, participants attended two familiarization sessions over a one-week period that lasted about 45–60 min per session. During these familiarization sessions, participants were properly fitted with the EKSO GT robotic exoskeleton before performing balance, walking-related tasks (e.g., sit-stand transfers), and walking on short distances with visual and verbal feedbacks while the certified therapist actuated each step (i.e., FirstStep mode). As the participants’ level of proficiency increased during the familiarization sessions, they learned to safely ambulate with the exoskelon at a self-selected comfortable speed using their own walking aid (i.e., rolling walker or forearm crutches) while taking control of each step triggered via anterolateral body weight shifts (i.e., ProStep mode) under the direct supervision of a certified therapist. Following these two familiarisation sessions, participants began the six-week progressive locomotor training program administered by a certified therapist that encompassed a total of 18 training sessions (three sessions/week; 60 min/session). Depending on the level of each participant’s proficiency, on the participant’s tolerance, and on the activities planned for the session (e.g., instructions and basic training to initiate sit-stand transfers, walking and turning with forearm crutches), the workload was periodically adjusted using walking distance, duration, and speed parameter progressions [7]. After each session, all training parameters and other relevant information (e.g. total standing time, total walking time, and total number of steps) were recorded.

Main outcome measures

Statistical analysis

Descriptive statistics (i.e., mean, standard deviation) were calculated for all demographics and clinical characteristics as well as for all outcome measures. After a Shapiro-Wilk test confirmed the normality of the walking speed measures, the pre- and post-training walking speed measures were compared using a paired Student t test for repeated measures with the significance level set at p ≤ 0.05. These statistics were computed using SPSS statistic software version 17.0 (IBM Corporation, Armonk, New York).

Recruitment

A summary of the recruitment process, along with the number of participants who successfully completed each stage of this process and reasons for excluding potential participants, are illustrated in Fig. 1. A total of 49 individuals with a SCI contacted the research team by phone or via email to express their interest in participating in the research project during an overall 11-month enrollment period split into two phases: October to December 2015 (3 months) and May to December 2016 (8 months). In reality, the overall enrollment period per se was shorter (about 8 months) since the last participants needed to be recruited no later than November 1, 2015 and 2016, respectively. Upon completion of the pre-screening interview of potential participants over the phone (N = 49), 19 individuals with a SCI were excluded for different reasons based on the answers provided to specific screening questions during this initial step. Out of the 30 potential participants who underwent clinical pre-screening, 11 individuals with SCI were excluded for different reasons during this second step. Finally, out of the last 19 potential participants who underwent clinical screening and completed the familiarization sessions, five individuals with SCI refused to participate in the study during this last recruitment step. Hence, a total of 14 individuals with SCI were enrolled in the study (recruitment rate = 28.6%). The most common reasons for excluding potential participants were the presence of limited passive dorsiflexion range of motion at the ankle (n = 13/30; rate = 43.3%) as well as time, transportation, or accommodation constraints linked to the program requirements (N = 9/30; rate = 30%). The most common reason for refusing to participate was the fear of developing a complication as a result of ambulating with the robotic exoskeleton system after having tried the robotic exoskeleton (n = 3/5; rate = 60%).

Attendance

Most participants (N = 11/14) completed all training sessions (attendance rate = 100%) whereas two participants were deprived of one (adjusted attendance rate = 94%) and four training sessions (adjusted attendance rate = 78%), respectively, since the program was temporarily suspended during the holiday season. Hence, the overall attendance rate was 97.9% (229 completed training sessions/234 planned training sessions). One participant was withdrawn from the study by the research team after one training session and was not accounted for in the attendance rate and learnability statistics (details provided in Adverse Events section). This was the only participant who dropped out of the study (n = 1/14; drop-out rate = 7.1%).

Learnability and performance

A summary of the progression of the standing time, walking time, and number steps taken per session is illustrated in Fig. 2. On average, during the locomotor training program, the standing time, the walking time, and the number of steps taken per session were 49.7 ± 12.7 min, 33.4 ± 12.5 min, and 1190 ± 561 steps, respectively. Between the start (mean of sessions 1 and 2) and the end (mean of sessions 17 and 18) of the locomotor training program, the standing time, walking time, and number of steps taken per session progressed by 45.3%, 102.1%, and 248.7%. The majority of participants (N = 10/13) were already self-initiating their steps via lateral and anterior bodyweight shifts toward the weight bearing lower extremity (i.e., prostep mode) at the first session of the training program. Two additional participants reached this level at the second training session whereas another participant reached it at the 8th session of the training program. Most participants (N = 10/13) developed the ability to ambulate with Canadian crutches after 3.5 ± 3.3 training sessions whereas the other 3 participants continued to use a rollator walker throughout the training program (Fig. 3). Most participants (N = 11/13) needed moderate assistance provided by one physiotherapist at the start of the training sessions whereas the two participants needed minimal and maximal assistance, respectively (Fig. 3). Upon completion of the training sessions, one participant needed moderate, four participants needed minimal, and four needed contact guard assistance provided by one physiotherapist, whereas four participants needed a physiotherapist to stand-by while walking (Fig. 3). All participants needed moderate to maximal assistance provided by one physiotherapist for all sit-stand transitions throughout the training sessions. As for the walking speed, it increased significantly (p ≤ 0.0001; + 66.8%) between the start (mean ± 1 SD = 0.15 ± 0.02 m/s) and end (mean ± 1 SD = 0.25 ± 0.05 m/s) of the training program. These last results do not include the data of one participant with C6 tetraplegia as he only took a limited number of steps at a very slow pace and needed maximal assistance of the certified therapist during the first week (i.e., invalid result for the 10MWT).

Adverse events

As mentioned previously, one participant was diagnosed with bilateral type I non-displaced fractures of the calcaneus after having completed the two familiarization sessions and the first training session. Uncertainties exist about the specific cause of the fractures. This participant was withdrawn from the study. Four participants reported exacerbation of pre-existing shoulder pain, stiffness, or discomfort whereas one participant developed soreness at the thumb over the course of the locomotor training program. This finding was unexpected as no previous study reported on upper extremity pain, stiffness, or discomfort. One participant failed to report a previously complete anterior cruciate ligament tear to the research team and developed severe knee hyperextension at the 10th training session that was solved by blocking knee extension to − 3° thereafter. Six participants experienced orthostatic hypotension with systolic blood pressure drops of ≥ 20 mmHg during a training session. No participant developed any soft tissue or skin problem nor fell during the locomotor training program. For the three certified trainers involved in the intervention, no adverse effect was documented. Last, a battery failure and a hip joint bearing assembly failure were directly linked to the robotic exoskeleton itself over the course of the locomotor training program.

Recruitment

The recruitment rate reached in this preliminary study (28.6%) was acceptable although it remains relatively low considering the number of potential participants (N = 49) who initially expressed their interest in participating into the proposed study and completed the different steps of the recruitment process. Moreover, considering that the study was conducted in a rehabilitation center hosting an ultra-specialized SCI rehabilitation program servicing the western part of the Province of Quebec, the fact that numerous strategies were implemented to overcome potential barriers (e.g., a dedicated research professional in charge of the recruitment, multiple recruitment strategies implemented, telephone pre-screening interview to minimize the number of visits, free parking, free training sessions), and that all participants were allocated to the locomotor training program with the robotic exoskeleton, a higher recruitment rate was anticipated (i.e., ≥50%). Nonetheless, this recruitment rate is 1.7 times greater than the one reported in other feasibility studies investigating locomotor training programs with a robotic exoskeleton in individuals with complete or incomplete SCI in England (17%) [15] and Germany (12%) [18]. Only one recent multi-center study investigating a single training session with a self-stabilizing robotic exoskeleton in individuals with SCI has reached a recruitment rate near 50% (i.e., 20 participants recruited among 46 screened for eligibility) [17]. Nonetheless, the recruitment rate of the present study compares relatively well to the rates reached in other feasibility studies investigating various task-specific gait-training programs offered to relatively homogeneous samples of individuals with neurological impairments (e.g., stroke = 6.7% [20], Parkinson disease = 11% [21]). In the present study, the most important reason (16 out of 35 potential participants = 46%) for not qualifying for the training program was due to musculoskeletal impairments with the leading cause being a reduced passive range of motion at the ankle, knee, or hip. For the same reason, other preliminary studies have excluded up to 77.8% of potential participants (7 out of 9 potential participants excluded) [16]. The second most important reason was linked to time constraints (7 out of 35 potential participants = 20%). Contrary to other preliminary studies [16, 22], transportation did not emerge as a major barrier to the recruitment process nor the dropout rate in the present study since only two potential participants based their decision on this criterion (2 out of 35 potential participants = 5.7%). Taking these reasons together, developing a home-based pre-training program with indirect supervision of a therapist that would target gains in passive range of motion at the lower extremity and progressive standing time prior to initiating the locomotor training program may be warranted.

Different strategies may need consideration to optimize recruitment rate and facilitate attendance in future clinical trials [e.g., offering training sessions during the evening and weekend; offering training sessions away from the main rehabilitation center affiliated with the project (e.g., other rehabilitation centres, community-based physical activity centers, living labs in shopping malls); adjusting the training schedule to best match participants’ availability with a minimum of two training sessions per week; proposing temporary housing alternatives for potential participants living further away who demonstrate an interest in participating]. Last, it is important to highlight that the recruitment rate may have been lower if potential participants had had a chance to be allocated to an alternative experimental group undergoing a different training program or a control group with no training program.

Attendance

The attendance rate reached in the present preliminary study (97.9%) was excellent. The high attendance with respect to the scheduled training sessions confirms the commitment of the participants who engaged into the locomotor training program. This is further supported by the fact that only one participant dropped out of the program following an adverse event (i.e. calcaneus fracture). Hence, a completion rate of 92.9% (n = 13 participants/14 participants) was reached and is greater than the 50% documented in another feasibility study proposing a comparable program [15]. The importance of the familiarization sessions needs to be highlighted since 5 out of the 19 participants (26.3%) decided not to engage into the locomotor training program at that time. Although not formally documented, these familiarization sessions allowed the research team, to some extent, to further screen potential participants who were hesitant to engage into the locomotor training program and potential participants to take an informed decision about their commitment based on a lived experience. Additionally, it supports the relevance of adopting a flexible approach when scheduling the training sessions to accommodate all stakeholders, especially the participants. Since the training sessions involved no or very limited socialization with individuals with similar sensorimotor impairments and functional disabilities (i.e., individualized approach), aside from the interaction with one or two therapists, the commitment of participants to complete the training session and, to some extent, their acceptance of the new technology, especially with regard to its perceived usefulness and ease-of-use, is also established.

Learnability and performance

Individuals with a complete motor SCI demonstrated a capability to quickly learn to ambulate overground with a robotic exoskeleton. Overall, the standing and walking time (including the number of steps/session) progressed at a faster rate during the first half than during the second half of the locomotor training program. Overall, the participants stood and walked at least 30 and 20 min, respectively, at the first training session which is recommended in clinical practice to anticipate beneficial effects among long-term manual wheelchair users with a spinal cord injury [23, 24]. The level of therapist assistance also rapidly decreased over the course of the locomotor training program with most participants requiring no more than minimal assistance after the 8th and 9th training sessions (halfway into the program) and only contact guard or stand-by assistance by the end of the program when walking. Additionally, most participants walked with forearm crutches, with or without the use of the self-controller that allows the user to drive few basic functions of the robotic exoskeleton (e.g., initiation of the first step, continuous walking in ‘prostep’ mode, and stops), by the end of the program. Overall, the learnability trajectory, illustrated for the first time in the present study using measures systematically collected at each training session, compares to some extent with the ones reported using only pre- and post-intervention measures in previous study using a similar or different robotic exoskeletons [3, 16, 18, 22].

The learning process also may have been facilitated by optional distinct auditory feedbacks automatically generated when the participant respectively reached the lateral and forward body weight shift targets required prior to initiating steps, especially early on during the learning stage [25]. Moreover, although not formally assessed in the present study, some participants periodically filmed their performance, especially at the beginning of the study, to facilitate their learning and complement the therapist’s subjective feedback (i.e., visual feedback-induced performance improvement) [26]. Hence, in addition to the adjustability of some exoskeleton parameters (e.g., reducing body shift amplitudes to initiate steps, reducing step height, increasing step length), numerous clinical strategies (e.g., reducing level of human assistance, changing walking aid) are also possible to adjust the level of challenges during overground walking with the robotic exoskeleton as the participant’s level of proficiency improves. Maintaining a level of challenge during learning is also known to positively impact a participant’s level of motivation and attendance, both of which are crucial in the context of any clinical trial in which participants are assigned to receive an intervention [27].

As for the performance, the walking speed was found to increase significantly between the start and end of the training program. In fact, the mean walking speed reached in the present study (i.e., 0.25 ± 0.05 m/s) is similar to the weighted mean gait speed of 0.25 ± 0.14 m/s reported in a recent meta-analysis investigating a heterogeneous group of individuals with a complete SCI who completed, with different models of overground robotic exoskeletons, various training protocols encompassing a wide range of training sessions [2]. Nonetheless, reaching faster walking speed after 18 sessions may still be possible with additional training since able-bodied adults, who have completed basic training with the robotic exoskeleton, reach on average a self-selected comfortable walking speed of 0.38 ± 0.09 m/s when they were asked to avoid all voluntary muscular contraction of their lower extremities (i.e., passive walking) [28].

Safety

Among all participants, one serious adverse event occurred during the study. One participant sustained bilateral type I non-displaced fracture of the calcaneus after completing the two familiarization and the first training sessions. Although uncertainties exist about the specific cause of the fractures, both the fragility fracture risk of the calcaneus and the elevated vertical ground reaction force, known to reach about 36% ± 15% of the bodyweight at heel strike when walking with an overground robotic exoskeleton [29], are potential explanatory factors. This participant was withdrawn from the study and referred to the medical team until the fractures were healed. These fractures occurred even though the screening process was thoroughly completed by an experienced research physiotherapist and the minimal standing time tolerance (i.e., ≥30 min), recommended by the manufacturer of the exoskeleton, was verified. Unfortunately, another preliminary study also reported a comparable fracture of the talus during a locomotor training program with another overground robotic exoskeleton [15] whereas a review recently suggested an overall incidence rate of bone fracture of 3.4% [3]. The fact that some studies have predominantly included individuals with recent SCI (≤ 1 year), a time period during which bone mineral density declines at the L/Es and distal vertebrae (i.e., infralesional osteoporosis) may not have yet stabilized at levels significantly below those of age and gender-matched able-bodied individuals [30], may explain why this risk may have been underestimated [16]. Further investigation will be needed to explore all potential causes to implement additional screening elements for severe osteoporosis into the process (e.g., fracture risk stratification algorithms for adults with SCI [31], threshold for bone mineral density or architecture at the ankle and foot) and to develop solutions addressing the complex challenges linked to physical activities performed in standing position in individuals with SCI in the future. Other minor adverse events, predominantly linked to exacerbation of pre-existing (N = 4) or the development of new (N = 1) musculoskeletal-related non-debilitating pain at the upper extremity, were also documented (n = 5/14; 35.7%) over the course of the locomotor training program. Yet, all these participants opted to continue the training sessions while exploring personalized solutions to alleviate or even eliminate pain over the course of the program (e.g., increased number and duration of rest periods during sessions; cushioning at the handle of the walking aid; recommendation of stretching exercises post-training; use of nonsteroidal anti-inflammatory drugs) with the certified trainer(s). Unexpectedly, no exoskeleton-related skin or soft tissue issue was observed in the present study although it affected up to 50% of participants in previous studies and typically leads to interruptions of the intervention or withdrawal of participants from the studies [15]. All the above-identified risks remain impossible to eliminate, warrant thoughtful consideration, and should be carefully explained within the informed consent form along with the strategies implemented to minimize or overcome them. Finally, two events linked to the robotic exoskeleton itself (device malfunction) occurred over the course of the locomotor training program. The first event was a battery failure that required its replacement while the second event was a mechanical problem with a hip joint bearing assembly malfunction due to a damaged bolt holding the proximal and distal joint components together. In both cases, the problems were solved within a 48-h period with the prompt assistance of the company’s customer service department and had minimal impact on the conduct of the study.

Limits of the study

Limitations in the present study were the small sample size of relatively homogeneous participants recruited at a single site as well as the absence of a control group. Uncertainties about the best research design and outcome measures to adopt in future clinical trials continue. Because the study only included long-term manual wheelchair users with a chronic SCI living in the community, the generalizability of the results beyond this reference population, such as in ambulatory individuals with an incomplete SCI, requires caution. Prudence is also suggested when inferring about participants’ acceptance and satisfaction, particularly when addressing attendance and learnability, as this dimension was not reported. Hence, the findings of the present study should be considered preliminary, but it is anticipated that they will stimulate interest in conducting future larger-scale level I or II clinical trials investigating the efficacy or effectiveness of locomotor training programs with an overground robotic exoskeleton in long-term manual wheelchair users.