This RCT showed after 20-session, PAAR, SCAR, and CT had significant within-group functional improvements in gait independency (FAC), balance (BBS), and walking speed (10MWT). Between-group comparison suggested robot-assisted training (PAAR and SCAR) could result in significantly greater improvement in functional independency than conventional training in usual care (CT). In particular, powered assistance in PAAR that actively moved the paretic ankle to facilitate subjects were able to walk faster with higher cadence in the 20-session robot-assisted training when compared with the ankle-locking swing-controlled robot in SCAR.
Our previous RCT on chronic stroke (n = 19) had compared PAAR with SCAR in similar experiment setting, which showed robot-assisted trainings were effective in chronic stroke, with FAC improved + 0.6 and walking speed + 0.07 m/s after 20-session training [18]. In the current study for sub-acute stroke, both PAAR and SCAR had + 1.4 improvement in FAC, with more than 56% of subjects turned from dependent walker (FAC < 4) at baseline to become independent walker (FAC ≥ 4) after intervention; while CT only had 29%. For walking speed, PAAR in the current study had + 0.32 m/s improvement, the greater proportion of sub-acute stroke subject walked faster than the minimal clinically important difference (MCID = 0.16 m/s), in PAAR (71.4%) vs CT (41.2%) (χ(1) = 5.290, p = 0.021) was in line with their improved gait independency [28]. These results agreed with several systematic reviews that supplementing conventional physiotherapy with electromechanical-assisted gait training in sub-acute stage would have greater functional improvement than chronic stage [4, 5, 19].
Rehabilitation robotics are capable of delivering intensive, repetitive and adjustable gait assistance patterns while sharing workload of therapists [21, 32]. Existing clinical application of high-intensity task-specific gait training often performed on treadmill or level ground [10, 33, 34]. The current study demonstrated that wearable robot-assisted training could even be implemented in simple stair environment as a feasible rehabilitation approach. Previous studies showed that mild stair training in chronic stroke could improve physical activity level [11], trunk stability and balance [10, 12], walking speed and endurance [10, 35]. The robot-assisted stair training described in the current study required only one skilled trainer walking alongside the stroke subjects for safety and verbal cueing, while the posture adjustment for foot drop correction could be handled automatically by the robot itself [32].
This study was one of the first clinical trials that applied robot assistance in stair training for sub-acute stroke survivors. Existing lower-limb rehabilitation robots were often limited by their device weight and portability for stair environment, so few of these devices could be evaluated and developed to the stage of commercialization and clinical application [22]. G-EO system was a commercialized end-effector robot that could simulate stair climbing in a treadmill-like environment by moving foot plates in cycle to reproduce step length and height of stairs, but the system was bulky and stationary. The RCT evaluated the stair version of G-EO system focused on balance training of chronic stroke subjects [36]. Portable-power ankle-foot orthosis (PPAFO) developed in the University of Illinois used a pneumatic bidirectional rotatory actuator to provide untethered ankle assistance on level ground and stairs. The robot and control algorithm were evaluated on healthy subjects (n = 5) as a technical feasibility test [37]. Recent development of ReStore exo-suit (ReWalk Robotics, USA) featured a soft garment-like design driven by Bowden cable, could offer potential solution to reduce device weight and bulkiness of robot at ankle joint [24]. However, few studies investigated how impaired subjects would response to these rehabilitation robots immediately during walking on stairs, and only few studies reported the therapeutic effects of these devices in multi-center RCT setting [21–23]. Our results demonstrated feasibility of intensive stair training using ankle robotics for stroke rehabilitation. More similar researches should be done in the future to confirm the value of intensive stair training in clinical application.
Effect sizes were computed for the three outcome measures (FAC, BBS, and 10MWT) to determine the strength of association for the statistically significant interactions (Table 3). Between-group comparison of FAC revealed a medium effect size between robot-assisted training and conventional training (PAAR vs CT 0.671, SCAR vs CT 0.610). This suggested a larger sample size would have possibly produced more statistically significant effect. The effect size difference between PAAR and SCAR were small in FAC (0.010), but the 4-week intervention showed a large effect size difference in 10MWT (0.752) and medium effect size difference in BBS (0.567). Hence, PAAR might be more favorable than SCAR toward functional improvement in walking speed and balance.
Comparison between PAAR and SCAR revealed an interesting finding about the effect of active powered assistance. During the 20-session robot-assisted training, PAAR could walk faster speed and higher cadence than SCAR when wearing the robot (Fig. 3), which implies PAAR that offered more active assistance to facilitate the ankle joint in dorsiflexion might be superior than SCAR that provided passive support to dropped foot for better foot clearance. These enhanced gait stabilities and walking speed in PAAR could be maintained even after removing the robotic assistance, as supported by the therapeutic effects in the Post clinical assessment scores. Hence, active robotic assistance might play an important role in the gait relearning; passive support, as in the SCAR device, offered relatively limited persistent gait improvement in term of walking speed. Results of another RCT suggested 26-week provision of passive AFO (similar to SCAR) did not have any effects on kinematic gait parameters of sub-acute stroke subjects (n = 26) [16]. More clinical trials and follow-up studies are required to generalize these results.
There were limitations in the current study. First, the sample size was based on the power analysis to go through the ethics committee in hospitals for this novel intervention RCT study, and it was relatively a small trial. Second, comprehensive gait analysis was not available due to the safety concern when the sub-acute subjects need to transfer from the hospital to the gait analysis laboratory.