Impact of the robotic-assistance level on upper extremity function in stroke patients receiving adjunct robotic rehabilitation: sub-analysis of a randomized clinical trial

In terms of the general background of the stroke rehabilitation, hemiplegia of the upper extremity (UE) is one of the most encountered conditions in hospital-admitted stroke survivors, affecting two-thirds of this population. Hemiplegia of the UE causes functional impairment of the arm and hand, thus negatively impacting the daily activities of stroke survivors [1, 2]. Rehabilitation therapy is considered the foundation of stroke treatment for improving the motor skills and quality of life of survivors. Further, repetitive training is an effective method to facilitate recovery from stroke and assist in restructuring the neural networks [3, 4].

Therefore, in stroke rehabilitation, robots have been widely utilized by clinicians, as they allow the user to perform repetitive movements consistently and precisely [5]. Robotic therapy enables patients to perform repetitive training with voluntary movements by providing mechanical assistance to the upper limbs, which can hardly move voluntarily due to the sequelae of stroke. In stroke rehabilitation, “active assistive training” is a traditional method that has been employed in many clinical settings (the patient attempts a voluntary movement while the therapist provides some form of limb support and mechanical assistance to complete the desired movement).

Reportedly, in terms of robotic therapy in patients with stroke, the guidelines stipulated by the American Heart/Stroke Association are effective for the treatment of UE paresis in stroke survivors [6]. However, whether robotic therapy can produce better UE recovery compared to conventional rehabilitation remains unclear. Further, a previous randomized controlled trial (RCT) showed no significant difference between conventional rehabilitation and robotic therapy in UE functionality improvement in stroke survivors [7‐9]. A systematic review also indicated that each robotic therapy and a high-intensity conventional approach produced similar effects on UE improvement in stroke survivors [10]. Additionally, another systematic review and meta-analysis concluded that robotic therapy could produce more significant improvement than active assistive training, but the effect size of robotic therapy appeared to be relatively small [11].

In the study of robotic therapy applications, one of the most developed paradigms is the investigation of robot-generated assistance. Recent robotic therapy using assistive controllers, such as force-field, intelligence difficulty (the exercise via audiovisual) [12], and feedforward controllers, allow stroke survivors to move their affected UE more independently. This exercise could be conducted in a similar way as “active assist” exercise, which is manually assisted by a rehabilitation therapist. The developed rehabilitation robot behaviour mode is composed of two concepts: error reduction and error augmentation. The error reduction concept indicates that the use of robot assistance can reduce the risk of movement error occurrence by allowing the performance of appropriate exercises, which can reinforce the process of motor learning [13, 14]. In contrast, the error augmentation concept stipulates that, to exaggerate movement errors, challenging-based assistive controller should be used in resistance training, which facilitates the process of motor learning [15]. Considering the nature of robotic treatment according to the two different types of concepts mentioned above, robotic treatment program in stroke rehabilitation has become more complex in recent years.

Some previous studies recommend the error reduction strategy using robotic assistance in UE training for improving UE function in stroke survivors. Rowe et al. [16] investigated the appropriate amount of robotic assistance using finger exoskeleton robotics and found that greater robotic assistance improved the UE function on the affected side compared to less robotic assistance in post-stroke practice (there was no significant difference, but there was a significant improvement in the secondary outcomes). On the other hand, several researchers have reported the possibility of negative effects from physical active assistive exercises [17, 18]. These studies suggest that excessive robotic assistance in intentional movements may contribute to the Slacking Hypothesis (decreased motor output, energy expenditure, and attention performed by the subject during the exercise). This would highlight the importance of providing “assistance-as-needed” in active assistive exercises [15]. These findings suggest that it remains unclear whether a positive rehabilitation outcome correlates with the level of robotic assistance provided for a stroke survivor to complete UE voluntary movement training.

Therefore, the purpose of this exploratory study was to examine the trend of the impact of robotic assistance level in order to derive data for future prospective controlled trial. In this paper, we defined “the level of robotic assistance” as “the amount of physical support generated by a robot when performing an active assistive exercise”. “Robotic therapy” was defined as a therapy using robotic assistance to improve the affected UE function of stroke patients with different severities of UE paralysis.

This research was performed to investigate how the level of robotic assistance and its adjustment could affect the functional improvement of the affected UE in two stroke survivor groups from the previous study [19]: one group with severe-to-moderate UE hemiplegia; and the other group with moderate-to-mild UE hemiplegia. This study analyzed the outcome of two groups (the intervention group receiving “considerable” robotic assistance and the control group receiving “minimal” robotic assistance). The findings in this study demonstrated that the intervention group comprising stroke survivors with severe-to-moderate stroke severity (FMA < 30) showed more significant improvement in UE hemiplegia and performance (measured using FMA-proximal, WMFT-PT, and WMFT-FAS) than the control group with the same severity of UE condition. Interestingly, the stroke survivors with moderate-to-mild stroke severity (FMA ≥ 30) in the control group showed significantly improved UE hemiplegia and UE performance measured using the same outcome tools compared to the intervention group.

Regarding the group × severity interaction, there was no significant difference found in the FMA score. This could be because of the fact that the ReoGo robot used in this study is a type of robot that immobilizes the paralyzed hand but assists in exercises involving the shoulder, elbow, and forearm only [23]. More specifically, we believe that FMA, which does not assess the outcome of the paralyzed shoulder, elbow, and forearm, unlike the FMA-proximal, might have indicated more reliable and accurate results considering the way we utilized the ReoGo in this study.

This result suggests that, to maximize the improvements of the affected UE function in sub-acute stroke survivors, the robotic assistance may need to be increased in patients with severe-to-moderate paralysis and decreased in patients with moderate-to-mild paralysis, to improve the performance of voluntary movements as much as possible.

The American College of Cardiology/Stroke Society [6] suggests that robotic therapy can allow stroke survivors with moderate-to-severe upper limb paraplegia to participate in intensive practice. Furthermore, assistance from a robot is suggested to be implemented only as needed, and this approach is proved to be effective for facilitating UE recovery, especially in those with moderate-to-severe UE impairment. Considering recent trends of motor learning and neurorehabilitation [24, 25], minimal robotic assistance has been preferred to be implemented in stroke rehabilitation. Some studies [24, 26] highlighted that providing minimal robotic assistance can facilitate improvement of UE function by reducing spasticity of the affected UE in stroke survivors. Additionally, passive repetitive robot-assisted training at a high-frequency level and conventional therapy were found to produce similar effects [24, 26] in terms of the improvement of UE function [25]. Therefore, based on the findings of the present and previous studies, the implementation of robotic assistance with an as-needed approach should be encouraged when during robotic therapy, with consideration of the level of voluntary movement that a stroke survivor with UE paraplegia can perform. However, few studies have investigated the effect of different levels of robotic assistance on UE function recovery.

One of the examples is a study by Rowe et al. [16] that compared the effects of high and low assistance levels generated by an “exoskeleton robot” aimed at improving the movement of the hand and finger on the affected UE in stroke survivors during the recovery phase. In this study, although there was no significant difference in the primary outcome of hand and finger recovery (measured using the Box and Block Test), the secondary outcomes of lateral pinch strength and UE (shoulder/elbow/forearm) function (based on FMA) appeared to significantly improve with higher assistance level from a exoskeleton robot. This study also reported that the amount of improvement in the affected hand function was greater in subjects with a more severe condition. This result was similar to some of our study findings. However, our study used an “end-effector robot” designed to improve UE (shoulder/elbow/forearm) function significantly improved the whole impairment in stroke survivors by providing appropriate level of assistance in each condition. A recent study involving patients with stroke reported that end-effector-type robotics were more effective for affected UE performance than exoskeleton-type robotics [27]. However, Moggio et al. [28] reported that more significant functional improvements were observed when exoskeleton-type robotics were used than when end-effector-type robotics were used for the affected finger in patients with stroke. Therefore, the effects of end-effector or exoskeleton-type robotics on the affected UE or affected fingers after stroke may vary. Thus, it is possible that the results of this study could be interpreted as novel as compared with those of previous studies.

Additionally, with regard to the detailed examination of the relationship between subjects’ severity of UE condition and robotic assistance in our study, the following findings were more evident in our study than in a previous study: the amount of assistance by the robot should be increased for subjects with severe-to-moderate UE impairment, and the amount of assistance by the robot should be reduced for subjects with moderate-to-mild UE impairment. The results of our study highlights the importance of considering “the ability to provide only the minimum assistance necessary”, which is one of the three hypotheses (high mechanical compliance, the ability to assist patients in completing desired movements, and the ability to provide only the minimum assistance necessary) stated by Wolbrecht et al. [29] for stroke survivors to perform effective movement practice. Our study findings support this hypothesis; therefore, maximal improvement of the paralyzed UE could be increased when clinicians provide the right amount of robotic assistance level, considering the severity of the UE paralysis.

One of the mechanisms of the effect of active assistive exercise using a robot is Hebbian plasticity, which enhances the proprioceptive input associated with spontaneous movement of the affected UE. Takahashi et al. [30] reported that active assistive exercise using a robot significantly improved the function of the affected UE after stroke. In the same paper, they also reported that functional MRI evaluation of the grasping task practiced during active assistive exercise with a robot showed a significant increase in the activation of the sensorimotor cortex during the intervention period.

The results of our study would suggest that stroke survivors with severe-to-moderate UE impairment can have a reduced risk of error occurrence by increasing the amount of robotic assistance. Our study also suggested that stroke survivors with moderate-to-mild UE impairment can have a reduced risk of experiencing the Slacking Hypothesis by appropriately reducing the robotic assistance. Our study results would support the hypothesis that the assistance-as-needed active assistive exercise of the paralyzed UE generated by the robot facilitated the proprioceptive input, which caused Hebbian plasticity. However, it is still early to confirm whether the robotic assistance-as-needed approach for performing active assistive exercise of paralyzed UE is associated with Hebbian plasticity. Further study needs to include neurological evaluation of Hebbian plasticity, including functional MRI.

This study had several limitations. Data from a previous work [9] were utilized in the present study, but no explanation was provided for the exclusion of 655 patients; thus, there may have been a bias in the data analysis. Another limitation of the study was the small sample size. The small study population may have affected the subgroup analysis. In addition, there were some scattered P-values of < 0.05 in this study, but these were not adjusted for multiplicity. Therefore, this result might be incidental. Further, since this study was a retrospective sub-analysis of a previous study; thus, it remains unclear whether an appropriate level of robotic assistance promoted the functional improvement or whether the appropriate level of robotic assistance was determined after functional recovery. Therefore, no clear causal relationship was established between the amount of robotic assistance and amount of functional improvement, and our results should be interpreted while considering these factors. In addition, the amount of robot assistance according to the severity of the UE paralysis was examined using only the evaluated values based on classical clinical tests, but no examination using robot-specific objective parameters was conducted.

Therefore, future randomized controlled trials that use state-of-the-art robotic devices and that evaluates a larger sample size need to be conducted to investigate how the level and methodology of robotic assistance could impact the improvement of UE function.

An optimal amount of robotic assistance was found to be a key to maximize improvement in post-stroke UE paralysis. Furthermore, the severity of UE paralysis is an important consideration when deciding the amount of assistance in robotic therapy. More specifically, to obtain maximum improvement with robotic rehabilitation in patients with UE paralysis after a stroke, it is important to set a lower robotic assistance level for mildly paralyzed patients and a higher robotic assistance level for severely paralyzed patients to encourage appropriate voluntary movements. In the future, it would be worthwhile to compare the effects of robot rehabilitation between low and high assistance groups through prospective randomized controlled trials.