Effects of Robot-Assisted Rehabilitation on Hand Function of People With Stroke: A Randomized,Crossover-Controlled,Assessor-Blinded Study

Design

This study had a randomized, crossover-controlled, assessor-blinded design. Participants were recruited from the Department of Physical Medicine and Rehabilitation of a medical university hospital with the approval of the ethics committee. All participants provided written informed consent.

Participants

Participants were recruited between February 1, 2018, and June 30, 2018. A total of 25 (17 men, 8 women) community-dwelling patients with subacute or chronic stroke were divided into two groups and participated in two phases of treatment: RT and conventional therapy (CT; Lins et al., 2018). Patients were included in the study if they had had a first stroke with hemiplegia, had subacute (3–6 mo) or chronic (>6 mo) stroke, could understand instructions, were in Brunnstrom Stages II–V of recovery, had sensory impairment (revised Nottingham Sensory Assessment [rNSA] Tactile score <2 and Kinesthetic score <3), and had muscle tone allowing movement (Modified Ashworth Scale score <3). Patients who were ages <20 or >75 yr, who were unable to clearly see or hear the feedback from the device, or who had other medical symptoms affecting movement were excluded.

Procedure

The following baseline data were collected from all participants (Table 1): gender, age, education level, affected side, etiology, site of stroke, poststroke duration, tactile and kinesthetic sensation function, Brunnstrom stage, muscle tone of the upper extremity, and cognitive level (Montreal Cognitive Assessment; Carson et al., 2018). Participants were randomly sorted into two treatment groups using a computer program with simple randomization. The groups received 60 min of RT or CT twice per week for 6 wk, with order of treatment counterbalanced after a 1-mo washout period. Participants’ performance was assessed at four time points: RT pretest, RT posttest, CT pretest, and CT posttest (Figure 1). The occupational therapist assessors were blinded to treatment assignment; they did not screen participants or provide intervention and went to the clinics only to conduct the four assessments. All participants continued regular therapy, including speech and physical therapy.

OPEN IN VIEWER

Figure 1.

Flowchart showing flow of participants through the study.

Outcome Measures

Data Analysis

A power calculation performed for a previous study indicated that 23 participants per group would provide 80% power with an α of .05 to detect a within-groups difference in FMA–UE scores (Hsieh et al., 2018). The data were analyzed using IBM SPSS Statistics (Version 20.0; IBM Corp., Armonk, NY). The significance level was set at .05. Demographic and baseline characteristics were evaluated using a χ² or Mann–Whitney U test.

The sequence and period effects of the clinical data were evaluated first. A Mann–Whitney U test was performed to evaluate the sequence effect between groups for the RT and CT posttest measures. A Wilcoxon signed-rank test was performed to evaluate the period effect within groups for the RT and CT posttest measures. A Mann–Whitney U test was performed between groups for RT and CT treatments, and a Wilcoxon signed-rank test was performed for within RT and CT treatments if the data did not exhibit a significant period or sequence effect. If the data exhibited a significant period or sequence effect, we performed a mixed linear model analysis to investigate differences between treatments at the endpoint. In the intention-to-treat analysis, the last score was used only for the participant who withdrew (see Figure 1).

Participants

A total of 25 participants were recruited. Two participants in the RT-first group withdrew from the study. One withdrew during the initial phase for medical reasons; thus, we excluded this participant’s data. The other participant withdrew before the CT phase as a result of a move; thus, we included this participant’s data and conducted an intention-to-treat analysis. The final number of participants was 24. No safety concerns or adverse events were related to study participation. Before treatment, there were no significant between-group differences in demographic, clinical, or EMG data.

Upper Extremity Sensorimotor Effectiveness

With the exception of EMG data, analysis of the clinical data identified no sequence or period effects. With respect to motor function, a significant time effect was observed in the RT-first group for proximal (p = .030) and total (p = .046) FMA–UE scores. Although both groups exhibited partial improvement in scores, nonsignificant time effects were observed for other outcomes, and nonsignificant group effects were observed for all measures (Table 2). With respect to sensory function, no significant improvement was observed in the light touch assessment for either group. Although improvement in proprioception was noted, no significant time or group effect was observed (see Table 2).

The EMG results for both treatment conditions are presented in Table 3. A mixed linear model analysis was performed to avoid period and sequence effects. A significant group effect was observed for the peak value of the EDC during Task 3 (the small-block grasping task; p = .05). No significant group effect was observed for the other EMG data.

Hand Function Effectiveness

The mean change in BBT score for the RT-first group was .88, and no significant effect in intragroup (p = .102) or intergroup (p = .790) comparison was observed. Analysis of grasp strength also did not show a significant intragroup (p = .550) or intergroup (p = .598) effect (see Table 2).

Activities of Daily Living Effectiveness

Regarding ADL function, a significant time effect was observed in the MBI scores of the RT-first group (p = .038). No significant time effects were observed in the CT-first group (p = .071) or between the groups (p = .867; see Table 2). In the RT-first group, the items on which participants improved the most were stairs (n = 8), dressing (n = 5), mobility (n = 3), and transfer (n = 3). In the CT-first group, the items on which participants improved the most were dressing (n = 3), stairs (n = 3), feeding (n = 2), and bladder mobility (n = 2).

Upper Extremity Motor Effectiveness

The FMA–UE results for the RT condition revealed significant improvements in the proximal (shoulder and elbow) and total scores, consistent with the results of previous studies (Balasubramanian et al., 2010; Hsieh et al., 2018; Hu et al., 2015; Hu, Tong, Song, Zheng, Lui, et al., 2009; Hu, Tong, Song, Zheng, & Leung, 2009; Lambercy et al., 2011; Veerbeek et al., 2017). Recovery is more difficult in the distal (hand) than in the proximal part of the UE. Moreover, participants in this study had severe sensory impairment and low FMA–UE scores. Early finger extension and intact finger proprioception predict better motor recovery (Rowe et al., 2017; Stinear, 2010). Intensive training to move the proximal part of the UE in the early stages after a stroke is common; perhaps for this reason, the function of the participants’ proximal UE was more efficient than that of the distal part. The proximal parts of the upper extremities are mainly used for stability and transport, and the distal parts are used for object manipulation (Hsieh et al., 2018). Even in distal hand training, the muscle activity of the proximal UE and coordination between proximal and distal movements remain necessary (Chae et al., 2000; Dewald et al., 2001; Takeuchi & Izumi, 2012).

Motor Improvement Evaluated Using Electromyography

Compared with the CT condition, results for the RT condition demonstrated a significantly smaller peak (p = .05) and mean (p = .09) amplitude of the EDC value during the small-block grasping task, which may indicate that grasping blocks was relatively easy (Sawada et al., 2017). The hand extensor muscle was more efficient in the RT condition during the small-block grasping task.

On the basis of a systematic review, Nordin et al. (2014) reported that kinematic and kinetic measurements are critical and widely used to provide an objective movement evaluation and to identify reduced dynamic behavior according to synergy patterns and interjoint (intralimb) coordination. Nordin et al. found that after RT, patients with stroke showed increased mean and peak speed, thus reducing the additional effort required to perform movements. These results were associated with the FMA–UE, muscle power, and Motor Status Scale scores.

Hu et al. (2015) conducted 20 robot-assisted wrist rehabilitation sessions, using sensory cues and guidance for correct muscle use during training. The extensor and flexor muscles were coordinated and movement smoothness was achieved after robot-assisted rehabilitation (Blank et al., 2014; Hu et al., 2015; Nordin et al., 2014). Thus, improving hand extensor muscle control could effectively reduce synergistic patterns in people with subacute to chronic stroke after RT.

Upper Extremity Sensory Effectiveness

Light touch and proprioception were not significantly different between the RT and CT conditions after 12 sessions. However, proprioception of the elbow improved after RT. Nordin et al.’s (2014) review demonstrated that position sense improved, but not significantly, among people with subacute to chronic stroke. This finding is consistent with our results.

The loss of somatosensation affects movement planning and interlimb coordination. Movement planning is attributed to feed-forward sensorimotor control, which influences target attainment strategies including initiation of movement and initial speed with which a person moves to reach the endpoint. Sensorimotor function has been found to be significantly correlated with FMA–UE scores (Nordin et al., 2014). This finding is consistent with our results, which showed that participants’ proprioception of the elbow and FMA–UE scores improved. Interlimb coordination affects the accuracy of limb positioning during movement and enables functional tasks, such as drinking, to be performed satisfactorily. Our results also demonstrated that after RT, participants achieved high efficiency during the small-block grasping task.

Hand Function Effectiveness

We found no significant differences in BBT scores and dynamometer grip strength between the conditions, results that are inconsistent with those of previous studies (Masiero et al., 2011; Takahashi et al., 2008). The mean total FMA–UE scores were 20.46 for the RT participants and 21.67 for the CT participants. According to Woytowicz et al. (2017), scores <28 indicate severe impairment. In that study, 75%, 16.7%, and 8.3% of participants had severe, moderate, and mild motor impairment, respectively. Our study participants had more severe impairment, which may have influenced their distal dexterity recovery.

Baseline BBT scores between 20 and 30 are 2–6 times more likely to be followed by clinically significant improvements in outcome measures compared with baseline BBT scores of <10 (Hsieh et al., 2014). Only 6 participants in our study could perform the small-block grasping task, and only 2 had a baseline BBT score of >10. The BBT requires grasping, transporting, and releasing objects, which require a combination of proximal and distal functional movements (Kontson et al., 2017). Although the motion of the proximal UE improved among participants in our study, insufficient distal function made the BBT task difficult.

In other studies (Fischer et al., 2007; Veerbeek et al., 2017), shoulder and elbow robotics exerted small but significant effects on muscle strength. Most studies used evaluation tools, such as the Medical Research Council Scale for Muscle Strength, Motricity Index Arm subscale, and Motor Power Scale (Veerbeek et al., 2017), that are different from those used in our study. Fischer et al. (2007) reported that improvement may have resulted primarily from developing new movement strategies or improving proximal arm control rather than from improving finger extension or strength, consistent with our results.

Activities of Daily Living Effectiveness

Participants’ ADL ability improved more after RT than after CT, especially on the Dressing, Mobility, Stair Climbing, and Chair/Bed Transfer indexes of the MBI. These results are consistent with those of a study of people who received 25 hr of robot exercise with MIT-MANUS; they experienced positive functional outcomes including increased proximal arm strength, reduced motor impairment in the shoulder and elbow, and improved ADL function (Volpe et al., 2000). Proximal hand improvement could transfer to ADL function, and training the arms may improve walking rehabilitation because of the activation of interneuronal patterning networks after a stroke (Kaupp et al., 2018). However, our results are inconsistent with those of Norouzi-Gheidari et al. (2012), who used the proximal part of upper extremity robotic devices. The intervention regions they used for RT (proximal UE) were different from those used in our study (distal UE). A Cochrane Review by Veerbeek et al. (2017) found that ADL improvement varies by the duration and amount of training, type of treatment, and differences in patient characteristics. Moreover, after RT, patients with stroke have shown improved quality of movement as assessed with scales related to ADLs such as the Motor Activity Log and Stroke Impact Scale (Balasubramanian et al., 2010; Hsieh et al., 2014, 2018). Further study is warranted.