Safety and Selection Criteria of Single-joint Hybrid Assistive Limb for Upper Limb Paralysis After Stroke: A Phase I Trial

Abstract

Robotic rehabilitation is a high-intensity intervention for upper limb paralysis after a stroke. This study explored the safety and feasibility of using a single-joint hybrid assistive limb (HAL-SJ) exoskeletal device for upper limb paralysis in patients with acute stroke. In total, 11 patients with stroke (6 with moderate paralysis and 5 with severe paralysis) were enrolled between October 2021 and October 2023 in a stroke care unit. The patients underwent HAL training 3 times a week for 6 sessions. No serious adverse events related to HAL-SJ occurred, and participants demonstrated significant improvements in the Fugl-Meyer Assessment Upper Extremity (FMA-UE) and Action Research Arm Test (ARAT) (pre- vs. post-intervention; p < 0.05). The minimal clinically important difference (MCID) for FMA-UE was surpassed in 5 patients (83.3%) with moderate paralysis and 2 (40.0%) with severe paralysis. Regarding ARAT, 4 patients (66.7%) with moderate paralysis exceeded the MCID, whereas none (0.0%) with severe paralysis did. These findings suggest that HAL-SJ is both safe and feasible, particularly for individuals with moderate paralysis.

Introduction

Motor paralysis of the upper limbs manifests in 55%-75% of individuals afflicted by stroke,¹⁾ with scant instances of full functional restitution.²⁾ The adverse impact of upper limb motor paralysis on the activities of daily living (ADLs) in patients with stroke and the habitual non-use of the paralysed upper limb detrimentally influences their overall quality of life,^3,4) necessitating the early and assertive pursuit of functional recovery.

Early-phase functional recovery post-stroke necessitates the activation of residual corticospinal tracts through regular exercise and facilitation of neuroplasticity via active engagement of the paralysed upper limb.^5-7) Additionally, utilisation of the paralysed upper limb ameliorates abnormal interhemispheric inhibition and fosters use-dependent plasticity (UDP) in patients with stroke.⁸⁾ In this context, interventions to secure the amount of movement of the paralysed upper limb, such as constraint-induced movement therapy, are recommended for the rehabilitation of upper limb paralysis.⁹⁾ However, standard rehabilitation for patients with subacute stroke typically involves 32 or fewer repetitions of upper limb movements,¹⁰⁾ necessitating more intensive intervention modalities.

Recently, robotic rehabilitation has emerged as a prominent approach for delivering high-intensity exercises aimed at treating upper limb paralysis. Robotic rehabilitation, representing a methodology supported by robust evidence, is distinguished into exoskeleton and end-effector types.¹¹⁾ End-effector robots generate movement through their most distal segment without a one-to-one joint correspondence.^12,13) Hence, these robots are suitable for patients with sufficient residual motor skills to control movement.¹²⁾ On the contrary, exoskeleton robots exhibit a structure analogous to the human upper limb, with their joint axes aligning with those of the human upper limb.¹²⁾ These devices are intended to function in conjunction with the human upper limb and thus can be affixed to various sites on the upper limb.¹³⁾ Therefore, they do not require greater mobility and can be used from the early stage. The single-joint hybrid assistive limb (HAL-SJ) is a type of exoskeleton robot designed to improve limb function through appropriate and interactive feedback.¹⁴⁾ The HAL-SJ has demonstrated efficacy in upper limb function and ADL^15,16) and its compact and portable design confers the advantage of bedside applicability in patients with acute stroke. The Stroke Rehabilitation Clinician Handbook 2020 suggested that exoskeletal robotics for upper limb rehabilitation may not demonstrate substantial improvement compared to traditional methods,¹⁷⁾ leading to uncertainty regarding their effectiveness.^12,18,19)

To the best of our knowledge, few studies have examined the safety and appropriate candidates for HAL-SJ usage. This study aimed to ascertain the safety of elbow training using the HAL-SJ for post-stroke upper extremity paralysis and to delineate the patient characteristics that most significantly contribute to favourable outcomes with this device. The minimal clinically important difference (MCID) served as an index of clinical significance to examine the differences.

Materials and Methods

Patients

The study included 11 patients with upper limb motor paralysis following acute stroke who received treatment at the stroke care unit of our university hospital between October 2021 and October 2023. Inclusion criteria were established to ensure patients met specific eligibility requirements, excluding those who deviated from these criteria. Patients were included if they met the following criteria: 1) first occurrence of stroke; 2) Brunnstrom recovery stage upper extremity II-IV; 3) ability to provide voluntary consent; 4) absence of significant cognitive deficits that could affect HAL-SJ evaluation or performance; 5) absence of skin disease or skin fragility at the application site; 6) absence of orthopaedic disease conditions affecting upper limb movement; 7) detectable bioelectrical signals (BES) in the biceps and triceps brachii; and 8) effective management of comorbidities.^15,20,21) A patient selection flow diagram is shown in Fig. 1. The study received approval from the Ethics Committee of Tokushima University Hospital (approval number: 1934) and was conducted in accordance with the Declaration of Helsinki. All patients provided informed consent before enrolment in the study.

Fig. 1.

Flow diagram and results of a phase I trial evaluating the safety and efficacy of HAL-SJ for upper limb paralysis after stroke. SCU, stroke care unit; BES, bioelectrical signals; BRS, Brunnstrom recovery stage; MCID, minimal clinically important difference; HAL-SJ, single-joint hybrid assistive limb

Data collection

The primary outcomes of the study included assessing the safety and feasibility of HAL-SJ usage. Adverse events (AEs) were categorized as symptoms unrelated to HAL training, such as exacerbation of neurological findings, fractures, infections, or thrombosis occurring after the stroke, and symptoms specifically related to HAL training, such as abrasions, sore muscles, contact dermatitis, or arthritis.

Secondary outcomes focused on motor impairment and function. In assessing upper limb paralysis post-stroke, the Fugl-Meyer Assessment Upper Extremity (FMA-UE) was employed to evaluate the body function and structure domain within the International Classification of Functioning, Disability, and Health. The Action Research Arm Test (ARAT) was used for the activity domain.²²⁾ The FMA-UE quantifies upper limb motor paralysis severity, with scores ranging from 0 to 66 points, categorized as severe (0-19 points), moderate (20-46 points), or mild (≥47 points).²³⁾ The ARAT evaluates ADL, featuring 4 sub-components: grasp, grip, pinch, and gross movement, with scores ranging from 0 to 57 points.²⁴⁾ Assessments were conducted before and 2 weeks after the intervention, supervised by 2 occupational therapists and a neurosurgeon proficient in evaluation techniques.

Methods

Robotic rehabilitation using HAL-SJ and occupational therapy

Since the average length of stay in acute care hospitals in Japan is 16 days, the duration of the intervention for this study was set as 14 days.²⁵⁾ The patient underwent robotic rehabilitation using HAL-SJ thrice a week for 40 mins per day, totalling 6 times, and received the usual occupational therapy for the others.²⁰⁾ HAL-SJ (Cyberdyne, Tsukuba, Japan) is a sophisticated rehabilitation support robot designed to facilitate joint movement through voluntary control based on the BES detected from the skin surface during muscle contraction (Supplementary Fig. 1).¹⁴⁾ The skin was wiped with alcohol before applying the surface electrodes to reduce skin impedance. Subsequently, reference electrode positions were applied to the biceps and triceps brachii, and the cybernic control mode (CVC-Gentle) was employed, facilitating a gradual modulation of assist torque. The occupational therapist individually calibrated the parameters of the HAL-SJ to modulate the degree of assistance (assist-gain 0-100 and assist level 1-20) and the equilibrium allocation between flexion and extension (BES-balance, flexion 5%-100%, and extension 5%-100%), facilitating the execution of high-repetition rates. The intervention was conducted in the sitting position while monitoring the BES signals on the display screen. Potential risks associated with HAL-SJ training include abrasions and contact dermatitis resulting from device utilisation, as well as muscular strain and arthritis due to excessive exertion. Additional potential hazards, such as undue joint stress resulting from device malfunction (including fractures and dislocations) and falls due to loss of device control, were meticulously monitored by the occupational therapist during each intervention.

Occupational therapy included range of motion exercises, muscle strengthening, task-oriented training, basic movement training, walking, ADL training, and aerobic exercises. Similarly, physical therapy was provided for 20 mins per session, 5 times per week, with no intervention targeting upper extremity function.

Statistical analysis

The Wilcoxon signed-rank test was employed to assess the difference in each outcome measure before and 2 weeks after the intervention, with subsequent computation of the effect size. Effect size (r) serves as a measure that quantifies the magnitude of the effect or difference observed, independent of sample size or statistical methodologies across studies. An effect size of 0.10 denotes a small effect, 0.30 indicates a medium effect, and 0.50 signifies a large effect.

All analyses were performed using IBM SPSS (version 22). Statistical significance was set at p < 0.05. Additionally, the MCID of FMA and ARAT reported in previous studies were used to determine the effect.^26,27)

Results

Primary outcomes

In this study, robotic rehabilitation utilising the HAL-SJ was provided to 11 patients who met the inclusion criteria. Baseline patient characteristics are shown in Table 1. No serious AEs occurred in any of the patients. One patient experienced sore muscles as a HAL-related AE, whereas another had worsened neurological symptoms due to fluctuating symptoms of branch atheromatous disease, unrelated to HAL. However, no AEs were observed in patients with sensory impairment (Cases B, F, and K), and all patients completed the protocol (Table 2).

Table 1.

Baseline characteristics of the combined HAL-SJ group

Variables	Patients with moderate paralysis							Patients with severe paralysis					Moderate Mean ± SD	Severe Mean ± SD
	A	B	C	D	E	F		G	H	I	J	K
Age	75	73	77	79	50	63		61	71	62	53	31	69.5 ± 11.1	55.6 ± 15.2
Sex	M	F	M	F	M	M		M	M	M	F	F	M: 4	M: 3
Diagnosis	IS	ICH	IS	IS	ICH	IS		ICH	ICH	IS	IS	IS	IS: 4	IS: 3
Portion	M1	Thal	CR	CR	PU	CR		PU	Thal	CR	CR	BG
Handedness	Right	Right	Right	Right	Right	Right		Left	Right	Right	Right	Right	Right: 6	Right: 4
Paralysis side	Right	Right	Left	Left	Right	Left		Right	Left	Right	Left	Left	Right: 3	Right: 2
Interval from onset	5	4	7	5	6	8		5	4	6	6	10	5.8 ± 1.5	6.2 ± 2.3
Initial NIHSS	6	4	6	7	2	5		6	10	7	5	6	5.0 ± 1.8	6.8 ± 1.9
FMA-UE	28	30	24	24	45	31		19	2	12	18	9	30.3 ± 7.8	12.0 ± 6.9
FMA-A	26	14	22	17	28	21		19	2	12	15	9	21.3 ± 5.3	11.4 ± 6.4
ARAT	4	27	6	3	16	13		4	0	0	3	0	11.5 ± 9.2	1.4 ± 2.0
Sensory impairment	No	Yes	No	No	No	Yes		No	No	No	No	Yes	Yes: 2	Yes: 1
Repetition frequency	358.3	358.3	94.0	153.3	325.0	176.0		300.0	80.0	277.0	296.7	91.7	244.2 ± 116.7	209.1 ± 112.9

ARAT: Action Research Arm Test; BG: basal ganglia; CR: corona radiata; F: Female; FMA-A: Fugl-Meyer Assessment category A; FMA-UE: Fugl-Meyer Assessment Upper Extremity; HAL-SJ: single-joint hybrid assistive limb; ICH: intracranial haemorrhage; IS: ischaemic stroke; M: Male; NIHSS: National Institutes of Health Stroke Scale; PU: putamen; SD: standard deviation; Thal: thalamus

Table 2.

Incidence of AEs during the trial

Type of AEs	Participants, n (%)
Total AEs	2 (18.2)
Total SAEs	0 (0.0)
AEs leading to withdrawal	0 (0.0)
Device failure	0 (0.0)
HAL-related AEs
Abrasion	0 (0.0)
Sore muscles	1 (9.1)
Contact dermatitis	0 (0.0)
Arthritis	0 (0.0)
Non-HAL related AEs
Exacerbation of neurological findings	1 (9.1)
Fracture	0 (0.0)
Infection	0 (0.0)
Thrombosis	0 (0.0)

AE: adverse event; HAL: hybrid assistive limb; SAE: serious adverse event

Secondary outcomes

Significant enhancements were observed in FMA-UE, FMA category A, and ARAT pre- and post-intervention (p < 0.05), with a r ≥ 0.5 for all parameters (Table 3). When comparing the severity of upper limb paralysis between moderate and severe cases, the change in FMA-UE was 20.2 ± 9.5 for moderate paralysis and 9.2 ± 8.7 for severe paralysis post-intervention, with a tendency towards higher changes in FMA-UE and ARAT observed in the moderate paralysis cohort (Table 4). The MCID of FMA-UE during the acute phase was exceeded in 5 patients (83.3%) with moderate paralysis and 2 (40.0%) with severe paralysis. Furthermore, the MCID of ARAT was surpassed in 4 patients (66.7%) with moderate paralysis and none (0.0%) with severe paralysis.

Table 3.

Change in each Indicator Before and after HAL-SJ

Indicators	Pre-HAL-SJ	Post-HAL-SJ	p-Value	r
FMA-UE	22.00 ± 11.88	37.18 ± 19.34	0.005	0.60
FMA-A	16.82 ± 7.57	23.82 ± 11.00	0.008	0.57
ARAT	6.91 ± 8.46	17.18 ± 16.70	0.012	0.54

ARAT: Action Research Arm Test; FMA-A: Fugl-Meyer Assessment category A; FMA-UE: Fugl-Meyer Assessment Upper Extremity; HAL-SJ: single-joint hybrid assistive limb

Table 4.

Changes in Evaluation Results Per Patients

Indicators	Moderate paralysis group								Severe paralysis group
	A	B	C	D	E	F	Mean ± SD		G	H	I	J	K	Mean ± SD
Pre-FMA-UE	28	30	24	24	45	31	30.3 ± 7.8		19	2	12	18	9	12.0 ± 7.0
Post-FMA-UE	52	64	39	40	52	56	50.5 ± 9.6		34	2	19	39	12	21.2 ± 15.3
ΔFMA-UE	24	34	15	16	7	25	20.2 ± 9.5		15	0	7	21	3	9.2 ± 8.7
Pre-FMA-A	26	14	22	17	28	21	21.3 ± 5.3		19	2	12	15	9	11.4 ± 6.4
Post-FMA-A	32	36	31	26	31	30	31.0 ± 3.2		28	2	11	25	10	15.2 ± 10.9
ΔFMA-A	6	22	9	9	3	9	9.7 ± 6.5		9	0	-1	10	1	3.8 ± 5.3
Pre-ARAT	4	27	6	3	16	13	11.5 ± 9.2		4	0	0	3	0	1.4 ± 1.9
Post-ARAT	25	45	20	6	30	43	28.2 ± 14.7		10	0	0	10	0	4.0 ± 5.5
ΔARAT	21	18	14	3	14	30	16.7 ± 8.9		6	0	0	7	0	2.6 ± 3.6

ARAT: Action Research Arm Test; FMA-A: Fugl-Meyer Assessment category A; FMA-UE: Fugl-Meyer Assessment Upper Extremity; SD: standard deviation

Discussion

All patients who underwent HAL-SJ in this study experienced no serious AEs, indicating the safety of the protocol. Statistical analysis revealed a significant improvement in upper limb function pre- and post-intervention (Table 2). Subsequently, the effectiveness of robotic rehabilitation using the HAL-SJ was examined by comparing the MCID of each index reported in previous studies. The MCID of FMA in the acute phase is 10 points,²⁶⁾ while for ARAT, it stands at 12 points for the dominant hand and 17 points for the non-dominant hand.²⁷⁾ Patients surpassing the MCID demonstrated notable enhancements in FMA-UE (moderate paralysis: 88.3% vs. severe paralysis: 40.0%) and ARAT (moderate paralysis: 66.7% vs. severe paralysis: 0.0%) among those with moderate paralysis.

In this study, patients with moderate paralysis, as indicated by FMA-UE scores ranging from 20 to 46 points, demonstrated significant improvements in motor function. Additionally, individuals who completed approximately 300 high-repetition movements (severe paralysis: Cases G and J) also exhibited improvement. Recent meta-analyses on robotic rehabilitation have emphasised the importance of a high exercise frequency for enhancing functionality.²⁸⁾ The findings of this study support prior research.^20,28) Moreover, optimising task difficulty and enhancing the excitability of remaining corticospinal tracts are documented as crucial factors for improving motor paralysis in patients with acute stroke.^5,29) Thus, frequent exercise using HAL-SJ from the acute stage could have activated the sensorimotor cortex³⁰⁾ and promoted UDP and Hebbian plasticity.^8,31) Investigating the ideal assistance strength and BES-balance is imperative for future research. One potential factor contributing to the limited improvement in patients with severe paralysis could be the challenge of adjusting the difficulty level adequately due to simultaneous contraction of the biceps and triceps muscles, hindering the performance of high-repetition exercises while wearing HAL-SJ. In this study, enhancements were observed in both FMA and ARAT scores among patients with moderate paralysis; however, a previous investigation documented a lesser degree of change in ARAT scores (moderate: 16.67 ± 8.94 points vs. 6.25 ± 8.81 points).²⁰⁾ This discrepancy could be attributed to variations in the baseline severity of motor paralysis (FMA: 30.33 ± 7.76 points vs. 24.92 ± 14.39 points), the number of intervention sessions (6.00 ± 0.00 vs. 5.08 ± 1.83), and the duration before the introduction of HAL-SJ (5.83 ± 1.47 days vs. 15.67 ± 9.17 days).²⁰⁾

To our knowledge, few studies have reported on the safety of HAL-SJ training in patients with acute stroke. No serious AEs were observed in the phase I study, suggesting that elbow HAL-SJ training is a safe rehabilitation tool for patients with upper limb paralysis after stroke, regardless of sensory impairment. In the safe operation of the device, it is essential to calibrate the exercise intensity appropriately and identify equipment malfunctions promptly through routine maintenance. Furthermore, HAL-SJ was most effective in patients with moderate paralysis, which is in line with previous study findings.²⁰⁾ Although reports suggest that robotic rehabilitation exhibits optimal efficacy in cases of moderate paralysis,³²⁾ the end-effector type has demonstrated effectiveness across a spectrum ranging from moderate to severe paralysis.³³⁾ Therefore, the most effective treatment for patients may differ depending on the device's control method, location, and concept.

This study had a few limitations. Firstly, the dual role of the therapist as both the HAL-SJ operator and evaluator raises concerns about potential bias. Secondly, the absence of a comparison group challenges the validity of our interpretations, especially considering the potential for spontaneous recovery effects during the acute phase. Therefore, future investigations should aim to expand the sample size and incorporate comparative studies with control groups.

In conclusion, we examined the safety of HAL training for upper limb paralysis in patients with stroke and identified key patient characteristics. This study elucidated the combination of HAL-SJ and occupational therapy for upper limb paralysis following acute stroke, yielding auspicious outcomes. Patients with moderate paralysis were the most suitable candidates for HAL-SJ, suggesting its potential for safe and effective implementation. Future studies should evaluate the benefit of this device with a control group.

Conflicts of Interest Disclosure

All authors have no conflict of interest.

Supplementary Material

Supplementary Figure 1

(a) Single type of Hybrid Assistive Limb. (b) HAL-SJ attached to upper limb. The elbow joint features a visual feedback mechanism on its exterior, manifesting colour alterations to red, green, and yellow in correspondence with upper limb movements.