Cybernic treatment with wearable cyborg Hybrid Assistive Limb (HAL) improves ambulatory function in patients with slowly progressive rare neuromuscular diseases: a multicentre, randomised, controlled crossover trial for efficacy and safety (NCY-3001)

To perform voluntary movements, humans need motor intention, which is transmitted via signals from the brain through the nervous system, and finally to effector muscles. In 1928, Sherrington used the term ‘motor unit’ for the complex of a lower motor neuron and the muscle fibres innervated by its axon [1]. For voluntary movements, the motor units are controlled by descending tracts of the central nervous system. There are a number of diseases that impair voluntary movements, and can occur at specific sites from the central nervous system to the motor unit. Previous studies, such as Ramón y Cajal [2], showed that in animal models, damaged neurons in the central nervous system (CNS) did not regenerate and synapses did not reconnect. Therefore, it has been thought that, in contrast to peripheral lesions, lesions of the CNS (including motor neurones) are permanent and cannot be regenerated even by rehabilitation. Hence, most rehabilitation programs for patients with motor impairment have not directly aimed to restore the function but rather only used the residual function to acquire compensatory activities [3].

Neuromuscular diseases are a broadly-defined group of disorders that involve injury to or dysfunction of the motor unit [4]. Especially in degenerative neuromuscular diseases, which are mainly caused by rare genetic variations, it has been considered that no therapeutic process is effective unless the underlying cause is removed at an early stage, owing to the dogma established by early on by studies such as that of Ramón y Cajal [2]. In these neuromuscular diseases, any treatment has been considered unlikely to be effective when advanced muscle atrophy and bone and joint deformities reach a progressed state. These progressive neuromuscular diseases are associated with a constant loss of the residual function and the difficulty in regaining compensatory activities. Moreover, excessive exercise conducted to improve the motor function has been known to lead to a further functional decline [5, 6]. Therapies that include muscle strength training and endurance training in neuromuscular diseases remain controversial [7‐11]. Hence, exercise therapy to restore the motor function has been considered dangerous and is not a common method of recovering function in these patients. For these reasons, there is a notable lack of safe and effective exercise programs for patients with neuromuscular diseases.

Therefore, until recently, only curative treatments based on the aetiology of neuromuscular diseases have attracted attention. Nevertheless, it has been shown that aetiologically-based treatments do not provide sufficient functional improvement in progressive neuromuscular diseases. For example, nusinersen—an antisense oligonucleotide—has been clinically tested for the treatment of spinal muscular atrophy (SMA), which is caused by a deficiency in the survival motor neuron (SMN) protein; the results showed that the effect of the treatment diminished as the disease progressed [12]. Furthermore, although leuprorelin—which inhibits the nuclear transportation of testosterone and the abnormal androgen receptor—has been clinically demonstrated to improve dysphagia in patients with bulbar spinal muscular atrophy (SBMA), there was no improvement in the ambulatory function [13]. While muscle biopsies in patients undergoing exon skip therapy to target the dystrophin gene with eteplirsen demonstrated the desired effect on protein synthesis in patients with the fatal neuromuscular disorder Duchenne muscular dystrophy (DMD), no sufficient relevant motor improvement was observed [14, 15]. Thus, even if drug developments are aetiologically successful, it is considered highly unlikely that lower motor neurons will re-innervate the muscle fibres and that the central nervous system will also regain control over the motor unit, including lower motor neuron. Therefore, the development of therapeutics that promote neuromuscular reconnection or remodulation has become an important research goal.

A member of our team, Dr Yoshiyuki Sankai, pioneered the innovative academic field known as ‘cybernics’, or the fusion of humans, robots, and information systems. Dr Sankai invented the wearable cyborg Hybrid Assistive Limb (HAL) based on the principle of cybernics [16‐20]. Cybernics technology links the human nervous system and a robot through bioelectric signals such as motor unit potentials, which results in a dynamic state in which the wearer and HAL are functionally and physically connected. HAL functions in accordance with the motor intention of the wearer and internal ideal movement patterns as if it were a part of the body. The process that realizes the fusion of a human and HAL via the mutual exchange of neural information (e.g., motor unit potentials and proprioception) and dynamic mechanical information (e.g., coordinates, velocity, and angular velocity in limb segments) between the wearer’s nervous system and HAL was named “interactive biofeedback (iBF)” by Dr Sankai. Since its inception, HAL has been implemented into the society (e.g., HAL lumbar type for labour support in the workplace) as well as the medical field as the world's first cyborg-type robot [21].

Despite visual similarities, HAL is completely different from other exoskeletal robots [22, 23], which are not cybernics based. These non-cybernic exoskeletal robots repeat mechanical patterns or are under the physical control of the wearer. In the case of HAL, the wearer performs voluntary movements, in which the neuromuscular and central nervous systems work to produce the intended movement of the integrated HAL and the wearer’s body. HAL is synchronized with the intention to move in accordance with the motor unit potentials produced by a command from the CNS. In this way, the iBF loop is created between the wearer and HAL [18, 24‐30] as follows: [brain ≫ spinal cord ≫ motor nerves ≫ (HAL + musculoskeletal system)] and [HAL ≫ musculoskeletal system ≫ sensory nerves ≫ spinal cord ≫ brain]. Using HAL, this loop can be easily repeated without increasing the neuromuscular system’s excessive load and fatigue. The physician or operator can intervene in this transmission loop by tuning the parameters embedded in HAL. Furthermore, it is believed that the return of sensory signals from the proprioceptors in the muscles, joints, and skin to the CNS is of great importance for neuroregeneration [31, 32].

In the present study, we propose that cybernics (i.e., the iBF created by the union of HAL and the nervous and musculoskeletal systems of the wearer) enables the regeneration of the impaired motor function. This neuroplasticity [33] would result in the reconnection of the CNS to the motor units [34] and the reconnection of motor neurons to muscle fibres. Thus, we hypothesized that the strengthening and adjusting of the synaptic connections in the neuromuscular and central nervous systems can be achieved via the iBF loop and may therefore result in neuroregeneration and regaining of the motor function, i.e., motor learning. At the beginning, we planned two clinical trials (RCTs) to assess the efficacy of iBF in neuroregeneration [30, 35, 36]: (1) NCY-2001, an RCT conducted to confirm whether iBF can induce neuroplasticity in case of the lesions of the CNS (i.e., lesions of the brain and spinal cord, located above the level of the motor units; spastic paraplegia), and (2) NCY-3001, an RCT conducted to confirm whether iBF is effective when motor units are damaged. When motor units are damaged, HAL needs to effectively detect weak and sparse motor unit potentials and compensate for very weak muscle power. Considering that HAL for medical use (lower limb type), which was originally developed for neuromuscular diseases, could also be helpful in a wider range of diseases and levels of progression (e.g., acute spinal cord injury and cerebrovascular disease to chronic disease), and given the lack of therapeutics that promote neuroregeneration, we sought to test the efficacy of HAL (NCY-3001) first in cases of slowly progressive rare neuromuscular diseases in which motor units were damaged (e.g., SMA, SBMA, muscular dystrophy, Charcot-Marie Tooth disease [CMT], distal myopathy, sporadic inclusion body myositis, congenital myopathy, and amyotrophic lateral sclerosis [ALS]).

Of the 33 patients who provided written informed consent, two were ineligible at the time of registration to the pre-observation period (Fig. 3). The evaluation of safety was conducted in the remaining 31 patients. One patient was withdrawn from the trial during the pre-observation period, leaving 30 eligible patients at the second time point of registration. After randomisation, six patients were excluded for the following reasons: in group A, one patient voluntarily withdrew from the trial prior to the last 2MWT during the treatment period, and one patient was excluded because of an unmet 2MWT; in group B, one patient withdrew from the trial before the last 2MWT owing to pneumonia, one was excluded owing to an unmet 2MWT, and two discontinued the trial prior to the last 2MWT owing to the investigators’ judgment that it was too difficult for these subjects to continue the trial. Therefore, for the analysis of the FAS, group A consisted of 13 patients and group B consisted of 11 patients. The PPS was identical to the FAS.

The distribution of patient-related characteristics and baseline measures is summarised in Table 1. Individual patient profiles of baseline characteristics are also shown in Table 2. The groups were well balanced in the majority of relevant factors, except that group A had a higher number of patients with an outpatient status during the treatment period compared with group B (p < 0.047). The baseline measures for the primary endpoint of 2MWT, Barthel index, and total MMT score were also comparable between the groups (Tables 1 and 2).

For the FAS, the 2MWT data for 24 individuals are shown in Fig. 3. Patients in Fig. 4 are displayed in order of the magnitude of the effect of the treatment 2 (HAL). A treatment effect of HAL over the control treatment effect was observed in 13 subjects (upward stripe bar). A pre/post effect of the treatment 2 (HAL) was observed in 18 subjects (upward black bar). Six patients (6-1*, 6-2*, 2-5*, 3-1*, 1-4*, 5-1*) with the negative effect (downward black bar) of the treatment 2 (HAL) were all the outpatients in the group A. While HAL was not effective in these six patients, four of them (6-1*, 6-2*, 1-4*, 5-1*) obtained sufficient improvement with the treatment 1 (hoist only) before starting the treatment 2 (HAL). The crossover analyses using the FAS are summarised in Table 3. The treatment effect with HAL-HN01 compared with hoist only was identified for the primary endpoint (2MWT), as reflected in the mean change in the 2MWT (10.066%, 95% confidence interval: 0.667–19.464; p = 0.037; Table 3). The carryover and period effects were not significant (p = 0.331 and p = 0.051, respectively), thereby confirming the validity of the assumptions in the crossover design.

Regarding other endpoints, the percent changes in cadence at 10MWT and in total MMT scores indicated significant improvement (p = 0.003 and p = 0.039, respectively) as a result of cybernic treatment with HAL in the crossover design. Differences in speed at 10MWT were substantial but not statistically significant (p = 0.059), while step length was not significantly different. Barthel index score, PRO total score (post–pre), PRO total score (post-then), and response shift total score (pre-then) exhibited no significant improvement in the treatment 2 (HAL) compared with the treatment 1 (hoist only) in the crossover design.

In the parallel group comparisons (V13–V4) between groups A and B in the first period (Table 4), the distance for 2MWT (15.577%, p = 0.044) and cadence (8.474%, p = 0.026) were significantly higher in treatment 2 (HAL) compared with treatment 1 (hoist only). Moreover, in the post-/pre-evaluations in the first period, 2MWT distance, gait speed, cadence, step length, and total MMT score were significantly higher only in treatment 2.

In V23–V4 (first + second treatment periods, Fig. 1, Table 4), the 2MWT distance, speed, cadence, and step length exhibited significant changes in comparison to the post-/pre-evaluation of both groups A and B. In this comparison, total MMT score showed significant changes only in group B. In V23–V4 (first + second treatment periods), the 2MWT distance for group A was significantly lower than that of group B (Table 4). The aforementioned trends lasted until the post-observational period in a manner comparable to that of all periods.

The PRO total score exhibited significant changes in comparison to post/pre and post/then in the first period and the first + second periods (Table 4). However, there was no significant difference in PRO between treatments 1 and 2 in the first period. The Barthel index showed no significant changes in comparing the post/pre evaluation and between groups in the first, first + second, or all periods.

For the 5th use of HAL-HN01 and later, the three point evaluation score improved significantly in comparison to both first time use in overall and all six criteria (Table 5). The response "difficult to use" was not found after the 7th use of HAL-HN01. The time taken to put on HAL at the 5th use had an average value of 323.5 s, a median of 233.5 s, a minimum of 110 s, and a maximum 1200 s, and exhibited a wide degree of variation between clinical sites. While the time taken to put on HAL tended to decrease each time, the differences were not significant.

Safety analyses were performed for all 31 subjects included in the ASaT analysis (Fig. 3). In one subject who wore HAL once only during the pre-observation period, no adverse event or device failure occurred. In the 30 randomised patients (Fig. 3), adverse events occurred in 55.2% (16/29 subjects, 31 events) in treatment 1 (hoist only) and in 80.0% (24/30 subjects, 42 events) in treatment 2 (HAL). There was no significant difference (p = 0.0539, Fisher’s exact test) in total adverse event's frequency between treatments 1 and 2. The incidence rate of adverse events associated with the use of HAL was 46.7% (14/30 patients), and a total of 19 HAL-related adverse events were recorded, mostly in treatment 2 (Table 6). All the adverse events were mild and were easily remedied. Of the adverse events, pain (myalgia, back pain, arthralgia, pain in extremity, as well as arthrosis deformans pain) associated with walking programs accounted for approximately a half of all recorded adverse events. Contact dermatitis, excoriation, erythema, and skin exfoliation were common owing to contact with electrodes and cuffs of HAL. Falls and contusions occurred when HAL was not in use, but they were reported by an investigator to be related to the HAL-HN01 because they happened during the treatment 2 period. In addition, there were no adverse events associated with a failure of the device.

The summary technical document containing the aforementioned data has been successfully reviewed by the PMDA. HAL-HN01 has been approved as a new medical device (approved number: 22700BZX00366000, November 25, 2015) by the Ministry of Health, Labour, and Welfare and is covered by general health insurance in Japan for the treatment of patients with neuromuscular diseases.

The present RCT sought to evaluate the cybernics treatment's efficacy and safety to improve the ambulatory function in eight rare neuromuscular diseases. The crossover analysis confirmed a significant effect of the cybernic treatment on the primary endpoint measure (2MWT), which showed a 10.066% increase compared with the control treatment (hoist only). The results of this trial confirmed that patients with ambulatory dysfunction owing to neuromuscular diseases can achieve significant improvement in the gait function by performing at least nine 40-min sessions of intermittent cybernic treatment with HAL-HN01.

The eligible participants were enrolled into the pre-observation period of the trial and randomised into one of the two groups, so FAS was determined (Fig. 3). In the case of the crossover design (Fig. 1), there are no statistical methods to properly complement the missing values. Therefore, it was appropriate to exclude subjects with partial completion of the 2MWT by the pre-defined criteria of FAS (Fig. 3). Groups A and B were balanced regarding all the baseline measures except for the outpatient status during treatment (Table 1). Eight subjects from group A underwent the trial as outpatients. As six of eight subjects exhibited a higher therapeutic effect of the control treatment than the HAL treatment (Fig. 4), we determined that the effect of treatment with HAL was not overestimated. There were no influences of baseline characteristics on the crossover analyses of treatment effects for groups A and B (Table 1).

Our study showed that the 2MWT was an appropriate indicator of ambulatory capacity in patients with neuromuscular diseases. While previous studies often used the six-minute walk test (6MWT) to assess the submaximal capacity for functional walking that is necessary for ADL across different disease states [37], the 2MWT can reduce high inter-test variability related to fluctuating motivation observed with the 6MWT and is a valid alternative to the 6MWT [37]. The 2MWT has been able to eliminate the problem of extreme muscle fatigue following 6MWT for subjects with neuromuscular diseases [5, 6].

As neuromuscular diseases are both progressive and incurable, and as aetiologically-based drugs for neuromuscular diseases have not been shown to stimulate neuromuscular reconnection or remodulation, our results demonstrating the successful intervention of cybernic treatment are truly ground-breaking [36]. The data for secondary endpoints in the present study demonstrated that cybernic treatment significantly improved cadence. Since we did not observe a significant change in walking speed, cybernic treatment may have alternatively influenced walking endurance as indicated by 2MWT. In addition, we also consider the significant improvement in the total MMT score after cybernic treatment to be groundbreaking. The present results dispel the notion that the use of HAL for neuromuscular diseases might be also an excessive and dangerous exercise therapy [5, 6]. Thus, to the best of our knowledge, HAL is the very first intervention to improve MMT in cases of neuromuscular disease [43]. Furthermore, we also observed substantial improvements in walking patterns in a visual gait analysis although differences were not significant. As ADL measured by the Barthel index did not significantly improve after nine cybernic treatments in this trial, further trials should investigate whether ADL can be improved by increasing the number of cybernic treatments. In addition, the PRO total scores did not significantly improve during our trial. As PRO scores were subjectively evaluated, recording these values at each of the four-time points, i.e., pre and post in the first and second periods in the crossover design, may have exceeded the limits of patients’ subjective measurement capacity. In the first period (V13–V4), parallel comparisons of groups A and B between the control treatment (treatment 1) and HAL (treatment 2) reaffirmed the primary crossover analyses’ robustness. Both the treatment 1 (hoist only) and treatment 2 (HAL) were performed for group A or B in the opposite order in V23–V4. The results showed that starting with the treatment 2 (HAL) first, followed by the control treatment (group B), led to a more significant improvement in 2MWT distance. These data may suggest that if the same number and content of cybernic treatment (HAL) sessions are performed, the earlier the cybernic treatment is given (such as in group B), the better results might be obtained. The PRO improvement effect of both the treatment 1 (hoist only) and 2 (HAL) on post–pre and post-then in the first period (V13–V4), and on post–pre in the first + second period (V24–V4) in group A and B was clearly significant. Excessive exercise therapy for patients with neuromuscular disease has been considered risky in the presence of overwork muscle weakness[5, 6]. Still, both treatment 1 and 2 improved patients' subjective ratings of the total PRO scores, indicating more comfort while walking, with no deterioration thought to be due to overwork. This may be because patients with neuromuscular diseases but a willingness to walk evaluate the walking exercise with hoist as safe and good, regardless of whether they use HAL or not. It might also indicate the absence of gait exercise in the present standard clinical care for patients with neuromuscular diseases.

The time needed to put on HAL was approximately 5 min on average, indicating that suiting up with HAL does not pose a substantial burden for patients and operators for clinical use. While the time necessary to put on HAL did not decrease over time, the ease of use steadily improved according to operator evaluations. The 5th time that HAL was worn, the time to put HAL on varied widely from a minimum of 110 s to a maximum of 1200 s, and no further improvements were noted by the 9th time, indicating that sufficient education/training before using HAL is necessary rather than natural learning.

In the present study, cybernic treatment with HAL did not cause serious adverse events or secondary muscle weakness owing to the excessive use of muscles in these patients. Myalgia and arthralgia, which were observed after cybernic treatment in the present study, are common after exercise and are manageable. As a general precaution when using HAL, it is necessary to observe and address the skin conditions that result from contact with electrodes and cuffs. In particular, it is necessary to pay attention to the effects of HAL on the skin in subjects with sensory impairment, as HAL electrodes are common products used for measuring bioelectric signals, such as in electrocardiogram, and there is little room for improvement in the electrodes themselves.

In this trial, we showed that nine cybernic treatments were safe and effective for patients with neuromuscular disease. Since iBF is a method of motor learning, it is generally considered that Hebb's law can be applied. However, it should be further explored at what point the learning effect of repeated cybernic treatments reaches a plateau. More specifically, it will be necessary to clarify how the number and frequency of treatments affect this learning curve. Observational studies to this end have already begun [44].

In order to elucidate the principle of effects of cybernic treatment, the underlying mechanisms of iBF should be examined for each component involved in locomotor function. The important components that control locomotor function include the skeletal system, motor unit, proprioceptive system, spinal locomotor central pattern generator [45], and other components of the central nervous system. The reactivation of the spinal locomotor central pattern generator will need to be investigated, as a mere repetitive walking exercise may not activate this centre.

It will be necessary to show proper motor innervation of muscle fibres when examining motor units following cybernic treatment. Stable walking under a variety of conditions requires the CNS to function based on voluntary motor intention [31, 32]. Therefore, it will be necessary to examine the reconnectivity between the CNS and motor units (i.e. neuroplasticity) [34] and to noninvasively elucidate the mechanisms responsible for improvements observed with the cybernic treatment [46]. We believe that the combined cybernic treatment with nusinersen therapy in SMA, leuprorelin administration in SBMA, and exon skip therapy in DMD would show synergistic effects and further improve the treatment effects compared with each monotherapy.

Even if aetiologically-based treatments, such as antisense nucleotide therapy, are performed several years after the onset of neuromuscular disease, the motor function has been shown not to improve despite improved biomarker levels and no further worsening of symptoms [12, 13]. In this case, aetiologically-based therapies combined with cybernic treatment may have significant pre/post effects.

In the targeted population in this trial, each disease entity is rare, and these patients are vulnerable. We combined eight neuromuscular diseases into a group as neuromuscular diseases that share the same motor unit involvement. This grouping allowed us to perform a randomised controlled study of HAL-HN01 for such patients. The limitations were a short-term trial, the small number of patients with each of the diseases, and the inability to stratify the allocation of patients across all the conditions. Thus, in the future, we would like to conduct an observational study of long-term efficacy and safety in a higher number of patients with each of the diseases.

The present study is the first RCT to show the effectiveness and safety of cybernic treatment for neuromuscular disease, and it provides hope for the effective treatment in those living with often debilitating and incurable neuromuscular disorders. This study showed significant improvements in gait function in patients with incurable neuromuscular diseases after nine iBF-based cybernic treatments with HAL-HN01 compared with walking exercises with a hoist only. We believe that the observed clinical effects, which are limited to improvement in gait function and total MMT score, stem from motor learning or functional regeneration by iBF. Cybernic treatment with HAL-HN01 has the potential to become standard treatment for walking impairment in patients with neuromuscular diseases, both with and without aetiologically-based therapies.

Lastly, the efficacy of iBF-based cybernic treatment will be undoubtedly verified by integrating the results of our next trial on spastic paraplegia caused by CNS lesions (NCY-2001) with those of the present study (NCY-3001). We consider that cybernic treatment can be used for a broad range of lesions and aetiologies, including those in the brain, spinal cord, and motor units, from acute to chronic stages [47]. Preliminary studies have also shown that HAL-HN01 is a useful medical device for improving the motor function. Future RCTs will be needed to prove efficacy and safety in other diseases, such as cerebrovascular disease [48‐50], spinal cord injury [51, 52], multiple sclerosis, Parkinson disease, spinocerebellar degeneration, and Guillain–Barre syndrome. Although the present RCT results are limited to only neuromuscular diseases, this study opens the possibility of applying cybernic treatment to other diseases. Our findings set a precedent for replacing existing walking exercise therapy using a hoist or treadmill alone with cybernic treatment using HAL.