Upper Limb Robotic Rehabilitation for Patients with Cervical Spinal Cord Injury: A Comprehensive Review

The upper extremities limitation represents one of the essential functional impairments in patients with cervical spinal cord injury. Electromechanics assisted devices and robots are increasingly used in neurorehabilitation to help functional improvement in patients with neurological diseases. This review aimed to systematically report the evidence-based, state-of-art on clinical applications and robotic-assisted arm training (RAT) in motor and functional recovery in subjects affected by cervical spinal cord injury. The present study has been carried out within the framework of the Italian Consensus Conference on “Rehabilitation assisted by robotic and electromechanical devices for persons with disability of neurological origin” (CICERONE). PubMed/MEDLINE, Cochrane Library, and Physiotherapy Evidence Database (PEDro) databases were systematically searched from inception to September 2021. The 10-item PEDro scale assessed the study quality for the RCT and the AMSTAR-2 for the systematic review. Two different authors rated the studies included in this review. If consensus was not achieved after discussion, a third reviewer was interrogated. The five-item Oxford CEBM scale was used to rate the level of evidence. A total of 11 studies were included. The selected studies were: two systematic reviews, two RCTs, one parallel-group controlled trial, one longitudinal intervention study and five case series. One RCT was scored as a high-quality study, while the systematic review was of low quality. RAT was reported as feasible and safe. Initial positive effects of RAT were found for arm function and quality of movement in addition to conventional therapy. The high clinical heterogeneity of treatment programs and the variety of robot devices could severely affect the generalizability of the study results. Therefore, future studies are warranted to standardize the type of intervention and evaluate the role of robotic-assisted training in subjects affected by cervical spinal cord injury.


Introduction
Spinal cord injury (SCI) represents one of the most disabling neurological conditions by complete or incomplete damage to the spinal cord with resulting detrimental consequences in motor, sensitive, and visceral controls [1][2][3][4].
The prevalence of SCIs widely varies among countries, ranging from 13.0 per million to 163.4 per million people [5]. Considering that most of the presentation involves young adults, both sanitary costs and lifetime assistance costs are highly burdensome, estimating a comprehensive cost of more than 1 million dollars per person [6]. SCIs might arise from mechanical damages (i.e., contusions, compressions or lacerations of the spinal cord) or non-traumatic events (e.g., degenerative cervical myelopathies, cancers, infections, intervertebral disc diseases, etc.) [6,7].
High-level spinal cord lesions could lead subjects to a high disability, considering the loss of arms and hands function related to detrimental consequences of functional impairment, reduced independence in activities of daily living (ADL), and a poor Health-Related Quality of Life (HRQoL) [1,2,7,8].
Rehabilitation might play a crucial role in the arm and hand functional recovery of patients affected by SCI, with a large variety of therapeutic options currently adopted. [7,9] It has been recently proposed that repetitive, task-specific, functional training could be considered effective in improving upper limb functions, even potentially interacting with the self-repair capacity of the spinal cord [10,11].
Among the new therapeutic options, robotic devices are well suited to produce intensive, task-oriented motor training that might enhance conventional rehabilitation facilitating the plasticity-related recovery by increasing sensory feedback and supporting the motor system [12].
These devices might perform arm or hand-assisted training, typically targeting either the shoulder and elbow, or the wrist and fingers. Robotic devices can be categorized as exoskeletons or end-effectors. Exoskeletons are devices that directly control the articulation of targeted joint(s), whereas robotic end-effectors contact users at the distal part of their limb [11,13,14]. Robotic devices are currently used in clinical practice to deliver an adequate intensity of training in terms of movement repetitions even in more severe subjects, which promotes functional recovery and may potentially facilitate adaptive plasticity [11,13].
In addition, robotic training provides the standardized rehabilitative training and monitors recovery of motor function in patients more objectively, thus reducing the subjective human influence [15]. Robotic rehabilitation aims to optimize learning strategies and to provide a patient-tailored rehabilitation plan [11]. Nowadays, more than 120 devices have been developed for upper limb rehabilitation of patients affected by neurologic disability [16].
To date, interest has been growing in the scientific literature, with several papers suggesting medical relevant features of robotic-assisted rehabilitation in functional recovery of patients affected by neurologic disability [14,[17][18][19]. However, despite these promising findings, there is not agreement on the effectiveness of this novel approach in the current clinical practice of the rehabilitation field. Moreover, even the expensive technology could limit the spreading of this advanced treatment in clinical settings and the evidence of its effectiveness in patients affected by neurological diseases of rehabilitative interest, including SCI. Therefore, this comprehensive review of systematic reviews and clinical studies summarizes the state-of-art on safety, clinical applications, and effectiveness of robotic rehabilitation in the integrated management of upper limb functional recovery in SCI patients.

Materials and Methods
The present study has been carried out within the framework of the Italian Consensus Conference on "Rehabilitation assisted by robotic and electromechanical devices for persons with disability of neurological origin" (CICERONE) [20].

Search Strategy
PubMed/MEDLINE, Cochrane Library, and Physiotherapy Evidence Database (PE-Dro) databases were systematically searched from inception to September 2021 for all the papers published following the SPIDER tool strategy [21], depicted by Table 1. This comprehensive systematic review of systematic reviews and clinical studies has been performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement. [22]

Selection Criteria
After the 'duplicates' removal, two reviewers (LP, LL) independently screened for inclusion title and abstract of all potentially relevant studies identified. In case of disagreement, a consensus was achieved by the decision of a third reviewer (AdS). Full-text studies were retrieved by the same two reviewers (LP, LL) and independently screened for inclusion. If consensus was not achieved by discussion between them, disagreements were solved by the decision of a third reviewer (AdS).
Randomized controlled trials were considered eligible if responding to the questions defined according to the following PICO model: (P) Participants: SCI patients in acute, subacute (≤3 months after injury), or chronic phase; (I) Intervention: Rehabilitation training with robotic-assisted devices for upper limb, with or without conventional therapy; (C) Comparator: Conventional rehabilitation; (O) Outcome measures: safety of robotic rehabilitation, the feasibility of robotic rehabilitation, upper limb strength, functioning, independence in ADL, and HRQoL.
We included systematic reviews, randomized controlled trials (RCTs), observational analytic studies, and case series. Exclusion criteria were: (1) papers involving animals; (2) language other than English; (3) case reports design; (4) participants with different neurologic disabilities from SCI; (5) robotic-assisted rehabilitation combined with other advanced technologies such as non-invasive brain stimulations (NIBS) or transcranial direct current stimulation (tDCS).

Data Extraction and Synthesis
All data were extracted from eligible full-text documents through Excel by two different authors. In case of disagreement, the consensus was achieved by the review of a third author.
All studies included were synthesized, describing both study characteristics and data extracted. A meta-analysis was not performed given the high clinical heterogeneity in design, intervention, and outcomes assessed in the different studies.

Study Quality
The five-item Oxford CEBM scale was used to rate the level of evidence (OCEBM website). The study quality included was assessed by the 16-item assessment of multiple systematic reviews 2 (AMSTAR 2) scale [23] for systematic reviews, and the 10-item PEDro scale24 for the randomised clinical trials. Regarding the PEDro scale, the risk of bias was rated as poor (0-3), fair (4)(5), good (6-8) and excellent (9)(10) in line with the PEDro scale. [24] Two different authors rated the studies included in this systematic review. If consensus was not achieved after discussion, a third reviewer was interrogated.
The studies included in this systematic review were published from 2012 [25] to 2020 [35], covering several Nations from all over the world; more in detail, seven studies were from the Americas (two from Canada [25,32] and five from USA [26,27,29,30,33]), two from Europe (one from Netherlands [28] and one from UK [35]), and two from Asia (Republic of Korea [31,34]).
The study cohort sample sizes were highly heterogeneous in the research studies, ranging from five (case series) [28] to 34 (RCT) [31] for clinical trials; nevertheless, the systematic reviews included larger samples (73 study participants by Singh et al. [32] and 88 by Yozbatiran et al. [33]). All the studies assessed patients of both genders, with ages ranging from 17 [26] to 76 years. [29] The study by Fitle et al. [27] did not report age.
Concerning the study quality of the clinical studies, we reported one good-quality [31], one fair-quality [34], according to the PEDro scale [24]. The two systematic reviews showed a low quality [31] and a critically low quality [33] according to AMSTAR 2 scale [23].

Main Findings of the Included Studies
All the case series [25][26][27][28][29] included in the present systematic review assessed the feasibility of robotic rehabilitation in SCI patients. Zariffa et al. [25] assessed both compliance and therapist timing, reporting that more rehabilitation exercises were performed with progressively less hands-on involvement by the therapist. Tolerance has been assessed by Francisco et al. [29], reporting no significant increase of self-reported pain and discomfort level during the therapy sessions. Accordingly, Cortes et al. [26] reported a high safety profile and tolerance without increasing pain and spasticity, and Vanmulken et al. [28] showed a discrete tolerance (Usefulness, Satisfaction and Ease-of-use questionnaire mean score of 66.1 ± 14.7%). Lastly, it should be highlighted that all papers included in this systematic review [25][26][27][28][29][30][31][32][33] did not report any major adverse event during robot-assisted training in SCI patients.
On the other hand, the RCT performed by Kim et al. reported significant differences between groups in terms of total SCIM-III score (7 [2 to 11] vs. 0 [−4 to 4]; p < 0.01). However, only the mobility (room and toilet) item significantly varied between groups (1 [0 to 3] vs. 0 [−1 to 1]; p = 0.02) in contrast with the other items not showing significant differences [31].
Both systematic reviews [32,33] reported that robot-assisted rehabilitation might be considered promising training to improve muscle function in SCI.
Lastly, the study by Frullo et al. [30] reported a significant improvement of normalized speed (p < 0.001), mean arrest period ratio (p = 0.001), and spectral arc length (p = 0.001) only in the assist-as-need group.

Discussion
Advancement in technology has been widely spreading in the rehabilitation field during the past two decades, and SCI patients might benefit from robotic rehabilitation. However, despite this approach being commonly adopted in the clinical practice, this systematic review showed that only a few studies assessed the effectiveness of roboticassisted training for recovering upper limb muscle strength and function in patients with SCI.
Taken together, our findings suggested that robotic devices for upper limbs might be considered safe, tolerable, and feasible in the complex rehabilitative management of SCI patients. However, to date, safety, tolerance, and feasibility of robot-assisted training have been primarily investigated in patients with other neurological diseases (i.e., stroke and multiple sclerosis) [36][37][38] and these outcomes should be deeply assessed in SCI patients, starting from the findings reported by the present systematic review.
Indeed, motor and sensory feedback stimuli are key components of task-oriented robotic training and might be more effective in patients with incomplete SCI than complete SCI [11][12][13][14]. Moreover, plasticity process can be elicited indirectly by sensory and motor afferent stimuli and directly through neuromodulation via non-invasive brain stimulation. More in detail, Yozbatiran et al. [39] suggested that modulating excitatory input of the corticospinal tracts on spinal circuits induced by tDCS combined with robot-assisted training could improve arm and hand functions in persons with incomplete SCI. This intriguing study has not been included in our systematic review, considering that the combination of robotic-assisted rehabilitation with other advanced technologies (i.e., NIBS and tDCS) was an exclusion criterion due to the limitation that they might affect the efficacy of robot-assisted training. However, we are aware that this combination should be deeply investigated in future studies on SCI patients.
Robotic training should be considered as an "add on" to conventional therapy in sub-acute SCI patients (≤3 months after injury); four studies included in this systematic review assessed the role of robotic-assisted rehabilitation combined with conventional physical therapy [25,28] and occupational therapy [25,31], probably due to complex scenario underpinning SCI management. In contrast, robotic treatment has been proposed as a stand-alone therapy in three case series out of four involving chronic SCI patients [26,27,29].
The present comprehensive systematic review showed a lack of evidence on differences between proximal (shoulder elbow) and distal (hand) training according to the robot design. More in detail, rehabilitation robots could be classified into two groups: end-effector based robots, which provide training capability encapsulating a large portion of the functional workspace, and exoskeletons, designed to resemble human anatomy with a structure enabling individual actuation of joints [40]. Therefore, we would like to highlight that future studies should involve enhanced control modes to allow additional treatment options in SCI patients; indeed, taking into account the different actions that the upper limb might exert (i.e., reaching and grasping), robotic devices might have a more targeted function with a more specific mechanical design in order to perform an adequate patient-tailored rehabilitation in subjects after SCI.
Concerning the type of intervention proposed, very high variability was recorded in terms of robot devices, the number of sessions per day, session duration, frequency, and joint involvement. This intrinsic limitation, probably related to the first phase of adopting new technology, severely affects the generalizability of these findings. In addition, it should be noted that the type of treatment intervention should be based on the SCI level, considering the clinical heterogeneity of functional disability occurring in cervical SCI. Future studies should focus on larger samples involving cervical SCI patients divided into subgroups to provide a patient-tailored robotic rehabilitative treatment.
In the literature, we found two similar systematic reviews investigating the role of robotic rehabilitation in SCI patients, albeit their quality was classified as low [32] and very low, [33] according to AMSTAR 2 scale [23]. Indeed, both Smith et al. [32] and Yozbatiran et al. [33] summarized the available literature on the robot-assisted training in upper limb rehabilitation of SCI patients, including even case reports and studies on the combination of robotic rehabilitation with other advanced technologies, severely affecting the homogeneity of data assessed and heavily influencing their results.
Nevertheless, by the present systematic review, the RCT performed by Kim et al. [31] was investigated first. This good-quality paper reported a significant improvement in terms of UEMS in the robotic training group compared to the control group (1 [0 to 3] vs. 0 [−1 to 1]; p = 0.03) in SCI patients; on the other hand, no significant changes in MRC scale were shown (p > 0.05). The authors suggested that significant improvement in muscle strength might have potential benefits in terms of short-distance mobility and electrical wheelchair manipulation. In line with these findings, significant improvements in SCIM-III scores (7 [2 to 11] vs. 0 [−4 to 4]; p < 0.01) in the robot-assisted rehabilitation group might have positive effects in terms of independence in the ADL [31].
Considering these findings, the present study might be viewed as the first systematic review performed by a large consensus panel of experts, including research studies specifically assessing the effects of robot-assisted training of the upper limb in patients with SCI. We showed that the current available literature on this topic might be defined as low-quality evidence. The lack of evidence might be partly due to the rapid evolution of advanced technologies with high costs that might not allow a standardization and reproducibility of single large-scale rehabilitation intervention.

Conclusions
Taken together, the present comprehensive systematic review summarized the stateof-the-art robotic-assisted rehabilitation treatments available for patients suffering from cervical SCI. Nowadays, robotic-assisted training is still experimental, but recent studies provided preliminary evidence showing intriguing positive effects on functional outcomes in SCI patients. We are aware that the high clinical heterogeneity of treatment programs and the variety of robot devices could severely affect the generalizability of the study results; therefore, future studies are warranted to standardize the type of intervention and evaluate the role of a robot-assisted training in the complex rehabilitation management of patients with SCI.