Vibrotactile-Based Rehabilitation on Balance and Gait in Patients with Neurological Diseases: A Systematic Review and Metanalysis

Postural instability and fear of falling represent two major causes of decreased mobility and quality of life in cerebrovascular and neurologic diseases. In recent years, rehabilitation strategies were carried out considering a combined sensorimotor intervention and an active involvement of the patients during the rehabilitation sessions. Accordingly, new technological devices and paradigms have been developed to increase the effectiveness of rehabilitation by integrating multisensory information and augmented feedback promoting the involvement of the cognitive paradigm in neurorehabilitation. In this context, the vibrotactile feedback (VF) could represent a peripheral therapeutic input, in order to provide spatial proprioceptive information to guide the patient during task-oriented exercises. The present systematic review and metanalysis aimed to explore the effectiveness of the VF on balance and gait rehabilitation in patients with neurological and cerebrovascular diseases. A total of 18 studies met the inclusion criteria and were included. Due to the lack of high-quality studies and heterogeneity of treatments protocols, clinical practice recommendations on the efficacy of VF cannot be made. Results show that VF-based intervention could be a safe complementary sensory-motor approach for balance and gait rehabilitation in patients with neurological and cerebrovascular diseases. More high-quality randomized controlled trials are needed.


Introduction
Balance is a complex multi-factorial system in which both motor and sensory components interact one with other [1]. The central nervous system integrates the information originate from visual, vestibular, proprioceptive and cognitive systems in a continuous sensorial re-weighting that ensure postural control in static and dynamic condition [1]. The weighting of the sensory inputs depends on both the environmental conditions and the motor task performed by the subject [2][3][4].
The integration of multi-sensory information is impaired in neurological diseases [4][5][6][7], leading to balance and postural control disorders and consequently to a high prevalence of falls and fear of falling [8]. It is known that among stroke survivors about 73% are reported to be fallers, 45-68% of people with Parkinson's disease (PD) fall each year and that up to 50% of the Multiple Sclerosis (MS) population are estimated to be fallers [8,9]. Postural instability and fear of falling represent one of the major causes of decreased mobility, limitation in physical activity, and social isolation resulting in reduced quality of life [10]. In recent years, rehabilitation strategies have been carried out considering an Brain Sci. 2021, 11, 518 2 of 15 active involvement of the patients during the rehabilitation sessions and a challenging task-oriented exercise [11].
Furthermore, the connection and the relationship between different aspects of cognitive and motor function is increasingly documented on motor and balance interventions as well as for the fear of falling in subjects affected by central nervous system diseases [24,25].
According to these approaches, new technological devices and paradigms have been developed to increase the effectiveness of rehabilitation by integrating multisensory information and augmented feedback [19,20,26] promoting the involvement of a top-down paradigm in neurorehabilitation, with an increase in the involvement of the cognitive functions [20,[25][26][27].
In this context, through vibrotactile feedback (VF), the vibratory stimulus was used as a peripheral therapeutic input, in order to provide spatial proprioceptive information to guide the patient during a motor task exercise [28,29]. VF is user-friendly and usable in the clinical rehabilitation setting [30], it needs a small actuator that generates a signal and supplies vibrational stimuli [21,31]. To date, many VF devices have been used to improve gait spatiotemporal parameters, facilitating a task-oriented rehabilitative approach in patients with neurological disorders [28,31,32]. Moreover, the purpose of the cue could be variable: it may represent information for alert, direction, spatial orientation and other communication, which should be made explicit to the performer before the training start [33]. It could be used alone or in combination with other or as a cognitive-motor task [33]. Furthermore, it could be combined with devices that make it feasible to assess the motor tasks performed by the patient and to adapt the vibrator stimulus to the performance [34]; thus, during the rehabilitation session, once an exercise has been assessed as being improperly performed, the patient can be provided with feedback in order to stimulate an adaptation and the motor task improvement [29] Although the VF has been shown to be useful for patients with neurological diseases across a variety of situations [29,35,36], there are no quantitative and qualitative reviews that systematically report the effect of VF on the motor functions in patients with neurological disorders. The present systematic review and metanalysis aimed to explore the effectiveness of the VF on balance and gait rehabilitation in patients with neurological diseases.

Materials and Methods
This systematic review and meta-analysis was performed in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement [37] and following the Cochrane Handbook for Systematic Reviews of Interventions [38]. The study's protocol was registered on PROSPERO International prospective register of systematic reviews website (registration number CRD42021217837).

Search Strategy and Eligibility Criteria
Electronic databases searched in November 2020 were MEDLINE (PubMed), PE-Dro (Physiotherapy Evidence Database). Search terms used were ("balance" AND "vibrotactile" OR "haptic" AND "neurological disorde*" OR "stroke" OR "parkinson" OR "multiple sclerosis" OR "traumatic brain injury"). Search terms were modified for each database and appropriate subheadings were used for each database searched (for detailed see Appendix A).
Controlled and non-controlled clinical trials (i.e., randomized or non-randomized trials), retrospective studies, case reports, case series, and observational studies, were included. No restrictions related to publication date, sex, and country were applied. Participants included in the studies presented static and/or dynamic balance disorders and a diagnosis of stroke, PD, MS and traumatic brain injury (TBI).

Study Selection and Data Collection Process
Duplicate records were identified and removed using the EndNOTE software. Study eligibility assessment and the data extraction process, carried out by two independent co-authors (SDA and AAP). In case of any disagreement, the opinion of a third author (MT) was used to reach accordance. The first selection of studies was initially conducted basing on the title and abstract and afterwards full-text articles were examined.
The summary of results was reported following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement [37]. Two authors (SDA and AAP) independently extracted the following relevant features of the included studies: name of primary author and publication year, participants, rehabilitative intervention, outcome measures.
The methodological quality of evidence was assessed using the PEDro scale [39] for the controlled trials and using a modified version of the Newcastle-Ottawa Scale (NOS) [14,40] for the observational studies. The assessment was performed by two authors (SDA and AAP); discrepancies were resolved by consensus with a third reviewer (MT) as an arbiter. The PEDro scale ranging from 0 to 10, the modified NOS ranging from 0 to 7. In both scales, the maximum score shows a better methodological quality.

Measures and Synthesis of Results
Data concerning qualitative synthesis are reported descriptively by using means, DS, percentages and ranges. A meta-analysis was performed using "Review Manager 5.3.5" (The Nordic Cochrane Centre, København, Dänemark, https://revman.cochrane.org/#/ myReviews. Only RCTs investigating comparable outcome measures were included in this quantitative synthesis. Alpha level was set at 0.05 to test for overall effect. For continuous measures ("pitch and roll" sway and angular velocity), standardized mean difference (SMD or Hedges's g) with 95% CI was obtained by using a random-effects model, in reason of the clinical and methodological variability detected among included studies. When available, absolute values (degrees for angles, degrees/seconds for angular velocity) were used; on the contrary, percentages were considered. An effect size ranging from 0.2 to 0.49 is to consider "small", from 0.5 to 0.79 "moderate" and a score of 0.8 or above represents a large effect. Heterogeneity was measured through I 2 statistics and explains how much of the variability among studies is due to heterogeneity rather than to chance. Values included in the range 0-40% may imply "no important" heterogeneity, 30-60% suggest "moderate" levels, 50-90% could indicate "substantial" and 75-100% "considerable" levels.

Results
Electronic searches identified 259 studies. Titles and abstracts were examined according to inclusion and exclusion criteria. The full texts of the articles were read to determine the eligibility. Furthermore, reference lists of identified articles were screened for additional relevant literature. Comparison of the retrieved titles identified five studies that were duplicates. The result consisted of 254 articles eligible for inclusion. After a full-text analysis 234 articles were excluded due to the following reasons: (a) did not concern balance and/or gait rehabilitation; or (b) did not provide active vibratory feedback. A total of 18 studies met the inclusion criteria and were included, just once, in the present systematic review ( Figure 1). Table 1 presents a narrative summary of results included studies with their associated characteristics and patient features. In particular, the following data are reported: first author's name, publication year, participants, intervention, and outcome measures.
The included studies were all published in English and were conducted in different countries. Six studies came from the USA, three from the Republic of Korea, and three from Japan; two studies were carried out in Germany, two in Switzerland, and two in The Netherlands; Italy, Spain, and China have contributed to this review with one study each one. Of the four investigated neurological diseases, 11 studies (61.1%) included patients with a diagnosis of PD; 6 studies (33.3%) concerning patients with stroke, specifically Brain Sci. 2021, 11, 518 4 of 15 three studies included chronic stroke patients (50%), two studies subacute-stroke patients (33,3%) and one study (16.6%) reported no data concerning the onset time; one study (5.5%) involved patients with a diagnosis of MS an and no studies including patients with TBI.
Brain Sci. 2021, 11, x FOR PEER REVIEW 4 of 16 Table 1 presents a narrative summary of results included studies with their associated characteristics and patient features. In particular, the following data are reported: first author's name, publication year, participants, intervention, and outcome measures. The included studies were all published in English and were conducted in different countries. Six studies came from the USA, three from the Republic of Korea, and three from Japan; two studies were carried out in Germany, two in Switzerland, and two in The Netherlands; Italy, Spain, and China have contributed to this review with one study each one. Of the four investigated neurological diseases, 11 studies (61.1%) included patients with a diagnosis of PD; 6 studies (33.3%) concerning patients with stroke, specifically three studies included chronic stroke patients (50%), two studies subacute-stroke patients (33,3%) and one study (16.6%) reported no data concerning the onset time; one study (5.5%) involved patients with a diagnosis of MS an and no studies including patients with TBI.
A total of 344 patients, of which there were 192 with neurological disease and presence of static and/or dynamic balance disorders, were included in the review. There were 127 participants with a clinical diagnosis of PD, 55 strokes, and 10 MS.
The Modified NOS scale was used to assess the quality of no RCTs. The NOS scale of the included studies ranged between 4 and 6 with a mean score of 5.36 points out of 7 ( Table 2). The PEDro score assessing the risk of bias of the included RCTs ranged between A total of 344 patients, of which there were 192 with neurological disease and presence of static and/or dynamic balance disorders, were included in the review. There were 127 participants with a clinical diagnosis of PD, 55 strokes, and 10 MS.
The Modified NOS scale was used to assess the quality of no RCTs. The NOS scale of the included studies ranged between 4 and 6 with a mean score of 5.36 points out of 7 ( Table 2). The PEDro score assessing the risk of bias of the included RCTs ranged between 4 and 7, showing two different levels of quality: high-quality studies (= PEDro score 6-10) and fair quality studies (= PEDro score 4-5). The mean PEDro score was 5.25 points out of 10 ( Table 3). None of the included studies reached the maximum score neither in Modified NOS nor in the PEDro scale mostly because of the participants' selection (low number and non-randomized).
The primary aim of the included studies was to evaluate the effect of VF on balance and gait in patients with static and dynamic postural impairments. The vibratory stimulus was provided as feedback to facilitate the patient's movement during static and dynamic tasks. The vibrotactile was supply both as unique augmented feedback or combined with others (haptic feedback different to the vibratory one), during different balance and gait tasks and using various devices and software. Moreover, not only one body application was used. Vibrotactile effectors were used mounted on belts or directly in contact with the patient's skin and placed in different body areas: head [35,41,42], sternum [43], waist level [44], at L4/L5 level [45][46][47], hip [48,49]; lower limb [50][51][52], anklets [32] and foot [53]. The experimental interventions lasted several days (maximum of 4 weeks) [54] or one single day and were mostly conducted in a hospital or laboratory setting. One study was carried out in a home situation [43]. Concerning the outcomes, both instrumental and clinical assessment were performed. The instrumental ones analyzed the static and dynamic parameters of postural control, while various clinical-scale tests and questionnaires were used to clinically assess balance, gait, risk of fall and the patient's self-assessment of balance disorders and satisfaction in the intervention. All the outcomes are detailed in Table 1.

VF Effects on Pitch Sway Angular Velocity
Three trials [35,41,49] were included in the analysis, with a sample size of 111. Forest plot indicates that VF resulted statistically significant in all these studies, when compared to control group. There was an overall large effect in favor of VF (SMD = −3.85 [−5.69; −2.00]); p < 0.0001; Heterogeneity is substantial and significant (I 2 = 82%; p = 0.004). The forest plot of comparison is shown in Figure 2.
The primary aim of the included studies was to evaluate the effect of VF on balance and gait in patients with static and dynamic postural impairments. The vibratory stimulus was provided as feedback to facilitate the patient's movement during static and dynamic tasks. The vibrotactile was supply both as unique augmented feedback or combined with others (haptic feedback different to the vibratory one), during different balance and gait tasks and using various devices and software. Moreover, not only one body application was used. Vibrotactile effectors were used mounted on belts or directly in contact with the patient's skin and placed in different body areas: head [35,41,42], sternum [43], waist level [44], at L4/L5 level [45][46][47], hip [48,49]; lower limb [50][51][52], anklets [32] and foot [53]. The experimental interventions lasted several days (maximum of 4 weeks) [54] or one single day and were mostly conducted in a hospital or laboratory setting. One study was carried out in a home situation [43]. Concerning the outcomes, both instrumental and clinical assessment were performed. The instrumental ones analyzed the static and dynamic parameters of postural control, while various clinical-scale tests and questionnaires were used to clinically assess balance, gait, risk of fall and the patient's self-assessment of balance disorders and satisfaction in the intervention. All the outcomes are detailed in Table  1.

VF Effects on Pitch Sway Angular Velocity
Three trials [35,41,49] were included in the analysis, with a sample size of 111. Forest plot indicates that VF resulted statistically significant in all these studies, when compared to control group. There was an overall large effect in favor of VF (SMD= −3.85 [−5.69; −2.00]); p < 0.0001; Heterogeneity is substantial and significant (I 2 = 82%; p = 0.004). The forest plot of comparison is shown in Figure 2.

VF Effects on Pitch Sway Angle
Two studies [35,41] were considered, with an overall sample of 20 subjects. The aggregate analysis shows significative effects for VF in both trials. Effect size is estimated as −1.49 (−2.21; −0.77), p < 0.0001; Heterogeneity is absent (I 2 = 0%) and not significative (p = 0.42). The forest plot of comparison is shown in Figure 3.

VF Effects on Pitch Sway Angle
Two studies [35,41] were considered, with an overall sample of 20 subjects. The aggregate analysis shows significative effects for VF in both trials. Effect size is estimated as −1.49 (−2.21; −0.77), p < 0.0001; Heterogeneity is absent (I 2 = 0%) and not significative (p = 0.42). The forest plot of comparison is shown in Figure 3.

VF Effects on Roll Sway Angular Velocity
The analysis included three trials [35,41,49] (sample size: 111), showing significant effects in favor of VF if compared to no feedback interventions. The overall effect size is estimated as large (−3.39 [−5.25; −1.54] p = 0.0003) and considerable levels of heterogeneity (I 2 = 90%, p < 0.0001) were detected. The forest plot of comparison is shown in Figure 4.

VF Effects on Roll Sway Angular Velocity
The analysis included three trials [35,41,49]

VF Effects on Roll Sway Angular Velocity
The analysis included three trials [35,41,49] (sample size: 111), showing significant effects in favor of VF if compared to no feedback interventions. The overall effect size is estimated as large (−3.39 [−5.25; −1.54] p = 0.0003) and considerable levels of heterogeneity (I 2 = 90%, p < 0.0001) were detected. The forest plot of comparison is shown in Figure 4.

VF Effects on Roll Sway Angle
Two RCTs [35,41] were included in this analysis, considering 20 subjects. Forest plot shows how both studies support the superiority of VTT in comparison to control treatments. The effect size is −1.95 (−3.66; −0.25) p = 0.02, and heterogeneity is substantial and significative (I2 = 77%; p = 0.04). The forest plot of comparison is shown in Figure 5.

VF Effects on Roll Sway Angle
Two RCTs [35,41] were included in this analysis, considering 20 subjects. Forest plot shows how both studies support the superiority of VTT in comparison to control treatments. The effect size is −1.95 (−3.66; −0.25) p = 0.02, and heterogeneity is substantial and significative (I 2 = 77%; p = 0.04). The forest plot of comparison is shown in Figure 5.

VF Effects on Roll Sway Angular Velocity
The analysis included three trials [35,41,49] (sample size: 111), showing significant effects in favor of VF if compared to no feedback interventions. The overall effect size is estimated as large (−3.39 [−5.25; −1.54] p = 0.0003) and considerable levels of heterogeneity (I 2 = 90%, p < 0.0001) were detected. The forest plot of comparison is shown in Figure 4.

Discussion
A systematic review and meta-analysis were performed to investigate the effectiveness of the VF on balance and gait rehabilitation in patients with neurological diseases. Results of the present systematic review suggest that VF programs are safe and could represent a shortterm beneficial intervention for neurological patients. However, it is difficult to generalize the results founded for the relatively few studies enrolled and for the heterogeneity of the interventions.
Protocols differ with respect to duration, required tasks, feedback localization, and type of vibrotactile devices.
In most studies, the vibrotactile vibrating motors were applied at the lower back level, probably because of the proximity with the center of mass (COM) position during quiet standing [56]. Moreover, this area provides a large, readily accessible surface with a relatively uniform shape that can be conveniently used to accommodate the vibrotactile device [57]. Lower limb application was used in five studies in order to directly influence the activity of the spinal locomotion centers during the gait [58,59]. Concerning the headmounted devices, as the authors themselves declared [35], they may be preferred over the other applications because the proximity of the cranial nerves to the cortical centers could eliminate potential errors and delays in sensory transmission and integration [60]. Despite this, the patients reported that they were more prompted to adjust the position of the head rather than the position of the whole body, as a result of this type of vibration.
VF was shown to be easily transmitted under clothing instead; no differences were found between the direct application of vibrating motors on the skin [46] or through a belt.
Due to the heterogeneity, it was not possible to detect if the application of the vibrotactile information in a determinate body area could be better than another one.
Changes in the level of displacements of the center of mass in static and dynamic position and changes during the gait were the most investigated outcomes. The instrumental assessment was carried out before the administration of the vibrotactile stimulus, during the execution of the motor tasks with VF application, and after. Few studies used clinical assessment of gait and balance.
Four RCTs [35,41,49,55] showed that the use of VF during a rehabilitation program could be an efficient method to reduce the body sway in PD e SM patients. Moreover, its effectiveness seems possible for several types of gait and balance disorders [49]. Actually, the results obtained from meta-analysis largely support the effectiveness of VF training if compared to no feedback programs. However, some issues such as the small number of included studies, several differences in study protocols (mostly assessment procedures) and considerable levels of heterogeneity invite to caution in interpreting these data. Unfortunately, there are not enough clinical evaluations to enable the assessment of VF training on the patient's quality of life and on daily living activity. However, it could be possible to speculate that the positive influence on the parameters of gait and balance shown by different studies, may also have a positive effect on these two not fully investigated aspects. In support of this, Rossi-Izquierdo et al. [48] showed that the group of patients with PD who performed the rehabilitation training supported by the VF had a significant reduction pre-and post-training in the standard balance deficit test (SBDT) composite score, associated with a significant improvement in Dizziness Handicap Inventory (DHI) and Activities-specific balance confidence scale (ABC) scores and associated with a significant reduction in the risk of falling. Moreover, Yasuda et al. [54] reported statistical significance in pre-post training analysis in Berg Balance Scale (BBS), Functional Reach Test (FRT), and in Timed-Up and Go test (TUG) in chronic hemiparetic stroke patients. Otis et al. [53] showed the use of VF as an enactive sole that uses a rhythmic vibrotactile cueing, while patients walked over different types of soil, could be useful in reducing the risk of falls in patients with PD. Another point in support of a possible positive implication on the quality of life of patients with PD of VF as supportive feedback in rehabilitation is given by Rossi et al. [32], who showed a reduction of the freezing of gate (FOG) duration.
The results showed that the VF can influence walking and gait parameters of patients with neurological diseases immediately and after a rehabilitation program lasting a few days. The lack of significant results at follow up [48,49] could lead to hypothesize that the VF has a greater efficacy by wearing it and in the short time after treatment than in the long term.
The positive effects of the VF in gait and balance parameters are reported to be greater if the vibratory stimulus is combined with other feedback as the haptic one [50][51][52] and the visual one [47].
The heterogeneity of the rehabilitation protocols and of the population does not make it possible to reach a conclusion on its effectiveness in rehabilitation but demonstrates the versatility of VF in different situations and conditions [49]. Furthermore, the heterogeneity of settings highlights the possibility to use VF by the patients independently, even in a domestic situation [45]. These characteristics make it a candidate to be considered as a supportive sensory stimulus in the context of the rehabilitation intervention focused on sensory-motor integration [55].

Strengths of the Systematic Review
To the best of the authors' knowledge, this review is the first aimed at investigating the effectiveness of the VF on balance and gait rehabilitation in patients with neurological disease. The strengths of this systematic review are: (i) to have highlighted that the VF is safe and could easily implement a standard rehabilitation program in patients affected by neurological disease; (ii) our methods were based on Preferred Reporting Items for Systematic reviews to minimize potential sources of bias; and (iii) inclusion and exclusion criteria were defined to minimize selection bias.

Study Limitations
Several limitations in the present review and meta-analysis are acknowledged. Firstly, the small number of available RCTs precludes the possibility of comparing the rehabilitative approach supported by the VF with other types of treatments. Secondly, methodological heterogeneity (e.g., study designs, outcome measures) restricted the number of studies eligible for the quantitative analysis. In addition to this, data reporting was frequently incomplete or not always provided in a useful way to perform meta-analysis; when possible, information was obtained by contacting authors via e-mail, or conversely, we properly managed them on the basis of available data.
Thirdly, the variability of the interventions does not allow to identify a single rehabilitative protocol that verifies the effectiveness. Moreover, the number of patients for each included pathology is small. The internal validity of studies is also limited, and the methodological quality is low to medium as a consequence of the study designs: lack of randomization and blinding, small or uncontrolled groups. Furthermore, a clinical and instrumental assessment of motor abilities should have been carried out to better clarify the effects on dynamic postural stability and gait parameters [61][62][63].

Conclusions
This review and meta-analysis highlights the effects of the VF provided during motor tasks, on gait and balance in patients with neurological diseases. Although the VF could be considered in the context of neurorehabilitation as a supportive sensory stimulus, not enough trials were found to establish the effectiveness of a rehabilitation program that includes the VF. Clinical practices recommendation on its efficacy cannot be made due to the lack of high-quality studies and heterogeneity of treatment protocols. Further studies focused on the evaluation of the VF effects on the quality of life and daily living changes are recommended.
Neurophysiological mechanisms linking VF intervention to enhanced balance functions should be explored after interventions to investigate possible neural mechanisms underlying the vibrotactile-induced improvements.
High-quality RCTs with cost-effective and long-term evaluations are necessary to influence clinical practice and the decision making process in neurorehabilitation.
This review may be considered as a starting point for future RCTs that could investigate the effectiveness of a vibrotactile training on balance, gait, daily life activities, and on the quality of life of patients with neurological and cerebrovascular diseases.