Introduction

Although the length of in-hospital stay following an episode of stroke has consistently decreased [1-3], many individuals return home with residual deficits. Balance and gait deficits are commonly observed in this population, leading to reduced ambulatory activity [4], limitations in activities of daily living and community participation [5,6], physical inactivity and subsequent deterioration in quality of life [7,8]. Therefore, rehabilitation efforts geared towards improving balance and mobility are important to facilitate functional independence and optimize community ambulation and participation. One of the promising intervention tools that is sought to facilitate this goal is virtual reality (VR) technology.

VR consists of a range of technologies that can be used to artificially generate sensory information in the form of a virtual environment (VE) that is interactive and perceived as similar to the real world [9,10]. Since VEs are interactive and game-like, they encourage active exploration, enhance engagement and provide motivation and enjoyment, thus allowing longer exercise sessions and improved treatment adherence [11-13]. VEs can be designed to resemble real-life scenarios including those encountered in the community [9,14]. It is not feasible to physically replicate realistic, community scenarios in the clinic or to safely train patients in the community. VR thus affords therapists with the unique opportunity to expose and train patients in these scenarios in a risk-free, graded manner, while providing intensive training and multi-sensory feedback [15,16]. These and other factors make VR-based intervention a useful adjunct or alternative to conventional therapy in re-training balance and gait dysfunctions post stroke. A review of the literature to explore the effect of VR-based interventions in re-training balance and gait and promoting independent community ambulation in this population is therefore important.

Several systematic reviews [17,18], meta-analyses [19,20] and a Cochrane review [21] have been undertaken to review the utility of VR technologies in retraining post-stroke individuals. Most of these reviews (with one exception [20]) had broader scopes of investigation and included upper limb retraining and/or cognitive rehabilitation. Further, these reviews considered only stronger study designs such as randomized controlled trials (RCT) for inclusion and thereby overlooked studies with different designs. We were, however, interested in examining the evidence on VR interventions on a targeted area (balance and gait post-stroke), with a broader and more flexible inclusion criteria as allowed in scoping reviews [22]. This allowed us to explore the added evidence to identify aspects of VR-based intervention that may prove useful in the treatment of balance and gait dysfunctions post-stroke. Further, we were interested in exploring with this scoping review, the utility of VR-based interventions in enhancing abilities required for community ambulation.

Community ambulation entails independent mobility outside the home [6] and involves dealing with environmental challenges such as low light, uneven terrain, external physical load, traffic, obstacles, time constraints etc. [23]. Various studies define diverse criteria for successful community ambulation [24]. For this review, we used one of the following criteria to identify results predictive of independent community ambulation:

1) post training gait speed ≥ 0.8 m/s, 2) functional ambulation category (FAC) of 5 (independent community ambulator) [25], 3) gait outcomes recorded in the community and, 4) outcomes related to negotiation of the environmental challenges (such as slope walking, obstacle negotiation etc.) [23].

The objectives of this scoping review were, therefore, to appraise the current state of information about the effects of VR intervention on balance and gait in post-stroke individuals and to explore the utility of VR-based interventions in facilitating independent community ambulation. The scoping review was conducted using the framework of Arksey and O’Malley [22], described in greater detail by Levac et al. [26].

Review

Search strategy

OVID MEDLINE, OVID EMBASE, PUBMED and PSYCINFO databases were searched using the terms “rehabilitation”, “virtual reality”, “stroke”, “balance”, “gait” etc. between the periods January 1950 to December 2013. The search strategies used in Ovid Medline and Pubmed databases are illustrated in Table 1. Similar strategies were employed for the other databases. Furthermore, cross references obtained from the included articles were also considered.

Table 1 Search strategies
Full size table

Selection of articles for review

1696 retrieved articles were first screened for relevance based on their titles and abstracts. Studies that were not published in the English language, available as abstracts only, that did not include post-stroke individuals or included a mixed etiology sample without separate description of outcomes related to the stroke sample, were excluded. Studies that - 1) used VR as a training tool for balance and mobility, 2) reported at least one outcome related to gait or balance and 3) published in the English language were included.

The included articles were scrutinised to extract information about VR systems, the training paradigms, outcomes and results. Twenty-four studies that met the inclusion criteria were retained for this scoping review. The subsequent sub-sections provide a synopsis of the study design, virtual environments (VE), outcome measures and the findings.

Study designs

Of the included studies, eleven were small RCTs (Level II evidence according to CEBM levels of evidence [27]) [28-38], four were controlled trials with concurrent control group (Level III evidence) [39-42], two studies had no [43] or a historical control group (Level IV evidence) [44] and six studies presented non-empirical evidence consisting of case reports, case series or proof-of-principle studies (Level V evidence) [10,45-50]. The proportion of RCTs has shown a consistent rise over the years, suggesting an increase in methodological rigor of studies in the field.

Participant characteristics

Some variability with respect to subject characteristics was observed among the included studies. Fifteen studies [28-35,38-40,42,45,47,49,50] recruited subjects with chronic stroke, seven [10,36,41,43,44,48] with sub-acute to chronic stroke, one [37] with acute stroke and one study did not provide information stroke chronicity [46]. Most studies included individuals between the ages of 50 and 80 years. Further, four studies [34-37] recruited in-patients receiving rehabilitation, twelve [10,28,30,38-40,42,45-48,50] recruited community-dwelling participants while seven studies [29,31,32,41,43,44,49] did not provide these details. A typical sample in most studies was thus middle-aged and old, community-dwelling chronic stroke survivors.

VR systems

Both custom-made and commercially available VR systems were used in studies (Table 2). Custom-made systems were usually laboratory specific and often combined other devices with the VR interface such as robotic devices [31,33,45] treadmills [10,30,34,40,43] and others. Among commercially available systems, two studies used the Interactive Rehabilitation and Exercise system, developed to comply with specific rehabilitation requirements and customization according to the patient’s needs [29,32]. Seven studies used over the counter gaming consoles such as the Nintendo Wii™, Sony Playstation™ or the Kinect™ systems [35,37-39,41,42,47]. These are designed for the general population but are being increasingly used for rehabilitation.

Table 2 VR systems and training protocols
Full size table

VR-based interventions

  1. a)

    VR tasks: The tasks used in the studies generally reflected the training objectives. For instance, studies aimed at improving balance utilized balance training tasks [32,35,37,38,41,46,47] while those aimed at improving gait utilized treadmill walking [10,28,30,36,40,43,49,50] or components of gait training such as ankle range of motion (ROM), strength [31,33,45] or appropriate activation/deactivation of ankle muscles [48].

  2. b)

    Training dosage: Most studies used training sessions lasting 40-60 min (Table 2). Some studies employed shorter (20 min) training sessions [30,34]. The training frequency varied from 2-5 times per week and total training duration lasted between 2-8 weeks. Consequently, the total VR intervention showed a wide variation ranging between 2 to 22 hours. Typical training doses comprised of sessions of 40-60 min duration, 3-5 times per week for 3-6 weeks.

  3. c)

    Feedback: Apart from the obvious intrinsic visual feedback perceived from the VE, additional intrinsic auditory, somatosensory or proprioceptive information were manipulated in some studies. For instance, Fung et al. [10] used a six-degree of freedom motion platform to simulate slopes in the VE to impart proprioceptive information congruent to walking on inclined surfaces, while Deutsch et al. [45] used haptic inputs to simulate turbulence or sensation of collision. This multisensory feedback could have acted as an important facilitator of intrinsic learning of the tasks, while enhancing engagement with the VE. Some studies also provided additional extrinsic feedback through knowledge of performance (KP) or knowledge of results (KR). Nine studies provided both KR and KP [10,29,31,32,38,43-46], three studies provided only KR [28,41,47], while others [30,34-37,40,42,48] did not provide or report on provision of KR/KP feedback. KP was provided either by the system [10,29,32,43,45,46] through graphs depicting an outcome or movement quality, or from verbal feedback (about movement quality, area of improvement etc.) by the therapists present during training [31,38,44]. KR was usually provided by the system as visual (e.g. success scores, placards) or auditory (e.g. cheering and other sounds) feedback.

Other motor learning principles such as motivation, variable practice, and attention through enhanced engagement were not addressed explicitly in most studies. [Table 2].

Outcome measures

The outcome measures utilized in the studies reflected body function and activities domain of the International Classification of Function (ICF [51]) and comprised of both laboratory measures and clinical tests. Balance assessment included center of pressure (CoP) measurements (CoP sway, sway velocity etc.) during static (quiet standing) [32,34,35,46], and dynamic postural tasks [34,36] as well as clinical tests such as the Berg Balance Score (BBS) [32,35-38,43,44,47]. Gait related outcomes were commonly measured during overground walking (reflecting transfer of VR training to overground gait) and included gait speed [28,30-32,34,36,38,39,42-44,48-50], spatiotemporal gait parameters (stride length, step length, cadence etc.) [28,31,32,34,36,49,50] and kinematic as well as kinetic gait parameters [33,48]. Clinical tests such as Timed-Up and Go (TUG) test [35,36,38,42,44,46,47], the 6-minute walk test [28,31,38,39,42,47] and Functional Ambulation Category (FAC) [29,31] were also reported.

Two studies measured ambulation in the community environment. Yang et al. [30] used the community walk test (time taken to walk 400 m in a community environment) while Mirelman et al. [31] reported on community walking activity (number of steps/day, average daily distance walked, speed, cadence etc.) using the Patient Activity Monitor. In addition, few studies reported outcomes reflecting challenges encountered during community ambulation. For e.g. Jaffe et al. [28] reported the obstacle test (the longest obstacle successfully crossed among a range of obstacle heights and lengths) and Fung et al. [10] reported on meeting time constraints, slope walking and obstacle avoidance. The details of all the outcomes reported can be found in Table 3.

Table 3 Summary of studies
Full size table

Effectiveness of VR-based interventions

Varying results were obtained in CoP measures comparing VR-based interventions with non-VR-based interventions. While no significant differences among VR-based and other interventions were found for CoP measures during static balance tasks, significant differences between interventions were noted in these measures during dynamic balance tasks. For example, Yang et al. [34] found a significant improvement in bilateral symmetry and CoP excursion under the paretic leg during the sit-to-stand task, while Kim et al. [32] reported a significant improvement in sway angles during a dynamic weight shift task only following a VR-based intervention. Similarly, for clinical measures of balance, five studies reported significant improvements in BBS scores [32,35,36,43,47] and two studies [37,47] reported an improvement on the Functional Reach Test following VR intervention as compared to other interventions.

Ten of the fourteen studies that reported gait speed found significant increases after VR-based intervention in comparison with other interventions. In addition, significant improvements in other spatiotemporal gait parameters such as cadence [32], step length [32,36,50], step time [32], stride length [36] and gait symmetry [50] as well as improvements in ankle, knee and hip ROM, and greater ankle moment and power were found following VR-based intervention [33,48] over other interventions. (Please see Table 3 for a detailed account of the results).

Further, we examined study outcomes using our criteria for independent community ambulation. Participants from four studies were able to achieve an average gait speed of ≥ 0.8 m/s following VR intervention [30,31,36,38]. Also, among participants who received VR-based intervention, 3/5 participants from You et al. [29] and 4/9 participants from Mirelman et al. [31] achieved an FAC of 5 placing them in the unlimited community ambulation category. In addition, when gait measures included community walking, Yang et al. [30] and Mirelman et al. [31] found significant increases in community ambulation time and community ambulation distance and speed respectively following VR intervention. Furthermore, Jaffe et al. [28] and Fung et al. [10] also found improvements in the ability to negotiate perturbations encountered in the community such as slopes and obstacles. An increase in gait speed of ≥ 0.8 m/s, improvement on FAC, improvement in community ambulation measures, as well as increased competency in dealing with environmental perturbations, could lead to independent community ambulation in some participants. These improvements were not seen following other interventions, suggesting an added advantage of VR-based interventions over other interventions in facilitating independent community ambulation.

One study [29] explored the effect of VR intervention on cortical re-organization post-stroke, wherein a significant shift from bilateral (pre-training) to ipsilesional (post-training) activation of the primary sensory-motor cortex during walking-like movements was found, suggesting that VR-based training can facilitate neuroplastic changes in the cortex.

Discussion

This scoping review was undertaken to appraise the impact of VR intervention on balance and gait in people post-stroke. Since our review included studies published up to December 2013, we were able to include additional papers as compared to the recent reviews [18,19]. Findings from this review indicate that VR-based interventions have an added advantage over conventional interventions in the improvement of balance while performing functional tasks, as well as in the improvement of gait speed and the quality of gait. Preliminary results also suggest that VR-based interventions may prove advantageous in promoting independent community ambulation. Some of the factors (e.g., repetitive variable practice, enhanced engagement, motivation, added feedback etc.) associated with the VR systems and the training paradigms used could be responsible for this added benefit.

VR systems

An accurate estimation of the VR system that yielded maximum gains could not be obtained from this review. Both customized and commercially available systems seemed to have equally beneficial effects over non-VR-based interventions. A recent review by Lohse et al. [19] also reported a similar view.

VR-based interventions

Intensive, task specific, variable practice in enriched environments with extrinsic (additional) feedback facilitates motor learning [52]. All of the studies included in this review used some or all of these components during VR training. However, some variability in the utilization of learning principles was found.

  1. a)

    VR tasks: The congruence of the study objectives, VR tasks and outcomes was an important factor that influenced results. For instance, outcomes related to static balance were not responsive to VR-based interventions that used dynamic balance training. Similarly, studies that used standing VR tasks failed to achieve improvements on gait-related outcomes [38,42,44]. However, tasks that trained walking-like activities or specific components essential for gait (ROM, strength etc.) did transfer to improved walking [29,31,45]. Task-specificity thus seems to be an important variable but not the only consideration in utilizing VR-based interventions.

  2. b)

    Training dosage: Considerable variability was observed in total training durations that ranged from 2 to 22 hours. Although, a higher number of repetitions and longer training times are known to have beneficial effects [52], study outcomes did not seem to depend exclusively upon the training durations. In fact, a combination of task-related training and dosage may have influenced outcomes. This was also observed by Fluet et al. [18] in their recent review.

  3. c)

    Feedback: VR-based interventions are inherently designed to provide rich visual feedback through the VEs. Further addition of auditory, haptic or proprioceptive inputs not only enhances engagement with the VE but also provides enriched environments for practice that may facilitate learning and underlying neuroplasticity [52,53]. In addition, most studies included in this review provided extrinsic feedback in the form of KP or KR. This may be especially important for learning in the post-stroke population as their intrinsic motor learning abilities may be compromised by the stroke [54]. However, the effect of this feedback on learning abilities in post-stroke individuals was not explicitly explored or reported in any study. Therefore, although it could be assumed that extrinsic feedback may have facilitated the performance, it was difficult to gauge the extent to which it may have impacted results following VR training.

In addition to the above factors, increased motivation and engagement due to game-like nature of these interventions and variable practice provided by the interactive simulations may have also contributed to the added benefits of VR observed in this review. However, even these aspects were not adequately addressed in most studies. This review, therefore, cannot comprehensively address benefits of VR-based interventions pertaining to facilitation of motor learning and dosing parameters.

VR-based interventions to promote community ambulation

As mentioned in the earlier sections, VR can be used to create scenarios simulating real-life situations including those that simulate community environments and its challenges. This provides therapists with a unique opportunity to train patients in community scenarios but in a risk-free, graded fashion. All of the studies in this review that used treadmill walking with VR utilized VEs simulating walking in the community such as in a park or a road intersection. However, only two studies used measures that assessed transfer of training to actual community ambulation, whereas most others utilized clinical measures like gait speed and FAC that could at best be considered relevant for community ambulation. Nevertheless, positive results indicating improved abilities to navigate in the community were observed suggesting that VR-based interventions could prove to be a useful tool to train independent community ambulation. Future studies should identify this unique advantage and explore utility of VR-based training in this area though use of robust study designs and appropriate outcomes.

Conclusion

Evidence from this scoping review suggests that VR-based interventions have the potential to become an effective tool in the treatment of balance and gait deficits post stroke. However, robust study designs that identify specific objectives and choose congruent and appropriate training tasks and outcome measures need to be employed in the future to ascertain appropriate intervention and dosing parameters and achieve optimal training of balance and gait in the post-stroke population.