Amy M. Dennett, BPhty(Hons) and Nicholas F. Taylor, PhD

From the Faculty of Health Sciences, La Trobe University and Allied Health Clinical Research Office, Eastern Health, Victoria, Australia

OBJECTIVE: To determine the effectiveness of computer-based electronic devices that provide feedback in improving mobility and balance and reducing falls.

DATA SOURCES: Randomized controlled trials were searched from the earliest available date to August 2013.

DATA EXTRACTION: Standardized mean differences were used to complete meta-analyses, with statistical heterogeneity being described with the I-squared statistic. The GRADE approach was used to summarize the level of evidence for each completed meta-analysis. Risk of bias for individual trials was assessed with the (Physiotherapy Evidence Database) PEDro scale.

DATA SYNTHESIS: Thirty trials were included. There was high-quality evidence that computerized devices can improve dynamic balance in people with a neurological condition compared with no therapy. There was low-to-moderate-quality evidence that computerized devices have no significant effect on mobility, falls efficacy and falls risk in community-dwelling older adults, and people with a neurological condition compared with physiotherapy.

CONCLUSION: There is high-quality evidence that computerized devices that provide feedback may be useful in improving balance in people with neurological conditions compared with no therapy, but there is a lack of evidence supporting more meaningful changes in mobility and falls risk.

Key words: falls; balance; mobility; virtual-reality; meta-ana­lysis.

J Rehabil Med 2014; 46: 00–00

Guarantor’s address: Amy Dennett, Department of Physiotherapy, Peter James Centre, Cnr. Mahoneys Road & Burwood Highway, Burwood East Victoria, 3151, Australia. E-mail: Amy.Dennett@easternhealth.org.au

Accepted Jul 7, 2014; Epub ahead of print Oct 7, 2014

Introduction

Falls are a major issue affecting older people, with one-third of people aged 65 years and over falling at least once each year (1). This has significant implications for the individual in terms of injury, reduced mobility and independence at home and in the community. The impact of falls also has a significant cost for health services, with the annual cost of falls-related injury in Australia expected to rise almost 3-fold to AUD 1.4 billion by 2051 (2).

Poor balance is an important risk factor for falls (3–5). Exercise programmes are often prescribed to improve balance, with the aim of reducing falls. Indeed, it has been shown that up to 42% of falls in older people living in the community can be prevented by well-designed exercise programmes that provide a moderate-to-high challenge to balance, that are undertaken for at least 2 hours per week (6, 7). These exercise programmes may be undertaken within a home or clinical setting (7, 8).

The use of technology through interactive gaming and virtual-reality is becoming increasingly popular with clinicians to create stimulating and challenging environments to improve balance and mobility within therapy sessions. The Nintendo Wii alone has been adapted for rehabilitation across health settings in many countries, with reports of 61% of urban stroke rehabilitation centres owning a console in Australia (9, 10). The advantage of such gaming and other virtual-reality systems is that they provide visual or other forms of feedback through a screen or a head-mounted device, which allows the user to interact with virtual environments to simulate balance and functional tasks at intensities that may be greater than traditional therapies (11–13). This may be important, given that visual and proprioceptive feedback are important components of balance control (14, 15). It is also thought that this technology may be more engaging and fun for patients as they can interact with programmes and may be less likely to focus on their health condition. In turn, this may assist with improved adherence to such therapy programmes.

Current evidence has demonstrated positive effects of virtual-reality rehabilitation in people with upper limb impairments post stroke (16, 17). Less is known about the effectiveness of computer-based technology in other populations due to lack of quality randomized controlled trials (8, 10, 18). To date, there has been no systematic reviews that we are aware of that have synthesized the available evidence on the effectiveness of computer-based interventions in terms of balance and mobility outcomes during rehabilitation.

The purpose of this systematic review was therefore to determine whether the use of computer-based technology that provides feedback is effective in improving balance, and whether any benefits carry over into improved mobility or reduced risk of falls.

Methods

Research design

A systematic review was completed in accordance with PRISMA guidelines (19, 20) and was registered prospectively through PROSPERO (CRD42013005726) (21).

Trials to be included in the review were identified though electronic database searching from the earliest available time until August 2013 through the following databases: MEDLINE, EMBASE and CINAHL. The search strategy used was based around synonyms and MeSH subject headings of the key concepts of: first, electronic devices and computer technology; secondly, balance; and thirdly, rehabilitation, combined with Boolean operators OR within each concept and the operator AND between the concepts (Appendix I). These terms were combined with the relevant filters (22) to identify randomized controlled trials and controlled trials. Studies from the EMBASE database were further limited by date (1990 to present) given the nature of the technology being studied. The search was supplemented by citation tracking and checking the reference lists of included studies.

Inclusion criteria

Title and abstracts of identified trials were assessed by 2 reviewers independently using predetermined inclusion and exclusion criteria (Table I). Full text articles were obtained and examined by both reviewers where inclusion could not clearly be determined by title and abstract alone. To be included in the review, trials were to use computer technology that provided feedback as a therapy tool such that any change could be attributed to the technology. Any device that was assistive or instructive was not included, as the investigation of the use of electronic feedback to improve balance was the primary aim of the current review.

Risk of bias assessment

Risk of bias of the included trials was assessed independently by 2 reviewers using the (Physiotherapy Evidence Database) PEDro scale. The 11 items on the PEDro scale are rated “yes” or “no”, with a maximum score of 10 achieved for internal validity as the first item is not rated. The PEDro scale is a valid measure of methodological quality of clinical trials that has undergone Rasch analysis and has moderate levels of inter-rater reliability (ICC 0.68, 95% CI 0.57, 0.76) (23, 24). Trials were deemed high quality if scores were 6 or above. Trials that scored 4 or less were deemed lesser quality (25).

Data analysis

Meta-analyses were completed using a random effects model for clinically homogeneous data using standardized mean differences for continuous data. Statistical heterogeneity was described with the I-squared statistic. Descriptive analysis was completed for data that could not be analysed with meta-analyses. Data included in the analysis were evaluated according to the population studied and comparing the intervention with no therapy or traditional physiotherapy interventions.

The GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach (26) was used to summarize the level of evidence for each completed meta-analysis. This approach involves downgrading the evidence from high to moderate to low to very low. Downgrading of the evidence would occur based on the following criteria: (1) the PEDro score was < 6 for the majority of trials in the meta-analysis, (2) there was greater than low levels of statistical heterogeneity between the trials (I2 > 25%), and (3) there were large confidence intervals. A 95% confidence interval of > 0.8 in the standardized mean difference of the meta-analysis was regarded as large, since this value could span the difference between an intervention having no effect and a large effect according to Cohen’s criteria (27). If there were serious issues with the methodological quality, such as all trials in the meta-analysis were < 6 PEDro score without allocation concealment and blinded assessors, then a double downgrade would occur (e.g. from high to low quality). Reasons for the grade applied to each meta-analysis were explained using footnotes.

Results

The search of electronic databases resulted in a yield of 3,208 articles, which was reduced to 58 after application of inclusion/exclusion criteria to title and abstracts. After further assessment of full-text articles, this was reduced to 28 trials. No further trials were identified from review of the reference lists of the already included studies. Two relevant trials were identified using citation tracking, resulting in a final library of 30 trials to be included for review (Fig. 1).

 

Fig. 1. Selection process.

Risk of bias

All of the articles included in this review were randomized controlled trials. The median score of the included articles was 5 out of 10 (mean 5.2, range 3–8, mode = 6) on the PEDro scale (Table II). Only one study (28) scored 8 on the scale, which is the highest possible score, given the nature of the intervention being studied, where it would not be expected to be feasible to blind patients or clinicians. The criterion for random allocation of participants was the only criterion met by all of the included trials. Fifteen trials reported assessor blinding (28–41). Concealment of treatment allocation occurred in only 5 trials (28, 35, 37, 39, 40) and intention to treat analyses was completed in 7 trials (28, 30, 33, 41–44).

Summary of data

The results of a total of 1240 people were included in the review. The majority of participants were women (60%) with a mean age of 61 years (SD 19 years). The main health conditions represented by participants included: acquired brain injury (6 trials (30–34, 40), multiple sclerosis (3 trials (29, 35, 36)), acute and chronic peripheral vestibular dysfunction (3 trials (42, 45, 46)), Parkinson’s disease (2 trials (41, 47)), and type 2 diabetes in 1 trial (48). Three trials investigated a healthy young adult population (49–51) and 11 trials investigated community-dwelling older adults (37–39, 43, 44, 52–57).

Interventions

All trials included in the review used computer-based technology to improve balance or mobility using electronic feedback. The most popular device was the Nintendo Wii, which was used in 15 trials (28–30, 35, 36, 38, 43, 44, 46, 49, 51–53, 55, 56). Other commercial video-game based devices were used in 4 trials (30, 37, 48, 49). Nine trials used customized electronic balance devices, where games could be individually tailored by the therapists (31, 39, 41, 42, 45, 47, 50, 54, 57). Two trials used virtual-reality based treadmill training (32, 40) and 2 trials used virtual-reality based robot training (33, 34). Most interventions were conducted in an outpatient clinic or laboratory setting, with two trials using a home-based setting (36, 37) and one completed in an acute hospital (28). The duration of each intervention varied from 10–60 min. Typically, sessions were completed 2–3 times per week for, on average, 6 weeks (range 1–20 weeks). The computerized intervention was compared with conventional physiotherapy interventions (termed physiotherapy for the purpose of this review) in 17 trials (28, 29, 31, 38, 39, 41, 44, 45, 47, 49–56). Physiotherapy interventions included traditional balance, seated exercise and home exercise programmes, which included functional mobility, static and dynamic balance, strength and flexibility exercises. Two studies included ocular mobility and optokinetic stimulation as part of the physiotherapy intervention (45, 46). Two trials compared the intervention with tai chi (55, 57). Two trials compared treadmill training with and without virtual-reality (32, 40) and 2 trials compared a robotic ankle device with and without virtual-reality (33, 34). Thirteen trials had a control group with no exercise intervention (30, 35–37, 41–44, 46, 48, 52, 54, 57) (Table II). However, 2 of these trials received education (48, 57).

Table II. Characteristics of included trials
Study PEDro Score (/10)	Subjects, n Females, % Age, mean (SD)	Health condition	Intervention	Comparison	Dosage	Setting	Outcome measures
Brumels et al., 2008 (49) 3	25 52% 19.6 (1.7)	Healthy young adults	Dance Dance Revolution Nintendo Wii Fit	Traditional balance programme	12–15 min 3 × /week 4 weeks	Laboratory	Static balance Dynamic balance	Single leg postural sway using force plate SEBT
Brichetto et al., 2013 (29) 6	36 61% 42 (11)	Multiple sclerosis	Nintendo Wii Fit	Traditional balance programme	60 min 3 × /week 4 weeks	Outpatient clinic	Static Balance Dynamic balance	Postural sway (BPEX platform) BBS
Daniel, 2012 (52) 3	23 40% 77 (5.3)	Community-dwelling older adults	Nintendo Wii Sports + weight vest	Seated exercise No therapy	45 min 3 × /week 15 weeks	Laboratory	Activity limitation Balance confidence	Senior Fitness Test: 30 s STS 30 s arm curls no. of steps in 2 min chair sit & reach 8 ft up & go 15 ft WT CHAMPS LLFDS ABC scale
Fitzgerald et al., 2010 (50) 4	28 55% 26.2 (2.7)	Healthy young adults	Exergaming using Dell Computers & Concept Wobbler	Traditional balance programme	15 min 3 × /week 4 weeks	Laboratory	Dynamic balance	SEBT DPSI
Franco et al., 2011 (53) 5	32 78% 78.27 (6)	Community dwelling older adults	Nintendo Wii Fit	Traditional balance programme	10–15 min 2 × /week 3weeks	Laboratory	Dynamic balance	BBS Tinetti Gait & Balance assessment
Fritz et al., 2013 (30) 7	30 Unknown 66.1 (9.7)	Chronic stroke	Nintendo Wii Fit & Sports PlayStation EyeToy2 & Kinetic	No therapy	50–60 min 4 × /week 5 weeks	Laboratory	Dynamic balance Activity limitation	Fugl-Meyer Assessment BBS DGI 6 min WT 3 m WT TUGT

Table II. Contd.
Study PEDro Score (/10)	Subjects, n Females, % Age, mean (SD)	Health condition	Intervention	Comparison	Dosage	Setting	Outcome measures
Gil-Gomez et al., 2011 (31) 6	17 35% 47.3 (17.8)	Chronic stroke TBI Cerebral neoplasm	eBaViR (modified WBB)	Traditional balance programme	60 min 3–5 × /week 4–6 weeks	Outpatient clinic	Dynamic balance Activity limitation	BBS BBA Anterior Reach Test Stepping Test Timed Stair Test 1 min WT 10 m WT TUGT 30 s STS
Hinman, 2002 (54) 4	88 63% 72	Community dwelling older adults	Biodex Balance System	Home exercise programme No therapy	20 min 3 × /week 4 weeks	Laboratory	Dynamic Balance Activity Limitation Falls Efficacy	BBS Timed 50 ft WT Modified FES
Jung et al., 2012 (32) 6	21 38% 62.1 (6.9)	Chronic stroke	Virtual-reality Treadmill Training (Mybud, Accupix, Gyeonggi-do)	Usual treadmill training	30 min 5 × /week 3 weeks	Laboratory	Activity limitation Balance confidence	TUGT ABC Scale
Laver et al., 2012 (28) 8	44 80% 84.9 (4.5)	Hospitalized older adults (mix of fractures, falls, medical and surgical)	Nintendo Wii Fit	Conventional physiotherapy	25 min 5 × /week duration of IP admission (usually 6 sessions ± 2)	Acute hospital	Dynamic Balance Balance Confidence Activity Limitation	Modified BBS ABC Scale TUGT FIM SPPB Timed IADL
Marioni et al., 2013 (42) 5	30 40% 45 (7)	Acute vestibular	VRS system (computerized posturography)	No therapy	30 min 1 × /week 5 weeks	Laboratory	Static Balance	Modified CTSIB LOS (computerized)
Melong & Keats, 2013 (51) 4	20 60% 19.7 (1.2)	Healthy young adults	Nintendo Wii Fit	Traditional balance programme	20min 3 × /week 4 weeks	Laboratory	Static Balance	Stabilometer
Mirelman et al., 2009 (33) 7	18 17% 61.4 (9.1)	Chronic Stroke	Rutgers Ankle Rehabilitation system	Rutgers ankle rehabilitation system without virtual-reality	60 min 3 × /week 4 weeks	Laboratory	Dynamic Balance Activity limitation	BBS Patient Activity Monitor Lower Extremity Fugl-Meyer Scale 6 min WT 7 m WT
Mirelman et al., 2010 (34) 6	18 17% 61.4 (9.1)	Chronic Stroke	Rutgers Ankle Rehabilitation system	Rutgers ankle rehabilitation system without virtual-reality	60 min 3 × /week 4 weeks	Laboratory	Activity limitation	Gait speed and kinematics (VICON)
Nilsagard et al., 2013 (35) 7	84 76% 49.7 (11.3)	Multiple sclerosis	Nintendo Wii Fit	No Therapy	30 min 2 × /week 6–7 weeks	Laboratory	Dynamic balance Activity limitation Balance confidence	FSST Dynamic Gait Index TUGT TUGT + cognitive task 25 ft WT Timed Chair Stand ABC Scale

Table II. Contd.
Study PEDro Score (/10)	Subjects, n Females, % Age, mean (SD)	Health condition	Intervention	Comparison	Dosage	Setting	Outcome measures
Pluchino et al., 2012 (55) 4	40 63% 72.5 (8.4)	Community dwelling older adults	Nintendo Wii Fit	Tai chi Traditional balance programme	60 min 2 × /week 8 weeks	Laboratory	Static balance Dynamic balance Activity limitation Balance confidence Falls efficacy Falls risk	One-Leg Stance Force plate COP Dynamic posturography Functional reach test TUGT Tinetti POMA ABC Scale FES FROP-COM
Prosperini et al., 2013 (36) 6	36 69% 36.2 (8.7)	Multiple sclerosis	Nintendo Wii Fit	No therapy	30 min daily 12 weeks	Home	Static balance Dynamic balance Activity limitation Falls frequency	Force plate COP (ProKin) FSST 25 ft WT Number of falls (reported)
Qutubuddin et al., 2007 (47) 3	15 0% 72.8 (6.8)	Parkinson’s Disease	Smart BalanceMaster	Conventional physiotherapy	30 min 2 × /week 4 weeks	Outpatient clinic	Static balance Dynamic balance	SOT Adaption test LOS (BalanceMaster) BBS
Rendon et al., 2012 (43) 6	40 39% 84.4 (5.2)	Community dwelling older adults	Nintendo Wii Fit	No therapy	35–45 min 3 × /week 6 weeks	Outpatient clinic	Activity limitation Balance confidence	8 ft Up & Go ABC scale
Rossi-Izquierdo et al., 2011 (45) 4	24 67% 51.7	Chronic Vestibular	Smart BalanceMaster	Optokinetic stimulation	15–20 min daily 5 days	Outpatient clinic	Static balance	SOT LOS (BalanceMaster)
Schoene et al., 2013 (37) 7	37 unknown 78 (5)	Community dwelling older adults	Dance Dance Revolution Stepmania	No therapy	15–20 min 2–3 × /week 8 weeks	Home	Dynamic balance Activity limitation Falls efficacy Falls risk	Choice Stepping Reaction Time Alternate Step Test INHIB stepping task TUGT TUGT Cognitive 5 × Sit to Stand FES PPA
Singh et al., 2012 (38) 6	38 100% 62.6 (4.8)	Community dwelling older adults	Nintendo Wii Fit	Traditional balance programme	40 min 2 × /week 6 weeks	Laboratory	Balance confidence Falls risk	ABC scale FallsScreen PPA
Singh et al., 2013 (56) 5	38 100% 62.6 (4.8)	Community dwelling older adults	Nintendo Wii Fit	Traditional balance programme	30 min 2 × /week 6 weeks	Laboratory	Static balance Dynamic balance Activity limitation	Postural sway (ProBalance) Ten Step Test TUGT
Sparrer et al., 2013 (46) 3	71 62% 43.5 (3.5)	Acute vestibular	Nintendo Wii Fit	No therapy	45 min daily 5 days	Laboratory	Static balance	SOT (EquiTest) One-leg stance (Wii)

Adverse events and adherence

Three trials reported that there were no adverse events attributed to the use of the computerized intervention (35, 42, 46). One trial (36) of the 30, reported adverse events attributed to the use of the intervention. In this trial, 3 participants developed low-back pain and 2 participants developed knee pain from the use of the computer-based intervention. The other 26 trials failed to report whether any adverse events occurred.

Four trials reported increased levels of enjoyment and motivation when using the computerized intervention (31, 49–51). However, 3 trials reported no significant difference in compliance with the computerized intervention compared with physiotherapy (28, 47, 51).

Effect of computer-based intervention vs physiotherapy in people without a primary health condition

For older community-dwelling people, meta-analysis of 4 trials with 132 participants provided low-quality evidence that there was no difference in dynamic balance (Fig. 2a) compared with physiotherapy. Meta-analysis of 5 trials with 153 participants provided low-quality evidence that there was no difference in falls efficacy or balance confidence (Fig. 2b). Furthermore, 3 trials with 79 participants demonstrated low-quality evidence that there was no difference in falls risk (Fig. 2c).

 

Fig. 2. Effect of computer-based intervention vs physiotherapy for people without a primary health condition. (a) Standard mean difference (SMD) (95% confidence interval (95% CI) of effect of computer-based intervention vs physiotherapy on dynamic balance for community-dwelling elders. GRADE = Low. GRADE working group grades of evidence (see Reason for downgrade). Reason for downgrade: all trials rated lesser quality (PEDro < 6) without blinded outcome measures and allocation concealment. (b) SMD (95% CI) of effect of computer-based intervention vs physiotherapy on falls efficacy and balance confidence for community-dwelling elders. GRADE = Low. GRADE working group grades of evidence (see Reason for downgrade). Reason for downgrade: 3 trials (51, 53, 54) rated lesser quality (PEDro < 6) without blinded outcome measures and allocation concealment. (c) SMD (95% CI) of effect of computer-based intervention vs physiotherapy on falls risk for community-dwelling elders. GRADE = Low. GRADE working group grades of evidence (see Reason for downgrade) Reason for downgrade: statistical heterogeneity (I2 = 48%), very large confidence intervals.

Trials on community-dwelling older adults that could not be included in the meta-analysis due to clinical heterogeneity also found no difference in static or dynamic balance, balance confidence or falls efficacy, or mobility compared with physiotherapy.

The 3 trials investigating use of computer-based balance training in young healthy adults could not be included in meta-analysis as there was insufficient data. Two of these studies did not show any significant differences in postural stability compared with traditional therapies (50, 51). One trial (49) did demonstrate significant improvement in reduction of postural sway compared with physiotherapy.

Effect of computer-based intervention vs physiotherapy in people with a primary health condition

Meta-analysis of 2 trials on 41 participants with a neurological condition demonstrated low-quality evidence that there was no difference for balance confidence (Fig. 3a) compared with physiotherapy. Meta-analysis of 4 trials with 76 participants demonstrated moderate-quality evidence that there was no difference for mobility (Fig. 3b) compared with physiotherapy.

Individual trials on participants with a neurological population that could not be combined in meta-analysis due to heterogeneity found no difference in static or dynamic balance.

One high-quality study investigating computer-based balance training on a sample of people with a variety of medical conditions on an acute hospital ward (28) also did not find any differences in mobility, dynamic balance or balance confidence after 6 sessions of therapy for the duration of their inpatient admission compared with physiotherapy.

 

Fig. 3. Effect of computer-based intervention vs physiotherapy for people with a health condition. (a) SMD (95% confidence interval (95% CI)) of effect of computer-based intervention vs physiotherapy on balance confidence for people with a neurological condition. GRADE = Low. GRADE working group grades of evidence (see Reason for downgrade). Reason for downgrade: statistical heterogeneity (I2 = 50%), very large confidence intervals. (b) SMD (95% CI) of effect of computer-based intervention vs physiotherapy on mobility for people with a neurological condition. GRADE = Moderate. GRADE working group grades of evidence (see Reason for downgrade). Reason for downgrade: very large confidence intervals.

Effect of computer-based intervention vs no therapy in people without a primary health condition

Meta-analysis of 4 trials with 130 community-dwelling older adults demonstrated low-quality evidence that there were no differences for dynamic balance (Fig. 4a) compared with no therapy. Meta-analysis of 4 trials with 139 participants also demonstrated low-quality evidence that there were no differences for falls efficacy and balance confidence (Fig. 4b) compared with no therapy. Meta-analysis of 4 trials with 183 participants also demonstrated low-quality evidence that there was no difference for mobility (Fig. 4c) compared with no therapy.

 

Fig. 4. Effect of computer-based intervention vs no therapy for people without a primary health condition. (a) SMD (95% confidence interval (95% CI)) of effect of computer-based intervention vs no therapy on dynamic balance for community-dwelling elders. GRADE = Low. GRADE working group grades of evidence (see Reason for downgrade). Reason for downgrade: 3 studies (43, 52, 53) rated lesser quality (PEDro < 6), without blinded outcome measures or allocation concealment. (b) SMD (95% CI) of effect of computer-based intervention vs no therapy on falls efficacy and balance confidence for community-dwelling elders. GRADE = Low. GRADE working group grades of evidence (see Reason for downgrade). Reason for downgrade: 2 studies (51, 53) rated lesser quality (PEDro < 6), without blinded outcome measures or allocation concealment. (c) SMD (95% CI) of effect of computer-based intervention vs no therapy on mobility for community-dwelling elders. GRADE = Low. GRADE working group grades of evidence (see Reason for downgrade). Reason for downgrade: 2 trials (52, 56) rated lesser quality (PEDro < 6), without blinded outcome measures or allocation concealment.

Individual trials that could not be included in the meta-analysis due to clinical heterogeneity found no difference in static or dynamic balance, or falls risk compared with no therapy.

No included trials compared the computer-based intervention with no therapy in young healthy adults.

Effect of computer-based intervention vs no therapy in people with a primary health condition

Meta-analysis of 4 trials with 213 participants with a neurological condition demonstrated high-quality evidence for improved dynamic balance immediately post intervention (Fig. 5a) compared with no therapy. Meta-analysis of 3 trials with 146 participants demonstrated high-quality evidence that there was no significant difference in mobility (Fig. 5b) compared with no therapy.

 

Fig. 5. Effect of computer-based intervention vs no therapy for people with a neurological condition. (a) SMD (95% CI) of effect of computer-based intervention vs no therapy on dynamic balance for people with a neurological condition. GRADE = High. GRADE working group grades of evidence. (b) SMD (95% CI) of effect of computer-based intervention vs no therapy on mobility for people with a neurological condition. GRADE = High. GRADE working group grades of evidence.

Individual trials that could not be included in meta-analysis due to heterogeneity demonstrated no difference in static balance or balance confidence in participants with acquired brain injury and multiple sclerosis (35, 36, 40) One trial on a vestibular population showed moderate effect in favour of the computer-based intervention for single-leg stance immediately post intervention, which was not sustained at 10 weeks (46). Another trial with a vestibular population demonstrated large effect in favour of the computer-based intervention for standing balance on foam with eyes open and closed immediately post and at 6 weeks post intervention (42).

One low-quality study investigating the computer-based intervention on a sample of older people with type 2 diabetes demonstrated a large effect in favour of the intervention for reducing postural sway after 10 weeks of training (48).

Discussion

The results of this systematic review of 30 randomized controlled trials provides high-quality evidence that computerized balance interventions that provide feedback improve dynamic balance in people with a neurological condition, and low-quality evidence that these devices improve static balance in people with vestibular conditions compared with no therapy. However, there was low-quality evidence that computerized balance interventions resulted in no difference in falls risk, falls and balance confidence or mobility in people who have a neurological condition compared with no therapy. In community-dwelling older adults and healthy young adults, there was low-quality evidence that computerized balance interventions resulted in no difference in balance, mobility or reduce falls risk compared with no intervention. Furthermore, compared with physiotherapy interventions used to improve balance, such as balance exercise programmes and tai chi, there was low-to-moderate-quality evidence that computerized balance interventions resulted in no difference in improving balance, falls or mobility in people regardless of age or whether a health condition was present.

Overall, these results question the usefulness of the use of computerized feedback interventions in training balance for people at risk of falling. One explanation for the lack of effect may be because of the uni-dimensional nature of the intervention. Exercise prescription, particularly in the complex area of falls and balance rehabilitation, needs to be multi-faceted and specific to the needs of individuals. Balance is a multi-system function that relies on the interaction of tactile, proprioceptive, visual and vestibular feedback in order to maintain the body’s centre of mass within the limits of our base of support during functional tasks (14, 15). It also relies on other factors, such as strength, pain, cognition and falls efficacy, to allow safe mobility (58, 59). Many of the devices included in this review predominantly focus on a single component of standing balance; the centre of pressure (18).

The results of this review may also be explained by the principle of specificity of practice; that is, the idea that you gain what you train (60, 61). Static and dynamic balance were the only outcomes that demonstrated improvement in our meta-analyses. Regardless of the type of device, their main objective is to improve balance. They do not necessarily provide a functional environment to improve other aspects of balance.

One interpretation of the results of this review is that it may not be the type of electronic device that is important, but more so the content of rehabilitation programmes that is most valuable in improving mobility and preventing falls. It appears that in this evolving world of technology, there may be too much focus on the marketing of these products for rehabilitation, with clinicians getting caught up in their popularity and not considering the evidence base for applying their use for therapy sessions. It is possible that getting back to basics and using well-designed exercise programmes and physiotherapy interventions may be more clinically effective at improving balance, mobility and reducing falls than relying on uni-dimensional computer-based interventions.

Previous research has demonstrated that, in order for exercise to be effective at reducing falls risk, it must be sufficiently challenging to balance and of high intensity (7). At least 50 hours of balance training is required to reduce falls, which equates to a twice weekly programme over 25 weeks (62). Of the trials included in this review, only 1 of 30 trials achieved this recommendation (36). The importance of exercise intensity using virtual-reality has been reported elsewhere (16). It is also possible that the interventions included in this review did not provide sufficient challenge to balance or mobility, particularly as many of the devices are designed predominantly for entertainment purposes and not for the purpose of rehabilitation (10). Furthermore, physiotherapy interventions are often multi-faceted and include falls education and environment modification, which are also important aspects for reducing falls risk (6, 63).

Our primary analysis did not consider the importance of novelty and enjoyment of using these computerized balance devices. Often balance training can become repetitive and mundane, but using these electronic devices provides a new and potentially exciting modality for therapy. It has been reported that in older, healthy populations found using computerized training was easy and enjoyable (54, 64). Three other studies on a young, university population also found higher levels of enjoyment in those who used the gaming device and subsequently found improved levels of adherence (49–51). This has important implications for adherence of balance exercise programmes in order to achieve the exercise dosage required for positive balance outcomes and maintaining these changes in clinical practice. A recent systematic review of the use of home-based virtual-reality and gaming systems demonstrated strong retention and adherence to these programmes (65). Despite this, the 3 trials in our review reporting adherence data reported no significant difference with the computerized intervention compared with physiotherapy (28, 47, 51).

The cost of purchasing these devices should also be considered when prescribing exercise for rehabilitation purposes. There is potential that their use may allow improved independence with therapy programmes and reduce need for therapist supervision, particularly in low-risk clients living in the community, as commercial gaming devices become more affordable and readily available. However, given the current lack of evidence to support the clinical advantages in using of computer-based devices for improving health outcomes compared with physiotherapy, a health economic analysis may be required.

Our findings are similar to that of previous reviews that focussed on a single intervention or single population (10, 16, 18). Our review added to the previous literature by including all computerized interventions that use feedback to potentially improve balance and did not limit the target population. Given the increasing frequency of use and development of this new technology, we were able to include 16 trials that have been published since 2011. This allowed us to conduct meta-analyses for our outcomes.

Our review followed PRISMA Guidelines (19, 20), which is the preferred reporting method for systematic reviews and meta-analysis. The GRADE approach (26) was also applied to the meta-analyses to help determine the level of confidence in the results for clinicians and policymakers. All of the studies included in the review were randomized controlled trials, which reduces risk of selection bias in increases confidence in our results.

Limitations

Limitations of this review include our search strategy, which did not include languages other than English and did not include all available databases. However, our search strategy was thorough as only 2 trials were identified with citation tracking. Another limitation is the quality of the available evidence. Much of the evidence in the meta-analyses was downgraded to low or moderate, due in part to the low-to-moderate quality of the individual trials in this review, as judged on the PEDro scale, and small sample sizes of the included trials contributing to relatively large confidence intervals around the estimated standardized mean differences. It should also be considered whether heterogeneity within subgroups of the meta-analysis as well as a variety of outcome measures that were used affected the results of the meta-analyses. For example, patients with a variety of neurological diagnoses were combined in the meta-analyses. However, the findings of individual trials in our review were consistent with the results of the meta-analyses.

Conclusion

There is high-quality evidence from 4 trials that computer-based interventions that provide feedback improve dynamic balance in people with a neurological condition compared with no therapy. There is low-quality evidence from 3 trials that computer interventions improve postural sway in people with a vestibular condition and in people with type 2 diabetes. However, there is high- to low-quality evidence that demonstrates no significant benefits for mobility, falls and balance efficacy and falls risk when using computerized interventions that provide feedback in young or community-dwelling older people, or people with other health conditions. There is low-to-moderate-quality evidence that demonstrates no significant benefits for the use of computerized interventions over traditional physiotherapy. Given the complex and multi-factorial nature of balance and falls, these results suggest that physiotherapy interventions should not be overlooked when considering the use of this emerging technology.

The authors declare no conflicts of interest.

References