Impact of Vestibular Rehabilitation and Dual‐Task Training on Balance and Gait in Survivors of Stroke: A Systematic Review and Meta‐Analysis

Abstract

Background

Evidence supports that vestibular rehabilitation therapy (VRT) improves the static and dynamic balance of survivors of stroke, yet VRT is rarely included in stroke rehabilitation guidelines. We aim to answer the question: What are the effects of VRT or dual‐task training (DTT) on balance and gait for reducing the risk of falls among survivors of late subacute and chronic stroke?

Methods

Following the Preferred Reporting Items for Systematic Reviews and Meta‐Analysis Statement guidelines, electronic databases PubMed, EMBASE, MEDLINE, Web of Science, and Scopus for English were searched to identify randomized controlled trials published within the past 10 years in the English language, investigating VRT for patients with late subacute and chronic stroke.

Results

Eleven studies (n=509 participants) were included in the systematic review, and 10 studies (n=413 participants) were included in a meta‐analysis. The average participant age was 60.9 years, with 62.11% male. On average, 36 months had passed since stroke onset. The pooled effect standardized mean difference suggests that VRT has a significantly large effect for improving balance (standardized mean difference, 0.64 [95% CI, 0.44–0.85], P<0.00001), particularly from balance‐specific training (standardized mean difference, 1.07 [95% CI, 0.70–1.45], P=0.002). Dual‐task training (DTT) moderately improved gait (standardized mean difference, 0.46 [95% CI, 0.18–0.74], P=0.001), with greater benefits from DTT compared with single‐task training.

Conclusions

Despite substantial heterogeneity across studies, the evidence supports that VRT, can probably improve balance, and DTT may improve gait outcomes among survivors of late subacute and chronic stroke. An optimal program for this population should focus on balance and DTT with subcomponents of gait and strength training. Further research is required to determine the optimal number of weeks, sessions/week, and duration (minutes) of VRT sessions.

Registration

URL:  https://www.crd.york.ac.uk/PROSPERO/; Unique identifier: CRD42023450254.

Nonstandard Abbreviations and Acronyms

BBS

Berg Balance Scale

CMI

cognitive–motor interference

DT

dual‐task

DTT

Dual‐task training

MCID

minimal clinically important difference

SMD

standardized mean difference

TUG

Timed Up and Go test

Research Perspective.

What Is New?

• This systematic review and meta‐analysis evaluates the effectiveness of vestibular rehabilitation therapy and dual‐task training in improving balance and gait outcomes in survivors of stroke, providing the first pooled estimate specifically targeting populations with late subacute and chronic stroke.

• The findings support the integration of vestibular rehabilitation therapy and dual‐task training into poststroke rehabilitation programs, emphasizing their potential to enhance dynamic balance and reduce fall risk in patients with stroke, thereby informing future clinical guidelines.

What Question Should be Addressed Next?

• Future research should explore the optimal dosing, duration, and delivery mode (in person versus telerehabilitation) of vestibular rehabilitation therapy and dual‐task training interventions to maximize their impact on functional recovery in stroke rehabilitation.

Stroke is the most common cause of adult disability in the developed world, with two thirds of survivors experiencing residual degrees of disability, ¹ , ² , ³ costing the UK health care system £25.6 billion per year. ⁴ As a global burden of disease, stroke prevalence is projected to increase 27% by 2047 across the European Union. ⁵

Balance and gait are essential components of functional movement, yet balance and mobility problems are among the most frequent and disabling effects of stroke, with 85% of survivors of stroke suffering from balance impairments. ⁶ , ⁷ Impaired balance and dizziness have consistently been identified as risks factors for falls, which has a negative influence on quality of life and wide‐ranging physical, psychosocial, and health care‐related consequence. ⁸ Balance problems among survivors of chronic stroke are multifactorial, especially during performance of complex tasks, when falls are most likely. ⁹ Balance is the ability to maintain the line of gravity within the base of support, which is complex, relying on integration of vestibular, visual, and somatosensory inputs to the central nervous system. ⁶ , ⁸

Previous studies have indicated that vestibular rehabilitation therapy (VRT) including balance physiotherapy, improves dynamic balance of survivors of stroke. ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ VRT consists of exercises that aim to improve balance, gait, somatosensory integration, and gaze stability through compensation, adaptation, habituation, or substitution of the vestibular ocular reflex or vestibulospinal reflex. ¹⁵ , ¹⁶ , ¹⁷ Multisensory exercises, consisting of activities simultaneously stimulating at least 2 sensory modalities, including visual, auditory, tactile, vestibular, and somatosensory systems, are effective for improving balance in people with balance disorders, including stroke. ¹¹ However, studies concerning the effect of multisensory VRT on balance among survivors of stroke is limited ¹⁴ and further research is needed to investigate the optimal strategy of multisensory VRT for stroke rehabilitation. ¹¹

Recent evidence from comprehensive systematic reviews and meta‐analyses have demonstrated a beneficial effect of VRT on gait and balance performance among survivors of stroke. ¹³ , ¹⁸ Despite this evidence, VRT is rarely included in stroke rehabilitation guidelines. ¹⁸ , ¹⁹

Cognitive–motor interference (CMI) is measured by dual‐tasking, which refers to the phenomenon in which performance deterioration of a cognitive or motor task occurs when both are performed simultaneously. ²⁰ , ²¹ Studies investigating dual‐task training (DTT) among survivors of stroke have shown promising results for improving gait performance and reducing CMI during functional mobility. ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ DTT involves the simultaneous execution of cognitive and motor tasks, which can be further categorized into dual‐motor task training and cognitive–motor DTT. Dual‐motor task training focuses on performing 2 motor tasks concurrently, whereas cognitive–motor DTT involves combining cognitive tasks with motor tasks. ²⁷ , ²⁸ , ²⁹ The distinction between these 2 types of training is essential as they target different aspects of cognitive–motor interactions and can have varying effects on gait performance and balance in survivors of stroke. ³⁰ , ³¹

Although some studies have demonstrated the positive effects of DTT on gait performance and balance abilities in patients with stroke, ²¹ , ²² , ²³ , ²⁶ , ³² , ³³ there is still limited research on the effects of DTT on balance control and fall risk reduction in this population. ²⁵ Improving gait ability in DT conditions among survivors of stroke is crucial, and more research is needed to better understand both the motor and cognitive effects of various DTT methods on balance and gait among patients with stroke. ³⁴

To our knowledge, this is the first systematic review to investigate VRT, including DTT effects on balance and gait in survivors of late subacute and chronic stroke. A recent systematic review and meta‐analysis investigated the effects of VRT on balance and gait in survivors of stroke ¹⁸ but included participants at any time post onset (ie, acute, subacute, chronic), did not consider DTT, and had no language or publication date restrictions. Through a comprehensive systematic review and meta‐analysis of randomized control trials (RCT), the aim is to address the lack of knowledge surrounding multisensory VRT and DTT for survivors of late subacute and chronic stroke with balance problems.

METHODS

The systematic review and meta‐analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta‐Analysis statement ³⁵ and was registered in the International Prospective Register of Systematic Reviews. The methods are briefly described here with more details provided in the Data S1 (Supplementary Material). Data not published within the article are available from the corresponding author on reasonable request. No ethical approval was required as per the Health Research Authority of England. ³⁶

Search Strategy

Electronic databases including PubMed, EMBASE (Ovid), MEDLINE (Ovid), Web of Science, and Scopus for English were searched to identify published RCTs of VRT for survivors of late subacute and chronic stroke. We limited our search to patients with late subacute and chronic stroke to reduce confounding factors commonly seen in earlier time points of acute and early subacute when more rapid and significant changes in brain function and physical abilities occur. By focusing on late subacute and chronic stages, we aimed to achieve a more stable and homogeneous patient population, enhancing the reliability of our findings. The Medline search was adapted to the syntax for EMBASE, Web of Science, and Scopus for English and is presented in Data S1 (Supplementary Material.) PubMed was used for exploratory searching only. English and German original language RCTs not older than 10 years (2013–2023) were searched, to ensure inclusion of the most recent and relevant studies. The searches were rerun on December 1, 2023 just before the final analyses with no further studies retrieved for inclusion.

Eligibility Criteria

Participants

Participant inclusion criteria were (1) adults >18 years of age, (2) diagnosis of stroke with onset >3 months (late subacute or chronic stroke) before study, (3) community dwelling, and (4) able to ambulate independently with or without a stick.

Exclusion criteria included participants with acute stroke (<3 months since onset) or an inpatient. Studies were excluded if time since stroke onset was not stated or could not be provided by authors.

Intervention

RCTs evaluating VRT or balance exercises with a primary aim of improving balance or gait or reducing fall risk were included. For this review, VRT studies aiming to improve balance or gait as the primary intervention, with or without other components including DTT were included.

Comparator

Control groups receiving usual care or conventional physical therapy or no intervention.

Outcome Measures

Studies with a primary outcome measure assessing static balance or gait or dynamic balance were included.

Screening Process and Data Extraction

Three reviewers (B.N., D.E.B., N.K.) independently reviewed studies using Rayyan. After removing duplicates, the initial screening process involved evaluating the article's title and abstract summary against the eligibility criteria. Articles that passed initial screening were examined in their entirety to determine their eligibility for inclusion in the final review. When a decision could not be reached between the 3 reviewers, a fourth reviewer (M.P.) was consulted to resolve any discrepancies.

One reviewer (B.N.) independently extracted data into predesigned Microsoft Excel data collection sheets. Any disagreement in the data extracted from studies was resolved through discussion between reviewers.

Critical Appraisal and Quality Assessment

The methodological quality of RCTs was evaluated using the Risk of Bias 2 tools as recommended in the Cochrane Handbook for Systematic Reviews of Interventions ³⁷ and has been suggested as the optimal approach for assessing risk of bias in physical therapy trials. ³⁸

The Grading of Recommendations, Assessment, Development, and Evaluation approach was used for rating the certainty of evidence (Table S1). ³⁹

Statistical Analysis

Statistical analysis was performed using Review Manager (RevMan), version 5.4. Data synthesis was conducted when at least 2 comparable trials were available. Postintervention means±SD were used to estimate the pooled effect. The baseline to postintervention change score was used to estimate the pooled effect for Berg Balance Scale (BBS) subgroup analysis; there were insufficient data to analyze change scores for the other outcome measures.

Continuous outcomes were analyzed using the inverse‐variance method under a fixed‐effects model, which assumes no between‐study variance and provides valid inference of the precision‐weighted average effect. ⁴⁰ , ⁴¹ For outcomes measured on different scales, results were pooled using standardized mean differences (SMD) with 95% CIs. For outcomes measured on the same scale, MDs were used. ⁴¹ , ⁴² SMDs were interpreted using Cohen's criteria: values <0.4, 0.4 to 0.7, or >0.7 representing a small, moderate, or large effect size, respectively. ⁴² A P value <0.05 was considered statistically significant if the 95% CI did not cross 0.

For missing data, ⁴² means±SD were estimated using ⁴¹ the skewness and mean variance methods ⁴³ proposed by Shi et al. (2023) ⁴⁴ , Wan et al. (2014), ⁴⁵ and Luo et al. (2018). ⁴⁶ Studies with incomplete statistical reporting or significantly skewed data ⁴⁴ where the mean±SD was either not available, could not be extracted, or could not be reliably estimated were excluded from the meta‐analysis ⁴⁴ . Studies with >1 intervention group were represented as individual studies in the meta‐analysis.

Heterogeneity was assessed using the I ² statistic of the chi‐square (X ² ) test, which quantifies the proportion of total variation due to between‐study differences but does not affect study weighting in a fixed‐effects model. To complement this, we report the range of SMDs across studies to assess variability in effect size. An I ² value of <40% is considered as low (“might not be important”) heterogeneity, whereas values of 40% to 60%, 60 to 80%, or >80% represent moderate, substantial, or considerable heterogeneity, respectively. If heterogeneity was considered as moderate to substantial, subgroup analysis was applied.

RESULTS

Results of Search/Study Selection

The initial search yielded 11 819 potentially relevant studies, with retrieval of 207 full‐text articles of which 11 met the criteria for inclusion in the qualitative systematic review and 10 studies met the criteria for the quantitative meta‐analysis as outlined in Figure 1.

Figure 1. Preferred Reporting Items for Systematic Reviews and Meta‐Analyses ³⁵ flow chart of identification and selection of studies; University College London.

Study Characteristics

All included studies were RCTs investigating the effects of VRT or balance rehabilitation on balance outcomes and fall risk within survivors of late subacute (>3 months) and chronic (>6 months) stroke. For additional details please see Table 1, which describes the included study characteristics and Data S2 (Supplementary Material).

Table 1.

Study Characteristics of Included Studies

Author (y)	Country	Study objectives	Participant inclusion criteria	Control arm	Intervention arm	Outcomes used
Correia et al. (2021)	Portugal	To assess the effect of a domiciliary program of oculomotor and gaze stability exercises on the incidence of falls and risk of fall in survivors of stroke.	68 patients with chronic stroke	Current institutional rehabilitation program for stroke84,85	Current program + supplementary domiciliary intervention program	BBS, Rhomberg Test, TUG, no. falls
Kim et al. (2013)	South Korea	Examine the effects of consecutive training in dual motor tasks on the gait ability 51,86of patients with chronic stroke.	29 patients with chronic stroke	Single motor task training	Dual motor task training	GAITRight, CM‐DT
Liu et al. (2017)	Taiwan	Assess the effects of cognitive and motor dual task gait training on dual task gait performance in patients with stroke.	28 patients with subacute or chronic stroke	Conventional physical therapy	Cognitive DT gait training, motor DT gait training	GAITRight, CM‐DT
Mansfield et al. (2018)	Canada	Determine if perturbation‐based balance training can reduce falls in daily life among individuals with chronic stroke.	88 patients with chronic stroke	Traditional balance training	Perturbation‐based balance training	BBS, Mini‐Balance Evaluation System Test, TUG, no. falls, ABC
Pang et al. (2018)	Hong Kong	Evaluate the effects of dual‐task balance/mobility exercise program on dual‐task interference during walking, fall incidence, balance self‐efficacy, participation in daily activities, and quality of life in individuals with chronic stroke.	84 patients with chronic stroke	Upper‐limb exercises	DT balance/mobility training group	TUG, CM‐DT, ABC, SS‐QOL, Frenchay Activities Index
Park et al. (2016)	Republic of South Korea	Determine whether multidirectional stepping training improves balance, gait ability, and falls efficacy in patients with stroke.	50 patients with chronic stroke	General physical therapy	Combined stepping exercise	BBS, TUG, 10MWT, activities‐specific balance confidence scale
Vahlberg et al. (2017)	Sweden	Evaluate the effects of progressive resistance and balance exercises on physical and psychological functions of individuals post stroke.	67 patients with chronic stroke	Regular activities as usual	Progressive resistance and balance exercises + motivational group discussions	BBS, 10MWT, 6MWT, MAS/modified MAS, Short Physical Performance Battery, EuroQol‐5D, FESf, Physical Activities Scale for Eldery, Geriatric Depression Scale‐20, Charlson Comorbidity Index
Qurat‐ul‐Ain et al. (2018)	Pakistan	Determine the effect of circuit gait training versus traditional gait training on mobility performance and quality of life in patients with subacute and chronic stroke	32 patients with subacute and chronic stroke	Standard treatment	Circuit training	BBS, FESf, SS‐QOL
Choi et al. (2015)	Republic of Korea	Determine whether a participant selected task‐oriented training program improved balance, activities of daily living performance, and self‐efficacy in patients with stroke.	20 patients with chronic stroke	Traditional rehabilitation therapy.	Task‐oriented training program	BBS, SES, Modified Barthel Index
Park & Lee (2019)	Republic of South Korea	Investigate the effectiveness of DT training using various cognitive tasks for the assessment of attention, executive function, and motor function in patients with stroke.	30 patients with chronic stroke	Conventional occupational therapy	DT training with different cognitive tests	BBS, Fugl‐Meyer Lower Extremity Motor Assessment Scale, modified functional reach, CM‐DT
Jarbandhan et al. (2021)	Belgium	To assess the feasibility and preliminary effectiveness of a home‐based semisupervised physiotherapy intervention to promote poststroke mobility in a low resource setting.	30 patients with chronic stroke	Usual care	Home‐based physiotherapy program (walking, functional and mobilization exercises) supervised in first 4 weeks and telesupervised in the past 4 weeks	BBS, 6MWT, SES

6MWT indicates 6‐minute walk test; 10MWT, 10‐meter walk test; ABC, Activities‐Specific Balance Confidence Scale; BBS, Berg Balance Scale; CM‐DT, cognitive–motor dual task; DT, dual task; FESf, Falls Efficacy Scale ‐ full version; MAS, Motor Assessment Scale; SES, Self‐Efficacy Scale; SS‐QOL, Stroke‐Specific Quality of Life Scale; and TUG, Timed Up and Go test.

Three (27%) of these studies ⁴⁷ , ⁴⁸ , ⁴⁹ investigated the effect of VRT on dynamic balance and gait outcomes, 3 (27%) studies ²⁰ , ⁵⁰ , ⁵¹ looked at dynamic balance outcomes alone, 1 (9%) study ⁵² looked at dynamic balance and falls outcomes, 2 (18%) studies ²¹ , ²² investigated gait parameters during DT training, and 1 (9%) study investigated functional gait and gait parameters during DT training. ⁵³

Participant Characteristics

This systematic review included 509 participants across the 11 studies included in. Participants had a mean age of 60.9 (range 50–87), with an average 62.11% male. Mean time since stroke across the 7/11 studies that reported it was 36 months (SD 34.5 months). Please see Table S2 (Supplementary Materials) for additional details.

Intervention Characteristics

A range of interventions were examined, with all studies including 2 or more forms of VRT (Table 2). Studies were classified according to the type of VRT: (1) gaze stabilization exercises or oculomotor or eye‐head movements, (2) balance specific exercises, (3) strength or functional exercises, (4) gait‐specific exercises, and (5) DTT.

Table 2.

Intervention Summary of Included Studies

Study reference	Gaze stabilization +/− oculomotor	Strength/functional exercises	Balance training	Gait training	Cognitive‐motor/gait dual‐tasking
Correia et al. (2021)	X
Kim et al. (2013)					X
Liu et al. (2017)					X
Mansfield et al. (2018)			X
Pang et al. (2018)					X
Park et al. (2016)		X	X
Vahlberg et al. (2017)		X	X
Qurat‐ul‐Ain et al. (2018)			X	X
Choi et al. (2015)		X		X
Park & Lee (2019)					X
Jarbandhan et al. (2021)		X		X

Balance‐specific training (n=4/11) ,strength (n=4/11), and DTT training (n=4/11) exercises were the most frequently assessed interventions, followed by gait training (n=3/11) and DTT (n=5), with 1 study including gaze stabilization or eye‐head movement exercises. A summary of intervention characteristics, exercise descriptions, and outcome measures is presented in Tables S3‐S5 (Supplementary Materials).

Interventions were delivered across an average of 6.08 weeks (SD=2.95, range=3–12), with an average of 4.18 visits per week (SD=2.94, range=2–14), for an average of 20.79 sessions (SD=5.78, range=12–34), with an average session duration session, of 46.67 minutes (SD=20.85, range 30–95 minutes). Except for 2 studies which adopted a semisupervised telerehabilitation ⁴⁷ and home program ⁵⁰ format, and 1 study ⁵¹ that did not report mode of delivery, majority (n=8) of the interventions were delivered in person, under supervision of a physical therapist. ²¹ , ²² , ⁴⁷ , ⁴⁸ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³

All studies included a control group in the form of conventional physical therapy or usual care (6/11 studies), ²² , ⁴⁷ , ⁴⁹ , ⁵¹ , ⁵² , ⁵⁴ the current institutional stroke rehabilitation program, ⁵⁰ single‐task training, ²¹ conventional occupational therapy, ²⁰ upper‐limb exercises, ⁵³ and 1 study where participants continued their usual activities. ⁴⁸

Outcome Measures

All studies included an assessment at baseline and post intervention. A range of outcome measures were used to evaluate the effects of VRT on balance in survivors of stroke, with details provided in the Tables S4 and S5 (Supplementary Material). Objective outcomes measures including were those that assessed dynamic and static balance, physical function, gait, CMI, number of falls, and general health.

Balance Outcomes

Eight studies evaluated balance using various outcome measures, ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³ , ⁵⁴ with the BBS as the most frequently used (8/11 studies) static balance outcome measure. ²⁰ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² , ⁵⁴

Cognitive–Motor Interference Outcomes

CMI was assessed in 4/11 studies, ²⁰ , ²¹ , ²² , ⁵³ with varying outcome measure tools of DT used including (1) walking while performing a cognitive task (serial subtraction), (2) Timed Up and Go (TUG) with verbal fluency or serial subtraction, (3) walking obstacle crossing with verbal fluency or serial subtraction, and (3) walking with auditory Stroop and clock tasks.

Gait Outcomes

The most commonly included gait/dynamic balance, outcome measures was the TUG test, followed by spatial and temporal gait parameters via the GAITRight analysis system. ⁴⁹ , ⁵⁰ , ⁵² , ⁵³ , ⁵⁵ , ⁵⁶

Falls

Number of falls was analyzed in 3 studies50, 52, 53 with all studies finding a reduced rate of falls in the VRT group compared with the control group. Mansfield et al. found that VRT decreases the risk of falls by 15.2%, ⁵² and Correia et al. reported a 37% estimated reduced risk of falling (risk ratio [RR], 0.37 [95% CI, 0.22–0.62]; P<0.001). ⁵⁰ Pang et al. showed that a DT program reduced the risk of falls and injurious falls by 25.0% (95% CI, 3.1%–46.9%, P=0.037) and 22.2% (95% CI, 4.0%–38.4%, P=0.023), respectively, which was maintained during the 6‐month follow‐up, compared with controls. ⁵³

Risk of Bias and Quality of Evidence

The risk of bias, assessed using the Risk of Bias‐2 tool, can be found in Figure S1 (Supplementary Material). Four studies (n=4/11, 28.6%) had low risk of bias, 4/11(28.6%) ²⁰ , ²¹ , ⁴⁹ , ⁵⁴ had high risk of bias, and 3/11 (21.4%) ⁴⁷ , ⁵⁰ , ⁵¹ had “some concerns” for risk of bias. High risk of bias arising from the randomization process was the most frequent reason (4/11 studies), ²⁰ , ²¹ , ⁴⁹ , ⁵⁴ with 1 study ⁵⁴ not reporting any information regarding randomization. All studies possessed some imprecision due to small sample sizes of <400 which contributed to downgrading the quality of evidence, as per Cochrane's recommendations. ⁵⁷ Please see Table 3 for additional details.

Table 3.

Summary of Findings for Quality of Evidence Using the Grading of Recommendations, Assessment, Development, and Evaluation System

Outcomes	No. of studies (participants)	Risk of bias	Relative effect (95% CI)	Quality of evidence (GRADE)†	Comments
Overall balance – pooled scores	10 (413)	Low	0.64 SMD (0.44–0.85)	Moderate a Due to inconsistency and imprecision.	VRT probably improves static balance and gait with a large effect.
Balance – BBS (post intervention)	7 (235)	High	3.98 mean difference (3.16–4.81)	Very low b Due to risk of bias, inconsistency, and imprecision.	It is uncertain whether VRT as measured by the BBS, improves static balance.
Balance specific training	3 (137)	High	1.07 SMD (0.70–1.45)	Very low c Due to risk of bias, inconsistency, and imprecision.	It is uncertain whether balance specific training improves static balance.
Strength/functional training	4 (151)	Some concerns	0.81 SMD (0.46–1.15)	Low d Due to inconsistency and imprecision.	Strength/functional training may improve static balance with a large effect.
Gait‐specific training	3 (74)	Some concerns	0.68 SMD (0.20–1.16)	Moderate e Due to imprecision.	Gait‐specific training probably improves static balance with a moderate effect.
Dual‐task training	4 (207)	Some concerns	0.46 SMD (0.18–0.74)	Lowf Due to inconsistency and imprecision.	DTT may improve gait, with moderate effect.
≤6 weeks of VRT	5 (149)	Some concerns	0.89 SMD (0.54–1.24)	Very Low g Due to risk of bias, inconsistency, and imprecision.	It is uncertain whether ≤6 weeks of VRT improves balance and gait.
>6 weeks of VRT	5 (264)	Some concerns	0.52 SMD (0.27–0.77)	Low h Due to inconsistency and imprecision.	>6 weeks of VRT may improve balance and gait with a moderate effect.
≤3 sessions/week	5 (270)	Some concerns	0.43 SMD (0.19–0.67)	Moderate Due to imprecision.	≤3 sessions/week of VRT probably improves balance and gait with a small effect.
>3 sessions/week	5 (143)	High	1.14 SMD (0.77–1.51)	Very low i Due risk of bias, inconsistency and imprecision.	It is uncertain whether > 3 sessions/week of VRT improves balance and gait.
≤30 minutes/session	5 (155)	Some concerns	0.92 SMD (0.57–1.26)	Low j Due to inconsistency and imprecision.	<30 minutes/session of VRT may improve balance and gait with a large effect.
>30 minutes/session	5 (234)	Some concerns	0.50 SMD (0.22–0.74)	Low k Due to inconsistency and imprecision.	>30 minutes/session of VRT may improve balance and gait with a moderate effect.

High=This research provides a very good indication of the likely effect. The likelihood that the effect will be substantially different ^‡ is low. Moderate=This research provides a good indication of the likely effect. The likelihood that the effect will be substantially different ^‡ is moderate. Low=This research provides some indication of the likely effect. However, the likelihood that it will be substantially different ^‡ is high. Very low=This research does not provide a reliable indication of the likely effect. The likelihood that the effect will be substantially different ^‡ is very high. BBS indicates Berg Balance Scale; DTT, dual‐task training; GRADE, Grading of Recommendations, Assessment, Development, and Evaluation; SMD, standardized mean difference; and VRT, vestibular rehabilitation training.

^‡ Substantially different=a large enough difference that it might affect a decision.

^† GRADE Working Group grades of evidence. ³⁹

^a Variability of outcome measures used to analyze overall balance.

^b Evidence of substantial heterogeneity (I ²=72%, P=0.002) and does not meet minimal clinically mportant difference.

^c Evidence of considerable heterogeneity (I ²=94%, P<0.0002).

^d Evidence of considerable heterogeneity (I ²=85%, P=0.0002) and CI crosses the line of no effect.

^e Evidence of low heterogeneity (I ²=37%, P<0.20) and small sample size.

^fVariability in outcome measures used to analyze gait and DTT.

^g Evidence of substantial heterogeneity (I²=73%, P<0.002) and wide CI.

^h Evidence of low heterogeneity (I²=35%, P<0.0007).

ⁱ Evidence of substantial heterogeneity (I²=73%, P=0.002).

^j Evidence of substantial heterogeneity (I²=73%, P=0.002) and a large CI.

^k Evidence of moderate heterogeneity (I²=41%, P=0.15).

Effects of VRT on Balance

All studies investigated the effect of various forms of VRT (excluding DTT), on static balance outcomes via BBS. One study ⁵⁰ found statistically significant between‐group differences with improved BBS and TUG scores with gaze stabilization exercises or eye‐head movements. Two studies ⁴⁹ , ⁵⁸ (14.3%) included strength/functional exercises with balance training in the VRT intervention, ⁴⁸ , ⁴⁹ and 2 other studies ⁴⁷ , ⁵¹ (14.3%) included gait training with strength/functional exercises in the VRT intervention. ⁴⁷ , ⁵¹ All 4 studies reported improved BBS scores, although these were not clinically significant as they did not meet the minimal clinically important difference (MCID) (Table S6 of Supplementary Materials). A single study ⁵⁴ investigated gait with balance‐specific training in a circuit fashion, which found clinically significant improvements in BBS (P=0.002) after 6 weeks. One study ⁵² (7.14%) investigated the effects of perturbation‐based balance training on balance and falls outcomes, reporting reduced falls compared with the control group (rate ratio, 0.62 [95% CI, 0.29–1.30]; P=0.20), with greater improvement in reactive balance control (Mini Balance Evaluation Systems Test) sustained 12 months post training, with improved BBS scores, although again, these were not clinically significant.

Effects of Cognitive–Motor Dual Task Training on Gait

Four studies ²⁰ , ²¹ , ²² , ⁵³ investigated DTT effects on gait/dynamic balance. ²¹ , ²² Two studies ²¹ , ²² found statistically significant improvements in spatial gait parameters of step length, ²¹ stride length, ²² and temporal gait parameters of gait speed, ²¹ , ²² cadence ²¹ step time, ²¹ and cycle time, ²¹ as assessed via the GAITRight system, following DT training. The MCID for change in gait speed among survivors of late subacute stroke was 0.175m/s ⁵⁹ ; change in gait speed was not clinically significant for any of the included studies that reported these data (Table S6 of Supplementary Materials). ²¹ , ²² One study ²⁰ found significantly stronger effects in the DT group for cognitive (trail making test, digital span test, Stroop test), and motor (Fugl‐Meyer Assessment, BBS, modified functional reaching test) tasks, compared with an occupational therapy program. Pang et al. ⁵³ found reduced DT interference in walking time (forward walking combined with verbal fluency, forward walking with serial‐3‐subtractions, and the TUG with verbal fluency) in the DT group, with these effects maintained 8 weeks post intervention. ⁵³ All 4 ²⁰ , ²¹ , ²² , ⁵³ studies support DT training compared with single‐task training for improved gait parameters, with less noticeable changes in cognitive task performance compared with motor task performance. ²² , ⁵³ Further details on individual study results can be found in the Supplementary Materials.

Meta‐Analysis

Ten studies, with 413 participants, were included in the meta‐analysis. Various outcome measures were used to assess balance, including BBS, Mini Balance Evaluation Systems Test, functional gait assessment, and Activities‐Specific Balance Confidence Scale, and gait, including a TUG, GAITRight analysis system, 10‐meter walk test, and 6‐minute walk test. The primary static balance outcome measure used in the analysis was the BBS, which was the most reported outcome measure (8/10 studies). When a study did not report BBS, TUG scores, ⁵³ or cadence (steps/min) ²¹ , ²² was used depending on which outcome the study reported. When investigating DTT effects on gait, TUG data were the primary outcome measure when reported, but if unavailable, cadence (steps/min) was used. One study ⁵² was excluded from the meta‐analysis due to statistical testing suggesting the data are significantly skewed away from normality as per the methods proposed by Shi et al. ⁴⁴ Table 3 provides a summary of findings.

Overall Balance and Gait

The pooled effect SMD (Figure 2) suggests that VRT has a large effect for improving static and dynamic balance/gait (SMD, 0.64 [95% CI, 0.44–0.85], P<0.00001), with evidence of substantial heterogeneity (I ²=62%, P=0.002) and moderate certainty of evidence (Table 3). The MCID for the BBS was met, although pooled MD scores for TUG data could not be calculated due to insufficient data for meta‐analysis from the included studies.

Figure 2. Meta‐analysis of overall effect of VRT on balance and gait of included studies.

CDTT indicates cognitive dual‐task training; DTGT, dual‐task gait training; IV, information value; MDTT, motor dual‐task training; STGT, single‐task gait training; and VRT, vestibular rehabilitation therapy.

Berg Balance Scale

Seven RCTs ²⁰ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵⁴ were included in a subgroup analysis of postintervention BBS outcomes. The between‐group MD post intervention is 3.98 [95% CI, 3.16–4.8] (P<0.00001) with low heterogeneity (I ²=26%, P=0.23) (Figure S2 of Supplementary Materials), meeting the MCID for patients with late subacute stroke and MID for patients with chronic stroke. Whereas the pooled MD change score from baseline to post intervention within‐groups was 1.70 [95% CI, 1.39–2.02] (P<0.00001) with considerable heterogeneity (I ²=85%, P<0.00001) (Figure S3 of Supplementary Materials), not meeting MCID for patients with either late subacute or chronic stroke (Table S6 of Supplementary Materials).

VRT Type

Subgroup analysis was performed to analyze the effects of different types of VRT on static balance and dynamic balance/gait outcomes (Figure 3). Only 1 study investigated the effects of gaze stabilization exercises on balance, and therefore a meta‐analysis could not be performed.

Figure 3. Meta‐analysis of VRT type on balance and gait of included studies.

CDTT indicates cognitive dual‐task training; DTGT, dual‐task gait training; IV, information value; MDTT, motor dual‐task training; STGT, single‐task gait training; and VRT, vestibular rehabilitation therapy.

Balance‐Specific Training Effects on Static Balance

Three RCTss ⁴⁸ , ⁴⁹ , ⁵⁴ investigated the effects of balance‐specific training that demonstrated a significantly large effect for improving balance (SMD, 1.07 [95% CI, 0.70–1.45], P<0.00001), although very low certainty of evidence (Table 3 and Figure 3).

Strength/Functional Training Effects on Static Balance

Four RCTs ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵¹ investigated the effects of strength/functional training, suggesting a large effect (SMD, 0.81 [95% CI, 0.46–1.15], P<0.00001) and low certainty of evidence (Table 3 and Figure 3).

Gait‐Specific Training Effects on Static Balance

Three RCTs ⁴⁷ , ⁵¹ , ⁵⁴ investigated the effects of gait‐specific training, which suggests a moderate effect size exists for improving balance (SMD, 0.68 [95% CI, 0.20–1.16], P=0.005), with moderate certainty of evidence (Table 3 and Figure 3).

Dual‐Task Training Effects on Gait

Four RCTs ²⁰ , ²¹ , ²² , ⁵³ investigated the effects of DTT on gait. Data were collected from 2 studies that reported TUG scores and 2 studies that reported cadence (steps/min) using the GAITRite gait analysis system. Two studies included 2 intervention groups, which were represented as individual studies in the meta‐analysis (ie, Liu et al. – cognitive DTT, Liu et al. – motor DTT). The findings suggest that DTT has a small to moderate effect on gait (SMD, 0.46 [95% CI, 0.18–0.74], P=0.001), with low certainty of evidence (Table 3 and Figure 3).

VRT Duration, Frequency, and Session Length

VRT Duration in Weeks

Five RCTs ²¹ , ²² , ⁴⁹ , ⁵⁰ , ⁵¹ investigated the effects of VRT interventions ≤6 weeks, which suggests a large effect exists for improving balance and gait outcomes (SMD, 0.89 [95% CI, 0.54–1.24], P<0.00001), although with substantial heterogeneity (I ²=73%, P<0.002) and very low certainty of evidence and. Five RCTs ²⁰ , ⁴⁷ , ⁴⁸ , ⁵³ , ⁵⁴ investigated the effects of VRT interventions >6 weeks, which found a moderate effect size for improving balance and gait (SMD, 0.52 [95% CI, 0.27–0.77], P<0.00001), with evidence of low heterogeneity (I ²=35%, P=0.18) and low certainty of evidence (Table 3). See Figure S4 (Supplementary Materials) for additional details. On the contrary, a sensitivity analysis of VRT duration based on BBS scores alone, showed a stronger effect for >6 weeks (SMD, 4.56 [95% CI, [2.32–6.80]) compared with ≤6 weeks (SMD, 3.89 [95% CI, 3.00–4.78] (Figure S5 of Supplementary Materials).

VRT Frequency (Sessions/Week)

Five RCTs ²⁰ , ²² , ⁴⁷ , ⁴⁸ , ⁵³ investigated the effects of ≤3 sessions/week of VRT, suggesting a moderate effect exists for improving balance and gait (SMD, 0.43 [95% CI, 0.19–0.67], P=0.0005) with moderate certainty of evidence. Five studies ²¹ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵⁴ investigated the effects of >3 sessions/week of VRT, suggesting a strong effect exists for improving balance and gait (SMD, 1.14 [95% CI, 0.77–1.51], P<0.00001), although with very low certainty of evidence (Table 3 and Figure S6 of Supplementary Materials). A sensitivity analysis of VRT sessions/week based on BBS scores alone, confirmed the finding with >3 sessions/week having a stronger effect compared with ≤3 sessions/week (Figure S7 of Supplementary Materials).

VRT Session Length

Five RCTs ²⁰ , ²¹ , ²² , ⁴⁹ , ⁵¹ investigated the effects of <30 minutes/session of VRT, suggesting a large effect for improving balance and gait outcomes (SMD, 0.92 [95% CI, 0.57–1.26], P<0.00001), with low certainty of evidence (Table 3). Five RCTs analyzed the effects of >30 minutes/session, which suggests a moderate effect for improving balance gait outcomes exists (SMD, 0.48 [95% CI, 0.22–0.74], P=0.0003), with low certainty of evidence (Table 3). See Figure S8 (Supplementary Materials) for additional details.

DISCUSSION

This systematic review and meta‐analysis evaluated the effectiveness of VRT and DTT for improving balance and gait outcomes among people with stroke, providing the first pooled estimate specifically targeting populations with late subacute and chronic stroke. The results suggest that in these populations, VRT likely improves balance, and DTT may enhance gait outcomes, both of which may contribute to fall risk reduction among individuals with stroke. ¹² , ⁴⁸ , ⁶⁰ , ⁶¹ As people fall mostly when in motion, improvements in dynamic balance and gait, as measured by TUG and cadence are likely more relevant to fall prevention than static balance measures such as BBS. ¹² , ⁶¹ , ⁶² , ⁶³

To our knowledge this is the first systematic review to assess the combined effect of VRT and DTT in participants with late subacute and chronic stroke including subcategorization of VRT and DTT for a more granular analysis of their effects. Additionally, this study is among the few to examine the impacts of VRT intervention length, timing and frequency. These findings support the integration of VRT and DTT into poststroke rehabilitation programs, highlighted their potential to enhance dynamic balance, reduce fall risk, and inform future clinical guidelines.

The large effect sizes for postintervention pooled BBS scores in 8/10 studies included in the meta‐analysis revealed a meaningful and statistically significant difference between groups, favoring the experimental VRT group. Alternatively, despite being statistically significant, within‐group change scores from baseline to post intervention did not meet the MCID for patients with either late subacute or chronic stroke, suggesting that although there was a measurable improvement, it may not be perceived as clinically meaningful by the patients.

However, due to substantial heterogeneity (I ²=62%), results should be interpreted with caution, as the substantial differences between studies could be due to variations in study populations, interventions, or methodologies. Furthermore, although heterogeneity was present, a fixed‐effects model was chosen to estimate the precision‐weighted mean effect across included studies to provide the most precise estimate for the available data, regardless of whether true effect sizes differ across studies. ⁴⁰ , ⁶⁴ , ⁶⁵ These findings highlight the need for further research to identify the factors contributing to this heterogeneity and to explore interventions that can achieve clinically meaningful improvements for patients with stroke.

Results also support that DTT probably improves gait parameters greater than single‐task training, although changes in gait speed were not clinically significant. ⁵⁹ However, it is important to note that due to heterogeneity between the studies, there is considerable variability in the effect sizes, which is due to true differences between studies rather than random chance.

Furthermore, it must be recognized that the outcome measure used will impact findings. Although the BBS is a commonly used assessment tool to evaluate static balance in survivors of stroke, it is more suitable for individuals with lower functional ambulation scores, potentially affecting the ability to detect meaningful changes and reach the MCID. ⁸ Moreover, the Mini Balance Evaluation Systems Test has demonstrated superior discriminative ability in categorizing patients with stroke into slow and fast walkers but was not included in this review. ⁶⁶ This highlights the importance of considering the choice of outcome measures in research and clinical practice, as different assessments may yield varying results and interpretations. ⁸

VRT of all types (balance‐specific training, strength/functional training, and gait‐specific training) included in the meta‐analysis improved static balance, and DTT was effective for improving gait, although with variable effect sizes and certainty of evidence as this was investigated in only 36% (4/11) studies. Balance‐specific training showed the strongest effect, followed by strength/functional training, although with very low certainty of evidence, and low certainty of evidence, respectively. Given the heterogeneity and small sample sizes, these results should be interpreted with caution and it cannot be assumed that these results are generalizable. Studies of larger sample sizes are required to improve precision and confirm these findings.

The results of this systematic review and meta‐analysis found that DTT, compared with single‐task training demonstrated a significantly larger effect on gait performance, as per TUG outcomes, whereas motor DTT training showed minimal difference in effect compared with cognitive DTT for cadence. There is no consensus on the best DT tests for gait among survivors of stroke, and the specific DTs that are most effective for assessing gait remain unclear. Further research is needed to determine the optimal dual‐task assessments that can effectively evaluate CMI and enhance gait performance in survivors of stroke.

Findings suggest that VRT interventions of both ≤6 weeks and >6 weeks produced improvements in balance and gait, with high and moderate effect sizes, respectively, although with very low and low certainty of evidence, respectively. Therefore, these findings are inconclusive, and further work is needed to identify optimal duration. There are many variables that contribute to a rehabilitation intervention's efficacy, and these findings do not consider training intensity, patient adherence, fatigue, and individual variability, which may each contribute to the current findings. Evidence also suggests that VRT interventions with >3 sessions/week produces stronger effects on static balance and gait, compared with ≤3 sessions/week, although with very low and moderate certainty, respectively. This is in agreement with the World Guidelines for Older Adults, which strongly recommend that exercise programs for fall prevention in community‐dwelling older adults include individualized and progressive balance challenging and functional exercises with 3 or more sessions taking place weekly for a minimum of 12 weeks for better outcome. ⁶⁷

However, the review findings should be interpreted with caution given the small number of included studies with relatively small sample sizes and half of the studies reflecting high to moderate risk of bias.

Interpretation of Results in the Context of Other Evidence

These findings align with the existing literature, with moderate certainty of evidence that VRT improves static balance, as per the BBS, and that DTT improves gait parameters (spatial and temporal), as per TUG and cadence scores, among survivors of stroke. ¹⁸ Previous systematic reviews' also support VRT programs as safe and easily implementable in patients with neurological disorders, ¹⁷ with beneficial effects on balance and gait performance in survivors of stroke, although with low certainty of evidence. ¹⁸ , ⁶⁸ VRT has consistently been recognized as a valuable intervention for survivors of stroke facing challenges such as dizziness, imbalance, and falls. ⁶⁰ Studies have shown that integrating VRT into poststroke rehabilitation protocols can enhance balance and gait performance. ¹⁷ , ⁶⁹

The existing literature also consistently demonstrates that DTT leads to improved gait performance and balance abilities in survivors of stroke. ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ³⁴ , ⁷⁰ These studies collectively highlight the importance of integrating cognitive and motor tasks into stroke rehabilitation programs. However, these studies also have several limitations that warrant further research and have highlighted the need for more exploration into the influence of task nature on DT interference magnitude and pattern in survivors of stroke. ¹³

Despite supporting evidence, there are limited high‐quality studies and more robust research, focusing on exercise type, intensity, frequency, duration, and DTT, is warranted to investigate an optimal VRT protocol for survivors of stroke.

Strengths and Limitations of Evidence

This review is the first to explore the effects of VRT, including DTT, on balance and gait among survivors of late subacute and chronic stroke. A recent review ¹⁸ explored VRT but excluded DTT training and included survivors of stroke of any onset.

Several limitations were addressed in this study, including the omission of DTT as a predefined search term, as our initial focus centered on VRT. Although this broader search strategy allowed DTT to be captured as an intervention in 4 of the included studies, explicitly incorporating DTT in the search terms could have enhanced the comprehensiveness of the review and identified additional relevant studies. Furthermore, the studies included a wide age range with varied stroke locations and types that were not stratified for. Future research should address these variables to allow for more individualized rehabilitation approaches according to stroke‐specific characteristics. Additionally, the variability in participant eligibility criteria and the poor reporting on ethnicity highlight issues of representation. Lastly, calculations for mean±SD of change scores were made when not reported, potentially introducing inaccuracies.

The small sample sizes across studies may limit the ability to detect true differences in the effects of VRT on balance and gait outcomes, and as such, findings should be interpreted with caution. ⁴⁰

CONCLUSIONS

Despite the substantial heterogeneity across studies, our review suggests that VRT is probably effective for improving static and dynamic balance/gait, with the novel finding that DTT may improve gait parameters among survivors of late subacute and chronic stroke.

In conclusion, DTT, whether focused on dual‐motor or cognitive‐motor DTs, plays a crucial role in addressing CMI and may improve balance and gait performance in survivors of late subacute and chronic stroke. By incorporating DT exercises tailored to the specific needs of individuals, rehabilitation programs may improve functional, motor, and cognitive abilities in this population.

Based on our comprehensive review and meta‐analysis, VRT programs for survivors of late subacute and chronic stroke should emphasize balance‐specific training and DTT, with components of gait and strength training included. However, the optimal duration and frequency of VRT interventions remain unclear and require further investigation.

To address current limitations and develop a more robust VRT protocol for survivors of stroke, high‐quality RCTs with larger sample sizes are needed. These studies should aim to standardize intervention protocols, reduce bias, and ensure findings are clinically meaningful and generalizable.

Sources of Funding

Supported by the European Union Horizon 2021 Research and Innovation Grant, The Chartered Society of Physiotherapists and the British Research Council.

Disclosures

The authors declare that there are no conflicts of interest relevant to this work. Author BN, DEB and MP are partially funded by the European Union’s Horizon 2021 research and innovation program (grant agreement no. 101057747).

Supporting information

Data S1–S2

Tables S1–S6

Figures S1–S8

PRISMA Checklist