Scholarly article on topic 'Biofeedback improves performance in lower limb activities more than usual therapy in people following stroke: a systematic review'

Biofeedback improves performance in lower limb activities more than usual therapy in people following stroke: a systematic review Academic research paper on "Medical engineering"

CC BY-NC-ND
0
0
Share paper
Academic journal
Journal of Physiotherapy
OECD Field of science
Keywords
{Stroke / "Physical therapy" / Biofeedback / "Systematic review" / Meta-analysis}

Abstract of research paper on Medical engineering, author of scientific article — Rosalyn Stanton, Louise Ada, Catherine M Dean, Elisabeth Preston

Abstract Question: Is biofeedback during the practice of lower limb activities after stroke more effective than usual therapy in improving those activities, and are any benefits maintained beyond the intervention? Design: Systematic review with meta-analysis of randomised trials with a PEDro score > 4. Participants: People who have had a stroke. Intervention: Biofeedback (any type delivered by any signal or sense) delivered concurrently during practice of sitting, standing up, standing or walking compared with the same amount of practice without biofeedback. Outcome measures: Measures of activity congruent with the activity trained. Results: Eighteen trials including 429 participants met the inclusion criteria. The quality of the included trials was moderately high, with a mean PEDro score of 6.2 out of 10. The pooled effect size was calculated as a standardised mean difference (SMD) because different outcome measures were used. Biofeedback improved performance of activities more than usual therapy (SMD 0.50, 95% CI 0.30 to 0.70). Conclusion: Biofeedback is more effective than usual therapy in improving performance of activities. Further research is required to determine the long-term effect on learning. Given that many biofeedback machines are relatively inexpensive, biofeedback could be utilised widely in clinical practice. [Stanton R, Ada L, Dean CM, Preston E (2016) Biofeedback improves performance in lower limb activities more than usual therapy in people following stroke: a systematic review. Journal of Physiotherapy 63: 11–16]

Academic research paper on topic "Biofeedback improves performance in lower limb activities more than usual therapy in people following stroke: a systematic review"

fp"S-295;No.ofP,SI,6

Journal of

PHYSIOTHERAPY

journal homepage www.elsevier.com/locate/jphys

Research

Biofeedback improves performance in lower limb activities more than usual therapy in people following stroke: a systematic review

Rosalyn Stanton a, Louise Adaa, Catherine M Deanb, Elisabeth Prestonc

a Discipline of Physiotherapy, The University of Sydney; b Department of Health Professions, Macquarie University, Sydney;c Faculty of Health, University of Canberra,

Canberra, Australia

ABSTRACT

Question: Is biofeedback during the practice of lower limb activities after stroke more effective than usual therapy in improving those activities, and are any benefits maintained beyond the intervention? Design: Systematic review with meta-analysis of randomised trials with a PEDro score > 4. Participants: People who have had a stroke. Intervention: Biofeedback (any type delivered by any signal or sense) delivered concurrently during practice of sitting, standing up, standing or walking compared with the same amount of practice without biofeedback. Outcome measures: Measures of activity congruent with the activity trained. Results: Eighteen trials including 429 participants met the inclusion criteria. The quality of the included trials was moderately high, with a mean PEDro score of 6.2 out of 10. The pooled effect size was calculated as a standardised mean difference (SMD) because different outcome measures were used. Biofeedback improved performance of activities more than usual therapy (SMD 0.50, 95% CI 0.30 to 0.70). Conclusion: Biofeedback is more effective than usual therapy in improving performance of activities. Further research is required to determine the long-term effect on learning. Given that many biofeedback machines are relatively inexpensive, biofeedback could be utilised widely in clinical practice. [Stanton R, Ada L, Dean CM, Preston E (2016) Biofeedback improves performance in lower limb activities more than usual therapy in people following stroke: a systematic review. Journal of Physiotherapy XX: XX-XX]

© 2016 Australian Physiotherapy Association. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

AUSTRALIAN PHYSIOTHERAPY

ASSOCIATION

KEY WORDS

Stroke

Physical therapy Biofeedback Systematic review Meta-analysis

Introduction

This is an update of a systematic review1 that examined the effect of biofeedback in training lower limb activities after stroke. Biofeedback is equipment that transforms biological signals into an output that can be understood by the learner, providing information to the learner that is not consciously available. That is, biofeedback takes intrinsic physiological signals and makes them extrinsic, giving the person immediate and accurate feedback of information about these body functions. Biofeedback can be delivered through various senses, such as visual, auditory and tactile systems, and can provide information about the kinematics, kinetics and/or electromyography of activities. Biofeedback is available from medical equipment (eg, electromy-ography, force platforms and positional devices traditionally used in clinical practice); or from non-medical equipment that is increasingly available and used in stroke rehabilitation (eg, recreational games such as the Nintendo® WiiTM). Biofeedback can be used in addition to verbal content; however, it also has the advantage that it can be set up for the patient to use when left to practise alone. However, biofeedback is not commonly used in stroke rehabilitation.2

The previous version of this review,2 which was published in 2011, examined biofeedback broadly in training lower limb

activities after stroke, including trials where any form of biofeedback was provided during practice of the whole activity (rather than part of the activity), with outcomes measured during the same activity. Twenty-two trials met the inclusion criteria and were included in the review; however, meta-analyses demonstrated significant heterogeneity that was best explained by the quality of the included trials. When analyses were limited to higher quality trials (PEDro score > 4), biofeedback had a moderate effect in the short term (10 trials, 241 participants, SMD 0.49, 95% CI 0.22 to 0.75) compared with usual therapy, which was maintained beyond intervention (five trials, 138 participants, SMD 0.41,95% CI 0.06 to 0.75), suggesting that learning had occurred. For a direct comparison of the effect of biofeedback interventions and usual therapy (which includes therapist communication), a post hoc meta-analysis was conducted of those trials where the amount of practice was equal in each group. That is, trials where the control group practised the same activity for the same amount of time as the experimental group, with the only difference being the substitution of biofeedback for therapist communication (presumably including feedback) in the experimental group. This meta-analysis demonstrated a moderate effect of a similar magnitude to the overall analysis (eight trials, 170 participants, SMD 0.51,95% CI 0.20 to 0.83), suggesting that biofeedback is superior to therapist communication.

http://dx.doi.org/10.1016/jjphys.2016.11.006

1836-9553/© 2016 Australian Physiotherapy Association. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/).

Since that review1 was published in 2011, a number of additional trials have been published that investigated the effect of biofeedback, warranting an update of the review. In particular, the potential of using recreational games in stroke rehabilitation has gained attention. The inclusion criteria for this updated review incorporated findings from the previous review. Specifically, this meant that the updated review would include any randomised trial investigating biofeedback from any signal (position, force, EMG) via any sense (visual, auditory, tactile), delivered concurrently during whole activity practice, compared with usual therapy that was practice of the same activity for the same amount of time in the control group with no biofeedback (but presumably with therapist communication), with outcome measures at the activity level and congruent with the activity trained. This ensures a true comparison of the effect of biofeedback compared with usual therapist communication. For the biofeedback intervention, inclusion in this update was based on whether the biofeedback delivered was concurrent rather than terminal feedback. This meant that commercially available recreational games would be included if the majority of the games played within the study delivered concurrent biofeedback, rather than inclusion based on the equipment itself. In order to make recommendations based on the highest level of evidence, this review included only randomised trials with a PEDro score > 4.

Therefore, the research questions for this systematic review were:

1. In adults following stroke, is biofeedback during the practice of lower limb activities more effective than usual therapy in improving those activities in the short term?

2. Are any benefits maintained beyond the intervention?

Method

Identification and selection of trials

Searches were conducted of: MEDLINE (1950 to September 2015); CINAHL (1981 to September 2015); EMBASE (1980 to September 2015); PEDro (to September 2015); the COCHRANE Library (to September 2015) and the PubMed databases (to September 2015) for relevant articles without language restrictions, using words related to stroke and randomised, quasi-randomised or controlled trials and words related to biofeedback (such as biofeedback, electromyography, joint position, and force) during lower limb activities (such as sitting, sit to stand, standing and walking) (see Appendix 1 on the eAddenda for the full search strategy). Titles and abstracts (where available) were displayed and screened by one reviewer to identify relevant trials. Full paper copies of relevant trials were retrieved and their reference lists were screened. The methods of the retrieved papers were extracted and reviewed independently by two reviewers (RS and EP) using predetermined criteria (Box 1). Disagreement or ambiguous issues were resolved by consensus after discussion with a third reviewer (LA).

Assessment of characteristics of trials

Quality

The quality of included trials was assessed by extracting PEDro scores from the Physiotherapy Evidence Database (www.pedro. org.au). Two trained raters independently carried out rating of trials in this database, and disagreements were resolved by a third rater. Where a trial was not included in the database, it was independently assessed by two authors who had completed the PEDro Scale training tutorial on the Physiotherapy Evidence Database. Only trials with a PEDro rating > 4 were eligible for inclusion in the review.

Participants

Trials involving adult participants of either gender, at any level of initial disability, at any time following stroke were included.

Box 1. Inclusion criteria.

Design

• High-quality randomised trial or quasi-randomised trial (PEDro score > 4/10)

Participants

• Adults

• Diagnosis of cerebrovascular stroke

• Any level of disability and any time after stroke

Intervention

• Experimental intervention includes biofeedback using any signal (EMG, force, position) via any sensory system (visual, auditory, tactile)

• Part of intervention must be biofeedback during practice of the whole activity

• Practice of whole activity must involve movement (such as reaching in sitting or weight shift in standing)

• Groups must practice the same activity for the same amount of time as the control practice (ie, only difference is feedback delivered)

Outcome measures

• Measures of lower limb activity (sitting, standing up, standing or walking)

• Measures congruent with the activity trained

• Measures of activity must involve movement

Comparisons

• Biofeedback versus usual therapy during the same activity

Age, gender, and time since stroke were recorded to describe the participants in each trial.

Intervention

The experimental intervention could be of any type of biofeedback, that is, using any signal (position, force, EMG) via any sense (visual, auditory, tactile). At least some of the intervention had to involve practice of the whole activity, and practice of the activity had to involve movement (such as reaching in sitting or weight shift in standing). The control intervention must have been the same activity, practised for the same amount of time, where the only difference between the groups was that the intervention group received biofeedback in addition to usual therapy (ie, therapist communication). Type of biofeedback, activity trained, and duration and frequency of the intervention were recorded to describe the trials.

Outcome measures

Measures of lower limb activity that were congruent with the activity in which biofeedback was applied were used in the analysis. Where multiple measures for one activity were reported, a measure was chosen that best reflected the aim of the biofeedback intervention (eg, step length). The measures used to record outcomes and the timing of measurement were recorded and compared to describe the trials.

Data analysis

Data were extracted from the included trials by one reviewer and crosschecked by a second reviewer. Information about the method (ie, design, participants, lower limb activity trained, intervention, measures) and data (ie, number of participants and mean (SD) of outcomes) were extracted. Post-intervention scores were used to obtain the pooled estimate of the effect of intervention in the short term (immediately following intervention) and in the longer term (some time beyond the intervention), as these were reported in a majority of studies. Since different outcome measures were used, the effect size was reported as

Research

Cohen's standardised mean difference (95% CI). A fixed-effect model was used initially, with random effects model planned in the case of significant heterogeneity (I2 > 50%). The analyses were performed using the MIXa program.3,4

Results

Flow of trials through the review

The electronic search strategy identified 3768 trials. After

screening titles and abstracts, 93 potentially relevant papers were

retrieved in full text. An additional 20 potentially relevant trials

were obtained following hand screening of the reference lists of

included trials and previous systematic reviews. After being

assessed against the inclusion criteria, 20 papers reporting

18 randomised trials were included in this review (Figure 1).5-24 Table 1 (see eAddenda) provides a summary of the excluded papers.

Characteristics of included trials

Eighteen trials (20 papers) including 429 participants investigated biofeedback as an intervention to improve activities of the lower limb following stroke. Activities trained included standing up (one trial), standing (eight trials) and walking (nine trials). The quality of included trials is presented in Table 2 and a summary of the trials is presented in Table 3. Additional information was obtained from the authors for two trials.15,16

Quality

The median PEDro score of the included trials was 6 out of 10 (mean 6.2, range 5 to 8). All trials were randomised, had similar groups at baseline, and reported the between-group difference, point estimates and variability for the groups. The majority of trials had < 15% loss to follow-up (94%) and assessor blinding (61%). Only some trials concealed allocation (28%), carried out an intention-to-treat analysis (28%), blinded participants (6%), or blinded therapists (0%).

Participants

Across the trials, the mean age of participants ranged from 47 to 66 years old. Overall, 61% of participants were male and 39% female. The mean time after stroke ranged from < 1 month to 10 years, with 53% of the trials carried out within 6 months after stroke.

Intervention

Biofeedback used in the experimental interventions included: weight distribution from a force platform or sensor (11 trials);

Titles and abstracts screened

• From electronic databases (n = 3768)

Papers excluded after evaluation of

full text (n = 93)"

• Research design not a (quasi-) randomised trial (n = 15)

• Participants not hemiplegic stroke > 18 years old (n = 1)

• Intervention not concurrent biofeedback (n = 14)

• Intervention not during whole activity (n = 36)

• Aim not to Improve at activity level (n = 5)

• No appropriate measure of activity (n = 10)

• Comparison with alternative intervention (n = 34)

• Same participants reported in other trial (n = 3)

• Amount of practice not equal between groups (n = 49)

• Quality not > 4/10 on PEDro scale (n = 45)

Figure 1. Identification and selection of trials.

a Papers may have been excluded for failing to meet more than one inclusion criteria.

muscle activity from EMG (three trials); linear gait parameters such as step width or length from foot sensors (three trials); and joint angle from a goniometer (one trial). Visual feedback was used in seven trials; auditory in seven trials; and a combination of both in four trials. The mean duration of intervention sessions was 33 minutes (SD 17), occurring with a mean frequency of 3.7 days per week (SD 1.6), and a mean duration of 5.2 weeks (SD 2.2).

Outcome measures

For standing up, the measure was weight distribution between the lower limbs (one trial). For standing, measures were the Berg

Papers excluded after screening titles/abstracts

Electronic databases (n = 3675)

Potentially relevant papers retrieved for evaluation of full text (n = 113)

• From electronic databases (n = 93)

• From reference lists (n = 20)

Papers Included In review (n = 20 Dariers. 18 trials^

Table 2

PEDro scores for included trials (n = 18).

Trial Random Concealed Groups Participant Therapist Assessor < 15% Intention-to-treat Between-group Point estimate Total

allocation allocation similar at baseline blinding blinding blinding dropouts analysis difference reported and variability reported (0 to 10)

Brasiliero 5 Y N Y N N Y Y Y Y Y 7

Byl 6 Y N Y N N N Y N Y Y 5

Cozean 7 Y N Y N N Y Y N Y Y 6

DeNunzio 8 Y N Y N N Y Y Y Y Y 7

Druzbicki 9 Y Y Y N N Y Y N Y Y 7

Engardt 10-12 Y N Y N N N Y N Y Y 5

Geiger 13 Y N Y N N N Y N Y Y 5

Grant14 Y N Y N N N Y N Y Y 5

Intiso 15 Y N Y N N Y Y N Y Y 6

Jonsdottir 16 Y N Y N N Y Y Y Y Y 7

Jung 17 Y N Y N N N Y N Y Y 5

Lee 18 Y N Y N N N Y N Y Y 5

Llorens 19 Y Y Y N N Y Y Y Y Y 8

Morris 20 Y Y Y N N Y Y N Y Y 7

Rao 21 Y N Y N N Y N N Y Y 5

Sackley 22 Y N Y N N Y Y N Y Y 6

Sungkarat 23 Y Y Y N N Y Y N Y Y 7

Yang 24 Y Y Y N N Y Y Y Y Y 8

PEDro scores from website www.pedro.org.au

Table 3

Summary of included trials (n= 18).

Trial Design Participants Lower limb activity Intervention Outcome measure used in analysis

Brasiliero 5 RCT n = 20 Age (yr) = 55 (SD 5.5) Gender=12M, 8 F Time since stroke = 2 to 3 yr Walking Exp = Step width + symmetry from foot sensors via visual feedback 20 minutes Con = no biofeedback during walking practice 20 minutes • • Walking speed (m/s) Timing=0, 1 day

Byl 6 RCT n= 12 Age (yr) = 63 (SD 5) Gender=4M, 8 F Time since stroke = 6 to 10 yr Walking Exp = Step length + width from foot sensors via visual feedback 90 min x 1 to 2/wk x 6 to 8 wk (12 sessions) Con = no biofeedback during walking practice 90 min x 1 to 2/wkx 6 to 8 wk (12 sessions) • • Walking step length (cm) Timing=0, 6 to 8 wk

Cozean 7 RCT n= 16 Age (yr) = 55 Gender=10M, 6 F Time since stroke = unknown Walking Exp = Ankle muscle activity from EMG via visual + auditory feedback 30 min x 3/wk x 6 wk Con = placebo biofeedback during walking practice 30 min x 3/wk x 6 wk Both = usual therapy • • Walking step length (cm) Timing = 0, 6 wk

DeNunzio 8 RCT n = 37 Age (yr) = 58 (SD 11)a Gender=19M, 18 F Time since stroke = unknown Standing Exp = Wt distr from force platform via visual/ auditory feedback 30 min x 6/wk x 2 wk Con = no biofeedback intervention during standing practice 30 min x 6/wk x 2 wk Both = usual therapy • • Unified Balance Scale (0 to 54) Timing=0, 2 wk

Druzbicki 9 RCT n = 50 Age (yr) = 60 (SD 11) Gender=32M, 18 F Time since stroke = 44 mth Walking Exp = Step length from foot sensors via auditory feedback 30 min x 5/wk x 2 wk Con = no biofeedback during walking training 30 min x 5/wk x 2 wk Both = usual therapy • • Walking speed (m/s) Timing = 0, 2 wk

Engardt 10-12 RCT n = 40 Age (yr) = 65 (SD 8) Gender=25M, 15 F Time since stroke = 1 mth Standing up Exp = Wt distr from force platform via auditory feedback 45 min x 5/wk x 6 wk Con = no biofeedback during standing up practice 45 min x 5/wk x 6 wk Both = usual therapy Load through affected leg during standing up (% BW) Timing = 0, 6 wk

Geiger 13 RCT n= 13 Age (yr) = 60 (SD 16) Gender=9 M, 4F Time since stroke = 4 mth Standing Exp = Wt distr from force platform via visual feedback 15 min x 2 to 3 /wk x 4 wk Con = no biofeedback during standing practice 15 min x 2 to 3 /wk x 4 wk Both = usual therapy • • Berg Balance Scale (0 to 56) Timing=0, 4 wk

Grant14 RCT n= 16 Age (yr) = 65 (SD 3) Gender=10M, 6 F Time since stroke = 1 mth Standing Exp = Wt distr from force platform via visual feedback 30 min x 5/wk (inpt) and 2/wk (outpt) x 8 wk Con = no biofeedback during standing practice 30 min x 5/wk (inpt) and 2/wk (outpt) x 8 wk Both = usual therapy • • Berg Balance Scale (0 to 56) Timing = 0, 8, 12 wk

Intiso 15 RCT n= 16 Age (yr) = 57 (SD 15) Gender=9M, 7 F Time since stroke = 10 mth Walking Exp = Ankle muscle activity from EMG via auditory feedback 30 sessions over 8 wk Con = no biofeedback during walking practice 30 sessions over 8 wk Both = usual therapy • • Walking step length (cm) Timing = 0, 8 wk

Jonsdottir 16 RCT n = 20 Age (yr) = 62 (SD 11) Gender=unknown Time since stroke = 4 yr Walking Exp = Ankle muscle activity from EMG via auditory feedback 45 min x 3/wk x 7wk Con = usual therapy 45 min x 3/wk x 7wk • • Walking step length (cm) Timing=0, 7, 13 wk

Jung 17 RCT n = 26 Age (yr) = 56(14) Gender= 14M, 7 F Time since stroke = 6.5 mth Walking Exp = Wt distr from force sensor (cane) via auditory feedback 30 min x 5/wk x 4 wk Con = no biofeedback during walking training 30 min x 5/wk x 4 wk Both = usual therapy • • Load through affected leg during walking (% BW) Timing = 0, 4 wk

Lee 18 RCT n = 24 Age (yr) = 47 (SD 12) Gender=16M, 8 F Time since stroke = unknown Standing Exp = Wt distr from force platform (WiiTM) via visual/auditory feedback 30 min x 5/wk x 6 wk Con = no biofeedback during standing training 30 min x 5/wk x 6 wk Both = usual therapy • • Functional Reach Test (cm) Timing = 0, 6 wk

Llorens 19 RCT n = 20 Age (yr) = 57 (SD 12) Gender=9M, 11 F Time since stroke > 1.5 yr Standing Exp = Foot placement from sensor (via camera) via visual feedback 30 min x 5/wk x 4 wk Con = no biofeedback during standing training 30 min x 5/wk x 4 wk Both = usual therapy • • Berg Balance Scale (0 to 56) Timing = 0, 4 wk

Morris 20 Q-RCT n = 26 Age (yr) = 64 (SD 11) Gender=12M, 14 F Time since stroke = 2 mth Walking Exp = Knee angle from goniometer via auditory feedback 30 min x 5/wk x 4 wk Con = no biofeedback during walking practice 30 min x 5/wk x 4 wk Both = usual therapy • • Walking speed (m/s) Timing = 0, 4, 8 wk

Research 5

Table 3 (Continued)

Trial Design Participants Lower Intervention Outcome measure used in analysis

limb activity

Rao21 RCT n = 28 Standing Exp = Wt distr from force platform (WiiTM) via • Fugl-Meyer scale (balance

Age (yr) = 59 (SD 13) visual/auditory feedback 3 sessions over 2 wk component)

Gender = 21M, 7 F Con = no biofeedback during standing training • Timing=0, 2 wk

Time since stroke = 13 days 3 sessions over 2 wk

Both = usual therapy

Sackley22 RCT n = 26 Standing Exp = Wt distr from force platform via visual • Rivermead Motor Assessment (gross

Age (yr) = 66 (SD 11) feedback 20min x 3/wk x 4 wk function component) (0 to 13)

Gender = 20M, 6 F Con = placebo biofeedback during standing • Timing=0, 4, 12 wk

Time since stroke = 5 mth practice 20min x 3/wk x 4 wk

Both = usual therapy

Sungkarat23 RCT n = 35 Walking Exp = Wt distr from force platform via auditory • Load through affected leg during

Age (yr) = 53 (SD 9) feedback 30 min x 5/wk x 3 wk walking (% BW)

Gender = 24 M, 11 F Con = no biofeedback during walking training • Timing=0, 3 wk

Time since stroke=4.5 mth 30min x 5/wk x 5 wk

Both = usual therapy

Yang24 RCT n= 12 Standing Exp = Wt distr from force platform (WiiTM) via • Berg Balance Scale (0 to 56)

Age (yr) = 60 (SD 15) visual feedback 20min x 3/wk x 3 wk (same • Timing=0, 3 wk

Gender = 9 M, 3 F amount practice)

Time since stroke = 6 mth Con = no biofeedback during standing training

(mirror) 20 min x 3/wk x 3 wk Both = usual therapy

BW=body weight, Con = control group, EMG = electromyography, Exp = experimental group, LL = lower limb, M/F = male/female, Q-RCT=quasi-randomised clinical trial, RCT = randomised clinical trial, UT = usual therapy, wt distr=weight distribution. a Only the groups related to the current review objectives report.

Balance Scale (four trials), the gross function component of Rivermead Motor Assessment (one trial), the Unified Balance Scale (one trial), Functional Reach (one trial), and the balance component of the Fugl-Meyer (one trial). For walking, measures were speed (four trials), step/stride length (three trials), and load

through the affected leg (two trials). Outcomes were measured after intervention (18 trials) and from 1 to 3 months after cessation of intervention (four trials).

Effect of biofeedback

SMD (95% CI) Study Fixed

Brasileiro5 ---

Cozean 7 ---

DeNunzio 8 - -

Druzbicki 9 ---

Engardt 10"12 ___

Geiger 13---

Grant14___

Intiso 15 ---

Jonsdottir 16 - -

Jung17 ---

Lee 18 ___

Llorens 19 - -

Morris 20 ---

Rao21 ___

Sackley 22 __

Sungkarat 23 - -

Yang 24__

Pooled ^

i-1--1-1-1-1

-2-101234

Favours Con Favours Exp

Figure 2. SMD (95% CI) of the short-term effect of biofeedback on lower limb activities immediately after intervention by pooling data from 17 trials (n = 417).

The short-term effect of biofeedback on activity limitations was examined by pooling data immediately following the intervention from 17 trials, comprising 417 participants, using a fixed-effect model. One study6 was not included in the meta-analysis because post-intervention data were not reported. Biofeedback improved lower limb activities compared with usual therapy (SMD 0.50, 95% CI 0.30 to 0.70, I2 = 31%) (Figure 2; see Figure 3 on the eAddenda for a detailed forest plot).

The long-term effect of biofeedback on activity limitations could not be examined because only four trials, comprising 84 participants, reported data beyond the intervention. This represented less than 25% of the trials included in the meta-analysis immediately after intervention.

Discussion

This systematic review demonstrated that biofeedback has a moderate effect25 in improving activities of the lower limb, such as standing up, standing and walking, in the short term when compared with usual therapy (ie, therapist communication) during the same amount of therapy. The results suggest that information feedback from biofeedback is superior to therapist communication for improving performance in people following stroke. During usual therapy, whilst therapists may provide some feedback, observational studies suggest that that the content of therapist communication during rehabilitation is more likely to be motivational statements rather than feedback.2,26,27 Thus, biofeedback is likely to be more effective than therapist communication due to the objective and accurate information feedback available to the patient during practice. With the inclusion of an additional 10 high-quality trials, this updated review provides a consistent, but more precise, estimate of effect in the short term compared with the equivalent analysis in the previous review.1

However, the longer term effects are less clear. Only four of the 18 trials measured outcomes at follow-up, with none of the new trials included in this updated review including measurement of outcomes at follow-up. Given that this is only 22% of all the

included trials, this was considered to be unrepresentative, and no meta-analysis was conducted on the effect in the long term. Further research (ie, large, well designed trials including outcomes at follow-up) is required to clearly estimate the long-term effect of biofeedback; that is, the effect on learning.

The mean PEDro score for the 18 trials included in this review was 6.2 out of 10. Given that the maximum achievable score for these types of trials is 8 (because it is difficult to blind therapists and participants to physical interventions), this represents moderately high quality and contributes to the credibility of the conclusions. There was some clinical heterogeneity in these trials. Participant characteristics of age and gender were similar, and the time since stroke was mixed between subacute (53%) and chronic (47%). There was a range of duration of intervention (one session to 8 weeks); however, the majority of trials were of 4 to 8 weeks in duration. Taken together, this suggests that the findings are credible and can be generalised cautiously.

This review had some potential limitations. Several of these limitations may have led to an overestimate of the effect of biofeedback. First, there was a lack of blinding of participants and therapists because this is not always possible in trials of biofeedback. Second, even after only including high-quality trials in the meta-analysis, the results were potentially affected by small trial bias, with an average number of 24 participants per trial (range 12 to 50 participants). Additionally, as is usual with trials of complex interventions, the outcome measures were not the same. This meant that a standardised mean difference had to be calculated from the meta-analysis, which is less clinically useful than a mean difference. Finally, only a small proportion of the trials measured the outcomes some time beyond the intervention, limiting conclusions of the effect of biofeedback on learning. There is a need for large, high-quality trials with adequate power and follow-up to investigate the effect of biofeedback in this population.

In conclusion, this systematic review provides evidence that augmenting feedback through the use of biofeedback is superior to usual therapy (ie, therapist communication) at improving the performance of lower limb activities in people after stroke in the short term. Further research is required to determine the effect of biofeedback on learning. Given that biofeedback is used infrequently, and that many biofeedback machines are relatively inexpensive and easily available, biofeedback could be utilised more widely in clinical practice.

eAddenda: Figure 3, Table 1 and Appendix 1 can be found online at doi:10.1016/j.jphys.2016.11.006 Competing interests: Nil. Source(s) of support: Nil.

Acknowledgements: The authors gratefully acknowledge Tien-Hsin Chang, Jia Han, Oktay Irmak, Eleni Pavlidou, Emily Peelgrane, Helen Preston, Sven Seibert, Juliana Symonds, J Rebecca Winbom, and Nikki Yang for assistance with translation. We would also like to thank Domenico Intiso, and Johanna Jondottir for providing additional information and data.

Provenance: Not invited. Peer reviewed. Correspondence: Rosalyn Stanton, Discipline of Physiotherapy, Faculty of Health Sciences, The University of Sydney, Sydney, Australia. Email: rsta3846@uni.sydney.edu.au

References

1. Stanton R, Ada L, Dean CM, Preston E. Biofeedback improves lower limb activities following stroke: A systematic review. J Physiother. 2011;38:135-142.

2. Stanton R, Ada L, Dean CM, Preston E. Feedback received while practicing everyday activities during rehabilitation after stroke: an observational study. Physiother Res Int. 2015;20:166-173.

3. Bax L, Yu LM, Ikeda N, Tsuruta H, Moons KG. Development and validation of MIX: comprehensive free software for meta-analysis of causal research data. BMC Med Res Methodol. 2006;6:50.

4. Bax L, Yu LM, Ikeda N, Tsuruta H, Moons KGM. MIX: comprehensive free software for meta-analysis of causal research data. 2008; Version 1.7 http:// mix-for-meta-analysis.info.

5. Brasileiro A. Gama G, Trigueiro L, Ribeiro T, Silva E, Galvao Eaue, et al. Influence of visual and auditory biofeedback on partial body weight support treadmill training of individuals with chronic hemiparesis: a randomized controlled clinical trial. Eur J Phys Rehabil Med. 2015;51:49-58.

6. Byl N. Zhang W, Coo S, Tomizuka M. Clinical impact of gait training enhanced with visual kinematic biofeedback: Patients with Parkinson's disease and patients stable post stroke. Neuropsychologiia. 2015;79:332-343.

7. Cozean CD, Pease WS, Hubbell SL. Biofeedback and functional electric stimulation in stroke rehabilitation. Arch Phys Med Rehabil. 1988;69:401-405.

8. DeNunzio AM, Zucchella C, Spicciato F, Tortola P, Vecchione C, Pierelli F, et al. Biofeedback rehabilitation of posture and weight-bearing distribution in stroke: a center of foot pressure analysis. Funct Neurol. 2014;29:127-134.

9. DruZbicki M1. Guzik A, Przysada G, Kwolek A, Brzozowska-Magoü A. Efficacy of gait training using a treadmill with and without visual biofeedback in patients after stroke: A randomized study. J Rehabil Med. 2015;47:419-425.

10. Engardt M, Ribbe T, Olsson E. Vertical ground reaction force feedback to enhance stroke patients' symmetrical body-weight distribution while rising/sitting down. Scand J Rehabil Med. 1993;25:41-48.

11. Engardt M. Long-term effects of auditory feedback training on relearned symmetrical body weight distribution in stroke patients. A follow-up study. Scand J Rehabil Med. 1994;26:65-69.

12. Engardt M. Rising and sitting down in stroke patients. Auditory feedback and dynamic strength training to enhance symmetrical body weight distribution. Scand J Rehabil Med. 1994;31:1-57.

13. Geiger RA, Allen JB, O'Keefe J, Hicks RR. Balance and mobility following stroke: effects of physical therapy interventions with and without biofeedback/forceplate training. Phys Ther. 2001;81:995-1005.

14. Grant T, Brouwer BJ, Culham EG, Vandervoort A. Balance retraining following acute stroke: a comparison of two methods. Canadian J Rehabil. 1997;11:69-73.

15. Intiso D, Santilli V, Grasso MG, Rossi R, Caruso I. Rehabilitation of walking with electromyographic biofeedback in foot-drop after stroke. Stroke. 1994; 25:1189-1192.

16. Jonsdottir J, Cattaneo D, Recalcati M, Regola A, Rabuffetti M, Ferrarin M, et al. Task-oriented biofeedback to improve gait in individuals with chronic stroke: motor learning approach. Neurorehabil Neural Repair. 2010;24:478-485.

17. Jung K, Lim Y, Cha Y, In T-S, Hur Y-G, Chung Y. Effects of gait training with a cane and an augmented pressure sensor for enhancement of weight bearing over the affected lower limb in patients with stroke: a randomized controlled pilot study. Clin Rehabil. 2015;29:135-142.

18. Lee HY, Kim YL, Lee SM. Effects of virtual reality-based training and task-oriented training on balance performance in stroke patients. J Phys Ther Sci. 2015;27: 1883-1888.

19. Lloréns R, Gil-Gómez JA, Alcañiz M, Colomer C, Noe E. Improvement in balance using a virtual reality-based stepping exercise: a randomized controlled trial involving individuals with chronic stroke. Clin Rehabil. 2015;29:261-268.

20. Morris ME, Matyas TA, Bach TM, Goldie PA. Electrogoniometric feedback: its effect on genu recurvatum in stroke. Arch Phys Med Rehabil. 1992;73:1147-1154.

21. Rao N, Zielke D, Keller S, Burns M, Sharma A, Krieger R, et al. Pregait balance rehabilitation in acute stroke patients. Int J Rehabil Res. 2013;36:112-117.

22. Sackley CM, Lincoln NB. Single blind randomized controlled trial of visual feedback after stroke: effects on stance symmetry and function. Disabil Rehabil. 1997;19: 536-546.

23. Sungkarat S, Fisher BE, Kovindha A. Efficacy of an insole shoe wedge and augmented pressure sensor for gait training in individuals with stroke: a randomized controlled trial. Clin Rehabil. 2011;25:360-369.

24. Yang Y-R, Chen Y-H, Chang H-C, Chan R-C, Wei S-H, Wang R-Y. Effects of interactive visual feedback training on poststroke pusher syndrome: A pilot randomized controlled study. Clin Rehabil. 2015;29:987-993.

25. Cohen J. Statistical Power Analysis for the Behavioural Sciences. 2nd ed. New York: Academic Press; 1988.

26. Durham K, van Vliet PM, Badger F, Sackley C. Use of information feedback and attentional focus of feedback in treating the person with a hemiplegic arm. Physiother Res Int. 2009;14:77-90.

27. Johnson L, Burridge JH, Demain SH. Internal and external focus of attention during gait reeducation: an observational study of physical therapist practice in stroke rehabilitation. Phys Ther. 2013;93:957-966.

What is already known on this topic: Previous reviews of the effect of biofeedback in stroke have been favourable, but have included trials with low methodological quality and trials where the amount of therapy time in the biofeedback and control group were not matched.

What this study adds: Although this review included only high-quality trials that compared equal amounts of therapy either with or without biofeedback, it was able to include many new trials. Overall, it provides a robust estimate that biofeedback has a moderately greater benefit on the performance of lower limb activities than usual rehabilitation.