Effects of (music-based) rhythmic auditory cueing training on gait and posture post-stroke: A systematic review & dose-response meta-analysis

Gait dysfunctions are common post-stroke. Rhythmic auditory cueing has been widely used in gait rehabilitation for movement disorders. However, a consensus regarding its influence on gait and postural recovery post-stroke is still warranted. A systematic review and meta-analysis was performed to analyze the effects of auditory cueing on gait and postural stability post-stroke. Nine academic databases were searched according to PRISMA guidelines. The eligibility criteria for the studies were a) studies were randomized controlled trials or controlled clinical trials published in English, German, Hindi, Punjabi or Korean languages b) studies evaluated the effects of auditory cueing on spatiotemporal gait and/or postural stability parameters post-stroke c) studies scored ≥4 points on the PEDro scale. Out of 1,471 records, 38 studies involving 968 patients were included in this present review. The review and meta-analyses revealed beneficial effects of training with auditory cueing on gait and postural stability. A training dosage of 20–45 minutes session, for 3–5 times a week enhanced gait performance, dynamic postural stability i.e. velocity (Hedge’s g: 0.73), stride length (0.58), cadence (0.75) and timed-up and go test (−0.76). This review strongly recommends the incorporation of rhythmic auditory cueing based training in gait and postural rehabilitation, post-stroke.

Despite this compelling evidence, a joint consensus concerning the influence of auditory cueing-based therapy and effective training dosages for recovering gait post-stroke are still warranted.
To the best of our knowledge, five systematic reviews and meta-analyses till date, have evaluated the effects of rhythmic auditory cueing on gait recovery post-stroke [27][28][29][30][31] . Even though, all of the included reviews reported beneficial effects of auditory cueing on gait performance, we observed substantial methodological limitations in these reviews: a) A limited number of controlled clinical trials were included b) The search for the studies was performed across few academic databases c) Ambiguity in the meta-analysis approach was observed i.e. no sub-group analysis or heterogeneity tests were performed d) The search for relevant literature was limited to few languages. Therefore, interpretation of results from these reviews from both a qualitative and quantitative perspective might indicate a bias. Moreover, till date, no meta-analysis has synthesized the current state of literature for determining specific training dosages with rhythmic auditory cueing for recovering gait and postural stability post-stroke. Therefore, in this present systematic review and meta-analysis an attempt has been made to address these shortfalls, by focusing on two main objectives: 1. Analyze the influence of training with rhythmic auditory cueing on spatiotemporal gait and postural stability parameters in individuals post-stroke. 2. Determine appropriate training dosages with auditory cueing that allows substantial enhancements in gait and postural stability.
Findings from this review shall help augment the predictive power concerning a patient's response to auditory cueing interventions, thereby guiding researchers, clinicians and patients themselves in their choice of an optimal rehabilitation intervention.

Methods
This review was conducted according to the guidelines outlined by Preferred Reporting Items for Systematic Reviews and Meta-analysis: The PRISMA statement 32 . A PRISMA checklist has been provided in Supplementary  Table 3.

Data sources and search strategy. Nine academic databases were searched from inception until
December 2017: Web of science, PEDro, EBSCO, Scopus, MEDLINE, Indian citation index, Cochrane central register of controlled trials, EMBASE and PROQUEST. A sample PICOS search strategy for EMBASE academic database has been provided ( Table 1).
An inclusion criterion was determined by two reviewers (S.G, I.G) for the systematic review procedure. The inclusion criterion for the studies were (i) The studies were either randomized controlled trials, cluster randomized controlled trials or controlled clinical trials (ii) The studies evaluated music-based auditory cueing interventions (any training duration, treatment setting) (iii) The studies evaluated spatiotemporal gait parameters (gait velocity, cadence, stride length, stride time, single/double-limb support duration) 33 (iv) The studies evaluated static or dynamic aspects of postural stability (Berg balance scale, Fugl-Meyer lower body assessment, Timed-up and go test, Timed sit-to-stand test, Activity-specific balance confidence scale) 34 (v) The studies included a subjective analysis of stroke outcome (optional) (vi) The studies scored ≥4 points on PEDro quality scale (studies scoring <3 considered of "poor" quality with high risk of biasing excluded 35 ) (vii) The studies were conducted on human participants affected from stroke (any age, disease duration and type) (viii) The studies were published in peer-reviewed academic journals or conference proceedings (un-published "grey" literature was not included) (ix) The studies were published in English, German, Hindi, Punjabi or Korean languages.
The two reviewers (S.G, I.G) duplicated the study selection, data extraction and quality assessment of the included studies. After selection of the articles, following data were extracted from each study i.e. author, journal name, publication year, selection criteria for participants, total sample size, description of the participants (gender, age, health status, duration of stroke, comorbidities), applied treatment intervention, characteristics of applied auditory stimuli, treatment interventions for the control group, dual-task application (if any), outcome measures, results, conclusions and special notes by authors. The data were then summarized and tabulated (Supplementary Table 2). In case of lack of quantitative data in the manuscript, the reviewers (S.G, I.G) made attempts to contact respective corresponding authors for data.
Quality and risk of bias assessment. The quality of the reviewed studies was assessed using the PEDro methodological quality scale 36 . This quality scale consists of 11 items which address external validity, internal validity, and interpretability. The scale can effectively detect potential bias with fair to good reliability 37 , and validity 36 . A rating of the methodological quality of the studies was carried out by both the primary (S.G) and secondary (I.G) reviewer. Ambiguous issues were discussed between the reviewers and a consensus was reached. The interpretation of the rated studies were that studies scoring 9-10 were considered of "excellent", 6-8 of "good", 4-5 of "fair", and <3 of "poor" quality 38 .
Data Analysis. A within-group i.e. pre-post meta-analysis approach was incorporated in the review to develop a quantitative interpretation of the auditory cueing interventions 39 . The meta-analyses were conducted using CMA (Comprehensive meta-analysis V 2.0, USA). The data in this analysis was distributed and separately analyzed for each outcome measure such as gait velocity, stride length, cadence, and timed-up and go test. Here, the use of either random/fixed effect meta-analysis was dependent upon the presence/absence of heterogeneity during the group analysis, respectively 40 . Moreover, forest plots with 95% confidence intervals were plotted. The effect sizes were adjusted and reported as weighted Hedge's g 41 . A positive effect size would represent a favorable outcome of the intervention and vice versa for the negative effect. Further, the thresholds for the interpretation of weighted effect sizes are as follows: an effect size of 0.2 is considered as a small effect, 0.5 as a medium effect and 0.8 as a large effect 42 . Further, heterogeneity between the studies was computed using I 2 statistics 42,43 . The interpretation of heterogeneity via I 2 statistics is as follows: 0-25% is considered as negligible heterogeneity, 25-75% as moderate heterogeneity and ≥75% as substantial heterogeneity, respectively. In cases where substantial heterogeneity was observed sensitivity analysis were performed to elucidate the "significant" cause of heterogenity 44 . In this analysis, the results were compared by either including or excluding results from studies that used inadequate randomization methods and/or differed in terms of applied intervention.
In the included studies rhythmic auditory cueing was subjected to patients according to their comfortable cadence. The evaluated parameters were spatiotemporal parameters of gait i.e. gait velocity, stride length and cadence. Furthermore, sub-group analyses were also performed to determine specific training dosages for application of rhythmic auditory cueing in a gait rehabilitation protocol. The main emphasis was laid to determine the duration of a training session and the number of days for which these sessions were performed during a week. Likewise, sub-group analyses were also conducted to analyze joint effects of treadmill training together with rhythmic auditory cueing on gait performance in stroke patients. This analysis was performed to analyze the joint influence of adjunct treadmill training with auditory cueing.
Details of weighted effect size, 95% confidence intervals, significance and heterogeneity have been reported for each outcome measure. Additionally, an analysis for publication bias was performed by Duval and Tweedie's trim and fill procedure 45 . This method involves imputation of the asymmetric studies from the left side to locate the unbiased effect and then re-fills the plot by reinserting the trimmed studies on the left and their imputed counterparts on the right to the mean effect 46 . The graph plots the evaluated weighted effect size i.e. Hedge's g values against standard error on a random effect model. The alpha level was set at 5%.

Results
Characteristics of included studies. The initial search across the academic databases, department's collection of articles and university's library repository (additional sources) yielded a total of 1,471 studies, which on implementing our inclusion/exclusion criteria, were reduced to 38 (Fig. 1). Thereafter, quantitative data was extracted from 25 studies. In the remaining studies where quantitative data was either mentioned in figures or not mentioned at all, attempts were made by the reviewers (S.G, I.G) to contact respective authors for relevant data. Qualitative data from the included studies have been summarized in (Supplementary Table 2). Of the 38 included studies, 11 were randomized controlled trials and 27 were controlled clinical trials. All the included studies reported that the stroke patients also received conventional physical therapy in addition to auditory cueing. Participants. A total of 968 participants were analyzed in the 38 studies. All the studies included mix gender patients affected from stroke. The included studies provided data on 322 females, and 529 males. Five studies did not specify the gender of the included patients [47][48][49][50][51] . Descriptive statistics relating to the age (mean ± standard deviation) of the participants were tabulated across the studies. Disease duration of stroke patients were also extracted (see Supplementary Table 2), however, five studies did not mention these details [47][48][49][50][51] .  Table 2).

Risk of bias.
Meta-analysis report. Gait velocity. Gait velocity was assessed among 25 studies. Additional data concerning different types of auditory stimulations 52 , and lesion sites 53 , in stroke patients was retrieved from two studies (Fig. 4). The analysis of studies revealed (Fig. 4)   preferred cadence and only allowed increments in tempo ranging from 1-3 bpm. On the contrary, Kim and Oh 48 subjected their participants to fixed tempo ranging from 20-100 bpm (Supplementary Table 2). Therefore, a comparison of different training dosages was not performed. Additionally, a comparative sub-group analysis for five studies analyzing effects of rhythmic auditory cueing without training (Supplementary Figure 3) revealed a comparatively smaller medium effect size in the positive domain (g: 0.33, 95% C.I: 0.12 to 0.54) and here as well no heterogeneity was observed in between the studies (I 2 : 0%, p > 0.05).

Discussion
The primary objective of this present systematic review and meta-analysis was to synthesize the current state of knowledge and determine the effects of rhythmic auditory cueing on gait performance and postural stability in stroke patients. The findings from the current meta-analyses suggest positive, medium-to-large standardized effects (pre vs post intervention effects) of rhythmic auditory cueing to enhance gait performance and dynamic postural stability post-stroke. The main findings are:  Several reasons can be affirmed to these observed gait and postural performance enhancements after training with auditory cueing. Firstly, from a neurophysiological perspective we presume that auditory cueing could have facilitated the deficit internal neural timing in stroke patients by bypassing the deficit fronto-striatal networks 56 , and the basal ganglia-somatosensory area motor loop 57 , through alternate pathways (see) [58][59][60] . Moreover, the enhanced sensorimotor synchronization developed between the perception of auditory cueing and gait execution might be due to enhanced periodic/phase corrections 61 . This development of enhanced temporal template/ prediction with the auditory stimulations could be due to pre-attentive "micro-timing", attentive "timescale" processing capabilities of the neural networks mediating phase, periodic corrections, respectively 62 . Secondly, training with auditory cueing could have facilitated re-organization of the deficit neural structures for instance, the stimulation could have increased the motor cortex excitability in the affected hemisphere further resulting in the motor recovery 21,29,63 . Thirdly, based on the findings of Fujioka, et al. 64 we expect that the auditory-motor co-activations could have facilitated neuroplasticity. According to the authors, auditory-motor training could  facilitate neuromagnetic β band oscillations (a functional measure representing auditory motor coupling and neuroplasticity 65 ) thereby assisting in motor recovery. In addition to these neurophysiological changes, rhythmic auditory cueing can impart multifaceted effects on musculoskeletal system as well [66][67][68][69][70][71] . Thaut, et al. 72 suggested that the recruitment and firing rate of motor neurons is determined by the firing rate of auditory neurons (central audiospinal facilitation 73 ), which in turn are stimulated with rhythmic entrainment. Likewise, in an electromyographic analysis during gait performance for post-stroke patients, Thaut, et al. 74 revealed that training with auditory cueing reduced muscular co-activation on the paretic side.
Moreover, we observed considerable enhancements in gait performance in studies incorporating training with auditory cueing as compared to direct application of auditory cueing in a single session i.e. gait velocity (training: 0.73, no training: 0.33), stride length (training: 0.58, no training: 0.25) and cadence (training 0.75, no training: 0.52). We presume that these enhancement in performance with training are due to an "entrainment effect" generated as a result of auditory-motor training 68,72 . This effect has been reported to facilitate movement regularity with repetitions (in this context cyclic movements of gait) further resulting in an enhanced "smoothened" learning pattern 26,[75][76][77] . Upon further sub-group analysis we observed differences in terms of performance because of shorter or longer training durations. Here, in a dose-response analysis we observed that a training duration of 20-45 minutes per session provided substantial increments in both the gait and postural performance as compared to shorter training sessions lasting for 8-10 minutes. These dose related findings are in line with a previously published review study reporting beneficial effects of auditory cueing on arm recovery following stroke 16 . Moreover, in light of recent neuroimaging and clinical studies these findings seem plausible 18,24,26 . Bangert and Altenmüller 24 , for instance reported auditory sensorimotor EEG co-activations after only 20 minutes of auditory-motor training. The authors reported this instantaneous plasticity in the cortex with right hemispheric anterior regions, which ideally represent audio-motor integration 24,25 . The authors further added, that this minimum time frame was vital for establishing stimulus response consistency between audio-motor signals. Similarly, Ghai, et al. 17 reaffirmed these findings and revealed enhanced proprioceptive performance 78,79 , after at least 30 minutes of auditory-motor training. According to the authors, this time frame is crucial for establishing an auditory-motor interfaced mapping resulting in a robustly learned skill set 80,81 .
In addition, we would like to point out some important gaps in the current state of literature which could be addressed by future studies. Firstly, importance of home-based interventions has been emphasized in several studies 70,82,83 . Home-based intervention can allow a patient to enhance their performance for daily life activities, and allow a patient to train for a longer duration in a cost-effective manner as compared to in rehabilitation centers 83 . None of the included studies in the current review elucidated the effects of auditory cueing as a home-based intervention. However, in our sub-group analyses we observed that using treadmill (a common home-based exercise modality) together with auditory cueing was an efficient way for enhancing spatiotemporal gait performance in patients with stroke (gait velocity: 0.15, stride length: 0.45, cadence: 0.39). Moreover, recently published review studies have recommended the positive influence of using auditory cueing as a home based intervention to facilitate gait recovery in neurological disorders such as, cerebral palsy 70 , and multiple sclerosis 84 . Therefore, based on the current state of evidence we strongly hypothesize that combining auditory-motor training in both rehabilitation centers and at home will further enhance the prognostic outcome of stroke patients.
Finally, our findings are in line with previously published "high-quality" systematic reviews and meta-analyses reporting medium-to-large positive effects of training with rhythmic auditory cueing on gait performance in stroke patients. This present study furthers the current state of knowledge concerning the efficacy of auditory cueing intervention for recovering gait, postural performance post-stroke. This review also addresses the limitations of the previously published reviews due to several of the following reasons. Firstly, the present review incorporates a higher number of experimental studies that support our conclusion i.e. 38 studies (968 participants) as compared to previously published reviews including ten (268 participants) 31 , eight (242 participants) 29 , seven (211 participants) 27 , and 2 (40 participants) 29 , studies. This large difference in the number of included studies could be affirmed to a higher number of relevant academic databases searched (with multiple languages) i.e. nine, and the inclusion of controlled clinical trials. Here, the inclusion of controlled clinical trials was justified based on the updated Cochrane guidelines for systematic reviews 85 . The guidelines recommend the addition of controlled clinical trials under the circumstances where data from randomized controlled trials is limited 86 . Secondly, this present review suggests specific training dosages with rhythmic auditory cueing for allowing enhancements in gait performance and postural stability. Thirdly, the present review provides evidence for the beneficial effects of auditory cueing training on dynamic postural stability i.e. timed-up and go test performance. Fourthly, this review provides evidence for the beneficial effects of direct application of rhythmic auditory cueing i.e. no training on gait performance in stroke patients. Lastly, this study provides evidence for the beneficial effects of adjunct training strategies like, treadmill training with rhythmic auditory cueing on gait performance in stroke patients.
Furthermore, we strongly recommend the reader to consider that it is not our intention to disregard the previously published reviews and meta-analyses. These reviews have addressed different factors in stroke recovery (quality of life, arm recovery, cognitive training, gait kinematics, applications by music therapist vs health care practitioner and more), which were not the objectives of the present review. Therefore, in our opinion interpretations should be drawn simultaneously from all the reviews to develop a better understanding of the influence of auditory cueing-based training strategies for stroke recovery.
There are four major limitations in this present review. First, this present systematic review and meta-analysis was not pre-registered in an international prospective register for systematic reviews, such as PROSPERO. Second, lack of descriptive statistics prevented us from including 13 studies in our meta-analysis i.e. out of 38 studies 25 were included. In order to address this limitation multiple attempts were made by the reviewers (S.G and I.G) to retrieve the data from the authors of the respective studies. Thirdly, this meta-analysis evaluated the effectiveness of auditory cueing training from a "pre-post intervention perspective". This is a major limitation of Scientific RepoRts | (2019) 9:2183 | https://doi.org/10.1038/s41598-019-38723-3 this study. We refrained from including a comparative analysis with the respective control groups due to limited data for the controlled groups mentioned in the studies. Fourthly, in the present meta-analysis a sensitivity analysis was performed to explore causes of heterogeneity instead of a meta-regression or stratified meta-analysis approach. The choice of this approach could raise concerns regarding the appropriateness to pinpoint the "significant" source of heterogeneity. We justify the choice of sensitivity analysis because it allowed us to simultaneously evaluate three moderators of training i.e. length of training session, number of training sessions per week and number of weeks for which training was performed. This however, was not possible with the use of a conventional meta-regression or stratified meta-analysis approach which only allows the evaluation of a single variable at a given instance.
In conclusion, rhythmic auditory cueing provides beneficial effects for enhancing gait performance and dynamic stability post-stroke. The present findings can be reliably interpreted as limited heterogeneity was ensured during the sub-group analyses, and the included studies had a "fair" overall quality i.e. 5.5. This review strongly suggests the incorporation of rhythmic auditory cueing based training post stroke for enhancing gait performance and postural stability. The review suggests a training duration for at least 20-45 minutes and for at least 3-5 times per week [87][88][89][90][91][92][93][94] .