The Effectiveness of Virtual Reality for Rehabilitation of Parkinson Disease: An Overview of Systematic Reviews and Meta-Analyses

Background: An increasing number of systematic reviews (SRs) and meta-analyses (MAs) of clinical trials have begun to investigate the effects of virtual reality (VR) in patients with Parkinson disease (PD). The aim of this overview of was to systematically summarize the current best evidence for the effectiveness of VR therapy for the rehabilitation of people with PD. Methods: We searched SRs/MAs based on randomized controlled trials (RCTs) for relevant literature in PubMed, Embase and Cochrane library databases from inception to December 5, 2020. The methodological quality of included SRs/MAs was evaluated with the Assessing the Methodological Quality of Systematic Reviews 2 (AMSTAR-2), and the evidence quality of outcome measures with the Grading of Recommendations, Assessment, Development and Evaluation (GRADE). Results: A total of nine SRs/MAs were included. The evaluation with AMSTAR-2 showed that all included SRs/MAs but one were rated as low or critically low quality studies. The GRADE criteria revealed 19 studies with very-low-quality evidences, 20 with low-quality evidences, 8 with moderate-quality evidences, and 1 with high-quality evidence. Effectiveness evaluation showed that VR therapy had greater improvement of stride length compared with control groups. However, there were inconsistent results (better or similar effects) regarding to gait speed, gait ability, balance, global motor function, activities of daily living, quality of life, postural control, cognitive function, and neuropsychiatric symptoms. Conclusions: VR therapy appears to be a promising and effective treatment for PD, but there is still a lack of high-quality evidence. In the future, rigorous-designed, high-quality RCTs with larger sample sizes are needed to further verify the effectiveness of VR therapy in the treatment of PD. VR therapy compared with AT (WMD = -0.18, 95% CI = -1.37 to 1.00, 4 RCTs, 120 patients[19]; WMD = -2.86, 95% CI = -5.60 to -0.12, 5 RCTs, 144 patients[20]). Only one SR/MA[18] found that VR therapy was superior than AT or PT in the gait ability using 10-Meter Walk Test (10-WMT) (WMD = 0.13, 95%CI = 0.02 to 0.24, 4 RCTs, 174 patients). One SR/MA[20] showed no signicantly greater increases in walk distance using 6-Minute Walk Test (6-WMT) when VR therapy compared with AT (WMD = 8.91, 95% CI = −43.32 to 61.13, 2 RCTs, 45 patients). Moreover, one SR/MA[17] indicated that Nintendo Wii (SMD = 2.44, 95% CI = 0.48 to 4.38, 4 RCTs, 125 patients) had immediate positive effects on functional locomotion (measured by gait speed and the time that a person takes to complete certain locomotion tasks: TUG and 10-WMT), but Xbox Kinect (SMD = -0.32, 95% CI = -0.98 to 0.35, 3 RCTs, 111patients) didn’t observe this phenomenon.


Background
Parkinson disease (PD) is the most common progressive neurodegenerative disease worldwide. PD prevalence is increasing with age and PD affects 1% of the population above 60 years [1]. It is estimated that by around 2030, the number of PD patients in China will reach 5 million, accounting for about 50% of the total number of PD patients in the world [2].PD is characterized by motor symptoms such as rest tremor, bradykinesia, rigidity and postural instability, which affect gait, balance and movement quality, leading to di culty in performing basic daily activities and quality of life and placing a heavy burden on families and society [3]. Multidisciplinary input is increasingly recognized as important in the treatment and management of PD [4].Currently, drugs and surgical approaches were main treatments of PD. Clinically approved drug treatments for PD mainly include levodopa, dopaminergic receptor agonists, and monoamine oxidase-B inhibitors. Levodopa is considered as " rst line" drug, but the long-term use of it leads to many complications [5]. Deep brain stimulation may be an effective treatment in PD patients; however, clinical trials have shown that it may have cognitive and psychiatric side effects [6]. Rehabilitation treatment is considered as an adjuvant to pharmacological and surgical treatments for PD to improve many dysfunctions and self-care ability, even delay the progression of the disease.
Virtual reality (VR) has emerged as a promising technology for researching complex impairments in people with PD and for providing personalized rehabilitation [7]. This technology typically combines real-time motion detection within a virtual environment in the context of a (video)game. The user can perceive, feel and interact with virtual environments, viewing an avatar (a character or graphical representation of the user) that mimics the user's movements [8] by multiple sensory channels such as sight, sound and touch [9]. Immediate feedback about performance and success is provided both concurrently (during game play) and terminally (at the end of the game). VR therapy attempt to promote activity-dependent neuroplasticity and motor learning [10,11]. Recently, numerous systematic reviews (SRs) and meta-analyses (MAs) based on randomized controlled trials (RCTs) regarding the clinical effectiveness of VR therapy in the treatment of PD have been published. However, the overall results have remained mixed or inconclusive and their quality is uneven. An overview of SRs/MAs is a relatively new method that aims to support clinical decision-making by synthesizing the ndings, critically appraising the quality and attempting to resolve discordant outcomes. Therefore, we conducted an overview of SRs/MAs to identify and summarize the existing evidence and to systematically determine the clinical effectiveness of using VR therapy to treat PD.

Methods
The overview was completed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [12] and the guidelines recommended by the Cochrane Collaboration [13]. The PRISMA checklist can be found in Additional le 1. The protocol was not prospectively registered.

Search Strategy
We systematically searched PubMed, Embase and Cochrane library databases from inception to December 5, 2020. We used a combination of Medical Subject Headings with Entry Terms, or EMTREE with keywords as follows: Parkinson Disease,virtual reality exposure therapy, systematic review, and metaanalysis. In addition, to ensure a comprehensive data collection, references of relevant reviews were searched manually to identify additional eligible studies. The search strategy for the PubMed database is shown in Additional le 2.

Eligibility Criteria
Types of studies: The overview included reviews, which had to be clearly identi ed by the authors as a 'systematic review' or 'meta-analysis' in either the title or abstract of the review. SRs/MAs having a systematic search strategy, covering RCTs and published in English were included.
Types of participants: Participants involved in reviews were clinically de nite diagnosis of PD and were de ned by the UK Parkinson's Disease Society Brain Bank or other diagnostic criteria. We had no restrictions on gender, age, drug dosage, duration and severity of PD. We included reviews reporting an intervention carried out in a mixed sample of participants if data for participants with PD were provided separately.
Types of interventions: Intervention groups were VR therapy alone or in combination with active interventions (AT) or passive interventions (PT). AT involved conventional exercise therapy, sensory integration balance training, neurodevelopment treatment, functional electrical stimulation, stationary cycling, maintenance of usual activities, cognitive training or usual care. PT included either health education or no intervention.
Types of controls: Control groups was de ned as either AT or PT without a VR component.
Types of outcome measures: reviews that assessed the motor and non-motor symptoms of PD as the outcome indicators were considered eligible.
The exclusion criteria included: (1) studies which had mixed samples (PD, stroke, multiple sclerosis, cerebral palsy or other neurological disorders) cannot extract data separately; (2) studies where PD patients all used VR without control group or control group was healthy individuals; (3) studies where PD patients with different symptoms (freezers vs. non-freezers) underwent the same VR therapy; (4) reviews, guidelines, conference abstracts, surveys, commentaries, editorials, letters, and notes.

Study Selection
All titles and abstracts were initially screened by two independent investigators (L.Y.Q and G.Y.G) after automatically removing duplicate results to identify potentially relevant studies for inclusion. At this stage, we excluded studies that were not focused on effects of VR therapy on PD patients or not described as SRs/MAs. Further, full-text articles were reviewed and selected according to eligibility criteria. We excluded reviews that did not present summary statistics for outcomes (effect size with 95%CIs). Final relevant studies were shortlisted. In case of discrepancies, consensus was achieved by discussion. If consensus could not be reached, a third reviewer (Y.Y.S) was consulted.

Data Extraction
Two investigators (L.Y.Q and G.Y.G) extracted the following basic characteristics from each eligible review: the rst author, publication year, country, the number of included studies, sample size, interventions (experiment interventions and control interventions), outcomes, quality assessment tools, main conclusions. Differences between the review authors were settled by discussion, and a third reviewer (Y.Y.S) was consulted if differences persisted. The study authors were contacted with the aim of acquiring additional information on the data presented.

Quality assessment
Quality of methodology of the each included review was evaluated using the Assessing the Methodological Quality of Systematic Reviews 2 (AMSTAR-2) tool [14]. AMSTAR-2 is a comprehensive critical appraisal tool for SRs/MAs of randomized and non-randomized studies that focuses on weaknesses in critical domains but not an overall score. The tool assesses 16 items, among which 7 are critical domains (item 2, 4, 7, 9, 11, 13 & 15). The evaluation is reduced to three options, "Yes", "Partial Yes" and "No". AMSTAR-2 classi es the overall con dence on the results of the review into four levels: high, moderate, low, and critically low.
The quality of the evidence was rated by the Grading of Recommendations Assessment, Development and Evaluation (GRADE) guideline according to ve items (inconsistency, imprecision, indirectness, quality, and publication bias) [15]. Results are divided into four levels: high, moderate, low and very low.
Two independent investigators (L.Y.Q and G.Y.G) conducted the assessment process of methodology quality and evidence quality. Discrepancies between investigators were resolved by discussion or by a third reviewer (Y.Y.S) in cases when a consensus was not reached.

Statistical Analysis
We carried out descriptive analyses of the methodology quality and evidence quality of the included SRs/MAs. We further described the summary measures (standard mean difference (SMD) or weighted mean difference (WMD), both with 95% CI, number of RCTs and patients) of the effectiveness of VR therapy for

Search Results
A ow diagram of study screening and selection procedures is illustrated in Fig. 1. Our electronic search yielded 92 potentially relevant publications. After automatic removal of duplicates, 54 records were screened on the basis of the title or abstract. Of the remaining 37 reviews, 28 reviews were excluded: participants were not PD (n = 6), intervention was not VR (n = 1), SRs/MAs were not based on RCTs (n = 8), conference abstracts lacked full text (n = 3), necessary data were not extracted (n = 9), and SRs/MAs were not English language (n = 1) after reading the full text. Finally, nine SRs/MAs [16][17][18][19][20][21][22][23][24] met the inclusion criteria and were included in this overview.

Study Characteristics
The characteristics of the nine SRs/MAs included in our nal analysis are summarized in Table 1. All studies were published between 2015 and 2020, including four articles from China [18][19][20]22], and one from UK [16], Italy [17], Brasil [21], Belgium [23], Japan [24] each. The number of apposite studies included in each review ranged from 2 to 16, and the sample sizes ranged from 74 to 574. The interventions in the experiment groups were mainly VR therapy alone or VR therapy plus AT, and the control groups were either AT or PT. The methodological quality assessment scales varied across the included SRs/MAs: ve SRs/MAs used the Cochrane Collaboration's tool, and four SRs/MAs used the PEDro scale. The conclusions from these SRs/MAs differed, but at least showed that VR therapy can achieve similar even better performance in some outcomes in patients with PD. Detailed information on the methodological quality of included SRs/MAs was provided in Table 2. AMSTAR-2 score showed that one (11.1%) review [22] was of moderate quality, three (33.3%) [19,20]were low, and that of all the others [16-18, 21, 24] (55.6%) were critically low. The key factors affecting the quality of the literature included item 2 (only two reviews [22,23] had registered and had a protocol before performing the review.), item 4 ( ve reviews [16,19,20,23,24] used a comprehensive literature search strategy with searching references of relevant reviews or searching relevant gray literature), item 7 (only one review [23] provided a list of excluded studies and justi ed the exclusions), item 9 (all reviews [16][17][18][19][20][21][22][23][24] reported risk of bias use a satisfactory technique), item 11 (all reviews [16][17][18][19][20][21][22]24] conducted statistical combination of results using appropriate methods), item 13 (all reviews [16][17][18][19][20][21][22][23][24] accounted for risk of bias in the primary studies when interpreting the results of the reviews), and item 15 (three reviews [19,20,22] carried out an adequate investigation of publication bias study and discuss its impact on the review).  Q15: If they performed quantitative synthesis did the review authors carry out an adequate investigation of publication bias (small study bias) and discuss its likely impact on the results of the review?
Q16: Did the review authors report any potential sources of con ict of interest, including any funding they received for conducting the review?

Evidence Quality of Outcomes
The results of the GRADE assessment were summarized below in Table 3. Figure 2 provided a graphic representation of the evidence base. Among these 48 outcome indicators, there were one (2.1%) high-quality evidence, eight (16.7%) moderate-quality evidences, 20 (41.7%) low-quality evidences, and 19 (39.6%) very low-quality evidences. Reasons for downgrading the quality of evidence to moderate or low even very low included poor methodological quality as a large   The had a large bias in random, distributive ndings or was blind. : The con dence intervals overlapped less, the P-value for the heterogeneity test was very smal : The con dence interval was not narrow enough. : Fewer studies were included, and publication bias cannot be carried out.

Effectiveness Evaluation
Gait Gait (composite measure). Only one SR/MA [23] involving four RCTs with a total of 129 participants with PD found no signi cant difference between VR therapy and AT (SMD = 0.20, 95%CI = -0.14 to 0.55) in gait as a composite measure.
Perceived con dence in balance. Two SRs/MAs [19,22] reported the perceived con dence in balance, of which one SR/MA [19] reported ABC scale changes and illustrated that VR resulted in no signi cant difference when VR was compared with AT (WMD = 1.69; 95% CI = −2.62 to 6.01, 2 RCTs, 115 patients). The other SR/MA [22] included three RCTs with 104 patients compared VR with AT or PT indicated that VR had a signi cant effect on perceived con dence in balance than AT or PT (SMD = -0.73, 95%CI = -1.43 to -0.03).

Activities of Daily Living
Three SRs/MAs [16,18,22] reported on activities of daily living, two of which concluded that there was no statistically signi cant difference between VR therapy and AT in activities of daily living, assessed by Section II of the Uni ed Parkinson Disease Rating Scale (UPDRS-II) (SMD = − 0.13, 95%CI = − 0.82 to 0.57, 1 RCT, 32 patients [16]; SMD = 0.25, 95%CI = -0.14 to 0.64, 4 RCTs, 103 patients [22] The possible mechanism of therapeutic effect of VR therapy for PD in stride length are as follows: (1) A decrease in stride length is an example of deterioration of motor automaticity in PD patients. PD patients have to rely on attentional control helping to perform motor skills to bypass automatic control mechanisms [27]. VR therapy provided more accurate and complete motor feedback and therefore enabled better stride amplitude correction than traditional physiotherapy [7]. (2) PD patients revealed the typical gait asymmetry with a signi cant difference between stride lengths of both legs. Dissociation of the visual and proprioceptive inputs can be manipulated in VR so that people with PD step to a target that is visually perceived to be of a smaller range of motion than is actually achieved, thereby training their motor systems to produce larger movements during subsequent trials [28].
In addition, we investigated potential causes of inconsistent results for outcome as follows: (1)  lacked a uni ed standard, for instance, balance performance was measured using BBS, DGI, FGA, ABC, etc. While BBS is considered to be a robust measure of balance performance [29], it is also characterized by substantial oor and ceiling effects [30].
From this overview, we found that the methodological quality and evidence quality of included SRs/MAs were unsatisfactory. According to AMSTAR-2, the methodological quality of all included SRs/MAs but one [22] were low or critically low. Only two SRs/MAs [22,23] had been registered protocols. The lack of registration may result in a great adjustment of the study process than expected, increasing the risk of bias and affecting the rigor of the SRs/MAs. Four SRs/MAs [17,18,21,22] didn't take into account the gray literatures or the references of included studies in the retrieval process, which may cause publication bias and language bias. All included SRs/MAs but one [23] did not provide a list of excluded trials with reasons for exclusions, which may undermine the transparency of the SRs/MAs and affect the reliability of their results. The sources of funding are not reported in all included SRs/MAs except for one [23], which may reduce the credibility of the research results due to potential con icts of interest. Six SRs/MAs [16-18, 21, 23, 24] didn't carry out an adequate investigation of publication bias and discuss its likely impact on the results of the review, which may affect the judgment of the validity of the analysis results.
Future studies could avoid these obvious de ciencies to improve methodological quality. According to GRADE, evidence quality of each outcome measurement was between moderate and very low. Most SRs/MAs (81.3%) were with low or very low-quality evidence. The most common among the downgrading factors in the included SRs/MAs was imprecision, which was re ected in the wide con dence interval and the small sample size. Next was publication bias, limitations and inconsistency, which was re ected in insu cient search strategy, large defects in the method design of random, blinding and allocation concealment, larger heterogeneity, etc. Based on this, there may be a certain degree of difference between the conclusions of the included MAs and the true results.

Strengths And Limitation Of The Overview
To the best of our knowledge, our study is the rst overview of MAs to explore the effect of VR therapy on PD rehabilitation, which may have certain reference value for the clinical practice. In addition, the ndings of this overview were based on relatively recent evidence, as all studies were published in the last ve years. Moreover, this overview included MAs of RCTs using strict inclusion standards, and excluded non-RCTs, observational cohort studies or MAs without extracting data in order to reduce the risk of bias. However, this study has several limitations. First, the methodological quality and evidence quality of the included MAs were generally low; thus, results based on primary studies should be interpreted with caution. Furthermore, we only searched English databases, so MAs published in other languages that met the inclusion criteria may have been missed.

Conclusion
The current evidence based on nine SRs/MAs suggests that VR therapy may be a promising complementary treatment for PD patients. This conclusion must be interpreted cautiously, given the generally low methodological quality and low evidence quality of the included SRs/MAs. Rigorous-designed, high-quality RCTs with larger sample sizes are needed to further verify the effectiveness of VR therapy in the treatment of PD.