Dropout rate in randomised controlled trials of balance and gait rehabilitation in multiple sclerosis: is it expected to be different for virtual reality-based interventions? A systematic review with meta-analysis and meta-regression

Virtual Reality Dropout rate in randomised controlled trials of balance and gait rehabilitation in multiple sclerosis: is it expected to be different for virtual reality-based interventions? A systematic review with meta-analysis and meta-regression --Manuscript Draft-- Manuscript Number: VIRE-D-22-00088R2 Full Title: Dropout rate in randomised controlled trials of balance and gait rehabilitation in multiple sclerosis: is it expected to be different for virtual reality-based interventions? A systematic review with meta-analysis and meta-regression Article Type: Original Research Keywords: Dropout rate; Multiple sclerosis; Adherence; virtual reality; Attrition. Corresponding Author: Cristina García-Muñoz Universidad de Cádiz Cadiz, Andalucía SPAIN Corresponding Author Secondary Information: Corresponding Author's Institution: Universidad de Cádiz Corresponding Author's Secondary Institution: First Author: María Jesús Casuso-Holgado First Author Secondary Information: Order of Authors: María Jesús Casuso-Holgado Cristina García-Muñoz Rocío Martín-Valero David Lucena-Anton Jose A. Moral-Munoz María-Dolores Cortés-Vega Order of Authors Secondary Information: Funding Information: Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation Dear Daniel Ballin Editor in Chief Virtual Reality Ref.: Submission ID VIRE-D-22-00088R1 Manuscript entitled: "Dropout rate in randomised controlled trials of balance and gait rehabilitation in multiple sclerosis: is it expected to be different for virtual reality-based interventions? A systematic review with meta-analysis and meta-regression". Dear Daniel Ballin, On behalf of my co-authors, I would like to express our gratitude for the comments and suggestions made by the reviewers on our manuscript entitled “Dropout rate in randomised controlled trials of balance and gait rehabilitation in multiple sclerosis: is it expected to be different for virtual reality-based interventions? A systematic review with meta-analysis and meta-regression”. In line with the reviewer’s comments, we have thoroughly reviewed the text to solve the repeated information and errors of in-text citations. Also, the manuscript has been professionally proof-reader to solve the typographical errors and unclear expressions. The professional proofreading certificate was attached to the submission platform. Finally, we remain at your disposal for any further information you may need. Yours sincerely, Prof. Cristina García-Muñoz Response to Reviewer Comments ID VIRE-D-22-00088R1 Manuscript entitled: "Dropout rate in randomised controlled trials of balance and gait rehabilitation in multiple sclerosis: is it expected to be different for virtual reality-based interventions? A systematic review with meta-analysis and meta-regression". Guest Editor’s comments: Thank you for your hard work on the revised mansucript and thorough answers to all reviewers comments. The manuscript has been accepted for publication, however, one of the reviewers brouht to our attention that the revised manuscript contains several typographical errors, unclear expressions and some errors of in-text citations. I therefore would like to ask you to conduct a thorough review of the final wording and citations and to provide a proper proof reading before publication of your manuscript. Response to Guest Editor´s Comments: On behalf of all my co-authors, I would like to thank you and the reviewers your nice work and comments that helped us to improve our manuscript. A tracked changes and clean version of the manuscript was attached to the submission platform. In line with the reviewer’s comments, we have thoroughly reviewed the text to solve the repeated information and errors in-text citations. Also, the manuscript has been professionally proof-reader to solve the typographical errors and unclear expressions. The professional proofreading certificate was attached to the submission platform. Thank you. Reviewer #2: I am happy to see that all my comments and suggestions have been considered and have improved the manuscript. Response to Reviewer#2: Thank you, your comments helped us to improve our manuscript. Reviewer #3: I think the authors have done a very good job in reacting to all the comments. Response to Reviewer#3: Thank you, your comments helped us to improve our manuscript. Reviewer #4: The manuscript has improved with the revisions submitted and I believe the authors have adequately addressed the reviewers' comments. However, there are some language and typographical errors in the manuscript which need to be corrected. Some sentences are unclear and there are instances where information is repeated between sentences. In addition, some errors are still present with the in-text citations. A thorough review and proof-reading of the manuscript is required. Response to Reviewer#4: On behalf of all the authors we are grateful for your comments, which helped us to improve the manuscript, and it to obtain the current form. In line with your comment, we have thoroughly reviewed the text to solve the repeated information and errors in-text citations. Also, the manuscript has been professionally proof-reader to solve the typographical errors and unclear expressions. The professional proofreading certificate was attached to the submission platform. 1 Title page Title Dropout rate in randomised controlled trials of balance and gait rehabilitation in multiple sclerosis: is it expected to be different for virtual reality-based interventions? A systematic review with meta-analysis and meta-regression Running Head Dropouts in multiple sclerosis for virtual reality Author’s information: María Jesús Casuso-Holgado 1. https://orcid.org/0000-0002-4217-6827 Cristina García-Muñoz2. https://orcid.org/0000-0003-2621-2098 Rocío Martín-Valero3. https://orcid.org/0000-0002-1664-3647 David Lucena-Anton2. https://orcid.org/0000-0003-2441-5342 Jose A. Moral-Munoz2, 4. https://orcid.org/0000-0002-6465-982X María-Dolores Cortés-Vega1. https://orcid.org/0000-0002-9514-8811 1. Department of Physiotherapy, University of Seville, Sevilla, Spain 2. Department of Nursing and Physiotherapy, University of Cadiz, Cadiz, Spain 3. Department of Physiotherapy, University of Malaga, Malaga, Spain 4. Institute of Research and Innovation in Biomedical Sciences of the Province of Cadiz (INiBICA), University of Cadiz, Cadiz, Spain Correspondence to: Dr. Cristina García Muñoz; [email protected] Manuscript clean Click here to access/download;Manuscript;Manuscript CLEAN.docx 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 2 ABSTRACT (258 words) Aim: To assess and meta-analyse the pooled dropout rate in absolute and comparative terms of randomised control trials using virtual reality for balance or gait rehabilitation in people with multiple sclerosis. Design: A systematic review of randomized control trials with meta-analysis and meta-regression. Data sources: A search was conducted in PubMed, Scopus, Web of Science, the Physiotherapy Evidence Database, the Cochrane Database, CINHAL, LILACS, ScienceDirect, and ProQuest. It was last updated in July 2022. Review Methods: After the selection of studies, a quality appraisal was carried out using the PEDro Scale and the Revised Cochrane risk-of-bias tool for randomised trials. A descriptive analysis of main characteristics and dropout information was performed. An overall proportion meta-analysis calculated the pooled dropout rate. Odds ratio meta-analysis compared the dropout likelihood between interventions. The meta-regression evaluated the influence of moderators related to dropout. Results: Sixteen studies with 656 participants were included. The overall pooled dropout rate was 6.6% and 5.7% for virtual reality and 9.7% in control groups. The odds ratio (0.89, p = 0.46) indicated no differences in the probability of dropouts between the interventions. The number, duration, frequency, and weeks of sessions, intervention, sex, multiple sclerosis phenotype, Expanded Disability Status Scale score, and PEDro score were not moderators (p > 0.05). Adverse events were not reported and could not be analysed as moderators. Conclusions: Dropouts across the virtual reality and control comparators were similar without significant differences. Nonetheless, there is a slight trend that could favour virtual reality. Standardisation in reporting dropouts and adverse events is recommended for future trials. PROSPERO database, registration number ID CRD42021284989 Keywords Dropout rate; multiple sclerosis; adherence; virtual reality; attrition. Statements and declarations The authors declare no conflicts of interest. Competing interests This research received no external funding. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 3 1. Introduction Different types of virtual reality technology (e.g., non-immersive, semi-immersive, or fully immersive) have emerged as an useful tool in neurorehabilitation with promising results for physical and cognitive rehabilitation (Voinescu et al. 2021). In this way, virtual reality-based interventions have been enhanced as a technological solution for telerehabilitation at the time of the COVID-19 pandemic (Matamala-Gomez et al. 2021). Furthermore, previous literature has proposed that virtual reality strategies present higher adherence in patients with neurological disorders (Asadzadeh et al. 2021; Dalmazane et al. 2021). Multitask training, patient motivation, safety, and the low cost of commercial devices are some of the benefits of using virtual reality for neurological rehabilitation (Forsberg et al. 2015; Gustavsson et al. 2021; Moan et al. 2021). Nonetheless, some undesired effects (e.g., headache, sickness, or nausea) (Massetti et al. 2018), as well as the difficulty of transferring the complex skills trained in virtual environments to the real world and the lack of ecological validity in a neurologically-impaired population (Levac et al. 2019), were reported. Specifically for balance training, the time of latency, the underestimation of perceived distances, and the dependence on specific systems (e.g., balance board) and virtual contexts were proposed as potential weaknesses of virtual reality environments (Morel et al. 2015). Multiple sclerosis is a global neurodegenerative disease affecting approximately three million people in the world (Tafti et al. 2022). Balance disorders, gait impairments, and fatigue are the main symptoms in patients with multiple sclerosis that obtain positive effects with physical therapy intervention (Amedoro et al. 2020; Abou et al. 2022). Particularly, virtual reality-based physical rehabilitation showed benefits for balance and gait training (Casuso-Holgado et al. 2018; García-Muñoz et al. 2021; Nascimento et al. 2021); however, fatigue is a significant barrier to participation in physical activity, which influences the participants’ adherence (Moore et al. 2022). A recent systematic review has summarised dropout data from randomised control clinical trials about exercise interventions in people with multiple sclerosis, concluding that mean age, the proportion of females, and intervention duration were moderators inversely associated with adherence (Dennett et al. 2020). Therefore, these findings could impact the sample size calculation, promoting an under- or overestimation. Furthermore, this could influence the differential dropout rate, which is how the degree of dropout differs between the intervention and comparator conditions after randomisation (Crutzen et al. 2015). It might affect the power of research and could present a risk of bias for randomised control clinical trials (Cooper et al. 2018). In view of this background, setting accurate expected dropout rates in virtual reality studies for rehabilitation in multiple sclerosis could help future trials to avoid problems in their internal or external validity. In addition, the identification of factors specifically associated with dropout in virtual reality trials could help clinicians when translating research into practice. As far as we are concerned, no previous systematic reviews were found reporting dropout in virtual reality interventions for balance and gait rehabilitation in this population. Thus, the present systematic review and meta-analysis aimed to: (1) systematically assess and meta-analyse the overall pooled dropout rate of randomised controlled trials using virtual reality as an intervention for balance or gait training in people with multiple sclerosis in both absolute and comparative terms; (2) analyse whether any participant or intervention factors are related to dropout; and (3) identify adverse events that could be the reason for dropouts. 2. Methods 2.1 Data sources and search strategy This systematic review was carried out following the 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Moher 2009). The review protocol was registered in the PROSPERO database (Registration number: CRD42021284989). 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 4 Two independent reviewers (M.J.C.-H., C.G.-M.) conducted an electronic search in MEDLINE (PubMed), Scopus, Web of Science (WOS), the Physiotherapy Evidence Database (PEDro), the Cochrane Database of Systematic Reviews (CDSR), CINHAL, LILACS, ScienceDirect, and ProQuest. The search was performed between July and November 2021. Neither language nor date filters were applied in the different databases. Key terms concerning intervention (‘virtual reality’, ‘game’, ‘gaming’, ‘exergaming’, and ‘interactive’), balance (‘balance’ or ‘postural control’), gait (‘gait’, ‘walking’, and ‘ambulation’), and ‘multiple sclerosis’ were combined as search terms in the strategies. The search strategy is shown in detail in Supplemental Material 1. 2.2 Research question and study selection The participants, interventions, comparisons, outcomes, and study design (PICOS) model was considered to set the following research questions: what dropout data are reported during the intervention and followup period by randomised control clinical trials conducting virtual reality intervention to improve balance or gait in multiple sclerosis and what are the possible moderators affecting dropout in these studies? Participants included in the review were female or male, aged between 18 and 65 years old, with any diagnosis of multiple sclerosis phenotype meeting the revised McDonald criteria (Thompson et al. 2018). Walking ability was preservedaccording to the Expanded Disability Status Scale (EDSS) score (EDSS ≤ 6). Included interventions involved any type of virtual reality systems aimed at improving balance or gait compared to other interventions based-on physical activity with or without external aid use. Furthermore, studies that reported dropout event information were included. 2.3 Data extraction and quality assessment First, two independent reviewers (C.G.-M. and M.J.C.-H.) identified potential articles in databases to be included in the systematic review through the title and abstract information. Next, duplicates were removed, and an exhaustive analysis of articles was carried out based on their full-text reading. This step was particularly focused on the selection criteria assessment, ensuring that the inclusion criteria were met before selecting suitable studies. In the case of disagreement, a third reviewer (M.-D.C.-V.) was consulted to decide on the inclusion of the documents. Once articles were selected, the quality assessment was conducted using the PEDro scale (Maher et al. 2003) and the Revised Cochrane risk-of-bias tool for randomised trials (RoB-2) (Higgins et al. 2019). PEDro is a reliable tool of 11 items that evaluates the inner validity of a clinical trial. If studies score above 6 points, they are classified as level I evidence (6–8: good; 8–10: excellent). If the score is below 5, they are classified as level II (4–5: deficient; <4: poor). ROB-2 allows the evaluation of bias in randomised control trials, comprising five domains (bias arising from the randomisation process, due to deviations from the intended interventions, to missing outcome data, in the measurement of the outcome, and in the selection of the reported result) that are qualified as a low or high risk of bias with some concerns (Sterne et al. 2019). Next, reviewers recorded the data for qualitative and quantitative synthesis. The extracted data were country, multiple sclerosis phenotype and disability status, female and male percentages, age, experimental and comparator group intervention characteristics, number of participants recruited and analysed, retention rate, dropout rates (for the experimental and control groups), reasons for dropout (in each group), and adverse events. Disagreements in data were solved by consensus with a third reviewer. Information provided by the included studies allowed us to calculate dropout rates in all cases, so no corresponding authors were contacted. 2.4 Data analysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 5 Dropout rate was calculated as the number of participants who did not complete the intervention and followup period divided by the total number of participants that underwent the randomisation process. Moreover, retention rate was the total number of participants that concluded the intervention, showing the adherence rate to treatment. For those studies that included more than two groups of intervention, comparison between groups was analysed separately two by two. To conduct the meta-analysis, the R Studio software (version 4.0.0) and its packages meta, metafor, and dmetar were used (Viechtbauer 2010; Balduzzi et al. 2019; Harrer et al. 2021). The proportion metaanalysis was performed through the metaprop function to determine the estimated dropout rate in virtual reality intervention, the control comparator, and all arms. Proportions were transformed using the logit transformation (Schwarzer et al. 2019). A binary meta-analysis based on odds ratios (ORs) was conducted to examine whether the probability of dropouts is higher in the virtual reality or in the comparator interventions. To assess the effect measure in binary outcomes, the OR with a 95% confidence interval (95%CI) was calculated, and the inverse variance method was used to adjust pooling estimations to sparse data (considering that dropouts are a rare event). Likewise, the Hartung-Knapp adjustment for a random effects model was implemented. Focusing on ORs, if the value is 1, there are no differences in dropouts between the experimental and comparator groups. In contrast, if the OR is greater than 1, a higher dropout rate was registered for the experimental group. The restricted maximum-likelihood estimator for tau2 was selected to estimate the between-study variance (Viechtbauer 2005). As some studies could present zero events in the experimental and/or comparator arm, a 0.5 continuity correction was added to all meta-analyses, as suggested by Gart and Zweifel (1967). Heterogeneity between studies was assessed through I2, tau2, and Cochrane’s Q (p < 0.05 indicates heterogeneity). When I2 presents a value above 50%, it means that large heterogeneity is found across studies (Higgins et al. 2021). A random effects model was employed considering the possible degree of heterogeneity between the included studies. Forest plots were used to show the outcomes of proportions and binary meta-analyses. The prediction interval was added as a red line to the forest plot to provide a measure of reliability of future treatment effects in new studies (Nagashima et al. 2019). Depending on the level of immersion of the subject within the virtual environment, virtual reality was classified as non-immersive, semi-immersive, and fully immersive for subgroup analysis. A sensitivity analysis was carried out to assess the influence of studies on the overall binary meta-analysis results. The influence was explored to detect the presence of outlier data and whether there were studies that contributed to heterogeneity or bias pooled results. A Baujat plot, a L’Abbé plot, and influence graphs were created to represent influential cases in meta-analysis. The influence graphs showed the studies that significantly influenced the pooled effect size in red. In addition, an exploratory graphical analysis of data was performed to examine whether there is a clear trend of effect size related to independent variables. Meta-regression was conducted to evaluate possible associations between participants or study characteristics which could vary in the presence of dropout events. Studies with no available data were excluded from the meta-regression analysis. Moreover, to run the meta-regression, at least three studies with the predictor were needed. The analysed moderators were interventions, number, duration, frequency and weeks of sessions, EDSS score, multiple sclerosis phenotype, and sex. Publication bias and small study effects were evaluated through a contour enhanced-funnel plot adjusted by the Duval and Tweedie trim and fill method (Shi and Lin 2020). Asymmetry in the funnel plot indicated the effect of small studies in the pooled results. To confirm the absence of asymmetry, a p-value greater than 0.05 must be reached in the Harbord’s test (Harbord et al. 2006) and the Egger bias test (Egger et al. 1997). 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 12 https://doi.org/10.7326/0003-4819-151-4-200908180-00135 Molhemi F, Mehravar M, Monjezi S, Monjezi S, Salehi R, Negahban H, Shaterzadeh-Yazdi MJ, Majdinasab N (2022) Effects of exergaming on cognition, lower limb functional coordination, and stepping time in people with multiple sclerosis: a randomized controlled trial. Disabil Rehabil 1–9. https://doi.org/10.1080/09638288.2022.2060332 Molhemi F, Monjezi S, Mehravar M, Shaterzadeh-Yazdi MJ, Salehi R, Hesam S, Mohammadianinejad E (2021) Effects of Virtual Reality vs Conventional Balance Training on Balance and Falls in People With Multiple Sclerosis: A Randomized Controlled Trial. Arch Phys Med Rehabil 102:290–299. https://doi.org/10.1016/j.apmr.2020.09.395 Moore H, Nair KPS, Baster K, Middleton R, Paling D, Sharrack B (2022) Fatigue in multiple sclerosis: A UK MS-register based study. Mult Scler Relat Disord 64:103954. https://doi.org/10.1016/j.msard.2022.103954 Morel M, Bideau B, Lardy J, Kulpa R (2015) Advantages and limitations of virtual reality for balance assessment and rehabilitation. Neurophysiol Clin 45:315–326. https://doi.org/10.1016/j.neucli.2015.09.007 Munari D, Fonte C, Varalta V, Battistuzzi E, Cassini S, Montagnoli AP, Gandolfi M, Modenese A, Filippetti M, Smania N, Picelli A (2020) Effects of robot-assisted gait training combined with virtual reality on motor and cognitive functions in patients with multiple sclerosis: A pilot, singleblind, randomized controlled trial. Restor Neurol Neurosci 38:151–164. https://doi.org/10.3233/RNN-190974 Nagashima K, Noma H, Furukawa TA (2019) Prediction intervals for random-effects meta-analysis: A confidence distribution approach. Stat Methods Med Res 28:1689–1702. https://doi.org/10.1177/0962280218773520 Nascimento AS, Fagundes CV, Mendes FA dos S, Leal JC (2021) Effectiveness of Virtual Reality Rehabilitation in Persons with Multiple Sclerosis: A Systematic Review and Meta-analysis of Randomized Controlled Trials. Mult Scler Relat Disord 54:. https://doi.org/10.1016/j.msard.2021.103128 Ozkul C, Guclu-Gunduz A, Yazici G, Atalay Guzel N, Irkec C (2020) Effect of immersive virtual reality on balance, mobility, and fatigue in patients with multiple sclerosis: A single-blinded randomized controlled trial. Eur J Integr Med 35:101092. https://doi.org/10.1016/j.eujim.2020.101092 Pagliari C, Di Tella S, Jonsdottir J, Mendozzi L, Rovaris M, De Icco R, Milanesi T, Federico S, Agostini M, Goffredo M, Pellicciari L, Franceschini M, Cimino V, Bramanti P, Baglio F (2021) Effects of home-based virtual reality telerehabilitation system in people with multiple sclerosis: A randomized controlled trial. J Telemed Telecare 1357633X2110548. https://doi.org/10.1177/1357633X211054839 Peruzzi A, Cereatti A, Della Croce U, Mirelman A (2016) Effects of a virtual reality and treadmill training on gait of subjects with multiple sclerosis: A pilot study. Mult Scler Relat Disord 5:91–96. https://doi.org/10.1016/j.msard.2015.11.002 Peruzzi A, Zarbo IR, Cereatti A, Dela Croce U, Mirelman A (2017) An innovative training program based on virtual reality and treadmill: effects on gait of persons with multiple sclerosis. Disabil Rehabil 39:1557–1563. https://doi.org/10.1080/09638288.2016.1224935 Phillips R, Hazell L, Sauzet O, Cornelius V (2019) Analysis and reporting of adverse events in randomised controlled trials: a review. BMJ Open 9:e024537. https://doi.org/10.1136/bmjopen- 2018-024537 Pitrou I, Boutron I, Ahmad N, Ravaud P (2009) Reporting of safety results in published reports of randomized controlled trials. Arch Intern Med 169:1756–1761. https://doi.org/10.1001/archinternmed.2009.306 Robinson J, Dixon J, Macsween A, van Schaik P, Martin D (2015) The effects of exergaming on balance, gait, technology acceptance and flow experience in people with multiple sclerosis: a randomized controlled trial. BMC Sports Sci Med Rehabil 7:8. https://doi.org/10.1186/s13102-015-0001-1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 13 Room J, Boulton M, Dawes H, Archer K, Barker K (2021) Physiotherapists’ perceptions of how patient adherence and non-adherence to recommended exercise for musculoskeletal conditions affects their practice: a qualitative study. Physiotherapy 113:107–115. https://doi.org/10.1016/j.physio.2021.06.001 Rose T, Nam CS, Chen KB (2018) Immersion of virtual reality for rehabilitation - Review. Appl Ergon 69:153–161. https://doi.org/10.1016/j.apergo.2018.01.009 Russo M, Dattola V, De Cola MC, Logiudice AL, Porcari B, Cannavò A, Sciarrone F, De Luca R, Molonia F, Sessa E, Bramanti P, Calabrò RS (2018) The role of robotic gait training coupled with virtual reality in boosting the rehabilitative outcomes in patients with multiple sclerosis. Int J Rehabil Res 41:166–172. https://doi.org/10.1097/MRR.0000000000000270 Schwarzer G, Chemaitelly H, Abu‐Raddad LJ, Rücker G (2019) Seriously misleading results using inverse of Freeman‐Tukey double arcsine transformation in meta‐analysis of single proportions. Res Synth Methods 10:476–483. https://doi.org/10.1002/jrsm.1348 Shi L, Lin L (2020) The trim-and-fill method for publication bias: practical guidelines and recommendations based on a large database of meta-analyses. Medicine (Baltimore) 98:e15987. https://doi.org/10.1097/MD.0000000000015987 Sterne JAC, Savović J, Page MJ, Elbers RG, Blencowe NS, Boutron I, Cates CJ, Cheng HY, Corbett MS, Eldridge SM, Emberson JR, Hernán MA, Hopewell S, Hróbjartsson A, Junqueira DR, Jüni Kirkham JJ, Lasserson T, Li T, McAleenan A, Reeves BC, Shepperd S, Shrier I, Stewart LA, Tilling K, White IR, Whiting PF, Higgins JPT (2019) RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ l4898. https://doi.org/10.1136/bmj.l4898 Tafti D, Ehsan M, Xixis KL (2022) Multiple Sclerosis. Treasure Island (FL): StatPearls Publishing Thompson AJ, Banwell BL, Barkhof F, Carroll, Coetzee T, Comi C, Correale J, Fazekas F, Filippi M, Freedman MS, Fujihara K, Galetta SL, Hartung HP, Kappos L, Lublin FD, Marrie RA, Miller AE, Miller DH, Montalban X, Mowry EM, Sorensen PS, Tintoré M, Traboulsee AL, Trojano M, Uitdehaag BMJ, Vukusic S, Waubant E, Weinshenker BG, Reingold SC, Cohen JA (2018) Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria. Lancet Neurol 17:162– 173. https://doi.org/10.1016/S1474-4422(17)30470-2 Tollar J, Nagy F, Toth BE, Torok K, Szita K, Csutoras B, Moizs M, Hortobagyi T (2020) Exercise Effects on Multiple Sclerosis Quality of Life and Clinical-Motor Symptoms. Med Sci Sports Exerc 52:1007–1014. https://doi.org/10.1249/MSS.0000000000002228 Torous J, Lipschitz J, Ng M, Firth J (2020) Dropout rates in clinical trials of smartphone apps for depressive symptoms: A systematic review and meta-analysis. J Affect Disord 263:413–419. https://doi.org/10.1016/j.jad.2019.11.167 Viechtbauer W (2010) Conducting Meta-Analyses in R with the metafor Package. J Stat Softw 36:. https://doi.org/10.18637/jss.v036.i03 Viechtbauer W (2005) Bias and Efficiency of Meta-Analytic Variance Estimators in the Random-Effects Model. J Educ Behav Stat 30:261–293. https://doi.org/10.3102/10769986030003261 Voinescu A, Sui J, Stanton Fraser D (2021) Virtual reality in neurorehabilitation: An umbrella review of meta-analyses. J Clin Med 10:. https://doi.org/10.3390/jcm10071478 Wright I, Mughal F, Bowers G, Meiser-Stedman R (2021) Dropout from randomised controlled trials of psychological treatments for depression in children and youth: a systematic review and metaanalyses. J Affect Disord 281:880–890. https://doi.org/10.1016/j.jad.2020.11.039 Yazgan YZ, Tarakci E, Tarakci D, Ozdincler AR, Kurtuncu M (2020) Comparison of the effects of two different exergaming systems on balance, functionality, fatigue, and quality of life in people with multiple sclerosis: A randomized controlled trial. Mult Scler Relat Disord 39:101902. https://doi.org/http://dx.doi.org/10.1016/j.msard.2019.101902 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 1 Title page Title Dropout rate in randomised controlled trials of balance and gait rehabilitation in multiple sclerosis: is it expected to be different for virtual reality-based interventions? A systematic review with meta-analysis and meta-regression Running Head Dropouts in multiple sclerosis for virtual reality Author’s information: María Jesús Casuso-Holgado 1. https://orcid.org/0000-0002-4217-6827 Cristina García-Muñoz2. https://orcid.org/0000-0003-2621-2098 Rocío Martín-Valero3. https://orcid.org/0000-0002-1664-3647 David Lucena-Anton2. https://orcid.org/0000-0003-2441-5342 Jose A. Moral-Munoz2, 4. https://orcid.org/0000-0002-6465-982X María-Dolores Cortés-Vega1. https://orcid.org/0000-0002-9514-8811 1. Department of Physiotherapy, University of Seville, Sevilla, Spain 2. Department of Nursing and Physiotherapy, University of Cadiz, Cadiz, Spain 3. Department of Physiotherapy, University of Malaga, Malaga, Spain 4. Institute of Research and Innovation in Biomedical Sciences of the Province of Cadiz (INiBICA), University of Cadiz, Cadiz, Spain Correspondence to: Dr. Cristina García Muñoz; [email protected] Formatted: Spanish (Spain) Formatted: Spanish (Spain) Formatted: Spanish (Spain) Manuscript with tracked changes Click here to access/download;Manuscript;MANUSC~2.DOC 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 2 1. ABSTRACT (2580 words) Aim: To assess and meta-analyse the pooled dropout rate in absolute and comparative terms of randomised controlled trials using virtual reality for balance or gait rehabilitation in people with multiple sclerosis. Design: A systematic review of randomized controlled trials with meta-analysis and meta-regression. Data sources: A search was conducted in PubMed, Scopus, Web of Science, the Physiotherapy Evidence Database, the Cochrane Database, CINHAL, LILACSilacs, ScienceDirect, and ProQuest. It was last updated in July 2022. Review Methods: After the selection of studies, a quality appraisal was carried out using the PEDro Scale and the Revised Cochrane risk-of-bias tool for randomised trials (ROB-2). A descriptive analysis of main characteristics and dropout information was performed. An overall proportion meta-analysis calculated the pooled dropout rate. Odds ratio meta-analysis compared the dropout likelihood between interventions. The meta-regression evaluated the influence of moderators related to dropout. Results: Sixteen16 studies with 656 participants were included. The overall pooled dropout rate was 6.6 % and 5.7% for virtual reality and 9.7% in control groups. The odds ratio (0.89, p = 0.46) indicated no differences in the probability of dropouts between the interventions. The number, duration, frequency, and weeks of sessions, intervention, sex, multiple sclerosis phenotype, Expanded Disability Status Scale score, and PEDro score were not moderators (p > 0.05). Adverse events were not reported and could not be analysed as moderators. Conclusions: Dropouts across the virtual reality and control comparators were similar, without significant differences. Nonetheless, there is a slight trend that could favour virtual reality. Standardisation in reporting dropouts and adverse events is recommended for future trials. PROSPERO database, registration number ID CRD42021284989 Keywords Dropout rate; Mmultiple sclerosis; Aadherence; Vvirtual reality; Aattrition. Statements and Ddeclarations The authors declare no conflicts of interest. Competing iInterests This research received no external funding. Formatted: No bullets or numbering Formatted: Indent: Left: 0", First line: 0" 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 3 2.1. Introduction Different types of virtual reality technology (e.ge.g., non-immersive, semi-immersive, or fully immersive) haves emerged as an useful tool in neurorehabilitation with promising results for physical and cognitive rehabilitation (Voinescu et al. 2021). MoreoverIn thisat way, virtual reality-based interventions have been enhanced as a technological solution for telerehabilitation at the time of the COVID-19 pandemic (Matamala-Gomez et al. 2021). Furthermore, pPrevious literature has proposed that virtual reality strategies present higher adherence in patients with neurological disorders (Asadzadeh et al. 2021; Dalmazane et al. 2021). Multitask training, patient motivation, safety, and the low cost of commercial devices are some of the benefits of using virtual reality for neurological rehabilitation (Forsberg et al. 2015; Gustavsson et al. 2021; Moan et al. 2021). Nonetheless, Despite these potential benefits, the recent literature has also highlighted some limitations of the use of virtual reality with rehabilitative purposes. These limitations are undesired effects (e.g., headache, sickness, or nausea) (Massetti et al. 2018), along withas well as the difficulty of transferring the complex skills trained in virtual environments to the real world or and thea lack of ecological validity in a neurologically-impaired population (Levac et al. 2019), were reported. Specifically for balance training, the time of latency, the underestimation of perceived distances, and or the dependence onto specific systems (e.g., balance board) and virtual contexts have beenwere proposed as potential weaknesses of virtual reality environments (Morel et al. 2015). Multiple sclerosis (MS) is a global neurodegenerative disease affecting approximately three million people in the world (Tafti et al. 2022). Balance disorders, gait impairments, and or fatigue are core the main symptoms in these peoplepatients with multiple sclerosis that, obtaining. The evidence is positive effects regarding the efficacy of with physical therapy intervention for the rehabilitation of patients with multiple sclerosis (Amedoro et al. 2020; Abou et al. 2022). Particularly, for virtual reality-based physical rehabilitation different types of virtual systems have shown showed benefits for balance and gait training (Casuso-Holgado et al. 2018; García-Muñoz et al. 2021; Nascimento et al. 2021; García-Muñoz et al. 2021);. Hhowever, fatigue is a significant barrier to participation in physical activity, which for this population that could influenceinfluencesing the participants’‘ adherence when they are enrolled to physical interventions (Moore et al. 2022). In that way, aAA recent systematic review has summarised dropout data from randomised controlled clinical trials about exercise interventions in people with multiple sclerosis, concluding that mean age, the proportion of females, and intervention duration were moderators are inversely associated with adherence (Dennett et al. 2020). However, no previous systematic reviews exist for dropout in virtual reality interventions for balance and gait rehabilitation in this population. Therefore, The aforementioned findings these findings could impact on the sample size calculation, promotinge an under- or overestimationed sample size calculation respectively and also differential dropout. Furthermore, this could influence the differential dropout rate, This occurs when in which is how the degree of dropout differs between the intervention and comparator conditions after randomisation (Crutzen et al. 2015).. The differential dropout rate might It might affect the power of research and could present a risk of bias for randomised controlled clinical trials (Cooper et al. 2018). In view of this background, Ssetting accurate expected dropout rates in virtual reality studies for rehabilitation in multiple sclerosis could help future trials to avoid problems in their internal or external validity. In addition, the identification of factors specifically associated with dropout in virtual reality trials could help clinicians when translating research into practice. As far as we are concerned, no previous systematic reviews were found reporting dropout in virtual reality interventions for balance and gait rehabilitation in this population. Thus, the present systematic review and meta-analysis has three aimsaimed to: (1) to systematically assess and meta-analyse the overall pooled dropout rate of randomised controlled trials using virtual reality as an intervention for balance or gait training in people with multiple sclerosis in both absolute and comparative terms; (2) to analyse whether any participant or intervention factors are related to dropout; and (3) to identify adverse events that could be the reason for dropouts. Field Code Changed Formatted: English (United States) Formatted: English (United States) Formatted: English (United States) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 4 3.2. Methods 2.13.1 Data sources and search strategy This systematic review was carried out following the 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Moher 2009). The review protocol was registered in the PROSPERO database (Registration Nnumber: CRD42021284989). Two independent reviewers (M.J.C.-H., C.G.-M.) conducted an electronic search in MEDLINE (PubMed), Scopus, Web of Science (WOS), the Physiotherapy Evidence Database (PEDro), the Cochrane Database of Systematic Reviews (CDSR), CINHAL, LILACSilacs, ScienceDirect, and ProQuest. The search was performed between July and November 2021. Neither language nor date filters were applied in the different databases. Key terms concerning intervention (‘“virtual reality’”, ‘“game’”, ‘“gaming’”, ‘“exergaming’”, and ‘“interactive’”), balance (‘“balance’” or ‘“postural control’”), gait (‘“gait’”, ‘“walking’”, and ‘“ambulation’”), and ‘“multiple sclerosis’” were combined as search terms in the strategies. The search strategy is displayed shown in detail in Supplementalry Material 1. 23.2 Research question and study selection To set the research question, The PICOS model (participants, interventions, comparisons, outcomes, and study design (PICOS) model) recommendations werewas considered to set the following research questionsas follows: what dropout data are reported during the intervention and follow-up period by randomised control clinical controlled trials conducting aan intervention through virtual reality intervention to improve balance or gait in multiple sclerosis and what are the possible moderators that affectaffecting dropout in these studies? Participants included in the review were female or male, aged between 18 and 65 years old, with any diagnosis of multiple sclerosis phenotype meeting the revised McDonald criteria (Thompson et al. 2018). Walking ability is was preservedconserved according to the Expanded Disability Status Scale (EDSS) score (EDSS ≤ 6). Included interventions involved any type of virtual reality systems aimed at improving balance or gait compared to other interventions based-on physical activity with or without an external aid use. Furthermore, those studies thatwhich reported dropout events information were included. 23.3 Data extraction and quality assessment First, two independent reviewers (C.G.-M. and M.J.C.-H.) identified potential articles in databases to be included in the systematic review through the title and abstract information. Following thisNext, duplicates were removedremoved, and an exhaustive analysis of articles was carried out based on their full-text reading. This step was particularly focused on the selection criteria checking assessment, and both reviewers ensured ensuring that the inclusion criteria were met before selecting suitable studies. In the case of disagreement, a third reviewer (M.-D..C.-V.) was consulted to decide on the inclusion of the documents. When Once articles were selected, thea quality assessment was conducted using the Physiotherapy Evidence Database PEDro Sscale (PEDro)(Maher et al. 2003) and the Revised Cochrane risk-of-bias tool for randomised trials (RoB-2) (Higgins et al. 2019). PEDro is a reliable tool of 11 items that evaluates the inner validity of a clinical trial. If studies are scored above 6 points on the PEDro Scale, they are considered to have aclassified as level I of evidence (6–8: good; 8–10: excellent)., and iIf while level II is when the score is below 5, they are classified as level II (4–5: deficient; <4: poor). ROB-2 is the updated and Formatted: Tab stops: 4.05", Left Formatted: Font: Italic 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 5 improved version of Cochrane ROB and allows the evaluation of bias in randomised controlled trials, . This tool is ccomprisinged of five domains (bias arising from the randomisation process, due to deviations from the intended interventions, to missing outcome data, in the measurement of the outcome, and in the selection of the reported result) that are qualified as a low or high risk of bias with some concerns (Sterne et al. 2019). Next, reviewers recorded the data for qualitative and quantitative synthesis. For both analyses extracted data wereThe extracted data extracted were country, multiple sclerosis phenotype and disability status, female and male percentages, age, experimental and comparator group interventions characteristics, number of participants recruited and analysed, retention rate, dropout rates (for the experimental and control groups), reasons for dropout (oin each group), and adverse events. Disagreements in data were solved by consensus with a third reviewer. Information provided by the included studies allowed us to calculatehe calculation of dropout rates in all cases, so none corresponding authors wereas contacted. 23.4 Data analysis Dropout rate was calculated as the number of participants who did not complete the intervention and followup period divided by the total number of participants that underwent the randomisation process. On the other handMoreover, retention rate was the total number of participants that concluded the intervention, showing the adherence rate to treatment. Rates for both conditions were calculated. For those studies that included more than two groups of intervention, comparison between groups have beenwas analysed separately two by two. To conduct the meta-analysis, the R Studio software (version 4.0.0) was selected and its packages meta, metafor, and dmetar were employed used (Viechtbauer 2010; Balduzzi et al. 2019; Harrer et al. 2021). The proportion meta-analysis was performed through the metaprop function, to determine the estimated dropout rate in virtual reality intervention, the control comparator, and all arms. Proportions were transformed using the logit transformation (Schwarzer et al. 2019). A binary meta-analysis based on odds ratios (ORs) was conducted to examine whether the probability of dropouts is higher in the virtual reality or in the comparator interventions. To assess the effect measure in binary outcomes, the OR with a 95% confidence interval (95%CI) was calculated, and the inverse variance method was used to adjust pooling estimations to sparse data (considering that dropouts are a rare event). Likewise, the Hartung-Knapp adjustment for a random effects model was implemented. Focusing on ORs, if the value is 1, there are no differences in dropouts and dropout exist between the experimental and comparator groups interventions. In contrast, if the OR is greater than >1, a higher dropout rate will bewas registered for the virtual reality interventionexperimental group. The restricted maximum-likelihood estimator for tTau2 was selected to estimate the between-study variance (Viechtbauer 2005). As some studies could present zero events in the experimental and/or/and comparator arm and in both, a 0.5 continuity correction was added to all meta-analyses, as suggested by Gart and Zweifel (Gart and Zweifel (1967). Heterogeneity between studies was assessed through the I2, tTau2,tau square and Cochrane’s Q (p < 0.05 indicates heterogeneity). When I2 presents a value above 50%, it means that large heterogeneity is found across studies (Higgins et al. 2021). A random effects model was employed considering the possible degree of heterogeneity between the included studies. To display the outcomes of proportions and binary meta-analyses, fForest plots were performedused to show the outcomes of proportions and binary meta-analyses. The prediction interval was added as a red line to the forest plot to provide a measure of reliability of future treatment effects in new studies (Nagashima et al. 2019). Depending on the level of immersion of the subject within the virtual environment, virtual reality was distinguished inclassified ason non-immersive, semi-immersive, and fully immersive Formatted: Font: Not Italic Formatted: Not Superscript/ Subscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 6 forr a subgroup analysis. However, immersive virtual reality and semi-immersive virtual reality interventions were excluded due to the low number of studies. A sensitivitye analysis was carried out to assess the influence of studies on the overall binary meta-analysis results. The influence was explored to detect the presence of outlier data and whether there will bewere studies thatwhich contributed to heterogeneity or bias pooled results. To represent influential cases in metaanalysis, a A Baujat plot, a L’Abbé plot, was carried out and an influence graphs wereas created to represent influential cases in meta-analysis. where studies which The influence graphslast showed the studies that significantly influenced the ed coloured in red the studies influencing significantly thea bias of pooled effect size in red were shown in red. AlsoIn addition, an exploratory graphical analysis of data was performed to examine whether there is a clear trend of effect size related to independent variables. Meta-regressions wasere conducted to evaluate possible associations between participants or study characteristics which could vary in the presence of dropout events. Studies with no available data were excluded from the meta-regression analysis. Moreover, to run the meta-regression, at least three studies with the predictor were needed. The analysed moderators were interventions, number, duration, frequency and weeks of sessions, EDSS Expanded Disability Status Scale score, multiple sclerosis phenotype, and sex. Publication bias and small study effects were evaluated through a contour enhanced-funnel plot adjusted by the Duval and Tweedie trim and fill method (Shi and Lin 2020). Asymmetry in the funnel plot will indicated the effect of small studies in the pooled results. To confirm the absence of asymmetry, a pvalue greater than > 0.05 must be reached in the Harbord’s test (Harbord et al. 2006) and in the Egger bias test (Egger et al. 1997). 4.3. Results 34.1 Study selection and methodological quality assessment In total, 7,024 articles were identified through the initial database search based on titles and abstracts. After that, duplicates were removed, reducing the overall number of results toobtaining 5,995 articles. Once the studies underwent the screening and eligibility steps, 16 randomised controlled trials that met the eligibility criteria were included for the qualitative synthesis and quantitative analysis. There was no disagreement between reviewers in the study selection process. Figure 1 displays showed the PRISMA flowchart which detailsdetailing the selection procedure [insert Figure 1.]. Excluded studies and theirits reasons are were detailed in Supplementalry Material 2. Regarding to the quality assessments, the PEDro scale results are shown in Supplementalry Material 3. PEDro scores were reported from the included studies:; thirteen with level I evidence (Lozano-Quilis et al. 2014; Hoang et al. 2016; Kalron et al. 2016; Calabrò et al. 2017; Peruzzi et al. 2017; Russo et al. 2018; Khalil et al. 2019; Munari et al. 2020; Ozkul et al. 2020; Tollar et al. 2020; Munari et al. 2020; Molhemi et al. 2021; Pagliari et al. 2021; Molhemi et al. 2022) and three with level II .(Brichetto et al. 2015; Robinson et al. 2015; Yazgan et al. 2020). Most of the studies were single blinded, in which thewith the assessor beingwas blinded to participant allocation. In addition, the ROB-2 summary overall score reported that most of the researchstudies presented some concerns, but only three studies (Robinson et al. 2015; Ozkul et al. 2020; Yazgan et al. 2020) had a ‘“high risk’” of bias (Fig.ure 2) [insert Figure 2.]. Disagreements between reviewers occasionally occurred sometimes for domain 2, but consensus was always reached without the participation of the third reviewer. Field Code Changed Formatted: Spanish (Spain) Formatted: Spanish (Spain) Field Code Changed Formatted: Font: Not Bold 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 7 34.2 Study design and population characteristics The main characteristics of the participants and the interventions can be foundwere shown in Table 1. The randomised pooled randomised population obtained from the reviewed studies reached a total of 656 participants with a mean EDSS scorexpanded Disability Status Scale of 4.22 (95%CI 4.15–4.30). From the overall population, 65.57% of subjects were female. The mean age was 45.12 (95%CI 44.66–45.59) and 65.57% of the population were female. All studies involved patients with relapsing-remitting type, eExcept for three studies which did not specify the phenotype of multiple sclerosis (Robinson et al. 2015; Kalron et al. 2016; Pagliari et al. 2021), all studies involved patients with relapsing-remitting type. Furthermore, eight studies (Lozano-Quilis et al. 2014; Brichetto et al. 2015; Hoang et al. 2016; Munari et al. 2020; Tollar et al. 2020; Munari et al. 2020; Yazgan et al. 2020; Molhemi et al. 2021, 2022) involved. pParticipants with any type of multiple sclerosis (relapsing-remitting, secondary progressive, and primary progressive) without subgroup analysise multiple sclerosis. were allocated to the two intervention groups in eight studies (Lozano-Quilis et al. 2014; Brichetto et al. 2015; Hoang et al. 2016; Tollar et al. 2020; Munari et al. 2020; Yazgan et al. 2020; Molhemi et al. 2021, 2022). Concerning the immersion of the virtual reality systems, From 16 studies, 14 studies employed nonimmersive virtual reality as the main experimental intervention, and four 4of them specifically carried out the intervention through used the Wii Fit system (Brichetto et al. 2015; Robinson et al. 2015; Khalil et al. 2019; Yazgan et al. 2020). Only two trials used the fully immersive virtual reality (Kalron et al. 2016; Ozkul et al. 2020). Most of studies compared the virtual reality intervention to improve balance or gait to conventional balance training (n = 13, 81.25%) (Lozano-Quilis et al. 2014; Brichetto et al. 2015; Robinson et al. 2015; Hoang et al. 2016; Kalron et al. 2016; Peruzzi et al. 2016; Kalron et al. 2016; Calabrò et al. 2017; Russo et al. 2018; Khalil et al. 2019; Ozkul et al. 2020; Molhemi et al. 2021, 2022; Pagliari et al. 2021; Molhemi et al. 2022), followed by robotic-assisted gait training (n = 3, 18.75%). (Calabrò et al. 2017; Peruzzi et al. 2017; Munari et al. 2020). The lowest number of sessions performed was 8 (Robinson et al. 2015), while the highest was 54 (Russo et al. 2018). Most authors proposed a frequency of intervention of 2 times per week with a minimum time per session of 30 minutes (Hoang et al. 2016; Kalron et al. 2016) and a maximum of 85 minutes (Calabrò et al. 2017). The mean number of for dropout events for the EG experimental group was 1.61 cases and 1.88 for the comparator group was 1.88 cases. TheA highestr number of dropouts in the virtual reality groups were registered by Hoang et al. (Hoang et al. (2016) and Pagliari et al. (Pagliari et al. (2021). The reasons reportedDropout reasons disclosed by the authors for dropout in both experimental and comparator groups were: difficulties reaching the research centre, transportation problems, schedulinge problems, moving to another city, refusal to participate, personal or familial issues, lack of motivation or time, loss of data due to administrative problems, exacerbation of symptoms, disease relapse, work intensity, and illness/medical reasons/hospitalisation not related to multiple sclerosis. Three of sixteen studies did not report any dropout events during the intervention or follow-up period (Brichetto et al. 2015; Calabrò et al. 2017; Russo et al. 2018). 34.3 Meta-analysis of proportions A total of 18 arms (k) from 16 studies were included in the proportion and binary meta-analysis; it is because, since one of the RTC randomised controlled trials presented three groups of interventionstudy groups (Tollar et al. 2020). From a total of 638 participants, a total of 63 cases of dropouts were reported. The forest plot indicated showed an overall pooled dropout rate of 6.6% (95%CI 3.2% –12.9%) not due towithout heterogeneity between studies (tau2 = 1.18, Q = 10.07, df = 17, I2 = 0%, 95%CI 0%–50%, p = 0.90) (Fig.ure 3) [insert Figure 3.]. The dropout rate for the virtual reality-based interventions was 5.7% (95% CI 2.3%–13.6%) against the 9.7% (95% CI 5.7%–-16.02%) registered for participants allocated in the comparator groups. The specific forest plots of the dropout rates for virtual reality and comparators are Field Code Changed Formatted: Spanish (Spain) Formatted: Spanish (Spain) Formatted: Not Superscript/ Subscript Formatted: Not Superscript/ Subscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 8 displayed in (Supplemental Material 4). Conversely, the final retention rate for the virtual reality and comparator groups were 94.3% and 90.3% of adherence, respectively. None of the prediction intervals calculated across the meta-analysis suggested that the treatment intervention would achieve the same effects in the future. 34.4 Binary meta-analysis (ORs) The likelihood of dropouts between the intervention of virtual reality and control comparators was evaluated by the binary meta-analysis through ORs. The main results showed a slightly lower probability that dropouts occurred in the virtual reality-based interventions than in the comparator groups, but a significant difference was not reached obtained (OR = 0.89, 95%CI 0.64–1.24, p = 0.46). No significant heterogeneity between studies was found (tau2 = 0, Q = 5.6, df = 17, I2 = 0%, 95%CI 0%–50%, p = 0.99) (Fig.ure 4) [insert Figure 4.]. The prediction interval failed to confirmed that the same effects would not happen in future studies. A subgroup meta-analysis sort by type ofaccording to the immersion level of the virtual reality was not carried out, because of the number of studies using immersive virtual realitysystems did not reach the minimum required (3 studies).of three studies. A post-hoc sensitive analysis using the L’Abbé and Baujat plots and influence graphs (Supplemental Material 5) showed that none of the included studies influenced heterogeneity or bias for the pooled effect size (see Supplementary Material 5), and . Also, no outliers were found. Additionally, no small study effects or publication bias were reflected shown in the contour-enhanced funnel plot (Fig.ure 5) [insert Figure 5], the Harbord test (p = 0.37), or theand Egger bias test (p = 0.34) [insert Figure 5]. 34.5 Metaregression The meta-regression revealed that the type of intervention, number, frequency, and duration of session, weeks of intervention, EDSSxpanded Disability Status Scale score, multiple sclerosis phenotype, sex, and methodological quality could not be related to the dropout events. A detailed description of the analysis is givenwas shown in Table 2. 5.4. Discussion A total of 16 randomised controlled trials that reportedreporting dropouts were subjected to meta-analysedis to calculate the overall pooled dropout rate of virtual reality-based interventions for the improvement of balance and gait in patients with multiple sclerosis. The main clinical implication highlighted by our resultsof the results of our study is was that the virtual reality-based training for balance and gait in people with multiple sclerosis is was highly accepted with a low dropout rate and high adherence duringto the studyies period. ButIn that way, as Torous et al. (Torous et al. (2020) suggested, that the retention in research contexts could change when experimental approaches are translated into a clinical setting. This could be especially important for long rehabilitation programmes in chronic conditions. A recent study (Hortobágyi et al. 2022) has reported a high adherence rate to a twoyears maintenance exercise program including exergaming in people with multiple sclerosis (Hortobágyi et al. 2022);. Hhowever, the sample size was very small, and more research about long-term adherence to virtual reality rehabilitation in this population is needed. Adherence is one of the main conflicts faced in rehabilitation; the therapeutic approach of multiple sclerosis is not an exception. As a result, looking for rehabilitation therapies that achieve higher participant compliance to treatment is vital (Arafah et al. 2017). If correct adherence is not achieved, the effectiveness of the rehabilitation might be limited and incur additional healthcare costs (Jack et al. 2010; Room et al. 2021). Accordingly, Pprevious literature has proposed that virtual reality strategies presented higher adherence in patients with neurological disorders (Asadzadeh et al. 2021; Dalmazane et al. 2021). Nonetheless, our results suggested lower dropout rates in virtual reality-based interventions, which may be Formatted: Not Superscript/ Subscript Formatted: Not Superscript/ Subscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 PRISMA 2020 flow diagram for new systematic reviews which included searches of databases and registers only Records identified from Databases (n= 7024) CINAHL (n= 58) Cochrane Library (n=179) PEDro (n= 10) ProQuest (n= 4986) PubMed (n=99) ScienceDirect (n=1328) Scopus (n= 162) WOS (n=202) Records removed before screening: Duplicate records removed (n= 1029) Records screened by title and abstract (n=5995) Records excluded based on title/abstract by a human (n=5943) Reports sought for retrieval (n =52) Reports assessed for eligibility (n = 52) Reports excluded (n= 36): Non-randomized clinical controlled trial design (n=15) Do not use virtual reality (n=3) Comparator is not based-on physical activity (n=5) Same information in another article (n= 1) Systematic review (n=4) Focus on upper limbs rehabilitation (n=4) Not in English language (n=1) Digital strategy (n= 1) Study protocol (n=2) Studies included in review (n= 16) Identification of studies via databases and registers Identification Screening Included Figure 1. Flow diagram of trials selection based on PRISMA 2020 guidelines. Figure 1 Click here to access/download;Figure;Figure 1 PRISMA Flowchart.docx PRISMA 2020 flow diagram for new systematic reviews which included searches of databases and registers only *Consider, if feasible to do so, reporting the number of records identified from each database or register searched (rather than the total number across all databases/registers). **If automation tools were used, indicate how many records were excluded by a human and how many were excluded by automation tools. From: Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71. doi: 10.1136/bmj.n71 For more information, visit: http://www.prisma-statement.org/ Figure 2. Cochrane risk of bias tool-2 summary. Figure 2 Click here to access/download;Figure;Figure 2 ROB-2.docx Figure 3. Forest plot of dropout rate for all groups of studies. Figure 3 Click here to access/download;Figure;Figure 3 Forest prop all.docx Figure 4. Forest plot of odds ratio comparing attrition from virtual reality intervention and other comparator interventions in people with multiple sclerosis to improve balance or gait. Figure 4 Click here to access/download;Figure;Figure 4 Forest OR all.docx Figure 5. Contour-enhanced funnel plot. Figure 5 Click here to access/download;Figure;Figure 5 Funnel plot.docx Table 1. Characteristic of studies included in the systematic review Study/ Country MS phenotype/ EDSS (mean; SD) Recruited /Analyzed (n) % Sex/ Age (mean ± SD) Experimental intervention Control group intervention Retention rate (%) Dropout rate (%) Reason for dropouts (EG/CG) Adverse events Brichetto et al. 2015 Italy 19 RRMS 9 SPMS 4 PPMS EDSS = 3.7 ± 1.2 EG: 16/16 CG: 16/16 F: 28.13% M: 71.88% 50.5 ± 11.6 12 sessions (60 min and 3 s/w, 4 weeks) Exergaming through Nintendo Wii Fit Balance Board, plus balance exercises in Balance Master Neurocom 12 sessions (60 min and 3 s/w, 4 weeks) Conventional balance training 100 % (32/32) DOEG: 0% (0/16) DOCG: 0% (0/16) --- NR Calabrò et al. 2017 Italy RRMS EDSS= 4.56 EG: 20/20 CG: 20/20 F: 62.5% M: 37.5 % 42.5 40 sessions (5 s/w, 8 weeks) Standard physical treatment (5 min of warning up, 5 min of strengthening, 20 min of postural control exercises) + 40 min of Lokomat + VR (avoid obstacles or catch objects on the trail) 40 sessions (5 s/w, 8 weeks) Standard physical treatment (5 min of warning up, 5 min of strengthening, 20 min of postural control exercises) + 40 min of Lokomat 100 % (40/40) DOEG: 0% (0/20) DOCG: 0% (0/20) ---- No adverse or harmful events during the intervention Table 1 Click here to access/download;Table;Table 1 Main characteristics.docx Hoang et al. 2015 Australia RRMS: 26 SPMS: 12 PPMS: 10 Unknown: 2 EDSS= 4.15 ± 1.3 EG: 28/23 CG: 22/21 F: 76% M: 24% 52.4 ± 11.75 24 sessions (30 min; 2 s/w, 12 weeks) Exergames (Stepmania and Choice stepping reaction time; Home step training system.) 24 sessions (30 min; 2 s/w, 12 weeks) Conventional balance training + stretching + strength exercises 88% (44/50) DOEG: 17.9% (5/28) DOCG: 4.5% (1/22) Discontinued intervention due to personal circumstances (EG), relapse (EG), health problems during reassessment not related to MS (CG) No adverse or harmful events during the intervention Kalron et al. 2016 Israel EDSS= 4.1 ± 1.3 EG: 16/15 CG: 16/15 F: 63.33% M: 36.67% 45.2 ± 11.6 12 sessions (30 min; 2s/w, 6 weeks) Immersive virtual reality system CAREN 12 sessions (30 min; 2s/w, 6 weeks) conventional balance training 10 min of stretching + 20 min of training (static postural control, weight shifting and perturbation exercises) 93.75 % (30/32) DOEG: 6.3% (1/16) DOCG: 6.3% (1/16) EG/CG: Difficulties in arrival to the MS center No adverse or harmful events during the intervention Khalil et al. 2018 Jordan RRMS: 40 EDSS= 3 ± 1.25 EG: 20/16 CG: 20/16 F: 68.75% M: 31.25% 37.38 ± 10.87 18 sessions (3s/w, 6 weeks) Exergame through Wii Fit and Microsoft Kinect sensor allows to 18 sessions (3s/w, 6 weeks) Home-base conventional balance training 80% (32/40) DOEG: 20% (4/20) DOCG: 20% (4/20) EG: Lack of family support, lack of outcome expectation, need to travel long distance, No adverse or harmful events during the intervention interact with six VR scenarios CG: not provided reason, lack of time and motivation Lozano- Quilis et al. 2014 Spain RRMS SPMS EDSS = NR EG: 6/6 CG: 6/5 F: 58.33% M: 41.67% 44.82± 10 10 sessions (45 standard rehabilitation + 15 min of virtual reality training; 1s/w, 10 weeks) RemoviEMVR system 10 sessions (60 min; 1s/w, 10 weeks) Conventional balance and gait training 91.67% (11/12) DOEG: 0% (0/6) DOCG: 16.7% (1/6) NR No adverse or harmful events during the intervention Molhemi et al. 2021 Iran RRMS: 30 SPMS: 9 EDSS: 4.8 EG: 19/19 CG: 20/20 F: 61.54% M: 38.46% 39.2 ± 8.4 18 sessions (35 min; 3s/w for 6 weeks) Exergame with Microsoft Kinect 18 sessions (35 min; 3s/w for 6 weeks) Conventional balance training 82.05% (32/39) DOEG: 15.8% (3/19) DOCG: 20% (4/20) EG/CG (Duringintervention): Difficulties in arrival to the research center, work schedules problems and transport problems, fall data EG: illness and exacerbation of symptoms Non adverse event during intervention Proofreading certificate Click here to access/download Supplementary Material CERTIF~1.PDF COVER LETTER Click here to access/download Supplementary Material Cover letter.docx Appendix Click here to access/download Supplementary Material Appendix.pdf PRISMA CHECKLIST Click here to access/download Supplementary Material PRISMA_2020_checklist.docx