Virtual Reality-Based Therapy Improves Fatigue, Impact, and Quality of Life in Patients with Multiple Sclerosis. A Systematic Review with a Meta-Analysis

Patients with multiple sclerosis (PwMS) have a high level of fatigue and a reduced quality of life (QoL) due to the impact of multiple sclerosis (MS). Virtual reality-based therapy (VRBT) is being used to reduce disability in PwMS. The aim of this study was to assess the effect of VRBT on fatigue, the impact of MS, and QoL in PwMS. Methods: A systematic review with meta-analysis was conducted through a bibliographic search on PubMed, Scopus, Web of Science, and PEDro up to April 2021. We included randomized controlled trials (RCTs) with PwMS that received VRBT in comparison to conventional therapy (CT) including physiotherapy, balance and strength exercises, and stretching or physical activity, among others; or in comparison to simple observation; in order to assess fatigue, MS-impact, and QoL. The effect size was calculated using Cohen’s standardized mean difference with a 95% confidence interval (95% CI). Results: Twelve RCTs that provided data from 606 PwMS (42.83 ± 6.86 years old and 70% women) were included. The methodological quality mean, according to the PEDro Scale, was 5.83 ± 0.83 points. Our global findings showed that VRBT is effective at reducing fatigue (SMD −0.33; 95% CI −0.61, −0.06), lowering the impact of MS (SMD −0.3; 95% CI −0.55, −0.04), and increasing overall QoL (0.5; 95% CI 0.23, 0.76). Subgroup analysis showed the following: (1) VRBT is better than CT at reducing fatigue (SMD −0.4; 95% CI −0.7, −0.11), as well as in improving the mental dimension of QoL (SMD 0.51; 95% CI 0.02, 1); (2) VRBT is better than simple observation at reducing the impact of MS (SMD −0.61; 95% CI −0.97, −0.23) and increasing overall QoL (SMD 0.79; 95% CI 0.3, 1.28); and (3) when combined with CT, VRBT is more effective than CT in improving the global (SMD 0.6, 95% CI 0.13, 1.07), physical (SMD 0.87; 95% CI 0.3, 1.43), and mental dimensions (SMD 0.6; 95% CI 0.08, 1.11) of QoL. Conclusion: VRBT is effective at reducing fatigue and MS impact and improving QoL in PwMS.


Introduction
Multiple sclerosis (MS) is a chronic, inflammatory, immune-mediated, and currently incurable disease that affects the central nervous system (CNS) [1]. It results in demyelination, glial reaction, and axonal loss [2]. MS is the leading cause of disability by chronic neurological disease in young adults [3], affecting more than 2.5 million people worldwide [4], with a prevalence of 36 cases per 100,000 people [5]. In Europe, MS shows a prevalence of 83 cases per 100,000 habitants, with an average annual incidence of 4.3 cases

Inclusion Criteria
The study selection stage was carried out by two authors (I.C.-P. and E.O.-G.) who independently screened the titles and abstracts of all studies retrieved by the search strategy from each database. Studies selected by at least one author were considered eligible for inclusion in this systematic review and were reviewed in detail. A third author (F.A.N.-E.) was consulted when a study raised doubts about its inclusion. The inclusion criteria applied were: (1) RCT or RCT pilot; (2) participants were PwMS; (3) the study design included at least two groups; (4) one group received an intervention with VR and the second group CT or NI; (5) the study aimed to assess the effect of VR on fatigue, MS-impact, or QoL; and (6) the study provided quantitative data about the variables of interest for the meta-analysis. The exclusion criterion was RCTs including different neurological diseases apart from PwMS. ). The following data were extracted: (1) authorship, publication date, study design, country, and funding received; (2) data related to participants (number of PwMS, age and sex); (3) experimental intervention characteristics (VR therapy length in weeks, number of sessions per week, and session time in minutes); (4) type of control intervention; (5) quantitative data obtained at the post-therapy evaluation (mean and standard deviation); and (6) follow-up time (immediate or long-term). Regarding quantitative data, when a study did not provide standard deviation, we estimated this measure using standard error, interquartile range, or range, using standardized transformations according to the Cochrane Handbook for Systematic Reviews of Interventions [33] and previous studies [18].

Outcomes
The outcomes assessed in the present review were: level of fatigue, impact of MS, and QoL. The selected studies provided data from different tests for each outcome (see outcomes section in Results).

Risk of Bias Assessment and Quality Evidence
The PEDro scale was independently used by two authors (M.S.-A. and Y.C.-C.) to assess the risk of bias and the methodological quality of the included studies. The PEDro scale comprises 11 items (item 1 is not used for the total score), with a score ranging from 0 (very low methodological quality and high risk of bias) to 10 (high methodological quality and low risk of bias) [34]. A study was considered high quality if it scored equal to or higher than 8 points [35].
In addition, the level of evidence of each meta-analysis was analyzed using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) metric. According to Meader (2014) [36], a level of evidence is conditioned by its risk of bias, inconsistency, imprecision, indirectness, and risk of publication bias. Inconsistency was assessed by estimating the level of heterogeneity (see statistical analysis section); imprecision was calculated from the number of participants per study and the number of studies in each meta-analysis, and indirectness was noted in articles in which the results were measured indirectly, registered as a "yes" or "no" [33]. Finally, the level of evidence was scored as follows: (1) high, if findings were robust; (2) moderate, if results might change after including new studies; (3) low, if the level of confidence in our pooled effect was very slight; and (4) very low, when any effect estimation was robust because some of Meader's items were not present in the studies included in the meta-analysis. Two authors (I.C.-P. and F.A.N.-E.) independently assessed the level of evidence of each meta-analysis and doubts were discussed with a third author (M.C.O.-P.).

Statistical Analysis
A meta-analysis was performed by two authors using Comprehensive Meta-Analysis version 3.0 (Biostat, Englewood, NJ, USA) [37] (E.O.-G. and I.C.-P.). The effect was estimated in a random effect of DerSimonian and Laird [38] using Cohen's standardized mean difference (SMD) [39] with a 95% confidence interval (95% CI), according the guidelines established by Cooper et al. [40]. Cohen's SMD can be interpreted as a four-level strength effect: no effect (SMD 0), small (SMD 0.2-0.4), medium (SMD 0.4-0.7) and large (SMD > 0.8) [41]. The result of each meta-analysis was displayed in forest plots [42]. Red diamonds represent the overall results of the meta-analysis, either from the subgroup analysis performed (subtotals) or from the set of all groups (total). The center of the diamond is the overall effect value and the width represents the overall confidence interval. The difference between the intervention and control groups can be considered statistically significant if the diamond is clearly positioned to one side of the reference line, but if it crosses it or just rubs it, no conclusions can be drawn from that point in one direction. The p-value for Egger's test (with p < 0.1 showing a risk of publication bias) [43], the visualization of the funnel plot [44] (which in cases of asymmetry indicates a possible risk of publication bias), and trim-and-fill estimation [45] were used to estimate the risk of publication bias. When the trim-and-fill estimation reported a variation higher than 10% with respect to the original pooled effect, the level of evidence was downgraded one level [46]. The level of heterogeneity was calculated by using the Q-test and its p-value (p < 0.1 indicates the existence of heterogeneity) and the degree of inconsistency (I 2 ) established by Higgins [47], where the level of heterogeneity can be rated as low (I 2 < 25%), moderate (I 2 between 25-50%), or large (I 2 > 50%) [47].

Sensitivity Analysis
The leave-one-out method (or one-study-removed method) was employed to assess the contribution or weight of each study to the global effect in each meta-analysis [33].

Subgroup Analysis
A subgroup analysis [33] was conducted to assess the effect of VR according to the comparisons conducted in the included studies. These comparisons showed the following: (1) VR vs. NI; (2) VR vs. CT; and (3) VR + CT vs. CT.

Study Selection
We identified 179 studies from different databases (PubMed n = 23, Scopus n = 75, WOS n = 60, and PEDro n = 21) and another eight additional records were identified from other sources. After duplicated studies were removed (n = 128), 59 studies were screened by title/abstract. Fourteen studies were excluded in the first screening and thirty-three were excluded afterwards, as they did not meet the inclusion criteria (reasons in Figure 1). Finally, 12 studies were included in the present systematic review with a meta-analysis [48][49][50][51][52][53][54][55][56][57][58][59]. Figure 1 shows the PRISMA flow chart for the study selection process.

Effect of Virtual Reality-Based Therapy on the Impact of Multiple Sclerosis
Five RCTs [54][55][56][57][58] provided data from 287 PwMS to assess the effect of VR-based therapy on the impact of MS in comparison to CT or NI. At first, an overall analysis showed low-quality of a low-to-medium effect favoring VR-based therapy (SMD −0.3; 95% CI −0.55, −0.04; p 0.02) on the impact of MS compared to CT or NI (Table 4, Figure 4), with a low level of heterogeneity (I 2 21%; Q-test = 8.9, df = 7; p 0.26) and a low risk of publication bias (p for Egger = 0.07 and 50% of variation after trim-and-fill estimation) (Supplementary Figure S2). Sensitivity analysis reported a variation of 41% with respect to the original effect, excluding a study by Tollár [58].

Effect of Virtual Reality-Based Therapy on the Impact of Multiple Sclerosis
Five RCTs [54][55][56][57][58] provided data from 287 PwMS to assess the effect of VR-based therapy on the impact of MS in comparison to CT or NI. At first, an overall analysis showed low-quality of a low-to-medium effect favoring VR-based therapy (SMD −0.3; 95% CI −0.55, −0.04; p 0.02) on the impact of MS compared to CT or NI (Table 4, Figure 4), with a low level of heterogeneity (I 2 21%; Q-test = 8.9, df = 7; p 0.26) and a low risk of publication bias (p for Egger = 0.07 and 50% of variation after trim-and-fill estimation) (Supplementary Figure S2). Sensitivity analysis reported a variation of 41% with respect to the original effect, excluding a study by Tollár [58].

Effect of Virtual Reality-Based Therapy on Overall Quality of Life
Six RCTs [49][50][51]53,57,59] reported data from 291 PwMS to assess the effect of VRbased intervention on overall QoL in comparison to CT or simple observation. An initial overall analysis provided moderate-quality evidence of a medium effect favoring VRbased therapy (SMD 0.5; 95% CI 0.23, 0.76; p < 0.001) (Table 4, Figure 6), without heterogeneity (I 2 2%; Q-test = 7.1, df = 5; p 0.21) and no risk of publication bias (p for Egger test 0.2) (Supplementary Figure S3). The one study removed showed a variation of 20% with respect to the original pooled effect when a study by Yazgan [53] was excluded.

Effect of Virtual Reality-Based Therapy on Overall Quality of Life
Six RCTs [49][50][51]53,57,59] reported data from 291 PwMS to assess the effect of VRbased intervention on overall QoL in comparison to CT or simple observation. An initial overall analysis provided moderate-quality evidence of a medium effect favoring VR-based therapy (SMD 0.5; 95% CI 0.23, 0.76; p < 0.001) (Table 4, Figure 6), without heterogeneity (I 2 2%; Q-test = 7.1, df = 5; p 0.21) and no risk of publication bias (p for Egger test 0.2) (Supplementary Figure S3). The one study removed showed a variation of 20% with respect to the original pooled effect when a study by Yazgan [53] was excluded.

Effect of Virtual Reality-Based Therapy on Overall Quality of Life
Six RCTs [49][50][51]53,57,59] reported data from 291 PwMS to assess the effect of VRbased intervention on overall QoL in comparison to CT or simple observation. An initial overall analysis provided moderate-quality evidence of a medium effect favoring VRbased therapy (SMD 0.5; 95% CI 0.23, 0.76; p < 0.001) (Table 4, Figure 6), without heterogeneity (I 2 2%; Q-test = 7.1, df = 5; p 0.21) and no risk of publication bias (p for Egger test 0.2) (Supplementary Figure S3). The one study removed showed a variation of 20% with respect to the original pooled effect when a study by Yazgan [53] was excluded.

Effect of Virtual Reality-Based Therapy on the Mental Dimension of Quality of Life
Three studies [49,50,59] reported data to assess the effect of VR-based intervention on the mental dimension of QoL (Table 4, Figure 9). Low-quality evidence of a medium effect favoring VR-based intervention (SMD 0.55; 95% CI 0.09, 1.01; p 0.018), without low heterogeneity (I 2 6.2%; Q-test = 2.1, df = 2; 0.35) and with a possible risk of publication bias, (p for Egger test 0.45 and 31% of variation after trim-and-fill estimation) was shown. In subgroup analysis, one study [59] showed low-quality evidence of a medium effect favoring VR + CT (SMD 0.6; 95% CI 0.08, 1.11; p 0.025) when compared with CT. Moreover, when VR-based intervention was compared with CT in two studies [49,50], low-quality evidence of a medium effect (SMD 0.51; 95% CI 0.02, 1; p 0.042) was shown favoring VRbased intervention, without heterogeneity (I 2 0%; Q-test = 1, df = 1; p 0.32).

Effect of Virtual Reality-Based Therapy on the Mental Dimension of Quality of Life
Three studies [49,50,59] reported data to assess the effect of VR-based intervention on the mental dimension of QoL (Table 4, Figure 9). Low-quality evidence of a medium effect favoring VR-based intervention (SMD 0.55; 95% CI 0.09, 1.01; p 0.018), without low heterogeneity (I 2 6.2%; Q-test = 2.1, df = 2; 0.35) and with a possible risk of publication bias, (p for Egger test 0.45 and 31% of variation after trim-and-fill estimation) was shown.

Discussion
In recent years, some studies have assessed the efficacy of different therapies to reduce the impact of MS and its muscle fatigue symptoms, as well as to increase the QoL of MS patients [60,61]. VR-based intervention is a novel therapy that is being used more in the treatment of neurological diseases [18,62], including MS. To date, a number of published reviews have assessed the effect of VR-based intervention in PwMS regarding balance, gait [29,30], and upper extremity recovery [31]. In addition, a recent review [63], which included a small number of studies, evaluated the effect of VR-based intervention on fatigue and QoL. For this reason, the present work was conceived to compile and analyze the more updated and recent evidence available thus far on the efficacy of VR-based therapy regarding such variables. Our systematic review with meta-analysis includes 12 RCTs published in the last 9 years that provide data from 606 patients with MS (42.83 ± 6.86 years), and evaluate the effect of VR-based intervention on fatigue (7 studies), the impact of MS (5 studies), and QoL (6 studies), differentiating between physical and mental dimensions. In the studies included we identified three different comparisons: (1) VR vs. NI; (2) VR vs. CT; and (3) VR + CT vs. CT. In an overall analysis, our findings showed that VR-based intervention reduces the level of fatigue and the impact of MS, and increases the QoL in PwMS. Specifically, the subgroup analysis revealed that: (1) compared with simple observation, VR-based intervention may be effective to minimize the impact of MS and to increase the overall QoL; (2) VR-based intervention reduces fatigue more than CT; and (3) VR-based intervention with CT is more effective that isolated CT to increase the physical and mental dimensions of the QoL.
Regarding fatigue, approximately 15% of PwMS consider fatigue as the most frequent and disabling symptom that reduces QoL and personal autonomy [64]. Therefore, it is important to find therapies that are able to reduce it. In this case, our overall analysis showed that VR-based intervention produces a low-to-medium effect in reducing fatigue in PwMS. In addition, when subgroup analysis was conducted, we found that VR was better than CT at reducing this variable. Our findings are in line with the recent review of Santos-Nascimento [63], although our meta-analysis on the effect of VR-based interventions on fatigue, includes seven studies (five more than the previous review), which means that our review is able to estimate the effect with data from different tests using Cohen's SMD, increasing the generalization and quality of evidence of the effect of VR vs. CT in reducing fatigue. However, no statistically significant differences were found when VR was compared with simple observation and when VR-based intervention was used with CT compared to CT alone. It is important to note that these two subgroups included a small number of studies (three and one, respectively) and it is possible that these results will change if new studies are included. It is accepted that in order to develop new therapies that improve muscle strength, muscle oxygenation, and heart function parameters, it is also important to reduce fatigue. Physical exercise has been postulated as an excellent therapy to reduce fatigue as consequence of the improvements in muscle resistance, heart rate, and respiratory frequency, all of which increase the physical condition of patients

Discussion
In recent years, some studies have assessed the efficacy of different therapies to reduce the impact of MS and its muscle fatigue symptoms, as well as to increase the QoL of MS patients [60,61]. VR-based intervention is a novel therapy that is being used more in the treatment of neurological diseases [18,62], including MS. To date, a number of published reviews have assessed the effect of VR-based intervention in PwMS regarding balance, gait [29,30], and upper extremity recovery [31]. In addition, a recent review [63], which included a small number of studies, evaluated the effect of VR-based intervention on fatigue and QoL. For this reason, the present work was conceived to compile and analyze the more updated and recent evidence available thus far on the efficacy of VR-based therapy regarding such variables. Our systematic review with meta-analysis includes 12 RCTs published in the last 9 years that provide data from 606 patients with MS (42.83 ± 6.86 years), and evaluate the effect of VR-based intervention on fatigue (7 studies), the impact of MS (5 studies), and QoL (6 studies), differentiating between physical and mental dimensions. In the studies included we identified three different comparisons: (1) VR vs. NI; (2) VR vs. CT; and (3) VR + CT vs. CT. In an overall analysis, our findings showed that VR-based intervention reduces the level of fatigue and the impact of MS, and increases the QoL in PwMS. Specifically, the subgroup analysis revealed that: (1) compared with simple observation, VR-based intervention may be effective to minimize the impact of MS and to increase the overall QoL; (2) VR-based intervention reduces fatigue more than CT; and (3) VR-based intervention with CT is more effective that isolated CT to increase the physical and mental dimensions of the QoL.
Regarding fatigue, approximately 15% of PwMS consider fatigue as the most frequent and disabling symptom that reduces QoL and personal autonomy [64]. Therefore, it is important to find therapies that are able to reduce it. In this case, our overall analysis showed that VR-based intervention produces a low-to-medium effect in reducing fatigue in PwMS. In addition, when subgroup analysis was conducted, we found that VR was better than CT at reducing this variable. Our findings are in line with the recent review of Santos-Nascimento [63], although our meta-analysis on the effect of VR-based interventions on fatigue, includes seven studies (five more than the previous review), which means that our review is able to estimate the effect with data from different tests using Cohen's SMD, increasing the generalization and quality of evidence of the effect of VR vs. CT in reducing fatigue. However, no statistically significant differences were found when VR was compared with simple observation and when VR-based intervention was used with CT compared to CT alone. It is important to note that these two subgroups included a small number of studies (three and one, respectively) and it is possible that these results will change if new studies are included. It is accepted that in order to develop new therapies that improve muscle strength, muscle oxygenation, and heart function parameters, it is also important to reduce fatigue. Physical exercise has been postulated as an excellent therapy to reduce fatigue as consequence of the improvements in muscle resistance, heart rate, and respiratory frequency, all of which increase the physical condition of patients [65,66].
However, the level of adherence is sometimes low, and PwMS face some barriers such as low functional capability, fear of falling, or difficulties in attending rehabilitation centers. VR-based interventions are based on the active performance of physical and functional exercises through video games that can be adapted in intensity, which means that the characteristics of a VR-based exercise can be adapted to take into account the patient's state of fatigue. This enables continuous training, even at home, increasing patient motivation and thus, probably, the effectiveness of VR therapy [67]. In addition, increasing patient motivation in VR therapy may increase their adherence to the therapy, thus favoring continuous training to improve muscular endurance and reduce movement fatigue [67]. This high level of adherence to VR therapy may be one of the causes of major improvements in fatigue in a VR group compared to a CT group. Sometimes, classical neurorehabilitation protocols are based on monotonous and passive CT exercises, while VR-based intervention allows for an enjoyable and customized therapy that involves the active immersion of patients and continuous work.
Our meta-analysis demonstrates that the use of VR-based therapies in neurorehabilitation protocols reduces the disabling impact of MS symptoms, although with a low effect. Subgroup analysis reported a medium effect of VR-based intervention compared to simple observation. However, no statistically significant differences were observed when VR-based intervention was compared to CT, as well as when VR-based intervention was used in combination with CT vs. CT alone. These results should be generalized with caution, due to the small number of studies included in each analysis. However, our results suggest that, in the absence of physical therapy, VR-based intervention alone can be used as an effective therapy to reduce the disabling impact of MS. In addition, it is important to remark that this study is the first review with a meta-analysis that provides information about the effect of VR-based interventions on the impact of MS.
Finally, we assessed the effect of VR-based interventions on global QoL. In a preliminary meta-analysis, VR-based intervention improves global QoL with a medium effect in PwMS. Subgroup analysis showed that VR and VR + CT are better than simple observation and CT, respectively, at improving overall QoL in PwMS. Compared to simple observation, our results are in line with the meta-analysis conducted by Santos-Nascimiento [63], although we include four more studies in our analysis. Both reviews are in favor of using VR-based therapy in CT protocols to improve the overall QoL. However, no statistically significant differences were found when VR-based intervention was compared with CT. We suggest that both therapies (VR or CT), when are used as single therapeutic options, are effective; but when these two therapies are combined, the effect is significantly greater than CT alone. Furthermore, we performed a subgroup analysis to assess the effect of VR-based intervention on the physical and mental dimensions of QoL, which was the first meta-analysis to have explored this factor. In both dimensions, the greatest effect of VRbased therapy was found when it was used in combination with CT compared to CT alone, with the important limitation that this analysis included one study. Although this result is limited, it is one of the most important findings of this review, as it reinforces the idea that VR-based intervention combined with CT can increase the effect of both therapies on QoL. In this sense, QoL can be improved thanks to the combination of these two therapies with two different objectives. As such, CT is a therapy more focused on analytical movements, while VR-based therapy allows PwMS to carry out functional movements and train activities for daily living with active exercise-based videogames integrated in sessions that are more ludic and motivating. Combining the use of customized CT techniques to restore specific joint, muscle, or balance disorders, together with VR-based intervention of different levels of difficulty and adapted to patient's preferences, can explain this large improvement in QoL metrics. Several studies have reported higher levels of fun and commitment in therapy in patients receiving a VR-based intervention compared to CT alone [68]. These positive results, along with improvements in mental overload during VR exposure, could be responsible for the increase in outcomes such as QoL [68]. Furthermore, in older adults, VR-based intervention can produce changes in the hippocampus and amygdala, which could be related to the control of negative emotions and, therefore, help reduce anxiety and depression; this approach could be applied to improve the QoL in PwMS [69]. The results of our review are in line with studies carried out for different neurological diseases [70][71][72], including MS [54], which show an improvement in QoL metrics after VR-based therapy. Finally, our systematic review includes a large number of studies that assess the effect of VR-based intervention on QoL, which increases the quality of evidence in comparison to other reviews [63].
At a neurophysiological level, VR-based interventions look to promote neuronal plasticity in the damaged brains of PwMS, with the aim of replacing or restoring missing functions. The brain is capable of adapting to environmental and pathological stimuli through neuronal or cerebral plasticity [73]. In MS, remyelination is essential to repair demyelination and recover from disabling symptoms [74], and neuronal plasticity is necessary to reorganize new synapses, with remyelination being responsible for clinical improvement in MS [75]. Neuronal plasticity decreases with age and with the duration of MS [76], so it is important to apply active and multisensory therapies in the first years of MS diagnosis, when patients are young. The multisensory experience produced by VR and the active participation of the patients to perform the therapy through videogames [23] could activate the mirror neuron system (MNS). The MNS, located in the frontal (inferior frontal gyrus) and parietal (inferior parietal lobe) lobes [77], is activated during the execution of a motor action and when an individual observes an action in other subjects [78]. Functional movements carried out with VR devices or the visualization of movements in VR devices could facilitate the activation of MNS in PwMS, possibly producing cortical and subcortical brain changes that stimulate synaptic remyelination and reorganization in motor brain areas. Furthermore, some studies have suggested that VR increases the motivation of MS patients during therapy, as well as their adherence to therapy [79]. In addition to active participation and enjoyment, the effect produced by VR-based intervention may be related to a distraction strategy. Previous studies have shown the distracting power of VR for the treatment of pain and anxiety in different situations due to immersion in virtual environments [80]. Compared other classical therapies, the distraction produced by a VR-based intervention focuses a patient's attention in the videogames, reduces negative emotions such as anxiety [81], and increases participation in the therapy. The power of distraction could be related to the effect of VR on the prefrontal cortex, which is responsible for blocking negative experiences and feelings [82]. The prefrontal cortex, specifically the dorsolateral prefrontal cortex and the inferior frontal gyrus, plays a crucial role in the inhibition of emotive responses and may be related to the regulation of emotions [83]. Thus, VR could be considered an excellent option to improve the mental dimension of QoL in PwMS who have difficulties adhering to CT.
This review presents updated practical implications for physical clinicians, such as physiotherapists or occupational therapists, as well as researchers. It also shows how VR-based therapy can reduce disabling symptoms of MS such as fatigue and increase QoL. The main advantage of VR-based intervention is the possibility of obtaining virtual environments that PwMS feel are similar to the real world, which leads them to perform functional tasks within these environments. VR-based intervention also has the added value of allowing participants to interact dynamically with objects or situations that would not be possible in the real world, promoting motor learning [84] with augmented feedback and multisensory inputs. Furthermore, VR-based intervention is a safe technique with few adverse effects reported in subjects with MS [85], and it offers the possibility of home treatment, a relevant advantage during the COVID-19 pandemic [86]. Various systematic reviews have demonstrated the efficacy of VR-based intervention as a home training method in the COVID-19 pandemic in patients with different neurological diseases, including MS [86,87]. Scientific evidence shows that home training based on VR is a therapy that provides motivation, and it can be useful in the rehabilitation of physical and cognitive function in PwMS [88]. The use of VR at home seems to have a positive impact as a method of support for traditional rehabilitation, especially during the COVID-19 pandemic, due to the difficulty of these patients accessing classical therapies in clinical centers [89]. For example, physical exercise improves resistance to fatigue and QoL, and can be practiced at home through VR-based videogames exercises. In this sense, the most analyzed systems used in VR neurorehabilitation are niVR devices, such as Nintendo ® Wii Balance Board ® or Nintendo ® Wii Fit ® , which are affordable and easier to transport and install at home. Other VR systems, such as BTS-Nirvana or Oculus Quest, allow full 360 • immersion in the virtual world, but also require a high level of spatial orientation and comprehension [90], and are more appropriate for use in clinical centers supervised by a clinician. In our review, the majority of studies (10 of 12) included niVR devices; thus, these results are more dependent on non-immersive virtual scenarios, which may be the most useful for clinical practice and home training in PwMS.
Finally, we must bear in mind that the present work has some limitations, and the results should be interpreted with caution. First, the low number of studies included in each meta-analysis and in the subgroup analyses, as well as the low number of participants per study, reduces the generalizability of our findings. Second, the low methodological quality of the included studies increases the risk of selection and classification bias. Third, the presence of publication bias and the variation in trim-and-fill estimations may distort the real effect of the therapy for different outcomes. Another limitation comes from the large variations observed in the sensitivity analysis, which may reduce the quality of our findings. In addition, the majority of the studies assessed the effect of VR-based intervention using niVR devices, so our results are more relevant to the effect of niVR devices. Finally, we must remark that all the assessments conducted in the included studies were performed immediately after intervention, which did not permit us to predict the effect of VR-based therapy in the medium and long term.

Conclusions
Our results showed that VR-based therapy is effective in reducing fatigue and the impact of MS, as well as increasing QoL in PwMS. Specifically, to reduce fatigue, VR-based intervention is better than CT. In terms of the impact of MS, VR-based intervention was better than simple observation. To increase overall QoL, VR-based therapy is better than simple observation and the combined use of VR-based intervention with CT is better than CT alone. Finally, VR-based intervention also showed a positive effect on the physical and mental dimensions of QoL, demonstrating a significant increase in both dimensions when the VR-based intervention was used in combination with CT, compared to CT alone. Nevertheless, further research is needed to assess the effect of VR-based intervention, both alone and when combined with other therapies.