Cerebral palsy (CP), the leading motor disability in childhood with a worldwide prevalence of 2 per 1000 live births, is a group of permanent movement and posture disorders caused by an early brain lesion during pregnancy, childbirth, or shortly after birth [1,2]. In addition to the motor consequences, sensory and cognitive functions affecting daily life activities may also be impaired [2]. From the rehabilitation perspective, motor skill learning–based approaches have shown encouraging efficacy in improving motor functions in children with CP in recent years [3]. Interventions based on motor skill learning principles (more intensity, performance of daily living activities, repeating functional tasks, individually tailored, progressively challenging, and motivating feedback) have been shown to promote persistent motor skill acquisition and neuroplasticity (practice-induced brain changes) [4,5]. In addition, it is now well established that sustaining children’s attention, interest, and enjoyment for a long time is one of the critical factors, but it is also challenging for optimizing rehabilitation outcomes and neuroplastic changes [6]. Consequently, researchers have focused on how to foster neuroplastic changes and functional improvements with more intensity and diversity of practice without a trade-off in terms of entertainment.

Thus, with the results of recent studies and the increase in access to computer-assisted technologies, the use of virtual reality (VR) as a means of neurorehabilitation for children with CP has been encouraged. VR refers to a computer-generated simulation that may vary in complexity and reality owing to the rapidly evolving nature of technology [7,8]. The VR environment can be experienced through a variety of display hardware, including standard monitors, flat screens, projection screens, and head-mounted displays. The method of interacting with the VR system may also vary from simple activation of computer keyboard keys, mouse, or joystick to more advanced motion camera interfaces [9,10]. Consequently, VR devices used in rehabilitation can be classified as immersive, semi-immersive, and nonimmersive depending on the level of immersion [11,12]. Furthermore, certain devices are specifically designed for rehabilitation purposes, whereas commercially available devices and games are also used in the rehabilitation of various disorders. A systematic review on the clinical utility of VR in neurorehabilitation reported that CP is the second most studied neurological disorder in VR rehabilitation [13]. In addition, You et al [14] and Golomb et al [15] showed cortical reorganization after VR therapy in children with CP and the association of the cortical changes with functional abilities. The advantage of using VR in rehabilitation is not only to increase entertainment with playful interactive games but also to have games with individually tailored, progressively adjustable difficulty levels and to provide active repetition in an enriched environment [9]. Thus, with these advantages, VR may well correspond to the challenges in rehabilitation such as the use of motor skill learning principles to develop long-lasting motor skill acquisition in children.

Several systematic reviews investigated the effectiveness of VR on upper and lower limb functions, gross motor functions, and postural control in children with CP in the short term, that is, immediately after treatment [16-18]. They showed a promising potential of adding VR into CP interventions for improving motor functions. However, the long-term effects of VR interventions on motor skill learning remain unclear, although durable retention of acquired skills is one of the crucial factors for rehabilitation objectives. Consequently, this systematic review and meta-analysis was conducted to examine the effect of VR on motor functions and daily life activities of children with CP from the short term (immediately after the intervention) to the medium term (≥1 week to <12 weeks after the end of the intervention) or long term (≥12 weeks after the end of the intervention).

We followed the guidelines of the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 statement to conduct this systematic review and meta-analysis and reported our results according to PRISMA 2020 checklist [19]. The protocol of this review was registered in the PROSPERO database (CRD42021227734).

Inclusion criteria of the studies were defined after discussion among all authors according to the PICO (Population, Intervention, Comparison, and Outcome) framework [20]. As a result, studies (1) with a population of children with CP aged <18 years; (2) using VR for main intervention; (3) comparing the experimental group (therapy involving VR) with any control group (another intervention except VR, nonintervention, placebo effect, or different cohorts); and (4) including baseline, postintervention, and follow-up assessments by using outcome measures on motor functions and daily life activities were included. We excluded studies if they (1) involved populations other than children with CP aged <18 years; (2) used VR only for assessment or focused on the development of VR devices; (3) used different outcome measures than on motor functions and daily life activities; (4) were not published in English; (5) were reviews, meta-analyses, commentaries, protocols, and conference abstracts; and (6) did not have follow-up assessment (ie, had only baseline and postintervention assessments). In this review, we decided to include studies involving commercially available VR devices or those specifically developed for therapy, regardless of the degree of immersive experience, type of display hardware, or the method of interaction with the VR system.

Two authors systematically and independently searched PubMed, ScienceDirect, Embase, and IEEE Xplore databases in September 2021 using predetermined keyword combinations, which were “virtual reality,” “virtual environment,” “videogaming,” “computer game,” “Kinect,” “Wii,” “PlayStation,” “cerebral palsy,” “CP,” “rehabilitation,” “therapy” and “motor skill learning.” The results of the search in the 4 databases were transferred to the EndNote (version 9; Clarivate), and duplicates were removed by 1 author. The search strategy used for each database is detailed in Multimedia Appendix 1. We did not apply any restriction regarding the date of publication for our search strategy.

After detecting and eliminating duplicated studies obtained from the searches, 2 authors first screened titles and abstracts according to the inclusion and exclusion criteria. After several discussions and resolving the conflicts between authors, full texts of studies that have the potential to be included were obtained. Then, the eligibility of the studies was assessed using full-text screening. All screening and study selection processes were performed by the same 2 authors who were blinded to each other using Rayyan web-based software (Qatar Computing Research Institute) [21].

The following information was extracted by 2 authors independently to a predesigned Microsoft Excel (Microsoft Corp) sheet: characteristics of the study (first author, year of publication, and study design), characteristics of the participants (number of participants, CP type, age, Manual Ability Classification System [MACS], and Gross Motor Function Classification System [GMFCS] level), characteristics of the VR interventions and comparison interventions (name, duration, frequency, length, total hour, priority focus, and motor skill learning principles involved in experimental interventions), characteristics of outcomes measures (timing and measures), and main outcomes. In addition, to collect the data from the included studies, another predesigned Microsoft Excel sheet was used, with the following content: mean and SD of experimental (VR) and control groups at postintervention and follow-up points and sample sizes of both groups. When there was no information about the means and SDs of results, they were calculated from SE, CI, t test, or P value as recommended by the Cochrane handbook [22]. Disagreements between the data collectors were resolved by a discussion.

Two authors independently assessed the methodological quality of the included studies using the modified Downs and Black checklist. This is a valid and reliable tool to assess the quality of both randomized controlled trials (RCTs) and non-RCTs [23]. Conflicts were resolved by discussion. This checklist comprises 27 items on the quality of reporting (10 items), external validity (3 items), internal validity (bias and confounding; 13 items), and statistical power (1 item).

The highest possible score for the original checklist was 32. However, as item 27 was complicated to score (from 0 to 5), a modified version of the scoring method was proposed, as observed in previous studies [17,24,25]. Thus, each item was scored as 0 (no) or 1 (yes), except for the fifth item that was scored as 0 (no), 1 (partially), or 2 (yes). Consequently, scores on the modified Downs and Black checklist corresponding to quality levels ranged between “excellent” (scores of 24-28), “good” (scores of 19-23), “fair” (scores of 14-18), and “poor” (scores of ≤13) [26,27].

We conducted a meta-analysis for the RCTs that met the criteria for eligible design and outcomes. The results from the studies on motor functions and activities of daily living were extracted into a Microsoft Excel sheet. For continuous data, standardized mean difference (SMD) with a 95% CI, which refers to the effect size, was measured using a random effects model because of the different clinical tests used in the included studies. For the interpretation of effect sizes, we followed Cohen rule in which 0.2, 0.5, and 0.8 represent small, medium, and large effect sizes, respectively [28]. Statistical significance was set at .05, and statistical heterogeneity was examined using the I² statistic. Heterogeneity was rated as low (I²≤25%), moderate (25%<I²<75%), or high (I²≥75%).

To explore whether the duration and intensity of VR training (more intensive: more therapy time per week, which is obtained by the higher frequency of sessions or longer duration of each therapy session) as well as the type of VR had some influence on the results, subgroup analyses were performed (when possible) according to the following criteria: duration (<30 min per session, ≥30 to <60 min per session, or ≥60 min per session), intensity (≤120 min per week or >120 min per week) and VR type (specifically developed for therapy or commercially available). The software Review Manager (version 5.4.1; Cochrane) for Windows was used to conduct the analyses of the included studies.

Our database search resulted in 1643 studies after removing 319 duplicate records. After title and abstract screening, we excluded 1492 records for the reasons specified in Figure 1. Full texts of the remaining 151 studies were assessed for eligibility, and finally, 9 studies met the inclusion criteria, 7 (78%) of which were RCTs.

All studies in this review (7 RCTs and 2 non-RCTs) were included in the methodological quality assessment. The total score of these studies ranged from fair to excellent. Of these 9 studies, 3 (33%) studies were classified as excellent-quality studies [32,34,36], 4 (44%) as good-quality studies [30,31,33,37], and 2 (22%) as fair-quality studies [29,35]. Table 4 presents the 4 main subscales and the total scores of the modified Downs and Black checklist. Scoring of all 27 items for the included studies is provided in Multimedia Appendix 2.

This systematic review and meta-analysis aimed to investigate whether VR-included therapies are effective on motor skill learning in children with CP compared with a control group, with special attention to the middle- and long-term effects. Overall, the results on upper limb function and performance showed the benefits of adding VR to CP rehabilitation in terms of postintervention improvements and supported that these acquisitions were mostly retained at follow-up. The meta-analysis results on upper limb functions and performance confirmed these benefits. In particular, Rostami et al [36] and Psychouli and Kennedy [37] used VR with mCIMT for 4 weeks, and greater improvements have been shown in the upper limb functions after VR with mCIMT compared with conventional therapy or mCIMT alone. Choi et al [33] matched VR with conventional occupational therapy for 4 weeks, and similarly, greater improvements have been shown in the upper limb functions following the intervention including VR compared with conventional occupational therapy alone [33]. In addition, these improvements were maintained at follow-up. It should also be noted that the contents of experimental interventions were shaped with motor skill learning principles (more intensity, the performance of daily living activities, repeating functional tasks, individually tailored, progressively challenging, and motivating feedback) in these 3 studies, which showed greater and long-lasting improvements in the VR groups (Table 2) [33,36,37]). Choi et al [33] and Rostami et al [36] applied these motor skill learning principles in VR game sessions, whereas Psychouli and Kennedy [37] used them mostly in mCIMT sessions and added a computer game as motivating feedback at the end of the intervention. Therefore, we concluded that incorporating motor skill learning principles into interventions including VR, through VR or therapy, induces greater benefits in motor function improvements in the long term. In addition, in the motor skill learning–based, VR-included interventions, VR has a substantially effect on maintaining motivation and eventually ensuring more intensity and repetition of child-initiated active movements during the interventions.

Furthermore, in line with our results from the subgroup analysis, applying more intensive VR training (involving more total hours in a shorter time, 4 weeks vs 6 weeks) had a significant influence on these larger improvements in the VR group both after the intervention and at follow-up compared with the other studies reporting no significant benefit of adding VR [32,34]. Thus, a higher intensity of VR training with either a longer duration or a higher frequency of training appears to be another essential factor for enhancing motor skill learning through VR. These findings were similar to the results of previous systematic reviews and meta-analyses [16,39-42]. They highlighted some principles of VR use in the neurorehabilitation of children with CP [13]. Rathinam et al [16] and Chen et al [39] also noticed in their reviews that more intensive VR training has more potential to result in improvements of functions in children with CP immediately after the intervention.

Furthermore, a study included in our review stated that parents were asked to encourage their children to use VR for 30 minutes daily. However, they played the games only for 7 minutes as a mean daily amount because of children’s low interest and motivation for the games in VR games [34]. This may explain why they did not find any improvement after VR-included therapy. In other words, maintaining interest and motivation also helps active participation in gaming in VR therapy and provides the expected therapy duration.

We also performed a subgroup analysis to determine whether VR types affected the upper limb function improvements. The results demonstrated that VR devices specifically developed for therapy have more potential benefits on the training of upper limb functions in children with CP when compared with the commercially available VR device (Nintendo Wii). In addition, VR devices specifically developed for therapy had a greater potential to make acquisitions persistent in the medium or long term. As explained by Chen et al [39], it seems more feasible to provide individually tailored interventions with active repetition in various contexts by integrating motor skill learning principles into VR devices when they are specifically developed for therapy purposes.

Concerning the outcomes of balance, trunk control, gross motor functions, and daily life functions induced by therapies combined with VR, we found only 2 studies that met the eligibility criteria of this systematic review and meta-analysis [30,31]. In addition, these 2 studies had different results. Decavele et al [30] found more improvement (though impermanent) after the VR combined period, whereas Pin and Butler [31] observed no significant difference between the groups. Even so, with the postintervention meta-analysis results of 2 good-quality studies, we showed that adding VR into therapy may offer additional benefits on balance and trunk control (SMD=1.04; P=.06) and likely in gross motor functions (SMD=2.85; P=.30) in children with BCP. Our findings were consistent with those of other reviews on balance and gross motor functions after VR interventions [18,43,44]. However, the very limited number of studies and small number of participants (<50) made it difficult to conclude the follow-up effects. Thus, it remains unclear whether there is more potential in VR-included therapy to facilitate long-lasting acquisitions in balance and gross motor functions in children with CP. The fact that both studies [30,31] only included children with BCP with GMFCS III-IV levels may also have affected the results. Further studies are needed to address the long-term effects of VR intervention on balance and gross motor functions in children with different severity of CP. The number of studies with long-term results that included daily functions was also very low (n=3) [30,33,34]. In fact, 2 out of 3 studies noted that changes in daily living activities after therapies with VR were significantly different between groups, with a tendency toward the VR group [30,33]. Similarly, the meta-analysis results were in favor of the VR group (SMD=0.29; P=.21). Considering that there are insufficient results on whether the acquisitions are long lasting, more studies on the daily function outcomes are also needed.

First, we found a limited number of studies and participants that met our inclusion criteria, especially for medium- and long-term outcomes. Only 7 RCTs and 2 non-RTCs were included in our systematic review, and most of them (7/9, 78%) had a relatively small sample size (<50). In addition, many studies did not report the MACS and GMFCS levels of the included children. The studies that reported the levels also included children with a relatively wide range of functional abilities. For instance, 1 study included children with GMFCS I-II level [29], whereas another study included children with GMFCS III-IV level [30]. As a result, this heterogeneity may have affected the results, but we could not perform a subgroup analysis because of the limited number of studies. Regarding the outcome measures, wide variations were used in the studies, and this heterogeneity of the outcome measures made it difficult to combine and interpret the results. Another limitation of this study is that if a study included in our review used different clinical tests to determine similar outcomes (upper limb functions, balance and trunk control, gross motor functions, and activities of daily living), all results were included in the analysis. Although the outcome measures were different, the intervention process of VR was identical in the same study. In addition, as one of the studies [30] had a crossover design, at the end of the study period, a control group (the group that did not receive VR) did not exist to allow comparison with the follow-up results of the VR group. Therefore, we were unable to conduct a meta-analysis of follow-up outcomes for balance, gross motor functions, and daily life activities, as there was only one study providing follow-up results on these outcomes.

This review shows that adding VR into upper limb rehabilitation of children with CP offers additional benefits in motor skill learning and enhances the maintenance of obtained improvements when the VR included training that cooperates with motor skill learning principles, with the inclusion of functional tasks as well as repetitive practice, more intensity, individual adjustment, progressive challenge, and motivation. For these reasons, it appears more appropriate to make use of VR devices specifically developed for therapy. In terms of balance, trunk control, gross motor functions, and daily life activity outcomes, combining VR with therapies can help induce more improvements in children with CP. However, the priority focus of most of the interventions with VR (7/9, 78%) in the included studies, and their outcomes, was upper limb functions. The effect size on motor skill learning of adding balance-focused VR devices into different therapies in children with CP is still not entirely clear; therefore, studies with long-term outcomes are needed.