Effectiveness of Mechanical Horse-Riding Simulator-Based Interventions in Patients with Cerebral Palsy—A Systematic Review and Meta-Analysis

Background: Mechanical horse-riding simulator (HRS) exercises are a type of therapy based on the use of robotic or mechanical devices that produces movement similar to a real horse with the aim of simulating hippotherapy. This review analyses the effectiveness of HRS therapies in patients with cerebral palsy (CP). Methods: A systematic review and a meta-analysis were carried out by searching studies in PubMed Medline, SCOPUS, Web of Science, CINAHL, PEDro and SciELO up until October 2022. We selected clinical trials that assessed the effectiveness of HRS therapy, compared to other interventions, in patients with CP. The main variables were gross motor function (its global score and dimensions, such as sitting ability), functional balance, spasticity, hip range of motion (ROM), posturographic balance and satisfaction. The risk of bias was assessed using the Cochrane Risk of Bias Tool. The pooled effect was calculated using Cohen’s Standardized Mean Difference (SMD) for a 95% confidence interval (95% CI). Results: Twelve studies were included in the systematic review, and 10 were included in the meta-analysis, providing data from 343 patients with spastic diplegic CP. Our findings revealed that HRS plus physiotherapy is more effective than physiotherapy in improving the total gross motor function (SMD 0.98; 95% CI 0.35–1.62), sitting ability of the gross motor function (SMD 0.84; 95% CI 0.32–1.36) and functional balance (SMD 0.6; 95% CI 0.1–1.08), and HRS therapy is better than sham to improve pelvic abduction ROM (SMD 0.79; 95% CI 0.21–1.37). Conclusions: Horse-riding simulator-based therapy is an effective therapy to improve gross motor function, functional balance and abduction pelvic ROM in children with CP, in comparison to physiotherapy or sham.


Introduction
Cerebral palsy (CP) is the most common physical disability in children, affecting 17 million people worldwide [1]. Currently, in high-income countries, the estimated prevalence is 1.6 cases per 1000 live births [2], which has decreased relative to previous prevalence data (two to three cases per 1000 in 2013) [3]. Cerebral palsy encompasses a group of permanent movement, posture and motor function disorders that changes with age as a result of damage to the developing fetal or infant brain [4,5]. Cerebral palsy is caused by brain damage that, among other effects, reduces activity in the motor cortex [6], producing inadequate design and execution of motor inputs and worse processing of corticospinal and somatosensory circuits [7]. Cerebral palsy is characterized by changes in musculoskeletal tissue, such as muscle weakness, muscle spasticity, decreased muscle as its safety, the possibility to exactly recreate the riding session in an indoor setting and its adaptability to the attributes of each patient [33].
To date, reviews have assessed the effect of HRS therapy in other neurological conditions such as stroke [35], musculoskeletal conditions such as back pain [37] and in older adults [38] with interesting findings. There is currently no systematic review or metaanalysis looking exclusively at the effect of HRS therapy compared to other therapies. In 2019, Dominguez-Romero et al. assessed the effectiveness of HRS therapy in patients with stroke and CP, including seven studies [35]. However, only four studies of all those included provided data from patients with CP and only two studies were used to perform the meta-analysis on gross motor function (total score) without finding statistically significant differences between HRS therapy and others. In this review, we found important restrictions, such as the use of the English language filter, the low number of studies included and that other variables such as balance or sitting ability were not assessed. In 2022, Heussen and Häusler, assessed the effectiveness of equine-assisted therapies for children with CP, including three studies that used HRSs [39]. This meta-analysis presents an important limitation; its literature search did not identify other studies that compare HRS therapy to other therapies in the outcome of interest. In addition, the generalization of these findings is difficult due to the low number of studies and the high risk of publication bias. To improve knowledge about the use of HRS in CP, the aim of this systematic review was to retrieve published evidence to assess the effectiveness of HRS interventions when comparing with other therapies in patients with CP on gross motor function, functional balance, spasticity, hip ROM, posturographical parameters and patients' satisfaction. Secondarily, we determined if the effect of HRS therapy was bigger when it is used alone or combined with physiotherapy.

Register and Guidelines
This systematic review with meta-analysis was carried out following the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [40], the A Measurement Tool to Assess Systematic Reviews (AMSTAR version 2) [41], and the Cochrane Handbook of Systematic Reviews of Interventions (Second Edition) [42]. Furthermore, the protocol of this systematic review with meta-analysis was previously registered in the PROSPERO database, obtaining the following registration number: CRD42022370252.

Literature Search
Two authors (D.M.-C. and E.O.-G.) independently carried out a bibliographic search in the following databases: PubMed Medline, Web of Science (WOS), Scopus, CINAHL Complete, Physiotherapy Evidence Database (PEDro) and SciELO. In order to find studies not published in these databases, we searched in the reference section of studies previously published and in the gray literature (congress abstracts, proceedings, and documents of experts, among others). The search strategy was developed based on the PICOS tool proposed by the Cochrane Library [42,43]: population (children with CP), intervention (HRS), comparison (other therapies apart from HRS), outcomes (gross motor function, functional balance, spasticity, ROM and posturographic spine and balance parameters) and study design (clinical trials). Our search strategy was carried out using keywords from the PubMed Thesaurus (MeSH) and CINAHL Subject Headings. The main terms employed were "cerebral palsy" and "horse riding simulator," and they were combined with other synonyms. The boolean operators were employed in our search strategy; "AND" was used to join the PICOS conditions selected, and "OR" to join related terms in each condition. Lastly, no language or publication date filters were used to perform the search. Any discrepancies related to the search were agreed with a third author experienced in literature searches (M.C.O.-P.). Table 1 shows the search strategy used in each database. SCOPUS (TITLE-ABS-KEY ("cerebral palsy") AND TITLE-ABS-KEY ("Horse riding simulator" OR "horse simulator" OR "simulator horse" OR "simulator equine")) Web of Science TOPIC: (*cerebral palsy*) AND TOPIC: (*Horse riding simulator* OR *horse simulator* OR *simulator horse* OR *simulator equine*) CINAHL Complete AB (cerebral palsy) AND AB (Horse riding simulator OR horse simulator OR simulator horse OR simulator equine) PEDro Cerebral palsy AND horse riding simulator SciELO Cerebral palsy AND horse riding simulator

Inclusion and Exclusion Criteria: Study Selection
The selection process of the studies included in this meta-analysis was carried out by 2 authors (D.M.-C. and I.C.-P.) independently, who were responsible for reviewing all the records found in each database by title and abstract. In addition, discrepancies in this phase were resolved by a third author (M.C.O.-P.). A study was only included in this review if it met all the inclusion criteria: (1) clinical trials; (2) that the study population was diagnosed with CP; (3) that the study had at least 2 intervention groups, 1 of which underwent HRS therapy and it was compared to another type of therapy different to HRS or no intervention; (4) studies that evaluated variables of interest for this study (see Section 2.5); and (5) studies that provided qualitative or quantitative data to perform the qualitative synthesis or metaanalysis. The following exclusion criteria were also established: (1) clinical trials where the sample comprised patients with different neurological diseases (not only CP); and (2) experimental studies with only 1 group (without a comparison group).

Data Extraction
The data extraction process of the included studies was carried out by two authors (D.M.-C. and E.O.-G.) independently, using a Microsoft Excel data collection form. All possible disagreements were resolved with a third author (M.C.O-P.). The following data were extracted from each study: (1) general characteristics (authorship, publication date, study design, country, setting and funding); (2) patient characteristics (total sample size, number of participants per group, age, sex, type of CP, disability and time since diagnosis); (3) characteristics of the experimental group and the control group (type of intervention, number of sessions, number of weeks, number of sessions per week, and duration of each session in minutes); (4) outcome data of the variables of interest (mean and standard deviation if a meta-analysis and qualitative synthesis and p-value for intra-groups and inter-groups comparisons); and (5) assessment time (post-intervention). When a study did not provide standard deviations, it was estimated using standardized transformations through the standard error, range, interquartile range and median [42,44].

Variables
The variables of interest to assess the effectiveness of HRS therapy in patients with CP were gross motor function and its five dimensions (lying and rolling, sitting, crawling and kneeling, standing up and gait ability), functional balance, spasticity, hip ROM and posturographic parameters and satisfaction.

Quality Assessment
The evaluation of the risk of bias in each study included and of the quality of evidence of the main findings was carried out by 2 authors (D.M.-C. and M.C.O.-P.) independently. Any doubts were resolved by a third author (E.O.-G.). At first, the Cochrane Collaboration Bias Tool Risk was used to assess the risk of bias in the studies included in the review. This scale assesses 6 bias domains (selection, performance, detection, attrition, reporting and others) through seven items (random sequence generation, concealment randomization sequence, blinding of participants, blinding of assessors, incomplete outcome data, selective reporting and others, ideally prespecified). Each item can be categorized as "+" (high risk of bias), "−" (low risk of bias), and "?" (uncertain risk of bias) [45]. Secondly, the quality of evidence of each meta-analysis was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) [46]. Furthermore, the recommendations of the checklist proposed by Meader [47] were followed to estimate the quality of the evidence, taking into account the risk of bias in each selected study, the inconsistency, the imprecision, the lack of directivity and the risk of publication bias. The quality of evidence was categorized as high (if our findings were robust); moderate (if our results changed when introducing new studies); low (if our results were very slight); and very low (when some elements were not present). The quality of the evidence for each meta-analysis was downgraded by one level for each factor found. When multiple limitations were found, the overall quality score was lowered by 2 levels.

Statistical Analysis
The meta-analysis was carried out by 2 authors (E.O.-G. and I.C.-P.) using the software Comprehensive Meta-Analysis v3.0 (Biostat, Englewood, NJ, USA) [48]. To perform the meta-analysis, we followed the recommendations of the Introduction to Meta-Analysis by Borenstein et al. [49] and of The Handbook of Research Synthesis and Meta-Analysis" by Cooper et al. [50]. According to the level of heterogeneity in each meta-analysis, we used a random or fixed effect model in accordance with Dersimonian and Laird [51]. Cohen's standardized mean difference (SMD) and its 95% confidence interval (95% CI) were used to calculate the pooled effect [52]. Effect size could be null (SMD 0), low (SMD 0.1-0.39), moderate (SMD 0.4-0.79) or large (SMD > 0.8) [53]. Additionally, when the same variable was measured with the same tests, we calculated the mean difference (MD) and its 95% CI, with the aim of comparing this result to the minimally clinically important difference (MCID) value for this test. Jaeschke et al. defined the MCID as "the smallest difference in score in the domain of interest which participants perceive as beneficial and which would mandate, in the absence of troublesome side effects and costs, a change in the patient's management" [54]. The findings of each meta-analysis were graphically represented in the forest plots [55]. The risk of publication bias was assessed taking into account three elements: the visualization of the funnel plots (asymmetry indicates the presence of risk of publication bias) [56]; the p-value for the Egger test (p < 0.1 indicates the risk of publication bias) [57]; and the trim-and-fill estimation [58,59]. If variations were found after trim-and-fill estimation that was larger than 10% of the original pooled effect, the quality of evidence would be downgraded by 1 level, even though the funnel plot was symmetric [60]. Finally, the level of heterogeneity was calculated using the degree of inconsistency (I 2 ) and the p-value for the Q-test (p < 0.01 indicates the risk of heterogeneity). The heterogeneity could be null (I 2 0%), low (I 2 < 25%), moderate (I 2 25-50%) or large (I 2 < 50%) [61,62].

Additional Analyses
To assess the contribution of each study to the overall pooled effect, we performed a sensitivity analysis using the leave-one-out method. In addition, we performed the following subgroup analyses: HRS plus PT (physiotherapy) vs. PT, HRS vs. PT, and HRS vs. sham.

Additional Analyses
To assess the contribution of each study to the overall pooled effect, we performed sensitivity analysis using the leave-one-out method. In addition, we performed the fo lowing subgroup analyses: HRS plus PT (physiotherapy) vs. PT, HRS vs. PT, and HRS vs sham.

Characteristics of the Studies Included in the Review
The studies included in this review were conducted between 1998 and 2022 in Indi [64,67], South Korea [65,66,70,71], Iraq [63], Spain [69], Brazil [73], Thailand [74] and th USA [72]. The included studies reported data from 343 patients with spastic diplegic C with ages between 2 and 16 years old, of which 56% were male versus 44% female. Th experimental group comprised 180 patients who received an HRS intervention alon [68,69,[72][73][74] or in combination with conventional therapy [63][64][65][66][67]70,71]. The contro group comprised 163 patients who received conventional therapy [63][64][65][66][67]70,71,73] o sham [68,69,72,74]. The duration of HRS therapy ranged from 1 to 12 weeks. The session were carried out one to three times per week, with the duration of each session rangin from 10 to 75 min. There was no follow-up in any of the studies, and all assessments wer performed at the end of the intervention. Finally, only one study received external fund ing [69]. Table 2 shows the characteristics of the included studies.

Characteristics of the Studies Included in the Review
The studies included in this review were conducted between 1998 and 2022 in India [64,67], South Korea [65,66,70,71], Iraq [63], Spain [69], Brazil [73], Thailand [74] and the USA [72]. The included studies reported data from 343 patients with spastic diplegic CP with ages between 2 and 16 years old, of which 56% were male versus 44% female. The experimental group comprised 180 patients who received an HRS intervention alone [68,69,[72][73][74] or in combination with conventional therapy [63][64][65][66][67]70,71]. The control group comprised 163 patients who received conventional therapy [63][64][65][66][67]70,71,73] or sham [68,69,72,74]. The duration of HRS therapy ranged from 1 to 12 weeks. The sessions were carried out one to three times per week, with the duration of each session ranging from 10 to 75 min. There was no follow-up in any of the studies, and all assessments were performed at the end of the intervention. Finally, only one study received external funding [69]. Table 2 shows the characteristics of the included studies.   Abbreviations: CP, cerebral palsy; GMFCS, gross motor function classification system; N, sample size; N e , number of participants in experimental intervention; N c , number of participants in control intervention; F, female; M, male; NR, not reported; HRS, horse-riding simulator; Min, minutes. Table 3 shows the Cochrane Risk of Bias Tool assessment for each study included in the review.

Risk of Bias Assessment
The risk of bias was high in three studies [65,66,70], medium in six studies [63,64,67,71,73,74], and low in three studies [68,69,72]. The most important risks for consideration were performance, detection and selection biases. The risk of performance bias was present in all studies due to the impossibility of blinding the participants. Detection bias appeared in seven studies (58% of all), and selection bias was present in six studies (50% of all studies).  [70] ? ?

Variables, Measurements and Synthesis
To assess gross motor function, the studies included reported data from the Gross Motor Function Measure-66 (GMFM-66), GMFM-88 and Gross Motor Function Classification System (GMFCS). To analyze gross motor function, we obtained data from the total score and/or its five dimensions (A: lying and rolling; B: sitting; C. crawling and kneeling; D: standing; and E: walking, running and jumping). Secondly, postural balance was assessed with data from the Pediatric Balance Scale (PBS). Thirdly, spasticity in different lower limb muscles was assessed with data from the Modified Modified Ashworth Scale (MMAS) and Modified Ashworth Scale (MAS). Later, hip and tilt pelvic ROM was evaluated with a goniometer.
Furthermore, other secondary variables included posturographic parameters of static balance using Pedoscan Sensor and F-mat sensor platform and F-scan system, posturographic spinal posture using ABW Mapper, seated trunk control with SATco and satisfaction with the therapy using the Autoquestionnaire Qualité de Vie Enfant image (AUQEI).
The results of this review are presented in two ways: first, a meta-analysis of the variables gross motor function, functional balance, spasticity and ROM and, then, a qualitative synthesis for pelvic tilt, posturographic balance assessment in stand-up or sitting positions (trunk control) and satisfaction with HRS therapy. Table 4 shows all the qualitative findings in the studies included. Both groups reported significant improvements (p < 0.001, respectively). The experimental groups show higher mean values than the control group. Statistically significant differences were found between groups (p = 0.028). Interaction analysis showed more improvement in the experimental group than the control group in each comparison (week assessment).
Choi, HJ et al., 2014a [65] Posturographic spinal posture ABW Mapper Statistically significant differences in interaction between groups and periods in trunk imbalance, pelvic torsion and pelvic tilt (p < 0.05).
Choi, HJ and Nam, KW 2014b [66] Posturographic static balance Pedoscan sensor No significant differences in interaction between the group and period (p > 0.05). Statistically significant differences between groups in ML sway (p < 0.05) No significant differences between groups and within groups in AP sway (p > 0.05).
Fernandes, LC et al., 2018 [67] Functional balance PBS Statistically significant differences in both groups (p < 0.0001 respectively). No statistically significant differences were found between groups in the post-intervention assessment (p = 0.4516) Statistically significant differences in both groups (p < 0.0001 respectively). No statistically significant differences were found between groups in the post-intervention assessment (p = 0.4516) Hemachithra, C et al., 2020 [68] Adductors spasticity MAS Statistically significant differences in the experimental group (p < 0.001). Statistically significant differences between groups (p < 0.001)

Hip ROM Goniometry
Statistically significant differences in the experimental group (p < 0.001). Statistically significant differences between groups (p < 0.001) Herrero, P et al., 2012 [69] GMF (Total and B dimension) GMFM-66 Both groups improved, although the HRS group reported greater scores than the control group in sitting and total GMFM For A, B and C dimensions, no statistically significant differences between groups and within groups (p > 0.05). For the D dimension, statistically significant differences were found in HRS (p = 0.03) but not between groups (p = 0.06).
For the E dimension, statistically significant differences were found in HRS (p =0.03) but not between groups (p = 0.19). For GMFM total score, statistically significant differences in the HRS group (p < 0.01) and between groups favors HRS (p < 0.01) Kang No statistically significant differences in both groups (p > 0.05). No statistically significant differences between groups (p > 0.05)

Seated trunk control SATco
Statistically significant differences in all groups. HR group reported more items with significant differences. The "Reactive control" item was statistically significant among the three groups (p < 0.05). Statistically significant differences were found in "reactive control" in the HR group vs. the SHS group comparison (p = 0.004).  Table 5 shows the main findings of the meta-analyses.

Gross Motor Function
For gross motor function, we assessed the effect of HRS therapy on each dimension (A, B, C, D and E) and on the global total score.
At first, two studies [70,71] reported data from 41 participants (20.5 per comparison) to assess the effect of HRS therapy on A, C, D and E dimensions of gross motor function. Our findings did not show statistically significant differences between HRS plus physiotherapy and physiotherapy on the A (SMD 0.25; 95% CI −0. 28 Table 5, Figure 2). The risk of publication bias could not be studied, and no heterogeneity was present.

Gross Motor Function
For gross motor function, we assessed the effect of HRS therapy on each dimension (A, B, C, D and E) and on the global total score.
At first, two studies [70,71] reported data from 41 participants (20.5 per comparison) to assess the effect of HRS therapy on A, C, D and E dimensions of gross motor function. Our findings did not show statistically significant differences between HRS plus physiotherapy and physiotherapy on the A (SMD 0.25; 95% CI −0. 28 Table 5, Figure 2). The risk of publication bias could not be studied, and no heterogeneity was present. Secondly, five studies [64,[69][70][71]74] reported data from 129 patients (25.8 per comparison) to assess the effect of HRS therapy on the B dimension (sitting ability) of gross motor function. Our findings showed a medium effect (SMD 0.52; 95% CI 0.15-0.9; p 0.006) favors HRS (Table 5, Figure 2). A low risk of publication bias was present (p for Egger 0.07) due to trim-and-fill and showed a variation of 15% (adjusted SMD 0.6; 95% CI 0.24-0.94) with respect to the original pooled effect ( Figure S1). The level of heterogeneity was moderate (I 2 41%; p 0.04). Subgroup analysis revealed that the use of HRS plus physiotherapy was better (SMD 0.84; 95% CI 0.32-1.36; p 0.002) than physiotherapy alone, showing an improvement in the sitting ability of the GMF-88 test of 7.64 points (95% CI 0.41-14.82; p 0.038). However, no statistically significant differences were found between HRS vs. sham (SMD 0.19, 95% CI −0.34-1.36; p 0.49; Table 5, Figure 3). Secondly, five studies [64,[69][70][71]74] reported data from 129 patients (25.8 per comparison) to assess the effect of HRS therapy on the B dimension (sitting ability) of gross motor function. Our findings showed a medium effect (SMD 0.52; 95% CI 0.15-0.9; p 0.006) favors HRS (Table 5, Figure 2). A low risk of publication bias was present (p for Egger 0.07) due to trim-and-fill and showed a variation of 15% (adjusted SMD 0.6; 95% CI 0.24-0.94) with respect to the original pooled effect ( Figure S1). The level of heterogeneity was moderate (I 2 41%; p 0.04). Subgroup analysis revealed that the use of HRS plus physiotherapy was better (SMD 0.84; 95% CI 0.32-1.36; p 0.002) than physiotherapy alone, showing an improvement in the sitting ability of the GMF-88 test of 7.64 points (95% CI 0.41-14.82; p 0.038). However, no statistically significant differences were found between HRS vs. sham (SMD 0.19, 95% CI −0.34-1.36; p 0.49; Table 5, Figure 3).

Functional Balance
Two studies [63,67] with three independent comparisons provided data from 68 participants (22.7 per comparison) to assess the effectiveness of HRS therapy on functional balance. Our findings showed low-quality evidence of a medium effect (SMD 0.6; 95% CI 0.1-1.08; p 0.018) of HRS plus physiotherapy in comparison to physiotherapy (Table 5, Figure 5). In addition, the combination of HRS plus physiotherapy increased the functional balance measured with PBS by 6.21 points (95% CI 1.14-10.62; p 0.015). No risk of publication bias or heterogeneity was found. Sensitivity analysis did not report substantial variations in the pooled effect when the studies were excluded.  Finally, six studies [63,67,[69][70][71]74] with seven independent comparisons provided data from 187 participants (26.7 per comparison) to assess the effect of HRS therapy on the total score of gross motor function. Our findings reported a medium effect (SMD 0.64; 95% CI 0.34-0.94; p < 0.001) in favor of HRS therapy (Table 5, Figure 2). The risk of publication bias was not present, and heterogeneity was low (I 2 Figure 4).

Functional Balance
Two studies [63,67] with three independent comparisons provided data from 68 participants (22.7 per comparison) to assess the effectiveness of HRS therapy on functional balance. Our findings showed low-quality evidence of a medium effect (SMD 0.6; 95% CI 0.1-1.08; p 0.018) of HRS plus physiotherapy in comparison to physiotherapy (Table 5, Figure 5). In addition, the combination of HRS plus physiotherapy increased the functional balance measured with PBS by 6.21 points (95% CI 1.14-10.62; p 0.015). No risk of publication bias or heterogeneity was found. Sensitivity analysis did not report substantial variations in the pooled effect when the studies were excluded.

Functional Balance
Two studies [63,67] with three independent comparisons provided data from 68 participants (22.7 per comparison) to assess the effectiveness of HRS therapy on functional balance. Our findings showed low-quality evidence of a medium effect (SMD 0.6; 95% CI 0.1-1.08; p 0.018) of HRS plus physiotherapy in comparison to physiotherapy (Table 5, Figure 5). In addition, the combination of HRS plus physiotherapy increased the functional balance measured with PBS by 6.21 points (95% CI 1.14-10.62; p 0.015). No risk of publication bias or heterogeneity was found. Sensitivity analysis did not report substantial variations in the pooled effect when the studies were excluded.

Abduction Pelvic Range of Motion
Two studies [68,72] with two independent comparisons provided data from 54 participants (27 per comparison) to compare the effectiveness of HRS vs. sham in increasing abduction pelvic ROM. Our findings revealed a large effect (SMD 0.79; 95% CI 0.21-1.37; p 0.008) that favored HRS therapy (Table 5, Figure 6), being able to increase it by 7.49 degrees (95% CI 2.45-12.5; p 0.004), compared to sham. The risk of publication bias could not be studied, and heterogeneity was moderate (I 2 37%; p 0.13).

Spasticity
The effect of HRS therapy in reducing spasticity was assessed in hip adductors, knee extensors and ankle dorsiflexors.

Abduction Pelvic Range of Motion
Two studies [68,72] with two independent comparisons provided data from 54 participants (27 per comparison) to compare the effectiveness of HRS vs. sham in increasing abduction pelvic ROM. Our findings revealed a large effect (SMD 0.79; 95% CI 0.21-1.37; p 0.008) that favored HRS therapy (Table 5, Figure 6), being able to increase it by 7.49 degrees (95% CI 2.45-12.5; p 0.004), compared to sham. The risk of publication bias could not be studied, and heterogeneity was moderate (I 2 37%; p 0.13).

Abduction Pelvic Range of Motion
Two studies [68,72] with two independent comparisons provided data from 54 participants (27 per comparison) to compare the effectiveness of HRS vs. sham in increasing abduction pelvic ROM. Our findings revealed a large effect (SMD 0.79; 95% CI 0.21-1.37; p 0.008) that favored HRS therapy (Table 5, Figure 6), being able to increase it by 7.49 degrees (95% CI 2.45-12.5; p 0.004), compared to sham. The risk of publication bias could not be studied, and heterogeneity was moderate (I 2 37%; p 0.13).

Spasticity
The effect of HRS therapy in reducing spasticity was assessed in hip adductors, knee extensors and ankle dorsiflexors.

Spasticity
The effect of HRS therapy in reducing spasticity was assessed in hip adductors, knee extensors and ankle dorsiflexors.

Qualitative Synthesis
Regarding the qualitative synthesis of the studies included in our systematic review, we were able to reach the following conclusions. Two studies [65,72] compared the effect of HRS therapy on pelvic tilt. Despite using different treatment protocols, the results of both studies concluded that there were statistically significant improvements in those patients who received HRS therapy (p < 0.05). Furthermore, we found two studies [65,74] that compared the effect of HRS therapy on trunk control during sitting. Both studies concluded that there were greater statistically significant improvements in children who were part of the HRS therapy group, especially in improving reactive trunk control (p 0.004).  [66,73] in the HRS therapy groups, while only Silva-Borges et al. (2011) found statistically significant differences between groups in anteroposterior sway [73]. Lastly, a single study [73] analyzed the satisfaction perceived by patients after treatment sessions. The study concluded that children who received treatment with HRS therapy perceived greater satisfaction compared to children who did not receive this therapy. In addition, this study reported that no child was unhappy with the use of the HRS, while 25% of children belonging to the CT group were unhappy with the therapy. Findings of the qualitative synthesis in each study included are shown in Table 4.  Finally, the spasticity of the ankle dorsiflexors and of the knee extensors was assessed in one study [63] with two independent comparisons (each one) (Table 5, Figure 7). No statistically significant differences were found between HRS plus PT vs. PT in the reduction of the spasticity of the ankle dorsiflexors (SMD −0.55; 95% CI −1.22-0.12; p 0.11) or of the knee extensors (SMD −0.12; 95% CI −0.77-0.52; p 0.71).

Qualitative Synthesis
Regarding the qualitative synthesis of the studies included in our systematic review, we were able to reach the following conclusions. Two studies [65,72] compared the effect of HRS therapy on pelvic tilt. Despite using different treatment protocols, the results of both studies concluded that there were statistically significant improvements in those patients who received HRS therapy (p < 0.05). Furthermore, we found two studies [65,74] that compared the effect of HRS therapy on trunk control during sitting. Both studies concluded that there were greater statistically significant improvements in children who were part of the HRS therapy group, especially in improving reactive trunk control (p 0.004).  [66,73] in the HRS therapy groups, while only Silva-Borges et al. (2011) found statistically significant differences between groups in anteroposterior sway [73]. Lastly, a single study [73] analyzed the satisfaction perceived by patients after treatment sessions. The study concluded that children who received treatment with HRS therapy perceived greater satisfaction compared to children who did not receive this therapy. In addition, this study reported that no child was unhappy with the use of the HRS, while 25% of children belonging to the CT group were unhappy with the therapy. Findings of the qualitative synthesis in each study included are shown in Table 4. Finally, the spasticity of the ankle dorsiflexors and of the knee extensors was assessed in one study [63] with two independent comparisons (each one) (Table 5, Figure 7). No statistically significant differences were found between HRS plus PT vs. PT in the reduction of the spasticity of the ankle dorsiflexors (SMD −0.55; 95% CI −1.22-0.12; p 0.11) or of the knee extensors (SMD −0.12; 95% CI −0.77-0.52; p 0.71).

Qualitative Synthesis
Regarding the qualitative synthesis of the studies included in our systematic review, we were able to reach the following conclusions. Two studies [65,72] compared the effect of HRS therapy on pelvic tilt. Despite using different treatment protocols, the results of both studies concluded that there were statistically significant improvements in those patients who received HRS therapy (p < 0.05). Furthermore, we found two studies [65,74] that compared the effect of HRS therapy on trunk control during sitting. Both studies concluded that there were greater statistically significant improvements in children who were part of the HRS therapy group, especially in improving reactive trunk control (p 0.004). found statistically significant differences in medial-lateral sway [66,73] in the HRS therapy groups, while only Silva-Borges et al. (2011) found statistically significant differences between groups in anteroposterior sway [73]. Lastly, a single study [73] analyzed the satisfaction perceived by patients after treatment sessions. The study concluded that children who received treatment with HRS therapy perceived greater satisfaction compared to children who did not receive this therapy. In addition, this study reported that no child was unhappy with the use of the HRS, while 25% of children belonging to the CT group were unhappy with the therapy. Findings of the qualitative synthesis in each study included are shown in Table 4.

Discussion
Although the use of HRSs in the management of motor, balance and gait disorders in different neurological and musculoskeletal diseases is increasing in physiotherapy approaches, only six reviews have compiled the published evidence about HRS therapy on chronic pain [33,75], autism spectrum disorder [76], stroke and cerebral palsy [35,39] or older adults [38]. Due to the high prevalence of CP in children and the fact that CP produces a high level of disability in them, it is necessary to assess the effectiveness of the therapies applied to them, such as HRS or physiotherapy, and to analyze what therapy can provide more improvement in its recovery. There is no review that compiles all published articles that have assessed the effectiveness of HRS therapy in improving gross motor function, balance, spasticity or hip ROM in children with CP. Two previous reviews [35,39] have reported on the effect of HRS therapy on gross motor function or balance in these patients, but the number of studies included was less than four in each case, so these results may not be generalizable and are susceptible to change when new studies are included in the meta-analysis, due to a possible risk of publication bias. Therefore, our systematic review is the first meta-analysis that includes the largest number of studies to date to analyze the effect of HRS therapy in children with CP. In addition, our review assesses, for the first time, variables such as spasticity, hip ROM, posturographic parameters and the level of satisfaction of the children with CP with HRS therapy. In addition, when it was possible, we provided subgroup analyses to assess if the effect of HRS therapy is large when it is used alone or combined with physiotherapy. The findings of our meta-analyses show that HRS therapy is effective in improving gross motor function, functional balance and hip abduction ROM in children with CP.
Regarding gross motor function, we assessed the effect of HRS therapy on the total score and on each dimension. We found statistically significant differences favoring HRS therapy in improving sitting ability and total gross motor function in comparison to other therapies such as physiotherapy or sham. In addition, the effectiveness of HRS therapy on sitting ability (SMD 0.84) and total gross motor function (SMD 0.98) was large when it was used in combination with physiotherapy compared to physiotherapy. Our meta-analysis identified two important benefits of HRS plus physiotherapy on these variables. On the one hand, the effect of HRS plus physiotherapy was unknown until this study was conducted. On the other hand, we calculated the mean difference between therapies, showing that HRS plus physiotherapy was able to increase the sitting ability and global total score in the GMF-88 test by 7.64 points and 11.21 points, respectively, compared with physiotherapy. Currently, no study has published the MCID for sitting ability in this test, so our data cannot be compared to this MCID, but we consider that an increase of almost 8 points for sitting ability may have relevance for clinical practice, helping these patients to better carry out their activities of daily living in the sitting position. However, on the global total score, our findings exceed the MCID value in GMF-88, calculated in 2020 by Storm et al. [77]. These two findings represent the most important findings of our review and establish that HRS therapy is effective for improving sitting ability and gross motor function in children with CP, but the effect was more pronounced when HRS was used with physiotherapy. These findings cannot be compared with the review of Heussen and Häusler, as it does not report specific data for the effect of HRS therapy unless it is integrated into an analysis of hippotherapy and therapeutic riding [39].
Secondly, our findings showed that the inclusion of HRSs in physiotherapy protocols is effective for improving functional balance in comparison to physiotherapy only (SMD 0.6). We determined that the combination of HRS plus physiotherapy increased the total score in PBS by 6.21 points. Chen et al. (2013) reported that the MCID value for the total score of PBS in children with CP was 5.83 points [78]. Our findings showed that the use of HRSs in physiotherapy for recovering functional balance is clinically relevant, and it exceeds the MCID for PBS in children with CP [78]. In addition, this data is supported by improvements obtained in balanced sitting posturographic parameters; therefore, HRS therapy increases the sitting trunk control [65] and reduces medial-lateral [66] and anteroposterior sway area [73] when assessed with static posturography. Our results cannot be directly compared with previous reviews since there are no studies that analyze the effect of HRS therapy on functional balance. Only, Dominguez-Romero et al. (2020) found that HRS therapy was better than conventional therapy interventions in improving functional balance (Berg Balance Scale) in patients with stroke, using data from two studies for the meta-analysis, agreeing with them that HRS therapy is effective for improving functional balance in patients with central nervous system diseases [35]. Previous studies have shown that balance training on unstable support surfaces produces activation of the trunk musculature and a continuous response of the back muscles to maintain a stable center of mass [79,80]. The new HRS devices can generate specifically three-dimensional slight movements of the trunk and pelvis per minute, similar to those experienced by the body riding a real horse in hippotherapy and favoring the training of postural reactions of the trunk [10].
The last meta-analysis assessed the effectiveness of HRS therapy on abduction hip ROM and lower limb spasticity. Our results showed that HRS therapy was effective in increasing the hip abduction ROM by 7.5 degrees in a goniometry assessment, in comparison to sham. Regarding spasticity, HRS plus physiotherapy was not better than physiotherapy only in reducing hip adductor, knee extensor and ankle dorsiflexor spasticity. However, in the hip adductor spasticity meta-analysis, our findings were underestimated as a result of publication bias, and when the pooled effect was calculated, taking into account this possible bias (trim-and-fill variation of 100%), we found that HRS therapy could have a large effect (adjusted SMD −0.81) in reducing spasticity in adductor muscles. This last result highlights the importance of performing future research with the aim of confirming this possible finding without the risk of publication bias. The improvement in hip abduction ROM and possible reduction in adductor spasticity can be explained by the continuous riding position of the equine simulator, in which both hips are abducted, keeping the adductor muscles continuously stretched, reducing its shortening and helping to prevent the neuromuscular hip dysplasia which is common in these children [81].
Finally, our review highlights an important finding in the study of Silva-Borges et al. (2011), in which the level of satisfaction in the HRS therapy was assessed in comparison to conventional therapy [73]. Children with CP that received HRS therapy reported more happiness, while in the conventional therapy group, approximately 25% of those children were unhappy with the therapy. This data highlights the need for therapies that attract children's attention, thus increasing their motivation and adherence to therapy to obtain better results. In previous studies, patients with various pathologies who have been subjected to conventional classical treatments have shown signs of monotony and lack of adherence [82,83], which could explain why they did not experience a clear improvement compared to other more active and striking therapies, such as HRS, virtual reality-based therapy or robot-assisted gait training devices.
HRSs have some benefits that allow their easy inclusion within the clinical practice, including the lower costs of maintenance of the machine compared to the costs of care and training of horses [33,75]. Moreover, the facilities where HRS therapy is carried out do not have to be large, unlike the facility that is needed to carry out therapy with real horses. Due to the large size requirements, many of the centers where hippotherapy is carried out are outside of urban centers, which means an extra cost for the trip to the facility for families. An HRS, being a device of relatively small size, can very easily be part of a hospital or neurorehabilitation clinic. There are also other aspects that favor therapy with HRSs, such as weather conditions, children's fear of riding the animal or potential allergic reactions that children may develop [75], although it also can appear due to plastic or metal materials of HRS being built.
Although the findings presented in this systematic review and meta-analysis are interesting and relevant for clinical practice, it is important to note some limitations. First, the small number of studies that are included for each variable in which the meta-analysis has been carried out should be highlighted. This is not a limitation of our literature search process unless, due to the scarcity of studies that assess the effect of HRS therapy in CP that have been published to date, they did not meet the inclusion criteria. The second is related to the precision of the results, which is derived from the number of participants. The included studies comprise small sample sizes, and this may lead to underestimation of the results when combining the studies with meta-analysis, as has occurred. The low number of studies conducted to date and their small sample size may make it difficult to generalize the results and reduces the quality of the evidence from the overall analyses; however, no other reviews have been published to date. Third, it is important to highlight the moderate risk of bias in the included studies. Furthermore, the heterogeneity in the HRS therapy protocols used (frequency, number of sessions, and devices used) does not help determine which type of HRS therapy is the most effective. It is also important to highlight the risk of publication bias present in some studies. However, at this point, it is necessary to point out that the risk of publication bias meant that there were no statistically significant differences between HRS plus physiotherapy vs. physiotherapy in increasing spasticity hip adductors, and the trim-and-fill estimation determined that, without publication bias, HRS therapy would improve this variable. Finally, the last limitation is that this review only assessed the immediate effect of HRS therapy because the studies included did not provide followup data.

Conclusions
This is the first systematic review with a meta-analysis that quantitatively assesses the effect of HRS therapy on gross motor function, functional balance, spasticity and hip ROM and reports a qualitative synthesis of other secondary outcomes such as posturographic parameters and satisfaction of the participants, in comparison to physiotherapy or sham. Our meta-analysis reported that HRS therapy is effective in improving overall gross motor function and sitting ability, functional balance and hip abduction ROM in children with CP. More specifically, we reported that when HRS is combined with physiotherapy, the improvements found in gross motor function, sitting ability and functional balance are higher, but no differences between these therapies were found to reduce spasticity in hip adductors, knee extensors or ankle dorsiflexors. More studies that assess the effectiveness of HRS therapy in these variables and in others, such as quality of life or functional independence, are needed to carry out in the future. An increase in the sample size of these studies will help obtain more robust and accurate results and enable the HRS therapy findings to be generalized. Finally, to guarantee the comparison between therapies in future studies, it would be necessary to homogenize the duration and intensity of the HRS protocols and to evaluate the variables in different follow-up times, in addition to immediate post-intervention, to verify the efficacy of HRS over time.