Motor effects of movement representation techniques and cross-education: a systematic review and meta-analysis

Abstract

INTRODUCTION

The objective was to assess the impact of movement representation techniques (MRT) through motor imagery (MI), action observation (AO) and visual mirror feedback (VMF) and cross-education training (CE) on strength, range of motion (ROM), speed, functional state and balance during experimental immobilization processes in healthy individuals, in patients with injuries that did not require surgery and in those with surgical processes that did or did not require immobilization.

EVIDENCE ACQUISITION

MEDLINE, EMBASE, CINAHL and Google Scholar were searched. Thirteen meta-analyses were conducted.

EVIDENCE SYNTHESIS

Regarding the immobilized participants, in the healthy individuals, MI showed significant results regarding maintenance of strength and ROM, with low-quality evidence. Regarding the process with no immobilization, VMF and MI techniques showed significant changes in maintaining ROM in patients with injury without surgery, with very low-quality evidence. Results had shown that MI demonstrated significantly higher maintenance of strength and speed in patients undergoing surgery, with low-quality evidence. No significant results were found in ROM. Low-quality evidence showed better results in AO plus usual care compared with usual treatment in isolation with respect to maintenance of functional state and balance. CE training demonstrated maintenance of strength in patients undergoing surgery, with moderate evidence; however, not in healthy experimentally immobilized individuals. VMF did not show significant results in maintaining ROM after surgery without immobilization, nor did MI in maintaining strength after surgery and immobilization.

CONCLUSIONS

MRT and CE training have been shown to have a significant impact on the improvement of various motor variables and on physical maintenance in general.

Introduction

After an orthopedic injury or a surgical procedure, immobilization or movement restriction of the injured limb, either due to pain, injury or the use of external immobilization, is a common scenario. Movement reduction or limb disuse can lead to neuroplastic, neurobiological and sensorimotor changes, both in neuromuscular function and at the supramedullary level.1, 2

Recently, research has found that loss of muscle size and strength occurs along with the presence of changes in neuromuscular function at the peripheral level and at the central level, with changes in muscle fiber excitability and contractility, as well as a reduction in spinal and corticospinal excitability and reduced central movement drive to the muscle.1 In addition, limb immobilization for at least 2 weeks has been found to be associated with a process of cortical reorganization in the thickness of the motor area primarily responsible for the specific body region, as well as its associated somatosensory cortex. A reduction in thickness was found in both brain regions. At the white matter level, a decrease in the corticospinal tract volume was also found. It therefore appears that cortical depression occurs during immobilization of a limb.2

These changes produced by the immobilization or disuse of a limb can lead to a reduction in functional variables of muscle strength, range of motion (ROM) or coordination, which has been associated with increased complications and recovery times.3, 4 Therefore, in recent years, research has been focused on developing new treatment alternatives that can reduce the potential impact of immobilization or disuse after a musculoskeletal injury and improve recovery processes. Some of these alternatives, such movement representation techniques or cross-education (CE) training, attempt to promote central nervous system activity to avoid cortical depression processes and thus prevent functional alterations. The main common factor in these techniques is that it is possible to influence the affected limb without the need for its active movement, leading to a great many possibilities for intervention when active movement is not possible.

These techniques have in common that they lead to an activation of the areas related to the planning, adjustment and automation of voluntary movement in a similar manner as to when the action occurs in a real manner. These techniques are motor imagery (MI), action observation (AO), mirror therapy and visual mirror feedback (VMF). MI is defined as a dynamic mental process that involves the representation of an action, in an internal manner, without its real motor output.5 AO, however, evokes an internal, real-time simulation of what the observer is seeing.6 VMF is defined as the reflective illusory movement perception in one limb upon viewing the moving opposite limb in a midsagittal mirror.7 On the other hand, CE, first described in the 19th century by Scriture et al.,8 is defined as an increased capacity to generate strength with the untrained limb as a result of training the other limb unilaterally.9 In this regard, previous research has suggested that a CE strength task could have led to cortical excitability and a motor learning effect that was reflected in improvements in performance in the untrained left arm.9, 10

It was therefore the main aim of this systematic review and meta-analysis to assess the impact of movement representation techniques and CE training on strength, ROM, speed, functional state and balance during experimental immobilization processes in healthy individuals, in patients with injuries that did not require surgery and in surgical processes that did or did not require immobilization.

Evidence acquisition

This systematic review and meta-analysis was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines described by Moher et al.11 The protocol of this systematic review and meta-analysis was registered (7^th February 2020) in an international register prior to starting the review (PROSPERO, CRD42020168372).

Inclusion criteria

The selection criteria used in this systematic review and meta-analysis were based on methodological and clinical factors, such as the Population, Intervention, Control, Outcomes, and Study design described by Stone.12

Population

The participants selected for the studies were older than 18 years, and were asymptomatic individuals or patients who met at least one of the following conditions: 1) experimental immobilization; 2) patients with fracture or injury, with or without immobilization; or 3) postsurgical patients with or without immobilization. The participant’s sex was irrelevant.

Intervention and control

The interventions were movement representation techniques (MI, AO or VMF) and CE strategy. The intervention could be given as an independent intervention, added to an intervention or embedded in an intervention (e.g., usual care or conventional treatment). Regarding MI, both visual and kinesthetic strategies as well as both perspectives of movement representations could be considered (first or third). Studies that used a combination of various movement representation techniques (e.g., graded motor imagery [GMI], consisting of implicit MI, explicit MI and VMF) were also included. Regarding the control group, the comparators were conventional intervention or usual care (physical therapy, exercise intervention) in combination or not with placebo interventions (cognitive task, relaxation).

Outcomes

The measures used to assess the results and effects were the following motor variables: ROM, balance, strength, functional state and walking speed.

Study design

Randomised controlled trials (RCTs), randomised parallel-design controlled trials (RPCTs) and prospective controlled clinical trials were selected.

Search strategy

The search for studies was performed using MEDLINE (PubMed), EMBASE, CINAHL and Google Scholar. The final search was run on 8^th February 2020 (Supplementary Digital Material 1: Supplementary Text File 1).

A validated search filter was used and adapted to all the databases.13-15 Based on international criteria, no restriction was applied with respect to the language of the studies.16 Using the same methodology, two researchers conducted the search for the studies independently. Consensus served to resolve possible differences between them. In addition, manual searching through journals that usually publish on the topic in question was used to include all available articles. In all the articles found in a first search, the ‘Introduction’, ‘Discussion’ and ‘Reference’ sections were reviewed in order not to miss any relevant articles. Mendeley’s appointment management software (Mendeley desktop v. 1.17.4, Elsevier, New York, NY, USA) was used to remove duplicate articles.17

Selection criteria and data extraction

First, a data analysis was performed by two independent reviewers (F.C.M and L.S.M), who assessed the relevance of the RCTs regarding the study questions and aims. This first analysis was made based on information from the title, abstract and key words of each study. If there was no consensus or the abstracts did not contain enough information, the full text was reviewed.

Second, the full text was used with the aim of assessing whether the studies met all the inclusion criteria. Differences between both independent reviewers were resolved by a process of consensus moderated by a third reviewer (R.L.T).18 Data described in the results were extracted by means of a structured protocol that ensured that the most relevant information was obtained from each study.19

Methodological quality assessment

The Cochrane Handbook for Systematic Reviews of Interventions version 5.1.0 was used to assess the risk of bias.19 This assessment tool covers a total of seven domains: 1) random sequence generation (selection bias); 2) allocation concealment (selection bias); 3) blinding of participants and personnel (performance bias); 4) blinding of outcome assessment (detection bias); 5) incomplete outcome data (attrition bias); 6) selective reporting (reporting bias); and 7) other biases. Bias risk was assessed as low, high or unclear.

Two independent reviewers (F.C.M and L.S.M) examined the quality of all the selected studies using the same methodology; disagreements between reviewers were resolved by consensus including a third reviewer (R.L.T). The concordance between the results (inter-rater reliability) was performed using Cohen’s kappa coefficient (κ): (1) κ>0.7 means high level of agreement between assessors; 2) κ: 0.5-0.7 is a moderate level of agreement; and (3) κ <0.5 is a low level of agreement.20

Qualitative analysis

The qualitative analysis was based on classifying the results into levels of evidence according to the Grading of Recommendations, Assessment, Development and Evaluation (GRADE), which is based on five domains: 1) study design; 2) imprecision; 3) indirectness; 4) inconsistency; and 5) publication bias.21 The assessment of the five domains was conducted according to GRADE criteria.22, 23

Data synthesis and analysis

The statistical analysis was conducted using meta-analysis with interactive explanation software (MIX, version 1.7).24 To provide a comparison between outcomes reported by the studies, the standardized mean difference (SMD) over time and corresponding 95% CI were calculated for the continuous variables. The statistical significance of the pooled SMD was examined as Hedges’ g, to account for possible overestimation of the true population effect size in small studies.25

The same three inclusion criteria were used for the systematic review and for the meta-analysis: 1) the results showed detailed information regarding the comparative statistical data of the exposure factors, therapeutic interventions and treatment responses; 2) the intervention was compared with a similar control group (e.g., usual care or conventional physical therapy protocol); and 3) data on the analyzed variables were represented in at least two studies.

The estimated SMDs were interpreted as described by Hopkins et al.;26 i.e., an SMD of 4.0 was considered to represent an extremely large clinical effect, 2.0-4.0 a very large effect, 1.2-2.0 a large effect, 0.6-1.2 a moderate effect, 0.2-0.6 a small effect and 0.0-0.2 a trivial effect. The degree of heterogeneity among the studies was estimated by the Cochran’s Q statistical test (a p value <0.05 was considered significant) and the inconsistency index (I²).27 I²>25% was considered to represent small, I²>50% medium and I²>75% large heterogeneity.28 The I² index is a complement to the Q test, although it has the same problems of power with a small number of studies.28 When the Q-test was significant (P<0.1) and/or the result of I² is >75%, this indicated that there was heterogeneity among the studies and the random-effects model was conducted in the meta-analysis. To detect publication biases and test the influence of each individual study, a visual evaluation of the funnel plot and exclusion sensitivity plot, seeking asymmetry, was performed. We also employed Egger’s regression tests to assess publication bias.29

Evidence synthesis

The study search strategy is shown in the form of a flow chart (Figure 1). A total of 34 articles that met the inclusion criteria were selected. Of the total number of articles included, 12 had performed complete immobilization due to surgery or an experimental condition. The remaining 22 articles included participants who had surgery or an orthopedic injury to the musculoskeletal system that restricted movement, although they did not receive external mobilization. The characteristics for which data were extracted (sample size, demographic characteristics, intervention, outcomes, main results, and conclusions) are presented in Supplementary Digital Material 2 (Supplementary Table I, Supplementary Table II).

Figure 1.

—PRISMA flow diagram.

Methodological quality analysis

The quality of all the studies was evaluated with the Cochrane assessment tool. Most of the studies had a low risk of selective reporting bias. The domain with the highest percentage of studies with a high risk of bias was the blinding of participants and personnel (performance bias). The risk of bias summary and risk of bias graph are shown in Figure 2 and Figure 3, respectively. The inter-rater reliability of the methodological quality assessment between assessors was high (κ=0.813).

Figure 2.

—Risk of bias summary. Review authors’ judgements about each risk of bias item for each included study (risk of bias scale).

Figure 3.

—Risk of bias graph. Review authors’ judgements about each risk of bias item presented as percentages across all included studies (risk of bias scale).

Study population characteristics

The total number of participants was 976. The number of participants included with immobilization was 313, and without immobilization 663.

Regarding immobilization studies, the nature of the immobilization was experimental in nine studies30-38 and due injury39 or surgery40, 41 in three studies. In relation to the no immobilization studies, five were nonsurgical procedures and included patients with orthopedic injuries: two ankle sprain,42, 43 one shoulder impingement,44 one adhesive capsulitis45 and one orthopedic hand injury.46 Seventeen studies were surgical procedures: four were anterior cruciate ligament reconstruction procedures,47-50 seven were knee surgery or knee replacement,51-57 three were hip replacement58-60 and three were orthopedic hand injuries.61-63

Interventions

Regarding the immobilization studies, six used MI or GMI as an experimental intervention30-32, 34, 40, 41 and six used a CE intervention.33, 35-39 Regarding the no immobilization studies, eight employed MI,42, 43, 47, 51-53, 58 five used VMF,45, 46, 61-63 four used AO,54-56, 59 and three used CE.48-50 Frenkel et al. had combined VFM with MI57 and Marusic et al. had combined MI and AO.60 Regarding control comparisons, all studies included standard rehabilitations as a control group, except the studies that used an experimental immobilization that used no intervention as a control group. In addition, Villafañe et al. and Cupal and Brewer47, 56 used a sham intervention.

Meta-analysis results

Experimental immobilization: effects of motor imagery on strength

The meta-analysis showed statistically significant differences for the MI intervention, with a very large clinical effect in two studies31, 34 (N.=46; SMD=2.73; 95% CI: 1.91-3.55; heterogeneity Q value 0.06; P=0.8) (Figure 4A). The shape of the funnel plot appeared to be symmetrical in the dominant model (Supplementary Digital Material 3: Supplementary Text File 2).

Figure 4.

—Synthesis forest plot. This forest plot summarizes the results of included studies (sample size, standardized mean differences [SMDs], and weight). The small boxes with the squares represent the point estimate of the effect size and sample size. The lines on either side of the box represent a 95% confidence interval (CI). A) Forest plot for experimental immobilization studies that used motor imagery intervention on strength outcome; B) forest plot for experimental immobilization studies that used motor imagery intervention on range of motion outcome; C) forest plot for experimental immobilization studies that used cross-education intervention on strength outcome; D) forest plot for surgery immobilization studies that used motor imagery intervention on strength outcome.

Experimental immobilization: effects of motor imagery on range of motion

The meta-analysis showed statistically significant differences for MI intervention with a moderate clinical effect in two studies30, 32 (N.=39; SMD=0.7; 95% CI 0.05-1.35; heterogeneity Q value 0; P=0.99) (Figure 4B). The shape of the funnel plot appeared to be symmetrical in the dominant model (Supplementary Digital Material 3).

Experimental immobilization: effects of cross-education on strength

The meta-analysis did not show statistically significant differences in the CE intervention in 3 studies33, 37, 38 (N.=51; SMD=1.85; 95% CI: -0.07 to 3.77; heterogeneity Q value 14.82; P<0.01; inconsistency I²=87%), and there was no evidence of publication bias in the meta-analysis (standard error [SE] 1.13; t=-3.04; P=0.2) (Figure 4C). The shape of the funnel plot appeared to be asymmetrical in the dominant model (Supplementary Digital Material 3). The sensitivity exclusion analysis suggested that two studies33, 38 significantly affected pooled SMD (Supplementary Digital Material 3). Egger’s test results suggested no significant evidence of publication bias for the analysis (intercept=1.93; t=4.26; P=0.15).

Surgery immobilization: effects of motor imagery on strength

The meta-analysis did not show statistically significant differences in MI intervention in two studies40, 41 (N.=61; SMD=0.13; 95% CI: -0.37 to 0.64; heterogeneity Q value 0.9; P=0.34) (Figure 4D). The shape of the funnel plot appeared to be symmetrical in the dominant model (Supplementary Digital Material 3).

Surgery without immobilization: effects of action observation on balance

The meta-analysis showed statistically significant differences in AO interventions with a moderate clinical effect in four studies54-56, 59 (N.=132; SMD=0.61; 95% CI: 0.18-1.03; heterogeneity Q value 3.92; P=0.17; inconsistency I²=24%), and there was no evidence of publication bias in the meta-analysis (SE=1.07; t=0.58; P=0.62) (Figure 5A). The shape of the funnel plot appeared to be symmetrical in the dominant model (Supplementary Digital Material 3). The sensitivity exclusion analysis suggested that one study55 significantly affected pooled SMD (Supplementary Digital Material 3). Egger’s test results suggested no significant evidence of publication bias for the analysis (intercept=-2.98; t=-0.03; P=0.98).

Figure 5.

—Synthesis forest plot. This forest plot summarizes the results of included studies (sample size, standardized mean differences [SMDs], and weight). The small boxes with the squares represent the point estimate of the effect size and sample size. The lines on either side of the box represent a 95% confidence interval (CI). A) Forest plot for surgery studies that used action observation intervention on balance outcome; B) forest plot for surgery studies that used action observation intervention on functional status outcome; C) forest plot for surgery studies that used visual mirror feedback intervention on range of motion outcome; D) forest plot for surgery studies that used cross-education intervention on strength outcome.

Surgery without immobilization: effects of action observation on functional state

The meta-analysis showed statistically significant differences in AO interventions with a moderate clinical effect in four studies54-56, 59 (N.=132; SMD=0.74; 95% CI: 0.34-1.14; heterogeneity Q value 3.54; P=0.32; inconsistency I²=15%), and there was no evidence of publication bias in the meta-analysis (SE=0.52; t=-0.5; P=0.67) (Figure 5B). The shape of the funnel plot appeared to be symmetrical in the dominant model (Supplementary Digital Material 3). The sensitivity exclusion analysis suggested that no study significantly affected the pooled SMD. Egger’s test results suggested no significant evidence of publication bias for the analysis (intercept=1.43; t=1.94; P=0.19).

Surgery without immobilization: effects of cross-education on strength

The meta-analysis showed statistically significant differences in the CE intervention with a moderate clinical effect in three studies.48-50 Two studies included different training dosage groups; thus, a total of five groups were included in the analysis (N.=163; SMD=0.65; 95% CI: 0.33-0.96; heterogeneity Q value 3.21; P=0.52; inconsistency I²=0%), and there was no evidence of publication bias in the meta-analysis (SE=0.73; t=-2.77; P=0.07) (Figure 5C). The shape of the funnel plot appeared to be symmetrical in the dominant model (Supplementary Digital Material 3). The sensitivity exclusion analysis suggested that no study significantly affected the pooled SMD. However, egger’s test results suggested significant evidence of publication bias for the analysis (intercept=2.02; t=3.67; P=0.03).

Surgery without immobilization: effects of visual mirror feedback on range of motion

The meta-analysis did not show statistically significant differences in the VMF intervention in four studies57, 61-63 (N.=141; SMD=0.46; 95% CI: -0.06 to 0.98; heterogeneity Q value 7; P=0.07; inconsistency I²=57%), and there was no evidence of publication bias in the meta-analysis (SE=2.62; t=-0.28; P=0.81) (Figure 5D). The shape of the funnel plot appeared to be asymmetrical in the dominant model (Supplementary Digital Material 3). The sensitivity exclusion analysis suggested that three studies significantly affected pooled SMD57, 61, 62 (Supplementary Digital Material 3). Egger’s test results suggested no significant evidence of publication bias for the analysis (intercept=7.57; t=0.45; P=0.7).

Surgery without immobilization: effects of motor imagery on strength

The meta-analysis showed statistically significant differences in MI interventions with a large clinical effect in three studies47, 52, 53 (N.=66; SMD=1.26; 95% CI: 0.71-1.8; heterogeneity Q value 2.07; P=0.36; inconsistency I²=3%), and there was no evidence of publication bias in the meta-analysis (SE=0.63; t=-3.34; P=0.19) (Figure 6A). The shape of the funnel plot appeared to be symmetrical in the dominant model (Supplementary Digital Material 3). The sensitivity exclusion analysis suggested that no study significantly affected pooled SMD. Egger’s test results suggested no significant evidence of publication bias for the analysis (intercept=1.32; t=5.37; P=0.12).

Figure 6.

—Synthesis forest plot. This forest plot summarizes the results of included studies (sample size, standardized mean differences [SMDs], and weight). The small boxes with the squares represent the point estimate of the effect size and sample size. The lines on either side of the box represent a 95% confidence interval (CI). A) Forest plot for surgery studies that used motor imagery intervention on strength outcome; B) forest plot for surgery studies that used motor imagery intervention on walking speed outcome; C) forest plot for surgery studies that used motor imagery intervention on range of motion outcome.

Surgery without immobilization: effects of cross-education on walking speed

The meta-analysis showed statistically significant differences in MI interventions with a moderate clinical effect in three studies58, 60, 64 (N.=71; SMD=0.56; 95% CI: 0.08-1.03; heterogeneity Q value 0.37; P=0.83; inconsistency I²=0%), and there was no evidence of publication bias in the meta-analysis (SE=1.73; t=-1.01; P=0.5) (Figure 6B). The shape of the funnel plot appeared to be symmetrical in the dominant model (Supplementary Digital Material 3). The sensitivity exclusion analysis suggested that two studies significantly affected the pooled SMD60, 64 (Supplementary Digital Material 3). Egger’s test results suggested no significant evidence of publication bias for the analysis (intercept=4.11; t=1.33; P=0.41).

Surgery without immobilization: effects of cross-education on range of motion

The meta-analysis did not show statistically significant differences in MI intervention in two studies51, 52 (N.=30; SMD=0.7; 95% CI: -0.89 to 2.29; heterogeneity Q value 3.42; P=0.06; inconsistency I²=71%) (Figure 6C). The shape of the funnel plot appeared to be symmetrical in the dominant model (Supplementary Digital Material 3).

Injury: effects of visual mirror feedback on range of motion

The meta-analysis showed statistically significant differences in VMF interventions with a large clinical effect in two studies45, 46 (N.=163; SMD=2.33; 95% CI: 0.33-4.34; heterogeneity Q value 6.76; P=0.01; inconsistency I²=85%) (Figure 7A). The shape of the funnel plot appeared to be asymmetrical in the dominant model (Supplementary Digital Material 3).

Figure 7.

—Synthesis forest plot. This forest plot summarizes the results of included studies (sample size, standardized mean differences [SMDs], and weight). The small boxes with the squares represent the point estimate of the effect size and sample size. The lines on either side of the box represent a 95% confidence interval (CI). A) Forest plot for injury studies that used visual mirror feedback intervention on range of motion outcome; B) forest plot for injury studies that used motor imagery intervention on range of motion outcome.

Injury: effects of motor imagery on range of motion

The meta-analysis showed statistically significant differences in MI interventions with a large clinical effect in three studies42, 44, 65 (N.=54; SMD=1.21; 95% CI: 0.11-–2.3; heterogeneity Q value 6.47; P=0.04; inconsistency I²=69%), and there was no evidence of publication bias in the meta-analysis (SE=-3.98; t=-4.42; P=0.14) (Figure 7B). The shape of the funnel plot appeared to be symmetrical in the dominant model (Supplementary Digital Material 3). The sensitivity exclusion analysis suggested that two studies44, 65 significantly affected the pooled SMD (Supplementary Digital Material 3). Egger’s test results suggested no significant evidence of publication bias for the analysis (intercept=1.66; t=5.69; P=0.11).

Qualitative analysis

With respect to experimental immobilization, and according to the GRADE recommendations, there was low-quality evidence regarding the effects of MI on strength and ROM, downgraded due to imprecision and a risk of bias. In relation to the no immobilization studies and post-surgery patients, there was low-quality evidence regarding the effects of AO on balance and functional status, downgraded due to imprecision and a risk of bias. In addition, there was low-quality evidence in relation to MI interventions for strength and walking speed, downgraded due to imprecision and a risk of bias. On the other hand, there was moderate evidence regarding CE training and strength improvements, also downgraded due to imprecision.

Finally, regarding the no immobilization studies and injury patients, there was very low-quality evidence in VFM and MI interventions for ROM outcome, downgraded due to imprecision, a risk of bias and inconsistency.

Discussion

The present systematic review and meta-analysis showed some relevant results regarding the behavior of some motor variables after the implementation of various movement representation techniques and CE training in various clinical scenarios or experimentally generated nonclinical settings.

The results of this systematic review and meta-analysis can be divided into 2 main groups: studies of people undergoing a process of immobilization and studies in which a process of immobilization was not implemented.

With respect to the immobilized participants, a total of four meta-analyses were included: 3 with healthy individuals immobilized experimentally and one with patients submitted to a surgical process. The results in this first group showed that in the healthy experimentally immobilized individuals, MI showed significant results with respect to maintenance of strength and ROM, with low-quality evidence. These results did not occur with the application of CE training. In patients undergoing surgery, MI did not show significant changes in strength maintenance.

Regarding to the group of studies that dealt with participants not undergoing an immobilization process, nine meta-analyses were included, which can also be divided into two groups: studies of patients with injury who did not require surgery and of patients with an injury who required surgery. First, the quantitative analysis of two meta-analyses showed that the application of VMF or MI techniques, in combination with usual care, showed statistically significant changes in maintaining ROM in patients with injury without surgery, with very low-quality evidence. Second, the quantitative analysis of seven meta-analyses showed that the MI technique, in combination with usual care, showed significantly higher maintenance of strength and speed in patients undergoing surgery, with low-quality evidence. However, this outcome was not found with respect to the ROM variable. In addition, low-quality evidence showed that AO plus usual care obtained significantly better results with respect to maintenance of functional state and balance compared with usual treatment in isolation, and the use of the VMF technique did not maintain ROM better than not applying the technique in surgical patients. Finally, the application of CE training showed a maintenance of strength in patients undergoing surgery, with moderate evidence.

Thus, regarding techniques, AO showed good results with respect to improvements in general functional state and with respect to improvements in balance in patients with injury undergoing surgery. CE training appears to have worked better in patients compared with healthy individuals who were immobilized experimentally. VMF appears to have worked better in injuries that did not require surgery compared with those that did. However, MI showed some results that should be further analyzed. With respect to ROM maintenance, it appears that MI worked better when immobilization was experimental or in patients who had injury but did not require surgery. However, in patients with injuries undergoing surgery, MI did not show significant results regarding ROM maintenance. With respect to strength, MI showed similar results, i.e. better results in healthy individuals with experimental immobilization and in patients with injury not requiring surgery, and poorer in patients undergoing surgery. Finally, MI showed good results with respect to maintenance of speed.

The maintenance of physical condition, and therefore the specific state of some motor variables after an experimental immobilization process or after a clinical process with or without surgery (e.g., regaining strength, motor relearning through the recovery of active ROM, maintenance of balance or speed), could indirectly reveal the state of brain region function in relation to the planning, automation and execution of voluntary movement, as well as those areas involved in the generation of strength (e.g., primary motor cortex, premotor cortex, supplementary motor area, base ganglia, cerebellum).66, 67

For example, Ranganathan et al.66 had shown that force recovery originates through an adaptive neuroplastic process in the activity performance of cortical regions leading to the motor units generating both higher intensity and the recruitment of a set of motor units that would normally remain without activity.

In relation to this, Moukarzel et al.52 has recently found that MI could be relevant to promoting motor relearning as well as motor recovery in patients with knee impairments. These authors, as well as other research groups,68, 69 have argued that the combination of muscle atrophy along with a deficit of neuromuscular activation are contributing factors to the reduction in muscle strength. Through MI, an adaptive neuroplastic process of cortical reorganization is likely to take place, thus improving movement readiness and resulting in increased motor recruitment and synchronization of motor units at the peripheral level.52 This result is what could explain and lead to an improvement in motor variables such as strength or active ROM.

Although quantitatively lower, due to the fact that movement representation techniques (AO, VMF and MI) have the ability to qualitatively activate the same areas at a cortical level as those activated during voluntary movement,7, 70, 71 it is likely that this explanation could be applied to any of the three techniques.

Therefore, as postulated by Moukazel et al.,52 it appears that movement representation techniques could increase, enhance and improve the readiness of voluntary movement through a process of reorganization at the cortical level, indirectly causing greater voluntary muscle activation and greater active ROM. Grangeon et al.72 also argued that improved accuracy of neurosensory motor control in a neurological patient undergoing surgery for tendon transfer with real practice plus mental practice should be associated with structural neuroplastic changes at the cortical level. This association was also claimed by Jackson et al.73 In this regard, it has been shown that mental unilateral movement provokes bilateral brain activity in similar brain regions as does physical movement.74

In fact, a large number of theories or explanations that aim to explain the effect of movement representation techniques on peripheral muscle activity have been proposed. The study conducted by Christakou et al.65 shows some of these explanations in an exceptional way. For example, they describe Carpenter’s ideomotor hypothesis from the end of the 19th century75 and Jacobson’s psychoneuromuscular theory in the 1930s.76 The latter proposes that the construction of gestural motor images could evoke neuromuscular responses in the muscles involved. This evocation was later proven by Hale and his research group.77 There is also the neural training hypothesis proposed by Sale and Enoka, which suggests that changes at a central level are those that cause an increase at a peripheral level of muscular activity.65, 78 Along these lines, Jowdy and Harris74 found a significant increase in muscle activity during movement representation tasks evaluated through surface electromyography. Finally, it has been found that the construction of movement images could lead to a better representation of the process of motor force generation at the central level, i.e., in the central programming and planning system of the cerebral cortex.67, 79 All this could explain why training through movement representation techniques might have an impact at a central level and consequently at a peripheral level.

Regarding ROM, attentional control theory suggests that participants who perform a mental movement process might be able to focus their attention on the appropriate muscles more easily, which could improve the learning of motor skills. It has also been proposed that other neurophysiological aspects, such as the modulation of corticospinal excitability or the involvement of the autonomic nervous system, could be related to motor changes following mental movement representation.80

Our results are consistent with other review studies. Yap and Lim81 found in their meta-analysis that MI was effective for the improvement in ROM among patients with chronic musculoskeletal pain. The systematic review and meta-analysis recently conducted by Peng et al.82 showed that AO improved a set of motor variables, including motor function, walking ability and gait velocity in neurologic patients. However, controversial results have also been found. For example, the recent meta-analysis conducted by Manochi et al.83 did not find evidence that movement representation techniques are effective in increasing strength in healthy individuals. We found different results. In addition, Paravlic et al.64 found that MI caused an increase in maximum voluntary force significantly greater than no intervention in healthy adults.

In regard to CE training, Lee and Carroll conducted a thorough review of the neural, spinal and peripheral adaptations that occur during this training.9 They proposed two non-exclusive hypotheses to explain its effects: modification of contralateral motor pathways and the relationship between CE and motor learning. This second hypothesis shares similarities with respect to the arguments above regarding the possible functioning of movement representation techniques. In this second hypothesis, the authors argue that the generation of neural adaptations that occur during CE is likely to be in areas related to the control and execution of the trained member’s voluntary movement. However, these modified neural circuits could be accessed during the untrained limb’s voluntary contractions, optimizing the descending command signals from the untrained hemisphere. This hypothesis appears to have been supported by Strens et al.84 In addition, the meta-analysis study conducted by Manca et al.85 found that sustained CE training caused a reduction in inhibitory mechanisms at the cortical level, suggesting that inhibitory phenomena occurring within the primary motor cortex could modulate corticospinal inhibition and excitability after contralateral training.

Regarding the comparison of results with other review studies, we found some data that could generate controversy with the current state of the art. For example, the meta-analysis conducted by Manca et al.86 showed that CE training caused significant changes in strength. These changes did not occur in the present meta-analysis because the study by Pawar et al. appears to have reduced the result to a non-significant value (Figure 5D). One of the main differences between our study and that of Manca et al.86 is that they excluded studies that dealt with individuals with an immobilization.

Limitations of the study

This study has some limitations. Although a systematic search strategy was followed, the risk of selection bias might still be present. Another limitation is the number of studies included in the meta-analysis, given this low number could represent inadequate statistical power and bias due to the sample size included in each comparison. In this regard, the low number of studies included could represent a bias in the interpretation of asymmetry in each forest plot; therefore, this situation should be interpreted with caution. Most of the studies did not include a placebo intervention in addition to usual treatment, which makes it difficult to determine whether effects were driven by movement representation techniques and not due to nonspecific effects. A clear limitation is that there is a high degree of heterogeneity. This should make us take the results obtained with caution. In addition, the article conducted bay Harput et al.,48 the CG is duplicated by evaluating two groups against it. This may lead to a double counting or multiplicity problem and should be considered as a limitation.

Conclusions

Movement representation techniques and CE training are a set of very low-cost techniques shown to have a significant impact on the improvement of various motor variables in particular, and on physical maintenance in general, during experimental immobilization processes in healthy individuals, in patients with injuries that did not require surgery and in surgical processes that did or did not require immobilization.

AO and CE training appear to benefit injured patients undergoing surgery, whereas MI and VMF appear to work better in healthy individuals undergoing experimental immobilization and in injuries not requiring surgery. However, the results of these techniques in maintaining physical condition were not significant in injuries requiring surgery. This study shows that movement representation techniques and CE training are valuable tools for physical maintenance, but further research is still needed due to several discrepancies. Future studies should contain a placebo group with the same intervention time as the experimental group(s); more studies combining mental and physical practice are needed; evaluation with a follow-up period and larger sample sizes based on adequate sample size calculation and minimizing type I and II errors are also needed.

Supplementary Digital Material 1

Supplementary Text File 1

Search strategies

Supplementary Digital Material 2

Supplementary Table I

Characteristics of the included studies on immobilization

Supplementary Digital Material 3

Supplementary Text File 2

Experimental immobilization