Impact of Constraint-Induced Movement Therapy (CIMT) on Functional Ambulation in Stroke Patients—A Systematic Review and Meta-Analysis

Constraint-induced movement therapy (CIMT) has been delivered in the stroke population to improve lower-extremity functions. However, its efficacy on prime components of functional ambulation, such as gait speed, balance, and cardiovascular outcomes, is ambiguous. The present review aims to delineate the effect of various lower-extremity CIMT (LECIMT) protocols on gait speed, balance, and cardiovascular outcomes. Material and methods: The databases used to collect relevant articles were EBSCO, PubMed, PEDro, Science Direct, Scopus, MEDLINE, CINAHL, and Web of Science. For this analysis, clinical trials involving stroke populations in different stages of recovery, >18 years old, and treated with LECIMT were considered. Only ten studies were included in this review, as they fulfilled the inclusion criteria. The effect of CIMT on gait speed and balance outcomes was accomplished using a random or fixed-effect model. CIMT, when compared to controlled interventions, showed superior or similar effects. The effect of LECIMT on gait speed and balance were non-significant, with mean differences (SMDs) of 0.13 and 4.94 and at 95% confidence intervals (Cis) of (−0.18–0.44) and (−2.48–12.37), respectively. In this meta-analysis, we observed that despite the fact that several trials claimed the efficacy of LECIMT in improving lower-extremity functions, gait speed and balance did not demonstrate a significant effect size favoring LECIMT. Therefore, CIMT treatment protocols should consider the patient’s functional requirements, cardinal principles of CIMT, and cardiorespiratory parameters.


Introduction
Stroke is the second leading cause of death, and its rising incidence, mortality, and disability impose a significant global burden [1]. Stroke is pronounced in low-and middleincome countries, diverse age groups, and gender [2]. Six and twelve months following a stroke, the mortality and disability rates increase dramatically (55.9% and 61.0%, respectively) [3]. The number of new cases, the prevalence of functional disability, and the mortality rate associated with stroke indicate the need for improved rehabilitation techniques that reduce the time to recovery [3].
LECIMT protocols and their effects on gait speed, balance, and cardiovascular parameters, which were prime elements contributing to functional ambulation.

Selection Criteria of Studies for This Review
RCTs that compared lower-limb training utilizing the principles of CIMT with conventional therapy or other neurorehabilitation techniques were considered for this review. Randomized controlled trials (RCTs), which included first-time or recurrent, hemorrhagic or ischemic poststroke subjects aged above 18 years, were considered. Studies that measured and analyzed the gait speed, balance, and cardiovascular parameters were primarily included in the review (Protocol registered under Prospero database with registration no: CRD42021260203).

Literature Search and Study Selection
Two independent reviewers (DRS and RSR) searched for pertinent English-language articles published between 2000 and 2022. In the initial phase, one reviewer (KG) reviewed recognized titles and references for their relevance. Later, studies were screened based on the selection criteria for their relevance. Finally, the expert reviewer (PKK) resolved the differences in the studies.
We searched the databases PEDro, Web of Science, PubMed, Scopus, EBSCO, MED-LINE, CINHAL, and Science Direct for relevant publications. The MeSH keywords included: stroke (chronic, subacute, acute), CIMT, forced use, restricted limb or extremity cerebrovascular accidents, hemiparesis, hemiplegia, gait speed, cardiovascular, balance, blood pressure, and percentage of heart rate maximum. Table 1 provides a summary of the search strategy's specifics.

Data Extraction
Two independent reviewers (SD and APG) were involved in the process of collecting relevant data from the included studies. The third reviewer (PKK) was invited to sort out discrepancies between the two reviewers in authenticating the data. The information extracted from the studies was based on 1. patients' features, 2. PEDro score, 3. sample size, 4. eligibility standards, 5. outcome dimensions, and 6. constraint applications.

Evaluation of Methodological Quality and Level of Evidence
PEDro scale, which consists of 10 domains, was used to assess the methodological quality of included studies. RCTs with a score of 9-10 on the PEDro scale were assigned as "excellent", 6-8 as "good", 4-5 as "fair", and less than four as "poor" methodological quality. Two independent assessors (JST and SD) used the PEDro scoring to evaluate the methodological quality of the studies. The third assessor was contacted to resolve the disparity in assigning the score.
Further to the criteria mentioned above, high-quality RCTs with PEDro ≥ 6 are considered as level 1 evidence and have been divided into two subcategories, level 1a (>1 study with PEDro ≥ 6) and level 1b (1 study with PEDro ≥ 6). Low-quality RCTs with PEDro < 6 designated as Level 2 evidence. This categorization was imbibed from Sackett et al.'s criteria outlined by Dixit and Gular [45]. Outcome measures in the current study were categorized as levels 1a, 1b, and 2 based on the number of quality trials supporting it.

Risk of Bias Assessment
Review manager 5.4.1 software (London, UK) was utilized to evaluate and synthesize the risk of bias among included studies. Two reviewers (KG and SD) independently rated individual studies on the following domains: 1. allocation concealment, 2. random sequence generation, 3. blinding of participants, 4. blinding of outcome data, 5. incomplete outcome data, 6. selective reporting, 7. discrepancy of intervention between the group, and 8. other biases. Usually, descriptors such as high risk, low risk, and unclear were used to report the domains. An expert reviewer (RSR) was invited to resolve the variance of opinion between the reviewers in designating the descriptors.

Data Synthesis
Meta-analysis was performed to determine the effect size of CIMT on gait speed, balance, and cardiovascular outcome measures, which were observed in the included studies. An outcome measure assessed in 2 or more trials was considered for meta-analysis.
When descriptive values were reported as the median and range in the included studies, the mean and standard deviations were calculated using the conversion formulae. If data were not reported, the relevant authors were contacted through email and requested to provide the information. Review manager 5.4.1 software was used to conduct meta-analysis.
The heterogeneity analysis (I2 statistics) was performed. Statistics were generated for the treatment effects on outcome measures if we found no clinical heterogeneity regarding the included subjects' characteristics. The present meta-analysis considered I2 statistics with more than 50% value as considerable heterogeneity [46]. A random-effects model was used for analysis to obtain results. A pooled standardized mean difference (SMD) was calculated for outcome measures.

Search Results
Ten studies were included in the review process. Details regarding the process of database searches and exclusion and inclusion of studies are provided in Figure 1.

Characteristics of Included Studies
Among the ten studies, 329 subjects participated in the trials (212 males and 117 females). The stroke controlled-trial population consisted of middle-aged and elderly adults aged 40 to 70 years. All participants involved in the trials suffered either ischemic or hemorrhagic strokes. Considering the mean and standard deviation for the stroke duration throughout the analyzed studies, one study included subacute strokes [47], six studies included subacute/chronic [48][49][50][51][52][53], and three studies included chronic stroke population [54][55][56] ranging from 1.5 to 6.5 years.
All included RCTs compared LECIMT with conventional physiotherapy or neurorehabilitation techniques. Six studies compared LECIMT with conventional physiotherapy [48,50,51,[53][54][55], one study compared LECIMT with neuro-developmental therapy (NDT) [42], and three studies compared conventional physiotherapy as an adjunct to LEC-IMT with conventional physiotherapy [50,52,56,57]. The type of constraint utilized in the controlled trials was either restraint of a non-paretic lower extremity or augmentation of a paretic lower extremity. Restraining of a non-paretic lower extremity was provided by applying an ankle mass in five studies [50,51,53,54], immobilization of the knee in addition

Characteristics of Included Studies
Among the ten studies, 329 subjects participated in the trials (212 males and 117 females). The stroke controlled-trial population consisted of middle-aged and elderly adults aged 40 to 70 years. All participants involved in the trials suffered either ischemic or hemorrhagic strokes. Considering the mean and standard deviation for the stroke duration throughout the analyzed studies, one study included subacute strokes [47], six studies included subacute/chronic [48][49][50][51][52][53], and three studies included chronic stroke population [54][55][56] ranging from 1.5 to 6.5 years.
All included RCTs compared LECIMT with conventional physiotherapy or neurorehabilitation techniques. Six studies compared LECIMT with conventional physiotherapy [48,50,51,[53][54][55], one study compared LECIMT with neuro-developmental therapy (NDT) [42], and three studies compared conventional physiotherapy as an adjunct to LEC-IMT with conventional physiotherapy [50,52,56,57]. The type of constraint utilized in the controlled trials was either restraint of a non-paretic lower extremity or augmentation of a paretic lower extremity. Restraining of a non-paretic lower extremity was provided by applying an ankle mass in five studies [50,51,53,54], immobilization of the knee in addition to a shoe insert in one study [56], and a negative kinematic restraint applied to a non-paretic lower extremity induced by robotic-assisted gait training (RAGT) in one study [55]. The mode of augmentation delivered to a paretic lower extremity among three clinical trials was diverse, with a compelling bodyweight shift to the paretic side (CBWS) [56], auditory feedback using a cane on the unaffected side [52], and enhancing the practice of the affected lower limb in training sessions and daily activities [58]. The total duration of constraint application was 20 min in two studies, 4.5 h in four studies, 90% of waking hours in one study, and applying while performing daily living activities in another study.
The duration of intervention in the clinical trials ranged from 20 min to 55 h. Two clinical trials observed the immediate effects of single-session LECIMT intervention (20 min) [54,55]. However, in seven studies, the participants were intervened for less than 20 h, and in three studies, the intervention was delivered for more than 20 h. A diverse duration of intervention was observed among the studies with the single session to 32 sessions, and each session ranged from 20 min to 2 h 45 min per session. Table 2 summarizes the details of the included subjects, intervention, and outcomes. Gluteus medius and vastus medialis oblique muscle activity, single support phase of the affected side (% GC) vertical peak force of the cane (% BW) and gait speed (m/s).
The experimental group showed significant improvement in muscle activation and gait speed. Gait training on a treadmill.
The nine training sessions, 30 min, two consecutive weeks.
Step length, hip, knee and ankle ROM, and gait speed(m/s).
Spatio-temporal and kinematic gait parameters improved in both groups equally.

Outcome Measures
Gait speed was pronouncedly observed in eight studies; compared to the control group, gait speed in the experimental group improved significantly in four trials [52,56,58,59] and showed similar improvements in four trials [48,50,54,55]. Berg balance scale in the experimental group showed marked improvements in two studies [56,59] and similar effects as a control group in one study [53]. Cardiovascular parameters, blood pressure, and percentage of heart rate maximum were reported in one study with the CIMT group showing no changes between pre-and post-sessions [50].

Methodological Quality and Level of Evidence
Among the comprised studies, three studies scored four [55,56,59], two studies scored five [47,50], one study scored seven [52], and three studies scored eight [50,51,53] on the Pedro scoring system. Therefore, based on the assessment, five studies were rated as fair, and five were rated as good. Details of the Pedro scores attained by the comprised studies are provided in Table 3. The gait speed and balance are supported by level 1b evidence, whereas cardiovascular outcomes with level two evidence. The level of evidence LECIMT on outcomes is summarized in Table 4.
Notes: "Y": yes; "N": no. The scores of the risk-of-bias assessment revealed that 50% of studies showed a high risk for concealed allocation and detection bias. In total, 75% of studies suffered from attrition bias and 100% of studies from participant bias. The included studies scored low risk for reporting bias and treatment imbalance. Details of the risk-of-bias assessment are given in Figure 2.

HR max Balance
• BBS and Postural symmetry Level 1b Good Notes: a. BBS: Berg balance scale; b. SPB: systolic blood pressure; c. DPB: diastolic blood pressure; d. % of HR max: percentage of heart rate maximum.
The scores of the risk-of-bias assessment revealed that 50% of studies showed a high risk for concealed allocation and detection bias. In total, 75% of studies suffered from attrition bias and 100% of studies from participant bias. The included studies scored low risk for reporting bias and treatment imbalance. Details of the risk-of-bias assessment are given in Figure 2.

Quantitative Synthesis
Due to their single session, two of the ten included clinical trials were excluded from the analysis. A meta-analysis was performed for gait speed and balance because these factors were observed in six and three RCT studies, respectively.
The CIMT group could not display a significant difference in gait speed compared to the control groups; the standard mean difference (SMD) was 0.13 at a 95% CI = −0.18-0.44, p = 0.42, with heterogeneity among the studies of I 2 = 4% at p = 0.39. Furthermore, on postfollow-up, the CIMT group was unable to display a significant difference in gait speed when compared to the control groups (SMD = 0.32 at 95% CI = −0.21-0.85, p = 0.24) with the heterogeneity among the studies being I 2 = 0% at p = 0.99 ( Figure 3).

Quantitative Synthesis
Due to their single session, two of the ten included clinical trials were excluded from the analysis. A meta-analysis was performed for gait speed and balance because these factors were observed in six and three RCT studies, respectively.
The CIMT group could not display a significant difference in gait speed compared to the control groups; the standard mean difference (SMD) was 0.13 at a 95% CI = −0.18-0.44, p = 0.42, with heterogeneity among the studies of I 2 = 4% at p = 0.39. Furthermore, on post-follow-up, the CIMT group was unable to display a significant difference in gait speed when compared to the control groups (SMD = 0.32 at 95% CI = −0.21-0.85, p = 0.24) with the heterogeneity among the studies being I 2 = 0% at p = 0.99 (Figure 3). The meta-analysis results for balance in the post-treatment CIMT group were unable to display a significant difference in balance when compared to the control groups; the SMD was 4.94 at 95% CI = −2.48-12.67, p = 0.19, with substantial heterogeneity among the studies (I 2 = 92% at p < 0.001). In addition, in studies reporting follow-up of treatment sessions, the CIMT group was unable to display a significant difference in balance when compared to the control groups (SMD = 3.84 at 95% CI = −2.33-10.01, p = 0.22); the hetero- The meta-analysis results for balance in the post-treatment CIMT group were unable to display a significant difference in balance when compared to the control groups; the SMD was 4.94 at 95% CI = −2.48-12.67, p = 0.19, with substantial heterogeneity among the studies (I 2 = 92% at p < 0.001). In addition, in studies reporting follow-up of treatment sessions, the CIMT group was unable to display a significant difference in balance when compared to the control groups (SMD = 3.84 at 95% CI = −2.33-10.01, p = 0.22); the heterogeneity among the studies was I 2 = 88% at p = 0.004 (Figure 4). The meta-analysis results for balance in the post-treatment CIMT group were unable to display a significant difference in balance when compared to the control groups; the SMD was 4.94 at 95% CI = −2.48-12.67, p = 0.19, with substantial heterogeneity among the studies (I 2 = 92% at p < 0.001). In addition, in studies reporting follow-up of treatment sessions, the CIMT group was unable to display a significant difference in balance when compared to the control groups (SMD = 3.84 at 95% CI = −2.33-10.01, p = 0.22); the heterogeneity among the studies was I 2 = 88% at p = 0.004 (Figure 4).

Discussion
This review discusses the qualitative and quantitative effects of LECIMT on the numerous parameters that determine functional ambulation. The current review is unique compared to previous reviews because it primarily observed gait and balance cardiopulmonary parameters exclusively among RCTs. The gait outcome measures were extensively reported in seven studies [50][51][52][54][55][56]58,59], followed by balance in four studies [53,56,58,59], cardiovascular fitness [50], functional mobility [53], functional ambulation

Discussion
This review discusses the qualitative and quantitative effects of LECIMT on the numerous parameters that determine functional ambulation. The current review is unique compared to previous reviews because it primarily observed gait and balance cardiopulmonary parameters exclusively among RCTs. The gait outcome measures were extensively reported in seven studies [50][51][52][54][55][56]58,59], followed by balance in four studies [53,56,58,59], cardiovascular fitness [50], functional mobility [53], functional ambulation [59], and motor functions [52]. In two trials, LECIMT showed superior effects for outcome measures when compared to its respective control group [58,59]. Eight studies exhibited an improvement in post and follow-up sessions similar to the control group for all outcome measures [48,51,52,[54][55][56], except for cardiovascular parameters, where there were no clear changes from pre-to post-sessions [50]. Similar findings in earlier reviews support the positive effects of LECIMT on gait speed, balance, functional mobility, functional ambulation, and motor functions [60]. However, compared to the control group, LECIMT protocols exhibited ineffectiveness in improving the speed of walking and functional balance in the present meta-analysis.
We found equivalent or better results in the experimental groups, including all or some of the core principles (intensive practice with repetitions, shaping, and transfer package) of CIMT. The beneficial effects of restraining a non-paretic upper limb in ULCIMT are due to the opportunity of engaging a paretic limb independently in functional task practice. In contrast, lower-extremity functions are typically bipedal, resulting in the paretic limb being used forcibly during LECIMT. This issue may explain why there were similar influences on the outcomes of the included studies regardless of the application of constraints.
Even though we expect the neuroplastic changes within six months following a stroke to enhance functional recovery [61], post-six months, functional improvements are not impossible [62]. Moreover, after six months of post-stroke intense practice simulating practical tasks, functional improvements are inevitable in stroke patients [42,63]. However, we observed that many studies did not include the core principles of CIMT, including shaping, hence emphasizing the progressive modifications among the treatment parameters, such as number, type, and complexity of the tasks [64][65][66]. Typically, therapists withhold treatment training due to a plateau in functional improvements in chronic stroke patients; however, with CIMT treatment, strategies including repetitive high-intensity training could produce neuroplasticity and functional improvements even after reaching a plateau in their improvements [67][68][69][70]. Many studies designed to maintain the same level of task training without progressing to the next level of challenge would have failed to achieve any better results than the conventional practice.
Lower-extremity motor recovery is more complex and delayed in stroke populations. Community ambulation is an integral component in improving quality of life (QOL), for which therapists require gait training, including multi-tasking, obstacle training, and training on different surfaces rather than simple task-specific training [42,63]. Some studies showed further improvements through LECIMT by emphasizing that stroke patients should perform multiple functional tasks at home and by promoting social support, selfmonitoring, and motivation for better adherence as a part of a "transfer package" [56,59].
The predictable walking ability, functional capacity, and QOL of stroke patients are strongly associated with cardiovascular efficiency [6,71,72]. Variations in blood pressure are an essential indicator of the possibility of hypertension in stroke, cardiovascular disease, high risk of recurrent stroke, renal failure, and all-cause mortality [73]. Hence, blood pressure variability should be used in clinical trials to evaluate the efficacy of treatment against cardiovascular-related mortality in stroke [73].
Promoting low-to-moderate physical activity is essential in stroke risk management. Moreover, aerobic exercise regimes increase the production of new blood vessels, nerve cells, and synaptic connections [74,75]. In the initial phases of the stroke, muscle strength and balance play a key role in determining functional ambulation. In later stages, cardiovascular capacity plays a significant role along these factors [23]. Despite being an essential component in stroke recovery, most studies included in the analysis surprisingly missed it as an outcome measure.
In this review, most studies preferred treadmills for gait training, improving cardiovascular capability. However, most studies missed setting the target level of intensity for cardiovascular conditioning and its progression, which is a crucial factor in improving cardiovascular fitness [76][77][78][79].
The gait speed and balance meta-analysis did not show any significant effect size. The type of gait training prescribed in experimental and control groups, heterogeneity in the restraint used, unequal treatment durations, small sample size, chronicity of the stroke, a smaller number of studies, and low methodological quality were to be credited for insignificant differences between the groups.
CIMT demands highly specific training for its efficient usage in clinical practice. However, therapists require expertise in identifying individuals' functional needs, determining and framing the functionally related tasks for practice, setting targets, progressing the intensity, monitoring improvements, and applying CIMT principles. We suggest that the in-expertise in the application of CIMT might contribute to similar improvements in both groups [42].
In some of the studies, many authors did not demonstrate improvements due to inefficient usage of the principles of CIMT, such as transfer package and shaping, lack of feedback, absence of functional task training, and not progressively increasing task complexity. Along with this involvement of various stroke populations with multiple phases and severities of stroke, variability in treatment durations, the intensity of training, and the type of tasks might be the reasons for equal effects in their counterparts.
The included RCTs could have been improved by providing more appropriate sample sizes and homogeneity in their intervention plans. In addition, most RCTs did not include various important outcome measures, such as QOL, functional mobility, and cardiovascular parameters. Due to the scarcity of RCTs, we included some RCTs with low methodological quality and with a single session of training. Moreover, further meta-analysis for other gait parameters, mobility and balance variables of lower-limb-related functions, and QOL could be warranted if sufficient RCTs are available.

Conclusions
In this meta-analysis, we observed that despite the fact that several trials claimed the efficacy of LECIMT in improving lower-extremity functions, gait speed and balance did not demonstrate a significant effect size favoring LECIMT. Moreover, forthcoming CIMT treatment protocols should consider the functional requirements of stroke subjects and the appropriate application of all its principles.