Effectiveness of Rehabilitation Interventions on Participation in Children and Adolescents With Unilateral Cerebral Palsy: A Systematic Review and Meta-Analysis

Abstract

Objective

To determine the effectiveness of rehabilitation interventions on participation outcomes in daily living, leisure, social, and educational contexts for children and adolescents with unilateral cerebral palsy (UCP), and to identify whether specific intervention categories yield superior participation improvements.

Data Sources

MEDLINE via PubMed, Physiotherapy Evidence Database, Cochrane Library, CINAHL, and Scopus were searched from inception to July 31, 2025, for randomized controlled trials (RCTs), non-RCTs, and single-arm pre–post studies involving participants aged 18 months to 18 years with UCP, rehabilitation interventions, and standardized participation measures.

Study Selection

After screening 417 records, 23 studies (19 RCTs and 4 pre–post designs; total N=840) met inclusion criteria.

Data Extraction

Two independent reviewers extracted study characteristics and participation outcomes. Risk of bias was assessed via Risk of Bias 2.0 and Risk Of Bias In Nonrandomized Studies of Interventions-I; certainty of evidence was judged using Grading of Recommendations Assessment, Development and Evaluation.

Data Synthesis

Pooled pre–post analyses using a random-effects model (Restricted Maximum Likelihood) showed a standardized mean difference of 1.44 (95% CI: 0.95-1.92), indicating strong within-group participation improvements; however, certainty of evidence was low because of methodological limitations and high heterogeneity (I²=95%). Secondary analyses showed no significant superiority for constraint-induced movement therapy (standardized mean difference=−0.09; I²=0%), technology-assisted interventions (standardized mean difference=0.15; I²=56%), or bimanual therapy (standardized mean difference=0.69; I²=88%).

Conclusions

Rehabilitation interventions yield substantial pre–post participation gains in children and adolescents with UCP; no specific modality showed clear superiority. Intervention characteristics may be key drivers. Future RCTs with active comparators and standardized, participation-focused protocols are needed.

Cerebral palsy (CP) is defined as a group of permanent developmental disorders of movement and posture that limit activity and are attributed to nonprogressive alterations in fetal or infant brain development.¹ CP is one of the most common causes of motor disabilities in the pediatric population and the leading cause of severe childhood disability.² In developed countries, the estimated global average prevalence of CP is 2-3 per 1000 live births,³ of which 20%-40% are classified as having unilateral cerebral palsy (UCP).⁴ This condition is characterized primarily by lateralized motor impairments on one side of the body affecting both the lower and upper limbs, although the upper limb is usually more prevalently affected.⁵ In addition, children and adolescents with UCP may have associated problems, including sensory impairment, intellectual disability, epilepsy, and vision, hearing, and speech disorders.⁶ Participation—long central to occupational therapy as occupational performance⁷—gained broader recognition when formalized it by the World Health Organization in the International Classification of Functioning, Disability and Health (ICF) Children & Youth Version⁸ as a core health and well-being outcome, particularly regarding opportunities in home, school, and community leisure activities.⁹ The understanding that child development and functioning emerge from dynamic interactions between the individual and their environments is rooted in ecological theory, which fundamentally reframes functioning as emerging not from individual characteristics alone, but rather through person-environment fit.¹⁰ According to the ICF,⁸ children and adolescents with UCP have impairments in structure and function, leading to activity limitations and participation restrictions, which in turn may decrease their quality of life. Although the ICF highlights the importance of addressing participation restrictions, therapeutic interventions have traditionally focused on other domains rather than participation outcomes.^11,12 The ICF also provides a framework for guiding assessments and therapeutic interventions that consider the expectations of the child and family, and suggests that the connections between its domains (body structure and function, activity and participation) are multidirectional.⁸ In this context, a clinical trial¹³ aimed to explore whether community interventions with participation-level goals also had effects on bodily function in young people (15-25y) with physical disabilities; although community participation goals showed no global effect on bodily function, all patients achieved significant improvements in ≥1 domain (motor function, cognition, affect, or performance). A recent systematic review examined factors affecting participation in children with CP and identified gross motor and manual function, CP type, and the home and community physical environment as the most influential determinants across activities of daily living, play, leisure, social participation, education, and rest/sleep.¹⁴

UCP presents a distinctly asymmetric motor profile compared with other CP types, with differential impacts on lateralized versus bilateral functions, motor learning capabilities, and participation patterns.¹⁵ Perinatal arterial ischemic stroke, a major etiology of CP, demonstrates that stroke laterality significantly influences neurologic outcomes, with bilateral strokes associated with substantially higher risks of motor impairment, epilepsy, and cognitive involvement compared with unilateral presentations.¹⁶ Given these distinct clinical presentations and outcomes, and the fact that participation outcomes in this specific population remain understudied, intervention research in UCP warrants focused and separate systematic investigation.

Previous reviews17, 18 have shown that exercise or education interventions could improve participation in children with different motor, developmental or intellectual disabilities, including children with different types of CP. Recently, a systematic review and meta-analysis of therapeutic interventions on participation outcomes in children with CP has been conducted.¹⁹ Their analysis revealed that interventions focusing specifically on participation significantly improved participation outcomes, whereas interventions targeting body functions, structures, or activities alone did not favorably affect participation. In addition, a randomized controlled trial (RCT)²⁰ in younger children with UCP (1-4y old) reported that intensive interventions such as hand and arm bimanual intensive therapy, including lower extremities (HABIT-ILE), significantly affected participation. Despite this growing evidence, many systematic reviews have investigated the effectiveness of different rehabilitation interventions on functional and activity outcomes for children and adolescents with UCP21, 22, 23, 24 but none have specifically focused on participation outcomes in unilateral CP. Moreover, Novak et al²⁵ conducted a comprehensive review using the Evidence Alert Traffic Light System to summarize best available evidence on interventions for children with CP, highlighting effective allied health strategies—including bimanual training, Constraint-Induced Movement Therapy (CIMT), goal-directed training, home programs, and occupational therapy approaches—which support the relevance of both physical and occupational therapy in optimizing participation outcomes. Given that children with UCP have distinct motor patterns and participation challenges compared with those with bilateral involvement, a specific examination of participation outcomes in this population is warranted.

Therefore, this systematic review and meta-analysis aimed (1) to determine the overall effects of rehabilitation interventions on participation outcomes in children and adolescents with UCP, as defined by the ICF framework and (2) to explore whether certain categories of rehabilitation interventions are associated with greater improvements in participation outcomes.

Methods

This systematic review and meta-analysis was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines for Systematic Reviews and Meta-Analyses²⁶ (supplemental appendix S1, available online only at http://www.archives-pmr.org/). The study protocol was registered prospectively on the PROSPERO register (registration number CRD42024571090).

Data sources and search strategy

To identify studies reporting the effects of rehabilitation interventions on participation among children and adolescents with UCP, 2 reviewers (G.S. and P.L.-M.) independently searched the electronic databases of MEDLINE via PubMed, the Physiotherapy Evidence Database, CINAHL (through EBSCO), Scopus, and the Cochrane Database of Systematic Reviews from their inception to July 31, 2025. Disagreements were resolved by consensus or discussion with a third reviewer (R.P.-C.). The literature search was conducted with the patients, intervention, comparator, and outcome approach and combined relevant terms related to (1) the clinical condition of the UCP, (2) the intervention of rehabilitation interventions, and (3) the different measurement of participation. The full search strategy is reported in supplemental appendix S1. To complete the systematic literature search, the references of eligible studies were reviewed to identify other relevant studies.

Eligibility criteria

The inclusion criteria included: design—RCTs, non-RCTs, pilot RCTs, and single-arm pre–post studies); population—children and adolescents diagnosed with UCP aged 18 months to 18 years; intervention—rehabilitation interventions, defined as structured interventions delivered by physical or occupational therapists aimed at improving motor function, mobility, functional abilities, or participation; comparison—control group, other nonrehabilitation interventions, or preintervention data; and outcome—participation, defined according to the ICF framework as involvement in life situations. Participation measures included standardized, validated instruments assessing engagement in daily activities, social roles, and community involvement.

Although some instruments, such as the Canadian Occupational Performance Measure (COPM), primarily assess perceived performance and satisfaction in individual daily tasks, they were included as participation outcomes when the identified goals reflected meaningful engagement in real-life settings. This decision is consistent with prior reviews and aligns with the ICF’s emphasis on person-centered, context-based functioning.²⁷ Thus, when individual goals established through the COPM explicitly pertained to participation domains (eg, social integration, recreation, engagement at school or in community contexts), these outcomes were considered valid participation measures, as previously applied in pediatric CP rehabilitation research.²⁷ For studies using COPM or Goal Attainment Scaling (GAS), outcomes were included only if they could be reasonably classified as participation-focused, based on the description of goals and activities in the study or supplementary materials (eg, parent-selected goals targeting daily activities, social participation, or community involvement).

In contrast, we excluded studies involving multidisciplinary interventions, defined as those combining physical or occupational therapies with other therapeutic approaches (eg, pharmacologic, psychologic, or educational programs such as conductive education or structured home exercise programs), to isolate the effects of rehabilitation interventions on participation outcomes and studies that did not report results in which participation outcomes were metanalized. There were no restrictions on language or year of publication.

When multiple reports referring to the same study/trial were identified, we prioritized the most detailed results or those from the study with the largest sample size. In instances where studies presented similar sample data, those employing a greater variety of specific participation measures were selected for inclusion in this review. This process ensured that only unique studies were included, avoiding duplication of participant data, in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 guidelines distinguishing between studies/trials and reports.

Data extraction

To examine the effects of rehabilitation interventions on participation in children and adolescents with UCP, 2 reviewers (C.L.-R. and A.T.-C.) independently extracted the following information from each included study: (1) characteristics of the studies: first author (y), country and type of study; (2) characteristics of the population: sample size, age (y), sex (female), affected side (right) and (3) functional classification descriptors: Gross Motor Function Classification System (GMFCS) level and Manual Ability Classification System (MACS) level, which provide standardized descriptors of gross motor and upper limb manual ability, respectively; (4) intervention characteristics: duration (wks), frequency (number of sessions/wk), time (h/session), and type of intervention; and (5) outcomes: participation scale. In addition, to specify the nature of the interventions, more details are provided: (1) details on the intervention provider, (2) description of the rehabilitation intervention, and (3) any additional interventions included in each study. Classification data (GMFCS/MACS) were not consistently reported across most of the studies, limiting our ability to stratify results by functional severity. Disagreements during data collection were settled by consensus.

Assessing risk of bias and certainty of evidence

The risk of bias of the included RCTs was assessed via the Cochrane Collaboration tool (RoB 2.0).²⁸ It is a tool that assesses the risk of bias in 5 domains: the randomization process, deviations from planned interventions, missing outcome data, outcome measurement, and selection of reported outcomes. Each domain was rated as low, some concerns, or high risk of bias, and an overall judgment was derived following the RoB 2.0 algorithm.

The methodological quality of the nonrandomized studies was assessed via the Risk Of Bias In Nonrandomized Studies of Interventions (ROBINS-I) risk of bias test.²⁹ ROBINS-I evaluates 7 domains: confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes, and selection of reported results. Domains were rated as low, moderate, serious, critical, or no information. We defined “no information” for any domain when key methodological details (eg, allocation process or outcome assessor blinding) were not reported and no supplementary data were obtainable from study authors.

The Grading of Recommendations Assessment, Development and Evaluation (GRADE)³⁰ was used to judge the certainty of the evidence. We judged the evidence for each outcome as high, moderate, low, or very low on the basis of trial design, risk of bias, inconsistency, indirect evidence, imprecision, and publication bias.

Assessments of the risk of bias and certainty of evidence were conducted independently by 2 reviewers (G.S. and P.L.-M.), and inconsistencies were resolved by consensus or discussion with a third investigator (R.P.-C.).

Statistical analysis

The Restricted Maximum Likelihood method (random-effects model)³¹ was used to estimate pooled effect sizes and their respective 95% confidence intervals (95% CIs) for participation outcomes. For RCTs comparing an intervention with a control group, standardized mean differences via Cohen’s d index³² were calculated using between-group comparisons of change scores, derived from pre–post intervention differences within each group (intervention and control). For single-arm pre–post studies, effect sizes were calculated using Cohen’s d based on within-group pre–post differences, with the mean and SD at baseline and postintervention entered into the meta-analysis software, which computed the standardized effect size accordingly. Positive values indicate improvements in participation after the intervention. Cohen’s d values of approximately 0.2 were considered indicative of weak effects, values of approximately 0.5 were considered indicative of moderate effects, values of approximately 0.8 were considered indicative of strong effects, and values above 1.0 were considered indicative of very strong effects.

Study heterogeneity was assessed via the I² statistic³³ and classified according to its values as unimportant heterogeneity (0%-40%), moderate heterogeneity (30%-60%), substantial heterogeneity (50%-75%), and considerable heterogeneity (75%-100%). Corresponding P values were considered, particularly when heterogeneity was found in the overlapping areas of these intervals.

The primary meta-analysis focused on the overall pre–post effect of rehabilitation interventions on participation. To explore whether study-level characteristics explained heterogeneity in participation outcomes, we conducted meta-regression analyses with mean participant age as a continuous moderator variable. In addition, 3 secondary meta-analyses were conducted as subspecific analyses to explore effects of particular intervention types. The categories—CIMT/Modified Constraint-Induced Movement Therapy (mCIMT), bimanual therapy (BIT/HABIT/HABIT-ILE), and technology-assisted interventions—were selected based on their frequency in the included studies and their prominence in pediatric UCP rehabilitation literature. When studies reported combined interventions (eg, CIMT+BIT delivered simultaneously to the same group), these studies were included in the primary analysis if participation outcomes were measured. In each secondary meta-analysis, “alternative therapy” refers to any rehabilitation intervention (physical therapy or occupational therapy) other than the primary intervention type being analyzed. This classification allowed comparison of each specific intervention approach against a pooled category of other active rehabilitation or control interventions.

To assess the robustness of the summary estimates and to detect whether any individual study accounted for a large proportion of heterogeneity, sensitivity analyses were performed. These included leave-one-out analyses, in which individual studies were sequentially excluded and new standardized mean differences with 95% CIs were calculated. Additional predefined sensitivity analyses were performed by excluding nonrandomized single-arm pre–post studies from the overall analysis, and by excluding studies applying HABIT-ILE in the bimanual therapy subgroup to isolate effects specific to upper limb bimanual interventions. Changes in pooled effect sizes, statistical significance, and heterogeneity (I²) were examined to determine the influence of study design and intervention type on the results. Finally, publication bias was estimated by visual inspection of funnel plots and the P value of Egger’s regression asymmetry test,³⁴ with P values <.10 considered statistically significant.

All the statistical analyses were performed with IBM SPSS Statistics (version 28.0.0.0).^a

Results

Study selection

A total of 417 studies were obtained. After 38 full text articles were read, 2335, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57 studies were included in this systematic review and meta-analysis (fig 1). The list of excluded studies after full text reading along with the reasons for exclusion are shown in supplemental appendix S2 (available online only at http://www.archives-pmr.org/).

Fig 1.

PRISMA 2020 flow diagram.

Study characteristics

The characteristics of the included studies and their main results are summarized in table 1. Nineteen RCTs35, 36, 37^,39, 40, 41, 42^,45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57 and 4 pre–post studies,^38,43,44,52 including a total of 840 participants, were analyzed. All studies involved a single intervention group and examined the effectiveness of rehabilitation interventions on participation in children and adolescents with UCP, published between 2011 and 2025. The countries of origin of the studies were heterogeneous, with 5 conducted in the USA,^{38,42,43,52,53} 4 in Turkey,^37,39,49,50 4 in Australia,^48,51,54,56 3 in Belgium,^36,40,55 2 in Taiwan,^47,57 and 1 each in Pakistan,³⁵ Brazil,⁴¹ the Netherlands,⁴⁴ Colombia,⁴⁵ and Canada.⁴⁶ Participants’ ages ranged from 19 months to 18 years (mean values across studies ranged from 5.2±2.7 to 12.3±2.7y), reflecting a wide developmental spectrum within the pediatric population. Among the 840 participants, 408 were women and 383 had right-sided hemiplegia, except for 2 studies in which the affected side was not reported.^46,53

Table 1.

Characteristics of included studies.

Study Characteristics			Population Characteristics				Intervention Characteristics					Outcome
First Author (Y)	Country	Study Design	n	Age (y) Mean±SD	Sex (Female)	Affected Side (Right)	Duration (wk)	Frequency (s/wk)	Time (h/s)	G1	G2	Tool	P Value
Afzal et al (2022)35	Pakistan	RCT	G1: 18 G2: 19	4-12 G1: 8.3±0.52 G2: 7.63±0.49	G1: 8 G2: 12	G1: 13 G2: 12	G1: 3 G2: 2	5	G1: 6 G2: 6+2	CIMT	mCIMT	CPQOL-Child	.019
Araneda et al (2022)36	Belgium	RCT	G1: 16 G2: 15	5-18 G1: 9.4±4.2 G2: 9.4±5.4	G1: 7 G2: 7	G1: 10 G2: 8	G1: 2 G2: 2	G1: 5 G2: 1-5	G1: 9 G2: 1-5	HABIT-ILE	Control PT/OT	COPM	.282
Atasavun et al (2016)37	Turkey	RCT	G1: 12 G2: 12	6-14 G1: 9.13±2.57 G2: 10.11±2.62	G1: 4 G2: 10	G1: 6 G2: 6	12	2	G1: ½+¾ G2: ¾	NW+PT	PT	COPM	.352
Bailes et al (2017)38	USA	Pre–post	11	6-17 9 y 11 mo*	3	7	4+12	7	6	Ness L300 Foot Drop System		COPM	.0006
Bingöl et al (2021)39	Turkey	RCT	G1: 16 G2: 16	6-15 G1: 10.75±2.95 G2: 10.12±2.96	G1: 6 G2: 9	G1: 7 G2: 7	10	3	3.5	mCIMT	BIT	CASP	<.001
Bleyenheuf et al (2015)40	Belgium	RCT	G1: 12 G2: 12	6-13 G1: 8.9±1.7 G2: 8.5±1.7	G1: 9 G2: 9	G1: 7 G2: 7	2	5	9	HABIT-ILE	PT/OT+wait list	LIFE-H	<.001
de Brito Brandão et al (2012)41	Brazil	RCT	G1: 8 G2: 8	3-10	G1: 2 G2: 4	G1: 5 G2: 2	3	5	5	CIMT	HABIT	COPM	.04
Ferre et al (2016)42	EEUU	RCT	G1: 12 G2: 12	2-12 G1: 5.2±2.7 G2: 5.8±2.3	G1: 7 G2: 7	G1: 7 G2: 8	9	5	2	HABIT	LIFT control	COPM	NR
Gardas et al (2025)43	USA	Pre–post	40	10.7±3.24	9	24	1	5	6	HABIT		COPM	.001
Geerdink et al (2015)44	Netherlands	Pre–post	20	8-18 9 y, 6 mo (8 y, 2 mo-17 y, 5 mo)†	9	10	1	5	CIMT: 3 BIT: 5	CIMT+BIT		COPM	<.001
Gordon et al (2011)45	Colombia	RCT	G1: 21 G2: 21	3.5-10 G1: 6.3±2.2 G2: 6.4±1.11	G1: 12 G2: 10	G1: 15 G2: 9	3	5	6	CIMT	HABIT	GAS	<.001
Hilderley et al (2025)46	Canada	RCT	G1: 55 G2: 44	6-18 G1:10.7±2.1 G2: 10.7±2.8	G1: 22 G2: 16	NR	2	5	7.5	tDCS+CIMT/BIT	Sham+CIMT/BIT	CASP	.40
Hsin et al (2012)47	Taiwan	RCT	G1: 11 G2: 11	6-8 G1: 6.9±0.6 G2: 6.9±0.6	G1: 6 G2: 6	G1: 6 G2: 6	4	2	3.5-4	HB-CIMT	HB-TR	CPQOL-Child	.350
James et al (2015)48	Australia	RCT	G1: 51 G2: 50	8-18 G1: 11 y 8 mo±2 y 4 mo G2: 11 y 10 mo±2 y 5 mo	G1: 25 G2: 25	G1: 23 G2: 30	20	6	½	Mitii	Control	COPM	<.001
Kara et al (2020)49	Turkey	RCT	G1: 15 G2: 15	7-16 G1: 12.3±2.69 G2: 11.8±2.85	G1: 8 G2: 8	G1: 8 G2: 8	12	3	30	MT+Power and Strength Exercises	OT	COPM	<.001
Kemer et al (2023)50	Turkey	RCT	G1: 11 G2: 9	6-12 G1: 11.00 (9/12)† G2: 9 (6/12)†	G1: 7 G2: 3	G1: 7 G2: 5	KT: 3-1-3 d PT: 4	PT: 2	PT: 40	KT+PT	PT	COPM	<.001
Pool et al (2015)51	Australia	RCT	G1: 16 G2: 16	5-18 G1: 10.11±3.10 G2: 10.5±2.8	G1: 7 G2: 8	G1: 11 G2: 12	8	6	4	FES	Control	COPM	<.001
Roberts et al (2020)52	EEUU	Pre–post	32	5-15 9 y 3 mo*	16	17	2	5	6	CIMT+VR+exoskeleton		COPM	<.001
Roberts et al (2025)53	USA	RCT	G1: 13 G2: 19	5-14 G1: 10 y 3 mo±3 y 0 mo G2: 8 y 7 mo±3 y 1 mo	G1: 7 G2: 12	NR	2	5	6	CIMT	CIMT+VR	COPM	.322
Sakzewski et al (2011)54	Australia	RCT	G1: 31 G2: 32	5-16 G1: 10.1 (5.7-16.7) G2: 10.1 (5.5-16.5)	G1: 15 G2: 15	G1: 16 G2: 15	2	5	6	CIMT	BIM	LIFE-H	<.001
Saussez et al (2023)55	Belgium	RCT	G1: 20 G2: 20	5-18 G1: 9.0±3.1 G2: 9.1±2.9	G1: 10 G2: 10	G1: 7 G2: 8	2	5	9	HABIT-ILE	HABIT-ILE + Reach Touch	COPM	<.001
Wallen et al (2011)56	Australia	RCT	G1:25 G2: 25	19 mo-7 y 10 mo G1: 48.8±21.9 mo G2: 48.4±20.5 mo	G1: 12 G2: 15	G1: 17 G2: 10	8	G1: 7 G2: 7	G1: 2 G2: 20´ flexible	mCIMT	Intensive OT	COPM	.61
Wang et al (2021)57	Taiwan	RCT	G1: 9 G2: 9	5-12 G1: 8.56±2.09 G2: 8.56±2.15	G1: 6 G2: 5	G1: 2 G2: 5	G1: 4+4 G2: 8	2	2.25	CIT-Wii	CIT	Participation ToP	.4

Abbreviation: BIT, Bimanual Intensive Training; CASP, Child and Adolescent Scale of Participation; CIMT, Constraint-Induced Movement Therapy; COPM, Canadian Occupational Performance Measure; CPQOL-Child, Cerebral Palsy Quality Of Live Children; G1, group 1; G2, group 2; GAS, Goal Attainment Scaling; HABIT, Hand Arm Bimanual Intensive Training; HABIT-ILE, Hand and Arm Bimanual Intensive Therapy Including Lower Extremity; HB, Home Based; LIFE-H, Assessment of Life Habits; KT, Kinesio Tape; LIFT, Lower limb intensive functional training; mCIMT, modified Constraint-Induced Movement Therapy; Miiti, web-based mobility program “Move it to improve it”; MT, Mirror Therapy; n, number of subjects; NR, no data reported; NW, Nintendo Wii; OT, Occupational Therapy; PT, Physiotherapy; tDCS, transcranial Direct Current Stimulation; TR, Traditional Rehabilitation; VR, Virtual Reality; s, session; ToP, Test of Playfulness; TS, Testing Sessions.

^⁎Mean.

^†Mean (range).

Across all analyses, the associated P values ranged from <.001 to .61, indicating that although several pooled effects reached statistical significance (P<.05), many others did not. This variability reflects the heterogeneity of study designs, intervention types, and outcome measures included in the review.

Interventions

The interventions described in the included studies were heterogeneous in terms of duration, frequency, daily dosage, and therapeutic approach. Total intervention duration ranged from ½ week to 20 weeks. Several studies implemented medium-length programs of 3-6 weeks,^35,36,40,41,43, 44, 45, 46, 47^,49,52, 53, 54, 55 whereas others applied highly condensed formats such as 2-week intensive phases or hybrid protocols combining shorter periods (eg, 4+12 wks or 4+4 wks) at different time points.^38,57 The frequency of sessions per week also varied substantially. Seventeen studies^34,36,38,40, 41, 42, 43, 44, 45, 46^,48,52, 53, 54, 55, 56 delivered interventions on consecutive days—daily schedules (5-7 sessions/wk), typically in camp or immersion formats—whereas the remaining studies provided 2-3 sessions/week in outpatient or school-based settings. The daily dose (h per session) ranged from very short 20-40 minute sessions^37,50,51 to highly intensive training blocks of 6-9 h/d.^{35,36,38,40,43,45,46,}52, 53, 54, 55

All RCTs compared 2 active rehabilitation interventions, and the pre–post studies applied a single rehabilitation protocol. No study included a nonrehabilitation intervention control group. The therapeutic approaches were diverse, although most studies involved intensive, goal-directed, task-specific training models. The most frequently applied interventions were CIMT, in either classic or modified forms, which appeared in at least 12 trials.^35,39,41,44,45, 46, 47^,52, 53, 54^,56,57 CIMT was sometimes combined with other modalities, such as bimanual intensive training (BIT/HABIT or HABIT-ILE, virtual reality, Nintendo Wii, or exoskeleton-assisted therapy. Bimanual intensive training, alone or in combination with CIMT, was implemented in at least 11 trials,^36,39, 40, 41, 42, 43, 44, 45, 46^,54,55 including HABIT-ILE variations targeting both upper and lower limbs. Some studies^38,51 evaluated lower limb–focused electrostimulation programs, such as functional electrical stimulation or the NESS L300 Foot Drop System, whereas others explored transcranial direct current stimulation for the upper limb.⁴⁶ Another protocol⁵⁰ investigated kinesio taping applied to muscles such as the gluteus medius, usually in combination with conventional physiotherapy. In addition, several interventions integrated technology-based components—virtual reality platforms, Wii gaming, web-based therapy, robotic or exoskeleton devices, and telerehabilitation—either as a complement to CIMT/HABIT or as stand-alone programs.^{37,48,52,54,55,57} Control or comparator arms in multiarm studies included standard physiotherapy or occupational therapy, wait list controls, conventional occupational therapy, or Lower Extremity Intensive Functional Training control programs.^36,37,40,42,47, 48, 49, 50, 51^,56 Detailed descriptions for each protocol are provided in supplemental table S1 (available online only at http://www.archives-pmr.org/).

Furthermore, the included studies assessed participation using a variety of standardized, validated instruments aligned with the ICF framework. The most frequently used measure was the COPM, applied in 15 studies.36, 37, 38^,41, 42, 43, 44^,48, 49, 50, 51, 52, 53^,55,56 The Cerebral Palsy Quality of Life Questionnaire–participation domain was used in 2 studies,^34,47 and the Child and Adolescent Scale of Participation in 2 studies.^39,46 Other tools included the Assessment of Life Habits (2 studies),^40,54 the GAS (1 study),⁴⁵ and the Test of Playfulness (1 study).⁵⁷ Although the GAS can assess goals across different ICF levels—including body functions, activities, and participation—it was included here when goals explicitly targeted meaningful engagement in life situations reflecting participation. Similarly, the Test of Playfulness, which evaluates children’s intrinsic motivation and engagement during play, was considered relevant as play is a recognized participation domain in children and contributes to social, cognitive, and emotional development. For all instruments, higher scores indicated better participation, reflecting greater autonomy, functional engagement, or satisfaction in the evaluated activities.

Risk of bias and quality assessment

The overall risk of bias assessment for the 19 RCTs showed that all except 1 study⁵⁷ presented a moderate risk of bias. This was mainly because of concerns in the domain of outcome measurement (D4), where most studies relied on subjective, self-reported questionnaires to assess participation. Furthermore, 21% of the studies presented “some concerns” regarding deviations from intended interventions (D2), typically because participant and therapist blinding is difficult in rehabilitation contexts. Overall, methodological quality was generally good across the remaining domains, with most studies showing low risk in randomization, missing outcome data, and adherence to planned interventions. Detailed ratings for each study and domain are presented in supplemental fig S1 (available online only at http://www.archives-pmr.org/).

Regarding the risk of bias assessment for the pre–post studies (supplemental fig S2, available online only at http://www.archives-pmr.org/), all were rated as having a serious overall risk because none included a concurrent control group or adjusted for confounding factors, resulting in serious bias in confounding (D1).

According to the GRADE classification, confidence in the pooled estimate of the effect on participation is low for rehabilitation interventions and CIMT/mCIMT and very low for bimanual and technology‐assisted interventions (supplemental table S2, available online only at http://www.archives-pmr.org/). Certainty was downgraded because of serious risk of bias in most trials and imprecision of the effect estimates; for bimanual therapy it was further lowered because of substantial heterogeneity (I²=88%).

Primary meta-analysis

The pooled effect size estimate for the overall pre–post effect of rehabilitation intervention on participation outcomes in children and adolescents with UCP was 1.44 (95% CI, 0.95-1.92), as shown in figure 2. This result is statistically significant and indicative of a strong effect. Heterogeneity among studies was considerable (I²=95%).

Fig 2.

Forest plot showing the standardized mean differences and 95% confidence intervals for pre–post changes in participation after rehabilitation interventions in children and adolescents with UCP.

The funnel plot showed marked asymmetry and Egger’s regression test was statistically significant (intercept= −2.13, P<.001), suggesting the presence of publication bias (supplemental fig S3, available online only at http://www.archives-pmr.org/).

Sensitivity analyses indicated that, when excluding the 4 single-arm pre–post studies, the pooled effect size for the overall pre–post analysis slightly decreased to 1.22 (95% CI, 0.79-1.65), remaining statistically significant and indicative of a strong effect. Heterogeneity remained considerable (I²=92%). In addition, excluding individual studies one by one generally did not substantially change the pooled effect size or heterogeneity estimates, indicating no single study disproportionately influenced the results. Notably, when the outlier study by Geerdink et al⁴⁴—which reported an unusually large effect size (Cohen’s d=−9.4)—was excluded, the pooled effect size decreased to 1.28 (95% CI, 0.88-1.68), indicating that this outlier had minimal influence on the overall conclusion. The exception was when the study by Sakzewski et al⁵⁴ was excluded, resulting in a pooled effect size of 1.59 (95% CI, 1.08-2.10) with heterogeneity remaining considerable (I²=95%), reflecting an increased effect estimate but persistent variability.

After controlling for age, residual heterogeneity remained substantial (I²=77.4%, τ²=4.03, H²=4.42), indicating that other study-level or intervention-level characteristics likely account for a greater proportion of the observed variability (as shown in fig 3).

Fig 3.

Meta-regression analysis examining the relationship between mean participant age and standardized effect sizes for participation outcomes.

Secondary meta-analysis

CIMT/mCIMT

Six studies including 193 participants were analyzed. The pooled effect size for the comparison between CIMT/mCIMT and alternative interventions was −0.09 (95% CI, −0.37 to 0.20), as shown in supplemental figure S4 (available online only at http://www.archives-pmr.org/). This result is not statistically significant, indicating no difference in participation outcomes between CIMT/mCIMT and other active interventions. Heterogeneity was negligible (I²=0%).

Egger’s regression intercept was not statistically significant (intercept=0.36, P=.59; 95% CI: −1.35 to 2.08), suggesting no publication bias in the meta-analysis comparing CIMT/mCIMT with alternative interventions. Sensitivity analyses indicated that the exclusion of individual studies, one by one, did not significantly alter the pooled effect size or its statistical significance.

Bimanual therapy

Eight studies including 256 participants were analyzed. The pooled effect size for bimanual therapy (HABIT/BIT/BIM/HABIT-ILE) compared with other active interventions or control was 0.69 (95% CI: −0.08, 1.46), as presented in supplemental figure S5 (available online only at http://www.archives-pmr.org/). Although the point estimate indicates a moderate to moderate effect favoring bimanual therapy, the CI results not statistically significant. Heterogeneity among the included studies was considerable (I²=88%).

Egger’s regression intercept approached statistical significance (P=.057), indicating no definitive evidence of publication bias. A sensitivity analysis excluding studies that applied HABIT-ILE—to isolate effects on upper limbs—yielded a weaker, nonsignificant pooled effect size of 0.21 (95% CI, −0.22 to 0.64), reinforcing the uncertainty of the effect of bimanual therapy on participation outcomes.

Technology-assisted physiotherapy interventions

Five studies including 215 participants were analyzed. The pooled effect size for technology-assisted physiotherapy interventions compared with control, conventional therapy, or other active interventions was 0.15 (95% CI, −0.30 to 0.60), as shown in supplemental figure S6 (available online only at http://www.archives-pmr.org/). This result is not statistically significant, indicating no difference in participation outcomes between technology-enhanced therapies and comparators. Moderate heterogeneity was observed among studies (I²=56%).

Egger’s regression intercept was not statistically significant (intercept=1.27, P=0.18; 95% CI, −1.05 to 3.60), indicating no indication of publication bias. Sensitivity analyses were performed by sequentially removing each study from the meta-analysis. The pooled effect size and overall conclusions remained unchanged, indicating that the results are robust and not driven by any single study.

Discussion

This systematic review and meta-analysis evaluated the effects of various rehabilitation interventions on participation in children and adolescents with UCP. The overall results showed a strong pre–post effect in improving participation, although considerable heterogeneity and the likely presence of publication bias suggest caution in interpretation. Specific interventions, such as CIMT/mCIMT and technology-assisted therapies, did not show significant differences when compared with other active rehabilitation approaches. In contrast, bimanual therapy yielded a moderate effect size; however, the CIs were wide and the result was not statistically significant. These findings suggest that strategies beyond isolated motor-focused interventions, such as parent coaching to facilitate participation in daily activities, environmental modifications, or structured play activities promoting social and educational engagement,^58,59 may be important to enhance participation outcomes. Such examples illustrate how participation-focused intervention frameworks can be operationalized in clinical practice.

To our knowledge, this is the first systematic review and meta-analysis to focus exclusively on participation outcomes in children and adolescents with UCP, thereby providing novel insights beyond previous reviews^15,18 that primarily addressed activity or motor outcomes. Consistent with a recent systematic review that identified gross motor and manual function, CP type, and the home and community physical environment as the most influential determinants of participation across multiple domains in children with CP,¹⁴ the studies included in our review similarly addressed participation in diverse contexts such as daily living, leisure, social participation, and education. This alignment underscores the multidimensional nature of participation and highlights the importance of considering motor function and environmental factors when designing and evaluating rehabilitation interventions aimed at improving participation outcomes in children with UCP. Ultimately, prioritizing participation as a rehabilitation goal is essential for enhancing quality of life, because it reflects children’s broader social, educational, and recreational inclusion. This underscores the importance of rehabilitation therapy referral for all children with UCP.⁶⁰

Interpretation of meta-analytic findings

The strong overall pre–post effect observed suggests that rehabilitation interventions hold substantial promise for enhancing participation in children with UCP, although the considerable heterogeneity indicates that these benefits may not be uniform across all approaches or populations. It is noteworthy that the Geerdink et al⁴⁴ study was an outlier in the meta-analysis; the large effect may be attributable to its intensive, goal-directed CIMT-bimanual training protocol combined with self-management training and substantial pre–post measurement sensitivity on the COPM—a patient-reported outcome measure particularly responsive to functional goal achievement. Nevertheless, sensitivity analysis confirmed that excluding this study had minimal influence on the overall pooled effect size demonstrating the robustness of our findings; the consistency of positive effects across studies supports the general conclusion of intervention effectiveness.

The lack of significant differences between CIMT/mCIMT and alternative interventions challenges assumptions about the superiority of constraint-based approaches for improving participation outcomes, despite their established efficacy in enhancing motor function.⁶¹ This discrepancy highlights the well-recognized gap between improvements in activity (eg, motor gains through CIMT or HABIT) and their actual transfer to meaningful participation in real-life contexts.

Although many interventions effectively target body functions and activities, participation requires additional factors such as sufficient intervention intensity, ecological validity, family involvement, and child motivation. This highlights the need for therapies that not only improve motor skills but are also explicitly designed to facilitate meaningful engagement in social, educational, and recreational environments, as emphasized in in recent scoping reviews that demonstrate that CP research predominantly focuses on body structure and function outcomes, with activity and participation outcomes remaining understudied.^62,63 Our conclusions regarding participation are based on theoretical considerations supported by the ICF framework rather than on direct evidence of superiority of specific therapy types. Participation, as a complex multidimensional construct involving social, environmental, and personal factors, necessitates therapeutic strategies beyond those targeting isolated motor impairments.²¹ The inconclusive results for bimanual therapy, despite its theoretical advantages for daily bimanual tasks,²² along with the lack of clear benefits from technology-assisted interventions, suggest that improvements in participation may depend more on intervention intensity, goal-directedness, and ecological validity than on specific therapeutic modalities. These patterns underscore the complexity of translating motor gains into meaningful participation and highlight the need to move beyond technique-specific approaches toward comprehensive, participation-focused intervention frameworks.⁶⁴

Methodological considerations

A critical consideration in interpreting our findings is the risk of bias and methodological limitations present in the evidence base. Egger regression test indicated statistically significant evidence of publication bias in the primary meta-analysis, although this finding should be interpreted with caution. In meta-analyses including a large number of intervention groups, as in the present case where 42 intervention arms from 23 studies were analyzed, statistical tests for publication bias have increased power to detect asymmetry, even when it may be of limited practical significance.⁶⁵ The tendency to preferentially publish studies with positive or significant results, particularly smaller studies, commonly leads to funnel plot asymmetries that are statistically detectable in large meta-analyses.

In addition, outcome measurement relied on subjective participation questionnaires—which, despite being validated and administered by trained assessors, may introduce bias from participants’ and therapists’ expectations and perceptions—and the impossibility of blinding participants and therapists in rehabilitation interventions can lead to subtle deviations in adherence and motivation.⁶⁶ Although these biases did not compromise statistical analyses or data integrity, they highlight the inherent challenges of bias control in nonpharmacological contexts. Furthermore, all single-arm pre–post studies were rated as having serious risk of bias in ROBINS-I Domain 1 because of the absence of a concurrent comparator and lack of adjustment for potential confounders (age, severity, concomitant treatments). This design limitation restricts causal inference and underscores the need to interpret these results with caution.

These methodological concerns directly translate into the low and very low certainty ratings observed across all intervention types. For CIMT/mCIMT, the low overall confidence reflects the serious risk of bias present in most trials and imprecision in pooled estimates, despite negligible statistical heterogeneity. For bimanual and technology-assisted therapies, very low certainty additionally reflects substantial inconsistency and wide, inconclusive CIs. These findings underscore that, although point estimates suggest potential benefits of intensive upper limb therapies on participation, our confidence in these effects remains limited by the inherent biases and design limitations identified.

Clinical and social implications

The findings of this meta-analysis have important clinical and social implications for children and adolescents with UCP. The strong overall pre–post effect observed suggests that rehabilitation interventions can meaningfully enhance participation.⁶⁷ The inconclusive evidence for CIMT/mCIMT and technology-assisted interventions with respect to participation outcomes suggests that clinicians should adopt more individualized, participation-focused approaches. Moreover, the considerable heterogeneity across studies underscores the need for personalized treatment plans that align with each child’s abilities, family priorities, and environmental contexts.¹¹ Clinicians are encouraged to design and refer children and adolescents with UCP to rehabilitation interventions that are intensive, goal-directed, and embedded in real-life contexts to optimize participation outcomes.

Study limitations

Rigorous risk of bias assessments using RoB 2.0 and ROBINS-I tools, combined with GRADE evaluations, provided a robust appraisal of evidence certainty. Focusing exclusively on participation outcomes in UCP addresses a critical literature gap and offers targeted clinical insights.

However, several limitations warrant caution. Regarding primary evidence, included studies exhibited variable methodological quality, intervention dosages, and participation measures—with some using specific participation scales and others employing instruments blending activity and participation (eg, COPM, GAS), potentially affecting accuracy. The predominance of COPM and other blended instruments, along with reliance on subjective, self-reports, and the impossibility of blinding participants and therapists introduced performance and detection biases. High heterogeneity driven by variable study designs, intervention dosing, and outcome measures further reduced precision. In addition, small sample sizes and potential publication bias decreased confidence. The heterogeneous and incomplete reporting of participant characteristics (particularly GMFCS or MACS levels) across included studies prevented us from conducting planned subgroup meta-analyses. From a review design perspective, the inclusion of both RCTs and single-arm pre–post studies introduced variability, and the absence of untreated control groups—because of ethical constraints—limits robustness and causal inference. Some studies comparing 2 active rehabilitation interventions (eg, CIMT vs mCIMT) could not be included in the quantitative synthesis because of insufficient pre–post data or lack of a common comparator, which may have led to the exclusion of potentially informative data. Finally, generalizability is limited by inconsistent or absent long-term follow-up data and the scarcity of studies across diverse geographic and clinical contexts.

Future research

Future research should prioritize rigorous designs with adequate sample sizes, blinded outcome assessment where feasible, standardized participant characteristics, standardized intervention protocols, and consistent long-term follow-up to reduce bias, improve precision, and resolve heterogeneity. In addition, future studies should focus on developing and testing interventions explicitly designed to enhance participation, such as participation-focused therapy programs, parent coaching, and environmental modifications. Only through such efforts can we definitively determine which rehabilitation interventions—particularly those tailored to individual goals and real-life contexts—may support participation in children with UCP.

Conclusions

This systematic review and meta-analysis indicates that rehabilitation interventions are associated with substantial pre–post improvements in participation for children and adolescents with UCP, enhancing engagement in daily life, leisure, social, and educational activities per the ICF framework. However, because these gains are based on within-group comparisons, the potential influence of natural maturation and concurrent therapies must be considered, despite sensitivity analyses supporting the robustness of these findings across diverse settings.

Furthermore, comparisons of intervention types showed no clear advantage for CIMT/mCIMT or technology-assisted therapies with respect to improving participation outcomes, and only a moderate, nonsignificant effect was observed for bimanual approaches. These results suggest that factors such as intervention intensity, goal-directedness, and real-world relevance may be more critical drivers of participation gains than specific modalities. Future research should focus on standardized, high-intensity, participation-focused protocols with active comparators to identify the most effective therapeutic elements.

Supplier

a. SPSS, version 28.0; IBM.

Disclosure

The investigators have no financial or nonfinancial disclosures to make in relation to this project.