Bimanual movements in children with cerebral palsy: a systematic review of instrumented assessments

This systematic review is reported according to the PRISMA guidelines. A PRISMA checklist was completed and the review protocol was published in Prospero (CRD42022308517).

Searches were conducted in English in the following databases: PubMed (1996 to March 2022), Scopus (2004 to March 2022), Cochrane Library (1995 to March 2022) and Web of Science (2004 to March 2022). To ensure the search was exhaustive, the following combinations of keywords were used: (1) keywords relative to children and adolescents: “child”, “adolescent”, “teen”, “infant”, “baby”, “newborn”; (2) “upper limb”, “upper extremities”, “upper body”, “arm”, “hand”, “bimanual”, “interlimb”; (3) “movement”,”motion”, “motor”, “biomechanical”, “kinematic”, “instrumented measurement”, “inertial”, “sensors”, “accelerometer”, “technologies”, “spatiotemporal”, “temporo-spatial”, “smoothness”, “fluidity”, “trajectory”; 4) “Cerebral Palsy”, “hemiplegia”, “hemiparesis”, “quadriplegia”, “tetraplegia”, “stroke”, “cerebrovascular accident” (the detailed equation is provided in Additional file 1). Search strings were formulated and tailored to the search syntax of each database to ensure a common search strategy.

Inclusion criteria were: (1) Design: full papers; (2) Participants: children and adolescents 0–18 years old with uni- or bilateral cerebral palsy; (3) Construct: exploration of bimanual movements (i.e. both hands used to achieve a goal) during symmetrical or asymmetrical tasks; (4) Type of instruments: quantitative instruments (3DMA, accelerometers, etc.); (5) Outcome measures: quantitative/instrumental measures of UL characteristics (spatiotemporal and/or kinematics and/or quality of movement variables and/or accelerometry) (see Additional file 2). Exclusion criteria were: (1) studies that evaluated spontaneous UL use (either unimanual or bimanual; no specific task) with no information on bimanual movements in the results section; (2) studies out of scope or not relating to motion analysis (e.g., electromyography or only force assessment); (3) conference papers; and (4) studies not published in English. Studies that included participants older than 18 years were also included, even if it was not possible to dissociate the results of the adults from those of the children and adolescents. Studies that included both typically developing children (TDC) and children with CP (unilateral or bilateral) were included, but the data from the TDC were not analyzed.

Two independent reviewers screened the titles and abstracts of the selected papers for inclusion. In case of disagreement, the full article was read and discussed until a consensus was reached.

The two reviewers extracted all the data independently.

A descriptive analysis of samples, type of instruments and outcome measures was performed. Quantitative results are expressed as mean (standard deviation [SD]) and categorical results as number (%). We did not conduct a meta-analysis because of the large number of protocols, heterogeneity of samples and varied outcome measures that prevented such an analysis.

The initial search identified 2015 papers published since 1995 after removal of duplicates. Following screening of titles and abstracts, 285 articles were deemed appropriate for full text screening. Of the 285 articles reviewed, 254 were excluded. The remaining 31 were included in the review (Fig. 1).

The aim of the studies was mostly to increase understanding of bimanual movement in different conditions (Fig. 2).

The mean Q-score of the articles included was 22/32 (SD 4.7). Seven articles had a score above 80% [23‐29], nineteen articles had a score between 60 and 80% [30‐47] and six articles had a score between 60 and 40% [48‐53]. Details of the Q-scores of each article are provided in Additional file 4.

In total, 452 children and young adults with CP, aged from 2 to 25 years (mean age 10.9 years [SD 3.2]) performed bimanual movements in the 31 studies and were included (Table 1). The majority had unilateral CP (uCP), only 3 children had bilateral CP [32, 41]. They were mainly female (318 females/241 males), with right-side impairment (161 left/181 right) and MACS level from I to III. Children under 5 years of age were included in 5/32 studies [23, 26, 35, 38, 49]). Children with uCP and TDC of same ages were included in 14 studies [23‐25, 31, 33, 34, 36, 37, 39, 41‐43, 47, 48]. One study compared children with uCP to an adult group [39].

The 31 studies used 3 different types of measurement systems: (1) 3DMA, including optoelectronic systems (Optotrack^®, VICON^®, Qualisis^®, MIE Medical Research^®) [24, 25, 27, 28, 30‐36, 38‐42, 44‐46, 49‐53] and electromagnetic systems (Polhemus^®) [26, 29]), (2) accelerometers (ActiGraph^®) [23, 26] and (3) other instruments (bespoke instrumented cube [43], Kinect device [48] or digitizer with two pens [47]) (Table 2). The 3DMA studies used models with 2 to 41 markers positioned on both ULs. Markers were always positioned on the wrists, they were also positioned on the trunk in 9 studies and on the hips in 1 study [44]. They used between 3 and 13 infrared cameras with a sampling frequency of 50–200 Hz. Few studies reported the biomechanical model used for the data analysis. The studies that used actimetry all used two accelerometers: one on each wrist and none on the trunk [23, 26].

Most of the protocols assessed functional tasks (Table 3). This included daily tasks (ex: placing a hat on the head, opening a drawer, picking up a box, food preparation, playing, etc.) performed in a laboratory setting (n = 13), that were mainly recorded with 3DMA systems [26‐30, 33‐38, 42, 48]. Child-friendly protocols were used, including game scenarios (n = 5), such as 3D virtual environment [48], videogames (Kinect [48], Nintendo Wii [32]) or projection of a scene on a screen [24, 25, 53]. The drawer opening task was the most frequently used (n = 7) [26‐28, 33‐35, 38], followed by bimanual reach to grasp tasks (n = 5) [36, 43, 48‐52], a box pick-up task (n = 3) [29, 36, 37] and bimanual circular movements (n = 3) [31, 44, 47]. The studies that used actimetry recorded the Assisting hand assessment (AHA) test session, rating the effectiveness of affected hand use during semi-structured, bimanual play activities [23, 26]. No studies evaluated spontaneous bimanual movement during free play or activities of daily living and none evaluated movement in the child’s home environment. Ten studies explored a combination of bimanual tasks (2–5 different tasks). In 5 studies, data from the bimanual tasks were only analysed for the impaired UL and not both ULs [24, 25, 30, 42, 48].

In most studies, the starting position was almost the same: sitting on a chair at a table with the knees and elbows flexed at 90° [24‐29, 31, 33‐35, 37‐40, 42‐53]. The tasks were performed under different conditions: object size, height and grasp-type [34, 39, 51, 53], target position [40, 41, 45, 46, 52], environmental feedback (ex: mirror or opaque screen) [31, 44‐46], each hand performance of asymmetrical bimanual tasks [26, 29, 33, 35, 47], or increasing difficulty [32]. Participants performed tasks at their self-selected speed, and 3 studies analysed maximum speed [33, 34, 53]. One to 5 trials were recorded, with at least 1 practice trial, except for the accelerometery measures that were recorded directly [23, 26]. The children and adolescents performed between 1 and 4 sessions of each protocol (most often 1 session [23, 25, 28, 29, 32‐34, 36, 37, 39, 42‐45, 47‐53]. Protocol durations were from a few minutes to 1 h.

Instrumented measurements were mainly used to calculate spatiotemporal variables during bimanual movements (26/32 studies) [27‐35, 37‐44, 48] (Table 4). Movement time, mean and peak velocity were the most frequently measured. Trunk and upper limb joint (mostly elbow and wrist) kinematics were evaluated in 15 studies [24, 25, 27, 27, 28, 30, 32, 34, 38, 39, 42, 48, 51‐53]. One study focused on hand movements [36]. Studies that evaluated movement quality analysed smoothness, trajectory (Index of Curvature (IOC) [24], accuracy [29] and circularity [47]. Smoothness was analysed by the number of movement units [30, 40, 41], spectral arc length (SPARC) [24, 43], number of velocity peaks [47, 48], normalized jerk [31], relative phase [45], and dysfluency of hand movement [53]. Four variables were calculated from accelerometery: mean activity count [23], asymmetry index [23], duration of limb use [26] and goal synchronization [26]. Bilateral movements were analysed with 6 specifically developed spatiotemporal variables: task completion time, goal synchronization of the hands, movement overlap time [27, 28, 33‐35, 38], interlimb coupling [29, 31, 49‐52], continuous relative phase [31, 44, 47] and asynchrony [53]. Task completion time, goal synchronization of the hands and movement overlap time were only analysed in the drawer-opening task [27, 28, 33‐35, 38].

In total, 22 studies evaluated one or more metrological properties of their instrumented measures [23‐31, 33‐43, 46, 48] (Table 5).

Spatiotemporal variables were evaluated both for research and clinical practice. Bilateral variables were used for research [28, 38, 44, 47, 49‐52] and clinical practice (pre-post intervention [27, 35, 37]) but only one study evaluated convergent validity: goal synchronization was correlated with the unimanual JTHFT assessment [33]).

The outcome measures that underwent the most validity assessments were ROM, MAX, APS, SPARC and IOC (convergent, discriminative validity and reliability) for research purposes [24, 25]. Accelerometery variables (mean activity count and asymmetry index) were used for convergent, discriminative validity and pre-post therapy, both in research and clinical settings [23, 26]. Kinematic (ROM), spatiotemporal variables (peak velocity, movement time, number of movement units and movement duration) [27, 30, 35, 37, 40, 41, 48] and quality of movement variables (smoothness) were used to assess the effect of therapy.

A wide variety of instruments has been used to quantitatively evaluate bimanual movements, depending on the objective or variables of interest (ex: accelerometers for actimetry). Most studies used 3DMA, probably because this was the first tool to be developed and because it has been shown to be consistent and accurate for gait analysis, for which it has become the gold standard [11]. However, the 3DMA system set-up for UL recordings appears to lack consensus with regard to the number of cameras, markers, etc. 3DMA is performed in a laboratory setting to measure standardized, reliable movements that may differ from those performed in daily life. We recommend the use of 3DMA as it provides an objective and accurate measure, which complements the clinical assessment but with more technical standardization. The results of this review also highlighted that other recently developed technologies, which involve smaller and less constraining systems, e.g. accelerometers, can be used [23, 26]. However, no studies have yet used these instruments to provide out-of-laboratory assessments and to investigate bimanual movements directly in daily situations (e.g. at home), likely because real-life measures are less standardized, require greater tolerance for longer recordings, and the analysis of bimanual data from the real-life setting is challenging [13]. Accelerometers or inertial measurement units should be increasingly used to measure UL performance in the home setting, i.e., how the child spontaneously uses their impaired UL in real life. Another pertinent suggestion is to adapt daily life objects to perform direct recordings in the child’s usual environment (e.g., toys, etc.) and to get away from the context of evaluation [41]. We recommend the development of tools that allow direct measurement of UL movements in real-life situations with as few constraints on the child as possible. This is necessary to improve understanding of the effects of therapies and to adapt them to the real difficulties encountered in daily life (Additional file 5).

The protocols lacked standardization, they involved different data collection procedures and, most importantly, different bimanual tasks. The choice of tasks for bimanual movement assessment is a real challenge because of the large variety of movements that can be performed with both ULs (symmetrical/asymmetrical, free/constrained, with/without object, proximal/distal, etc.). Hung et al. proposed a unique task, ‘the drawer-opening task’ [26‐28, 33‐35, 38], mainly to study coordination, however, most protocols used several tasks to provide a comprehensive and global assessment of bimanual movement. Based on the results of this review, we recommend the use of a set of 3 to 5 tasks both to provide an overview of the performance of different bimanual movements and to ensure precision, while also maintaining the attention and participation of the child during the assessment. To evaluate spontaneous movements as they are performed in daily life, tasks that involve interaction with objects should be evaluated: bimanual UL movement is often induced by the need to grasp, manipulate or hold objects. Some studies particularly focused on interaction with objects (size and shape) [34, 39, 51, 53] and factors that interfered with interaction and how information was processed (visual disturbance) [31, 44‐46] to evaluate all the situations that the child may face in daily life and to contribute to the understanding of bimanual function. In laboratory conditions, we recommend the assessment of simple daily movements (e.g., dressing, drinking and holding) or tasks within a game scenario [24, 25, 32, 48, 53] to better reflect spontaneous bimanual movements, whilst still being reproducible. The environment could be enhanced by video games or virtual reality [47]. The results of the review also support the need to directly assess bimanual movement in real-life situations (at home or school) (Additional file 5).

Spatiotemporal variables were the most often analysed. They were mainly used to describe the characteristics of bimanual movements and to evaluate the effect of therapies. Six bilateral variables were specifically developed to assess bilateral movements. Most bilateral variables were spatiotemporal and were designed for the ‘drawer-opening task’ [26‐28, 33‐35, 38] whereas other bilateral variables explored coordination between both hands [31, 44, 47, 53]. These variables may constitute important indicators since children and adolescents with CP use different motor coordination patterns during daily bimanual activities. Moreover, understanding coordination patterns could guide researchers and clinicians in the development of intervention programs that aim to improve bimanual hand coordination performance [6]. However, these variables have only been partially validated; reliability and responsiveness have yet to be determined. With regards to UL kinematics, similar variables to those used in unimanual assessments have been used to assess each UL separately, often to compare the ULs, however, no variables have been specifically developed to evaluate joint angles during bimanual movements. Quality of movement variables, i.e., smoothness or straightness, provide information that is difficult to obtain precisely with clinical assessments; this is a real advantage of motion analysis. However, there is also no consensus on the best way to measure some of these quality of movement variables (e.g. smoothness was measured using 6 different methods [24, 30, 31, 40, 41, 43, 45, 47, 48, 53]) and none have been completely validated. This finding highlights the difficulty of determining relevant UL variables that could be assessed across all types of bimanual tasks. According to the results of this review, we recommend that a comprehensive assessment of UL impairment should include spatiotemporal, kinematic and quality of movement parameters, including specific bilateral variables (goal synchronization, movement overlap time, etc.). This would help to better characterize pathological movements and to further determine those that are most relevant to guide the focus of therapies in clinical practice (Additional file 5).

The results of this review showed that most measures have not been fully validated. Reliability and responsiveness properties have not been sufficiently assessed, despite their importance. A representative example is the “drawer task”: it was used in 7 studies but only convergent and discriminant validity have been confirmed. Discriminant validity was the most evaluated: all the tools were able to differentiate between children with CP and TDC for all types of variables and bimanual tasks. Moreover, other types of discriminant validity were also evaluated, i.e. comparison between uni and bimanual movements [40, 41, 46, 47, 49‐52] and comparison of the affected and non-affected ULs during asymmetrical tasks [29, 33, 40, 41]. Future studies should include an evaluation of reliability since we found only 2 studies that assessed this property [24, 25]. Given the complexity of UL movements and the use of different instruments/markers, reliability must be evaluated because factors such as the child’s position could impact the accuracy and interpretation of the measure. In this review, we were unable to properly report on the responsiveness of the tools as it was not the main objective of the studies included. However, most studies demonstrated changes in the values of variables after different interventions, whatever the protocol and variables assessed. Therefore, instrumented assessments appear to be good indicators of therapy efficacy. Regarding convergent validity, the highest correlations were found for assessments of the impaired UL during bimanual tasks, the AHA and APS (a kinematic deviation index of the impaired UL), whereas bilateral variables were not correlated, demonstrating the pertinence of evaluating both ULs to better reflect bimanual function.

This review has some limitations. The lack of standardization of protocols and systems made direct comparisons of results between studies challenging. The quality of the articles included ranged widely (from 13 to 30/32): full descriptions of data acquisition and adherence to biomechanical recommendations (e.g., International Society of Biomechanics) were frequently missing. Furthermore, no studies reported a priori sample size calculations, therefore the statistical power is unknown. Among the articles included, 8 were from the same team [26‐28, 33‐36, 38]. The quality scale developed was provided to help interpretation but has not been previously validated, therefore, the quality results should be interpreted with caution. The age range of the children and adolescents who were compared in this review was very wide: analysis by age is necessary as toddlers and young adults may have different bimanual function. The results could not be generalized to all children and adolescents with CP as there was a lack of assessment in younger children and those with bilateral CP. Future research, involving younger children (pre-schoolers), and taking into account the different degrees of UL impairment (MACS 3–5) and diversity of CP type (bilateral CP, dyskinetic), will improve our understanding of bimanual movements and their development in these children and adolescents (Additional file 5).