Participants

In this study, 50 children with uCP (11y11m ± 2y10m, 27 males, 27 right-sided uCP) and 50 individually age- and sex-matched TDC (11y11m ± 2y10m, 44 right-handed) were included. Children with uCP were further classified according to their MACS-level (MACS-levels: l: N = 27, ll: N = 16, lll: N = 7). Technical problems caused missing data in some parameters. These children and their match were omitted from the analysis of the specific parameter in the comparison between children with uCP and TDC. Data analysis was conducted on the first level of the BOB and OH tasks using 50 matches, except for the BOB task's reaction time difference parameter (47 matches; MACS-levels: l: N = 27, ll: N = 16, lll: N = 6) and the OH task's hand speed bias (49 matches; MACS-levels: l: N = 27, ll: N = 16, lll: N = 7). The second level of the BOB task was analysed with 48 matches (MACS-levels: l: N = 26, ll: N = 16, lll: N = 7) and the circuit task with 46 matches (MACS-levels: l: N = 26, ll: N = 16, lll: N = 5). A full overview of the total number of included participants for each parameter can be found in  Additional file 2.

Comparison of bimanual parameters between children with uCP and TDC

First, we investigated the difference in bimanual coordination parameters of the three tasks between children with uCP and TDC. A visual representation of the effect sizes from the ANCOVA can be found in Fig. 3 and an overview of the descriptives of the ANCOVA with mean and 95% confidence interval shown in Table 2.

Fig. 3

Effect sizes (partial eta squared) with their 90% confidence interval from the ANCOVA investigating the difference in bimanual coordination between uCP and TDC for the main effect of group (A) and age (B). Partial eta squared is classified and presented in red (low effect size), yellow (medium effect size) and green (large effect size). A filled symbol (diamond or circle) represents a significant difference with p ≤ 0.05

Table 2 Overview of ANCOVA of bimanual parameters between children with uCP and TDCBall-on-bar task (BOB)

The results of the BOB task showed that children with uCP had worse bimanual coupling compared to TDC. Significant between group differences were found in two bimanual coupling parameters, with a moderate (mean bar tilt; level 1 and 2: p = 0.01, \(\eta_}}^\) =0.09) to high effect size (bar tilt standard deviation; level 1: p < 0.001, \(\eta_}}^\) =0.18; level 2: p < 0.001, \(\eta_}}^\) =0.20). No significant difference with only a small effect size was found for bar length variability (level 1: p = 0.14, \(\eta_}}^\) =0.03; level 2: p = 0.38, \(\eta_}}^\) =0.01). For interlimb differences, reaction time difference from the first level, hand speed difference from the second level and hand path length bias from both levels showed significant differences between groups with a large effect size, indicating greater differences between hands in children with uCP compared to TDC (level 1—reaction time difference: p < 0.001, \(\eta_}}^\) =0.36; level 2—hand speed difference: p < 0.001, \(\eta_}}^\) =0.19; both levels—hand path length bias: p < 0.001, \(\eta_}}^\) =0.21–0.23). Interlimb differences in hand path length are presented by higher path length of the dominant hand in children with uCP, whereas TDC exhibited longer hand path for their non-dominant hand (Fig. 5). Hand speed peak bias did not show a significant difference between groups with only a small effect size (p = 0.20–1.00, \(\eta_}}^\) =0.00–0.02). Regarding the effect of age, a main effect was found for reaction time difference, with a large effect size (p < 0.001, \(_^\)=0.34, Fig. 3B) and hand speed difference with a small effect size (p = 0.04, \(_^\)=0.05, Fig. 3B), indicating that older children exhibited less differences between hands for both groups.

Object-hit task (OH)

Bimanual coupling parameters of the OH task showed a significant difference between groups, but only with a low to medium effect size (hand transition: p < 0.001, \(_^\)=0.10; hand selection overlap: p = 0.05, \(_^\)=0.05), indicating that children with uCP crossed the workspace further with their dominant hand and required more space to use both hands effectively compared to TDC. Interlimb differences in the OH task showed that children with uCP hit more balls, moved more and faster with their dominant hand compared to their non-dominant hand, illustrated by large effect sizes (hand bias hits: p < 0.001, \(_^\)=0.27; movement area bias: p < 0.001, \(_^\)=0.22; hand speed bias: p < 0.001, \(_^\)=0.34), resulting in more differences between hands in comparison to TDC. No significant age effects were found with only low effect sizes for the bimanual parameters of the OH task (p = 0.14–0.90; \(_^\)=0.00–0.04).

Circuit task

The circuit task showed a significant difference in bimanual coupling between groups with a large effect size (bimanual coordination factor: p < 0.001, \(_^\)=0.31), showing that children with uCP performed the circuit task more sequentially compared to TDC. A significant influence of age was found for this parameter with a large effect size, showing that better coordination was present in older children in both groups (p < 0.001, \(_^\)=0.39).

Comparison of bimanual parameters between MACS-levels

We selected the parameters that exhibited large differences between children with uCP and TDC, with the goal of prioritizing parameters that reflect reduced bimanual coordination in children with uCP. This selection resulted in three bimanual coupling parameters (BOB and circuit task) and seven interlimb difference parameters (BOB and OH task). Figure 4 provides a summary of the effect sizes for this selection, while Fig. 5 presents the individual data points with post-hoc comparison. Due to three parameters (BOB task level 1 and 2: hand path length bias, OH task: movement area bias) exhibiting an interaction effect with age, a moderate regression analysis was conducted, leading to two separate analyses (ANCOVA and moderated regression) in this objective. Tables with the descriptive statistics of all the parameters are provided in the Additional files 3, 4, 5.

Fig. 4

Effect sizes (partial eta squared) with their 90% confidence interval from the ANCOVA (A, B) and moderated regression (C–E) investigating the difference in bimanual coordination in children with uCP across different MACS-levels for the selection. For the ANCOVA, the effect sizes for the main effect of MACS-levels (A) and age (B) is presented. Effect sizes are shown for the moderated regression for the interaction between MACS-levels and age (C) and main effect of MACS-levels (D) and age (E). Partial eta squared is classified and presented in red (low), yellow (medium) and green (large). A filled symbol (diamond, circle or square) represents a significant difference with a p ≤ 0.05

Fig. 5

Individual data points with back-transformed mean for the parameters of the ball-on-bar (A–F), object-hit (G–I) and the circuit task (J). Results of the differences between MACS-levels in children with uCP is presented in solid lines if a significant main effect of MACS-levels was present. TDC  typically developing children (blue), uCP  unilateral cerebral palsy, MACS  manual classification system (level l = turquoise, level ll = green, level lll = purple)

Ball-on-bar task (BOB)

The BOB task showed that for the bimanual coupling parameters, the bar tilt standard deviation of both levels, were influenced by the MACS-levels with high effect sizes. This indicates that children with lower manual ability or MACS level lll presented greater variability in bar tilt or reduced bimanual coupling (bar tilt standard deviation; level 1: p = 0.02,\(_^\)=0.18; level 2: p < 0.001, \(_^\)=0.36), compared to children with MACS l (p = 0.01) in level 1 and to MACS l (p < 0.001) and ll (p < 0.02) in level 2. For interlimb differences, level 1 of the BOB task demonstrated less difference between hands for reaction time with a high effect size (p < 0.001, \(_^\)=0.40) in MACS I compared to MACS II (p < 0.001) and MACS III (p < 0.001). No difference was found between children with a MACS level ll and lll (p = 0.62). Moreover, moderated regression indicated no moderation of age and no significant differences in hand path length bias between MACS-levels for both levels of the BOB task with only low effect sizes (interaction: p = 0.37, \(_^\)=0.04; main effect MACS-levels: p = 0.15–0.67, \(_^\)=0.01–0.08; main effect age: p = 0.05–0.19, \(_^\)=0.07–0.15). For the second level of the BOB task, hand speed difference was significantly different between groups with a high effect size (p < 0.001,\(_^\)=0.51), indicating that children with a MACS-level lll showed more difference in speed between hands compared to MACS l and ll (MACS I-II-III: p ≤ 0.02).

Object-hit task (OH)

For the OH task, the hand bias hits and hand speed bias did not show a significant difference between MACS-levels, despite moderate effect sizes (hand bias hits: p = 0.15, \(_^\)=0.10; hand speed bias: p = 0.20, \(_^\)=0.12). Figure 5G and I show a trend indicating that children with lower manual ability displayed greater interlimb differences. Lastly, moderated regression indicated that age significantly moderated the difference in OH task movement area bias across MACS-levels with a high effect size (interaction: p = 0.01, \(_^\)=0.16). This means that interlimb differences in movement area increased with age in children with MACS-level lll, while they decreased with age in children with MACS level I (b = − 0.04, t = − 3.13, p = 0.003). No relation with age was found in children with MACS level II (b = − 0.04, t = − 2.94, p = 0.01) compared to children with MACS lll. Furthermore, no moderating effect of age (b = 0.00, t = − 0.35, p = 0.71, \(_^\)=0.006) and no significant difference was found between MACS-levels l and ll (b = 0.02, t = 0.22, p = 0.82, \(_^\)=0.002), with only small effect sizes.

Circuit task

The bimanual coordination factor of the circuit task showed no significant difference between MACS-levels with a moderate effect size (p = 0.32,\(_^\)=0.07).

Relation between bimanual and unimanual or task execution parameters

Similar to previous objective, we utilized the same selection of parameters to emphasize those that exhibit a large difference between uCP and TDC. An overview of the correlation coefficients between the selected bimanual and unimanual and task execution parameters in children with uCP can be found in Fig. 6. Correlation matrix including all parameters can be found in additional file Additional file 6.

Fig. 6

Pearson and spearman's rank correlation coefficients of the unimanual and task execution parameters with the bimanual parameters of the ball-on-bar task (A), object-hit task (B) and circuit task (C) for the selected parameters in children with uCP. DH  dominant hand, NDH  non-dominant hand, s  spearman’s rank correlation

Relation between bimanual and task execution parameters

For the BOB task, bimanual coupling was moderately related with the task execution parameters, showing that higher bar tilt standard deviation was associated with a lower number of completed targets in both levels (rs = − 0.54–(− 0.63), p < 0.001), slower time to target (rs = 0.54, p < 0.001) in level 1 and more drops in level 2 (rs = 0.61, p < 0.001). More interlimb differences in path length between hands of children with uCP was moderately associated with a higher time to target in level 1 (r = 0.54, p < 0.001) and more difference in speed between both hands was moderately related to a lower number of completed targets (rs = − 0.63, p < 0.001) and a higher time to target (rs = 0.58, p < 0.001) in level 2. Lastly, more difference in reaction time between hands was highly related to reaching less targets (targets complete, rs = − 0.77, p < 0.001) and needing more time to reach the target (time to target, rs = 0.72, p < 0.001) in level 1.

For the interlimb differences of the OH task and bimanual coupling parameter of the circuit task, only no to low correlations were found (r and rs = − 0.46–0.44, p = 0.001–0.96).

Relation between bimanual and unimanual parameters

For bimanual coupling of the BOB task, no correlations were found between the bar tilt standard deviation and the unimanual parameters of both hands in both levels (rs = − 0.21–0.13, p = 0.15–0.95). Hand speed of the non-dominant hand was moderately associated to hand path length bias in the first level of the BOB task (r = − 0.60, p < 0.001) and lowly related in the second level (rs = − 0.48, p < 0.001), showing that a higher speed of the non-dominant hand was related to less interlimb differences for hand path length. Also, higher speed and more speed peaks of the non-dominant hand were only lowly related to less reaction time difference in the first level of the BOB task (rs = − 0.44–(− 0.33), p = 0.004–0.03).

For the OH task, the parameter hand bias hits was highly related to the hits of the non-dominant hand (r = − 0.75, p < 0.001) and movement area bias to movement area of the non-dominant hand (rs = − 0.62, p < 0.001), respectively. Hand speed bias was moderately correlated with all unimanual parameters of the non-dominant hand, where more interlimb asymmetry in hand speed was related to fewer hits (r = − 0.66, p < 0.001), slower hand speed (rs = − 0.65, p < 0.001) and less movement area (rs = − 0.50, p < 0.001) of the non-dominant hand.