Patients with TBPI were recruited from a database (Patroclo et al., 2019) maintained by the Laboratory of Neuroscience and Rehabilitation of the Institute of Neurology Deolindo Couto of the Federal University of Rio de Janeiro from June 2018 to September 2019. This database contains patients’ epidemiological, physical and clinical information collected by a multidisciplinary team through digital questionnaires. The following inclusion criteria were applied: age between 18 and 50 years, right-handed (Oldfield, 1971) before the lesion and with unilateral TBPI, diagnosed by clinical evaluation and/or complementary exams. Individuals were excluded if they presented other neurological injuries, a Mini-Mental State Exam (Folstein et al., 1975; Brucki et al., 2003) score below 24, visual loss or uncorrected visual impairment. Upper extremity function after a TBPI was measured using the version of the Disabilities of the Arm, Shoulder and Hand questionnaire (DASH) adapted to Brazilian Portuguese (Orfale et al., 2005). The DASH is composed of 30 questions that address the ability to perform upper extremity activities and the severity of symptoms (Hudak et al., 1996). Each question is scored on a scale from 1 to 5, with a total score rate from 0 to 100. A higher score reflects greater disability.

All participants were informed about the experimental procedures and provided a written consent form before the tests. The procedures were approved by the ethics committee of the Institute of Neurology Deolindo Couto, the Federal University of Rio de Janeiro (process number: 1.375.64).

Movements in three axes (mediolateral, X, antero-posterior, Y and vertical, Z) were recorded at 100 Hz using Vicon Nexus software version 2.2 (Vicon Motion Systems Ltd, Vicon, United States) and a seven-camera motion capture system with 1 Megapixel resolution (Vicon Bonita 10, Vicon, United States). The triaxial coordinate time series of the index finger reflective markers were recorded. The sequence of initial positions was generated using the software Presentation^® (Neurobehavioral System, United States). Data acquisition was synchronized with the stimulus presentation through a synchronizer (Vicon Lock).

Pre-processing was performed using Vicon Nexus 2.2 software. The index finger marker triaxial coordinates were reconstructed and analyzed offline using Matlab software (R2015a, Mathworks Inc., Natick, MA, United States). As the index finger describes a curved trajectory, we estimated index finger speed by its tangential velocity (v→) in relation to the path. For each instant of time, the tangential velocity (v→⁢(t)) is calculated by multiplying the sampling frequency (f_s) to the magnitude of index finger displacement vector (|Δ⁢r→|), which corresponds to the index finger displacement in 3-D space.

The 3-D reconstruction of the tangential velocity profile was estimated and subsequently filtered with a 5th-order low pass filter at 10 Hz cutoff. The movement was divided in two phases: (I) from the initial position to the target (apparatus or mouth) and (II) returning from the target to initial position. The onset of each phase was calculated using 5% of peak velocity (PV) as a threshold value. The same threshold was used to detect the end of each phase, i.e., when the tangential velocity dropped below the 5% of peak velocity (Esteves et al., 2016). Peaks and valleys were detected and marked along the velocity profile of each trial, and each phase of the movement consisted of two valleys and a peak. Phase I was separated for analysis and the tangential velocity profile was time-normalized by a linear interpolation of 200 points. Discrepant velocity profile shapes in each trial for each condition were detected through visual inspection, marked and excluded from the analysis.

All the analyzed kinematic variables displayed normal distribution (Kolmogorov-Smirnov calculated p ≤ 0.05). A paired T test was performed to compare the kinematic parameters collected from the right versus the left upper limb in the control group. No significant difference was observed (p > 0.05). Therefore, the kinematic values obtained for each arm (i.e., left and right) were averaged per control subject to perform statistical analysis. Kinematic analysis of the TBPI group corresponds to the uninjured upper limb recording, since only three participants succeeded in performing the tasks with the injured limb. A Mann Whitney U test was applied by comparing the kinematic measures of left and right uninjured sides to verify the existence of injury side effects.

A three-way mixed model ANOVA was performed with event sequence (fixed sequence vs. variable sequence), initial position (P1 vs. P2 vs. P3) and group (TBPI vs. control) as independent variables for each kinematic parameter. The level of significance was set at 5%. No main effects were observed for event sequence, so they were removed from the model, leaving a two-way model with the other two factors. A two-way mixed model ANOVA was applied with group and initial position as factors. Tukey’s post-hoc analysis was subsequently applied. All statistical analyses were performed using STATISTICA version 10 (Statsoft Inc., United States). Post-hoc power analyses were conducted using G*Power software (Erdfelder et al., 1996), with α = 0.05, two-tailed, to check the statistical power of our results.

From an initial group of 12 participants with TBPI, 11 individuals (2 females) matched the pre-established criteria. The TBPI group exhibited the following characteristics: median age, 35 years of age (range 22–43); median height 174 cm (range 152–184); and median weight 75 kg (range 39–105). Four participants had left arm and seven right arm TBPI (Table 1). A control group of 12 healthy individuals matched in gender, age and physical characteristics was recruited for the study with the following characteristics: median age, 28 years of age (range 19–48); median height 178 (153–190); and median weight 77 kg (52–101). Anthropometric measures and age were similar between the groups (Mann-Whitney tests p > 0.05 for all comparisons).

The characteristics of the individuals with TBPI are presented in Table 1. From eleven participants included in the TBPI group, four had complete (C5–T1) and seven had partial TBPI. Ten participants with TBPI had undergone surgical reconstruction of the brachial plexus and only one had not undergone any surgical procedure. The median time elapsed from the injury to participation in the experimental protocol was 48 (range: 4–116) months. The median DASH score was 45.0 (range: 25.0–59.5; Table 1). Only three individuals in the TBPI group were able to perform the tasks with the injured arm (data not analyzed).

The kinematic analysis of the uninjured upper limb of the TBPI vs. control participants was evaluated by means of a two-way ANOVA. This analysis was performed with group (TBPI patients vs. control) and initial position (P1, P2, and P3) as independent variables for each kinematic parameter. The group comparison results are shown for the tangential velocity profile, movement duration, trajectory length and index finger end-point parameters.

One-Way ANOVA was applied to compare those same kinematic parameters between the control and the TBPI groups in the cup-to-mouth task (Supplementary Table 2). The TBPI group was composed of the same eleven patients described in Table 1, and the control group was composed of nine out of the same twelve individuals tested in the pointing task. The TBPI group showed lower TPV (0.34 ± 0.04) compared to the control group (0.37 ± 0.02) [F(1, 18) = 6.91, p = 0.02]. No significant differences between groups were found for PV [F(1, 18) = 3.3, p = 0.09], MD [F(1, 18) = 3.68, p = 0.07], TL [F(1, 18) = 2.65, p = 0.12] and VFE [F(1, 18) = 0.12, p = 0.73]. With our sample size and effect size estimate for the TPV variable, the post-hoc power analysis revealed a 51% power to detect a large effect size (d = 0.95) between groups. No effect of injury side for the cup-to-mouth task was observed for any kinematic parameters.

For both tasks, a reduced TPV of the uninjured limb was found in the TBPI group compared to the control group. Shorter TPV values translate into a shorter period of acceleration followed by a longer deceleration period while performing the movement. This ratio provides information about the motor strategies used to control actions (Marteniuk et al., 1987; Papaxanthis et al., 1998; Sartori et al., 2011), since the deceleration phase duration increases with the task demand (Marteniuk et al., 1987). Rangel et al. (2021) have suggested that the plastic reorganization after TBPI is associated with modifications in motor planning. Changes in trunk representation, occurring bilaterally in M1 (Fraiman et al., 2016), in addition to balance impairment in adults with TBPI (Souza et al., 2016), could translate as greater motor cost to TBPI participants in performing both tasks in a standing position.

Hilt et al. (2016) showed that in healthy participants reducing the base of support leads to an increased deceleration phase in the TPV parameter. In the same way, the shortening of the acceleration phase and the lengthening of the deceleration phase in the TPV parameter found for both tasks in the present study could relate to the negative impact of TBPI on balance (Souza et al., 2016). A reduced TPV, favoring a lengthened control of the final portion of the reaching movement, could occur as a consequence of postural balance prioritization. Since the reduced TPV was also observed in the TPBI group in the cup-to-mouth task, then the mere control of upright posture could be challenging enough to impact the kinematics of the uninjured upper limb. In fact, although both tasks are performed in a standing posture, there are differences when it comes to their postural challenges.

The two tasks clearly differ in terms of the specific role of the trunk. In the whole-body task, there is a dynamic role for the trunk in assisting the upper limb movement (Hilt et al., 2016), whereas in the cup-to-mouth task the trunk functions rather as a stabilizer, allowing a relative independence of the upper limb movement (Flanders et al., 1999; Hilt et al., 2016). Crucially, the two tasks also clearly differ in the task demand imposed over postural control, being much higher for the whole-body task (Hilt et al., 2016). Thus, the augmented need of postural control imposed by the whole-body task could explain the higher number of altered kinematic parameters observed amongst TBPI participants.

Since a reduced TPV was found for both tasks in the TBPI group, it is reasonable to suppose that the longer deceleration phase would reflect an alteration in the motor plan related more generically to movements of the uninjured limb. Conversely, for the whole-body reaching task there could be a postural balance component associated more specifically with the other kinematic alterations, as discussed below.

In the whole-body reaching task, the participant has to choose a suitable hand trajectory toward the target while keeping postural balance (Flanders et al., 1999; Hilt et al., 2016). These task constraints can affect both movement planning and execution compared to reaching from a standing position without trunk displacement. Marteniuk et al. (1987) observed that reaching tasks demanding higher precision were executed with longer movement duration and a prolonged deceleration phase (when compared to the acceleration phase). Considering that reaching movements show an initial ballistic phase and a subsequent controlling phase in which fine adjustments are made to successfully achieve the task goal (Desmurget and Grafton, 2000), it is reasonable to think that tasks with higher precision requirements demand more sensory feedback control during the movement deceleration phase (Marteniuk et al., 1987).

Besides the effects of TBPI on the TPV parameter, we found lower PV values and longer MD in the TBPI group compared to the control group when performing the whole-body reaching task. These results are similar to those found by Callegari et al. (2018) who reported that during an upper limb pointing task, individuals in postural positions with less trunk stability showed longer movement duration and lower peak velocity. These effects might indicate greater difficulty in planning upper limb movements (Callegari et al., 2018).

Individuals with TBPI have also exhibited a reduction in TL of the index finger. A reduced trajectory length might reflect detriment in the correlation between trunk and limbs angular displacement (Stapley et al., 1999; Berret et al., 2009). This possibility seems supported by the fact that, although not statistically significant, TPBI end point appeared more compact and more precise than controls, suggesting that the task goal (virtual target to reach) is achieved despite a deterioration of coordination between the arm and the trunk. Moreover, upper limb kinematics was shown to be equilibrium-dependent (Stapley et al., 1999; Berret et al., 2009). For instance, Paizis et al. (2008) found a reduction of wrist trajectory length in young adults in equilibrium restriction conditions similar to those observed in older adult individuals without postural restrictions. The authors proposed that such a strategy becomes dominant in the motor plan of older adults (Paizis et al., 2008). Hilt et al. (2016) observed that the “optimizing balance” strategy was preferred when healthy individuals performed a whole-body-reaching task under reduced postural stability. According to the optimal control theory, motor control will always prioritize a lower cost to achieve the task goal (Todorov, 2004). Our results suggest that the cost to optimize motor planning to achieve the task goal successfully with the uninjured limb was greater for the TBPI group than for the control group.

In a free-endpoint whole-body reaching task the participants are free to choose their final hand position, which results in the greater involvement of implicit variables to guide action selection (Haggard, 2008; Andersen and Cui, 2009; Berret et al., 2011; Hilt et al., 2016). Interestingly, participants with TBPI were able to reach the final position as precisely as the control group, independently of the motor strategy used to achieve it. Furthermore, no significant differences were observed between the groups in respect of the finger endpoint parameter in the cup-to-mouth task.

Finger endpoint consistency for both groups, and the differences between the TBPI and control groups for the velocity and trajectory profiles in the reaching task could reflect the motor system’s efficacy in controlling the numerous degrees of freedom involved in the intended movement in order to achieve a same goal (Bernstein, 1967; Turvey et al., 1982; Schöner and Kelso, 1988; Latash et al., 2002; Bizzi et al., 2008). Chang et al. (2009) observed similar results in an animal model of peripheral nerve injury in the ankle extensor muscle; Although there was an alteration in the trajectories of individual joint kinematics, those of limb orientation and limb length remained largely invariant, even when considering the paralyzed ankle extensor muscles. These results suggest that changes in mean joint angles were coordinated as part of a long-term compensation strategy to minimize change in whole limb kinematics (Chang et al., 2009). This strategy may represent a fundamental compensation principle after peripheral injuries that allows the adaptation to the new condition with a minimal effect on overall motor function.

The differences observed between the starting positions in the whole-body reaching task for the majority of kinematic parameters confirm that they offer different mechanical constraints for movement planning and execution. Previous studies have shown that the differences in the initial positions of the upper limb in free-endpoint reaching tasks in sitting (Berret et al., 2011) or standing posture (Hilt et al., 2016) did not influence the target choice in healthy subjects. Similar results were observed in our present study, in which a small variability in the target choice was found for both groups independently of the initial mechanical constraints of the upper limb.

Current TBPI therapeutic objectives consist of improving muscle strength, range of motion and functionality (Milicin and Sîrbu, 2018; Rich et al., 2019; Chagas et al., 2021), as well as reducing the pain (Lovaglio et al., 2019) of the injured upper limb. Kinematic alterations shown here for the uninjured upper limb in the whole-body task in comparison to the cup-to-mouth task suggest, however, that the greater the postural challenge of the task, the greater is the motor cost to plan and perform it. Indeed, Souza et al. (2016) had already warned about the need to prevent and treat balance impairment in TBPI patients. The present work furthers this proposal by providing evidence in favor of including whole body and uninjured arm movements in TBPI rehabilitation programs. A better understanding of the motor plans changes after a TBPI upon the uninjured limb may help to improve the development of more accurate kinematic measures and more effective and customized rehabilitation programs.

The main limitations of the current study were the small sample size as well as its heterogeneity regarding the extent of the injury, the occurrence and type of surgical intervention and the time elapsed since the injury. We have reproduced a very well characterized task to impose trunk displacement (Hilt et al., 2016). This said, a higher number of participants might help to confirm the kinematic results. Also, a direct measurement of the trunk motion would be helpful to identify more precisely its contribution to the altered kinematic parameters found for the uninjured limb as compared to controls.