Grasping objects by former amputees: The visuo-motor control of allografted hands

2.1 Subjects

Five allotransplantation recipients (four men and one woman, mean age 33) and seven healthy right-handed subjects (six men and one women, mean age 32) were included in the experiment. Patients and control subjects were matched in age, gender and upper limb morphology (length from acromion to middle finger tip). The kinematic study and the clinical examination were part of the systematic patients’ post-transplantation follow-up. The protocol was approved by the local ethics board, and all participants gave informed consent before participating in the study that was conducted in accordance with the Declaration of Helsinki. All patients had suffered a traumatic bilateral hand amputation that occurred between the age of sixteen and twenty-nine. The bilateral transplantation was performed in a single surgical session that occurred from two and a half to five years after the amputation, resulting in a range from three to thirteen years prior to the study. For each patient, the surgical procedure included harvest of the two upper limbs of a brain-dead multi-organ donor after family’s consent for donation, preparation of the recipient’s forearm stumps (dissection, excision of pathologic tissues, identification of each tissular structure), union of grafts and recipient’s stumps via osteosynthesis of the ulna and radius (the appropriate length of the forearm bones was calculated with a formula based upon the length of the recipient’s humerus) and microsurgical anastomoses of arteries, veins, nerves, tendons or muscles (depending on the graft level), and finally skin closure (Gazarian et al., 2007; Hartzell et al., 2011; Herzberg, Weppe, Masson, Gueffier, & Erhard, 2008). The graft levels were at the middle or the distal third of the forearms in all patients. A lifelong immunosuppressive treatment is required to prevent graft rejection. According to the Edinburgh inventory (Oldfield, 1971), four patients were right-handed and one was ambidextrous before the amputation (the inventory was based upon patient’s recall from memory); after the transplantation, two of the previously right-handed patients had become left-handed, and the ambidextrous patient showed a slight preference for the right hand. This made impossible to find healthy subjects comparable in this respect, we thus chose right-handed healthy controls to at least match the initial hand preference of most of the patients. Each patient had altered active range of motion, motricity, strength, somatosensation and dexterity, resulting in daily life disabilities; one of them had an ankylosis of his right elbow as a consequence of his initial accident. All patients had experienced acute rejection episodes (see Table 1A and B for detailed anamnestic and clinical data of each patient).

2.2 Apparatus and procedure

Subjects sat facing a table, the trunk gently restrained against the seatback by a softened Velcro strap. They wore liquid-crystal goggles (PLATO, Translucent Technologies Inc.) with opening (allowing vision) and closing (preventing vision) mechanisms controlled by the experimenter via custom-made software. The two hands of each subject were assessed separately. In the “start” position, the hand rested on the table with the thumb and the index fingers held in pinch-grip position on a release-switch, positioned about 15 cm from the subject’s trunk along the sagittal axis. The objects used as targets for prehension movements were plastic cylinders (12 cm high) of three different diameters: 1.2 cm (small object), 3.1 cm (medium object) and 5.1 cm (large object). They were located on the table, 30 cm away from the starting point, at 45° of subject’s sagittal axis either on the ipsilateral or contralateral side with respect to the tested hand (see Fig. 1). The objects were presented one at a time at one of the two positions in a pseudo-randomized order (no more than two consecutive repetitions of the same condition). The goggles were closed before the beginning of each trial to preclude vision of the target object before the go signal. The go signal, a beep sound, was synchronized with the goggles opening. Subjects were instructed to rapidly and accurately reach and grasp to lift the object upon the go signal. In the Vision condition (with visual feedback), the goggles remained open during the entire movement; in the No-Vision condition (without visual feedback), the goggles were closed when the subject’s hand left the start position (releasing of the switch beneath it). This was aimed at preventing the visual online control of movement execution. Taken together, these four factors formed a set of 24 conditions (2 visual conditions×2 hands×2 object positions×3 object sizes). Eight trials were performed per condition for a total of 192 trials, divided in four blocks of 48 trials (one block per hand and per visual condition). Each block was preceded by a practice session during which participants grasped each object in each position once. If the participant failed to grasp the object, the trial was repeated at the end of the block. For each hand, the Vision condition block was recorded before the No-Vision block. To reduce participants’ fatigue, we allowed ample rest time between each block.

2.3 Movement recording and data processing

Upper limb movements were recorded using an optoelectronic VICON MX GIGANET system, composed of infrared stroboscopes and 200 Hz infrared cameras. Four passive infrared reflecting markers were placed upon the nails of the thumb and the index, the styloid process of the radius at the wrist, and the lateral epicondyle at the elbow. After recording and tri-dimensional reconstruction, the position data of each marker were filtered with a Butterworth filter, with a cut-off frequency of 6 Hz. The time-courses of wrist velocity, wrist acceleration, thumb-index grip aperture and grip aperture velocity were derived from these data. The movement parameters indicated below were determined on these profiles by a semi-automatic procedure with trial by trial manual verification.

2.4 Movement and kinematics parameters

The movement was divided in four components. The Reaction time component took place before the movement onset. The Transport, Grasp and Movement time components took place from the onset to the end of the movement. The onset of movement was defined as the first point of the first sequence of at least eleven consecutively ascending point, as derived from the wrist velocity curve. The movement end was determined visually on the grip aperture curve as the point of achievement of a stable grip on the object, before it was lifted.

2.5 Statistics

Trials affected by a loss of markers signal were excluded from the analysis (patients: 3.5% of trials; controls: 0.3% of trials). Reaction times could not be recorded in one patient and one control subject due to technical reasons. One block (Left hand, No-Vision) could not be performed in two patients, one for technical reasons, the other because the patient was unable to grasp the object in this condition. These patients were excluded from the analysis only with respect to the missing data block.

To calculate the participants’ success rate, we took in account only the first 48 trials of each block (8 trials per condition, 6 conditions per block), and not the trials repeated at the end of the block. The proportion of successful trials in each subject has been computed and used as observation in order to compare the two groups by means of a nonparametric permutation test. For the two patients with missing data regarding the Left hand No-vision block, this condition was excluded of the analysis.

A nonparametric multivariate analysis of variance (MANOVA) was conducted to compare the two groups (i.e. between-subject factor) and the effects of the four within-subject factors (Vision, Hand, Object Position and Size) on the movement components (in a global account of all parameters for each component). To take into account the intra subjects variability, a random effect model was defined (Basso & Finos, 2012; Finos & Basso, 2014) and a permutation-based approach was used to assess significance of tested hypotheses (Fisher combining function, 1000 random permutations). Only factors with significant (p≤0.05) effect were considered for parameters level (i.e. univariate) analysis. At the parameters level, a between groups analysis (patients, controls) and a complementary within group analysis were performed for each parameter and factor, using univariate nonparametric analysis of variance (ANOVA) as above.

The differences of intra-subjects variance of parameters and correlations between parameters were compared between groups. To this aim, the empirical within-subject variance and correlation within each subject was used as observation in a nonparametric permutation-based ANOVA test.

The correlations between clinical variables and movement parameters were analyzed for transplanted subjects with a Pearson correlation test evaluating the significance nonparametrically, by permutations. The assessed clinical variables were the patient age at the time of the amputation, the time between amputation and allograft, the duration of the surgical procedure, the ischemia duration of the transplanted limb during the surgery, the donor age, the time since allograft, the number of acute rejection episodes, the total active range of motion of the five fingers, the grip strength, the key pinch strength, the somatosensory performance in both the index fingertip in the Semmes-Weinstein monofilaments test and the two points discrimination test, and the dexterity performance in both the unimanual Purdue Pegboard Test and the Box and Block Test. Coherently with previous analyses, we – nonparametrically - combined the inference of all the correlations among clinical variables and movement parameters related to Transport phase, in such a way that a single p-value is provided to test the overall correlation between clinical variables and Transport phase. An analogous procedure is performed for the Grasping phase.

The significance level was set at p≤0.05 for all analyses.

3.1 Global performance

When subjects failed to achieve a stable grip or knocked down the object while trying to grasp it, the trial was characterized as unsuccessful. Videos of patients performing successful and unsuccessful trials are available in supplementary materials. Considering the first 48 trials of each performed block (i.e., excluding repeated trials at the end of the block), patients grasped successfully (i.e., grasped and lifted) the object in 96.6% (SD 2.4) of trials with visual feedback and 91.1% (SD 4.4) of trials without visual feedback. Controls succeeded in 100% (SD 0) and 95.7% (SD 3.1) of trials, respectively, bearing no significant differences between groups (vision: t = 2.71, p = 0.99; no-vision: t = 2.67, p = 0.98).

3.2 Group effect

The MANOVA revealed that controls and patients differed with respect to all movement components: the Reaction time (p = 0.007), the Transport component(p≤0.001), the Grasp component (p≤0.001) and the Movement time (p≤0.001). With respect to controls, patients exhibited longer reaction times and, after a higher number of acceleration peaks, they reached a reduced and delayed acceleration peak, as well as reduced and delayed velocity and deceleration peaks (see Table 2, Figs. 2, 3). The patients’ grasping component was further characterized by a reduced MVGA that, in turn, delayed a substantially reduced MGA (see Table 2, Figs. 3, 4). Finally, patients needed more time to perform the movement (see Table 2, Fig. 3). Noteworthy, the additional time cost was major in the final phase of the movement, from the maximal grip aperture to the achievement of a stable grip on the object, whereas the delay cumulated during the earlier movement phase (until the deceleration peak) was less important although significant (approximately 130 ms at the deceleration peak and 650 ms at the end of the movement) (see Fig. 3). Patients’ individual data aredisplayed in Figure S1 (see in supplementary materials).

3.3 Vision effect

The presence or absence of visual feedback during movement execution similarly impacted controls and patients movement components (no differencebetween patients and control subjects in the MANOVA). All in all, allografted patients were not additionally impaired when performing a prehension movement in the absence of visual feedback, with the exception of one patient who was unable to accomplish the task in this condition with his left hand, but succeeded with his right hand.

3.4 Hand effect

The use of either the right or left hand had no differential influence between groups as revealed by the MANOVA for all movement components.

3.5 Object position effect

The lateral position of the object did not interact with the factor Group in three of the four movement components; indeed only the Transport component was differently affected in patients and controls (p≤0.001). Patients tended to symmetrically reach toward ipsi and contra-lateral targets. By contrast, object position considerably affected controls’ Transport component: contralateral targets triggered more numerous and smaller acceleration peaks, and produced velocity and deceleration peaks of increased latencies and amplitudes with respect to ipsilateral targets (see Table 3A).

3.6 Object size effect

The object size affected the Transport component in controls and patients alike, but the other movement components revealed group differences.

Patients tended to exhibit longer reaction time for movements directed towards the small target compared to the medium target, whereas controls did not display such a trend (small versus medium: MANOVA p = 0.018, see Table 3B for details).

In the Grasp component, patients differed from controls when asked to grasp the largest object (interactions between groups and object size in the omnibus MANOVA for small versus large: p≤0.001 and medium versus large: p≤0.001). Both groups displayed the classical grip scaling as a function of object size, however the scaling was markedly reduced in patients as further indicated by the marginal significance of small versus medium comparison (see Table 3B, Fig. 4). In addition, the patients’ MGA latency was largely modulated by object size, with a striking increase for the largest object, whereas in controls it remained immune to the differently sized objects (see Table 3B, Fig. 4).

Object size also impacted Movement time differentially in controls and patients (significant interaction in the MANOVA between groups and the large size of the object compared with the small size: p = 0.006 and the medium size: p = 0.012). Allografted patients needed more time to fully achieve the prehension of the large object compared to the smaller ones. In sharp contrast, movement time remained constant across the different object sizes for the control group (see Table 3B, Fig. 4).

3.7 Correlations between movement parameters

To assess the extent to which the dynamical structure of patients’ movements was preserved, we ran correlation analyses between movement parameters, separately for each group. The groups did not show significant differences in the correlations between the transport parameters. The coupling between thetransport and the grasp parameters was altered in the patients group, as evidenced by significantly reduced correlations than in the controls group (acceleration, velocity and deceleration peaks latencies vs grip aperture onset latency and MGA latency; velocity peak vs MGA), and conversely, higher correlations than controls (acceleration and velocity peaks latencies vs MVGA) (see Table 4 for details). It is noteworthy that the temporal coupling of the grasp parameters among themselves was perturbed in allografted patients, as revealed by a reduced correlation than in controls between the grip aperture onset latency and the MGA latency, as well as the coupling between the movement time and the earliest temporal parameters of the movement (lower correlations than controls with grip aperture onset latency and acceleration, velocity and deceleration peaks latencies) (see Table 4).

3.8 Variability

There was no difference in intra-subject variability between the two groups concerning the Reaction time component. Variability was significantly different for the Transport component (p = 0.025), the Grasp component (p≤0.001) and the Movement time (p = 0.021): patients had a larger intra-subject variability than the control subjects for the deceleration peak latency (p = 0.010), the MVGA latency (p = 0.036), the MGA latency (p≤0.001) and the movement time (p = 0.016). On the contrary, they were less variable than controls for the MGA (p = 0.042).

3.9 Correlations between patients’ clinical variables and movement parameters

Among the clinical variables tested for correlation with the movement components, two are worth reporting: the ischemia duration of the transplanted hand during the surgical procedure and the key pinch strength.

There was a near to significant correlation between the ischemia duration and the transport component taken as a whole (r = 17.5, p = 0.069). At the level of the single parameters of the transport component, we observed that an increased number of acceleration peaks (r = 0.96, p = 0.025), delayed latencies of velocity (r = 0.96, p = 0.024) and deceleration (r = 0.93, p = 0,044) peaks were significantly correlated with a longer ischemia duration.

The key pinch strength also tended to be correlated with the grasp component (r = 9.36, p = 0.052) and the movement time component (r = 2.41, p = 0.089). At the single parameter level, a better key pinch strength was significantly correlated with a later grip aperture onset (r = 0.94, p = 0.027), and tended to correlate with a larger MGA (r = 0.96, p = 0.089) and a shorter movement time (r = – 0.98, p = 0.089). Finally, the correlations between a stronger key pinch and higher acceleration (r = 0.97, p = 0.027) and velocity peaks (r = 0.94, p = 0.028) were significant.

4.1 An impressive recovery of prehension movement in allografted patients

First of all, in about 95% of cases, patients who had been deprived of hands for several years have successfully grasped and lifted with each allografted hand differently sized cylinders, chosen to be representative of numerous objects of daily life like a can, or an AA battery. Such a result can be considered as a very positive functional outcome, first of all because these recovered abilities have positively influenced the patients’ daily life. All of them are independent for most of daily duties and they are active in a number of personal, social and professional activities (including a full-time cook job for one patient) and, compared to the pre-transplantation assessment their scores at the Disability of the Arm, Shoulder and Hand questionnaire had largely improved (Bernardon et al., 2015).

In addition to this qualitative evaluation, the kinematic approach highlights a substantial recovery of some of the critical features of the visuo-motor control. Among them, the preserved scaling of fingers’ maximal grip aperture according to the object size is noticeable as it stands as a fundamental property of the visuomotor control (Castiello, 2005; Jeannerod, 1981, 1984).

Moreover and remarkably, allografted patients were not more perturbed than controls by the suppression of the visual feedback during visuo-motor grasping. In the No-Vision condition, participants had to rely on their motor program, elaborated before movement onset, and on proprioceptive-dependent online control. Our findings underline two major aspects of patients’ recovery: they had the ability to produce an effective motor program, and they could rely on proprioceptive information of remarkably good quality. The only exception was one patient who failed to perform the task with his left hand but succeed with the right, suggesting that he did not rely solely on his motor plan. The discriminative touch performance of his left hand was among the poorest of the group, even if his proprioception was correct (see Table 1B, patient 1). This overall well preserved performance is far from obvious, since the somatosensory deficit clinically diagnosed in these patients would have rather led to the opposite prediction. Indeed, an important alteration of the prehension performance in the absence of visual feedback has been previously reported in hand replanted patients (Paysant et al., 2004). The contrast between the preserved performance of allografted patients newly reported here and the poor performance of replanted patients in a similar no-visual-feedback grasping task, could derive from milder proprioceptive impairments in allotransplantation recipients. Seventy-five percent of the replanted patients in Paysant et al’s study (eight right-handed patients replanted of their left hand, three to thirteen years prior to their study) had an altered sense of joint position. On the contrary, the allografted patients investigated here presented with mild somatosensory deficits that essentially concerned the superficial sensitivity (tactile sensitivity, spatial discrimination), whereas the discrimination of joint position under passive displacement of the wrist, index, thumb and little finger almost normal in most of them. The present kinematic findings also suggest that such a recovery of proprioception extends beyond passive sensation, to the dynamic kinaesthesic component of proprioception, likely contributing to the remarkable recovery of visuomotor control reported here. The use of Tacrolimus, a well-known molecule for enhancing nerve regeneration besides its immunosuppressive effect (Glaus, Johnson, & Mackinnon, 2011; Kuffler, 2009; Martin et al., 2005; Toll, Seifalian, & Birchall, 2011), in the immunosuppressive regimen of allografted patients could have contributed to this good proprioceptive recovery afterallotransplantation.

Finally, although the early parameters (acceleration, velocity) of the prehensile movements were slightly, but significantly altered compared to healthy controls, they were still correlated between them, similarly to those observed in healthy controls. In addition, other parameters (the latencies of grip aperture onset and MVGA) were fully comparable across groups. Together with the well preserved performance when acting in absence of visual feedback, these findings speak in favor of an effective visuomotor planning in these allotransplanted patients: although each movement parameter is regulated by both motor planning and online motor control, it has indeed been suggested that the early movement parameters depend more specifically on motor planning, whereas the later movement parameters rely predominantly on the online motor control (Desmurget & Grafton, 2000; Haaland, Prestopnik, Knight, & Lee, 2004). In this respect, the initial stage of the movement of allografted patients and thereby their motor program seems largelyspared.

4.2 A residual central motor deficit?

Despite the performance in the no vision condition may speak in favor of intact motor programming capabilities in hand allografted patients, two findings raise the question of some subtle impairments in these capabilities: the lengthening of the reaction time and the increased number of acceleration peaks. Reaction time is known to depend largely on motor planning (Sternberg, Monsell, Knoll, & Wright, 1978), and the altered pattern of the wrist acceleration – more numerous acceleration peaks – displaying in patients might also be ascribed to motor program alterations, as early parameters would in principle not benefit from feedback corrections (Desmurget & Grafton, 2000; Haaland et al., 2004). When tested in comparable conditions of object position and size, a similar perturbation of the acceleration phase has been reported in a patient suffering from optic ataxia (Roy, Stefanini, Pavesi, & Gentilucci, 2004), a sensory-motor deficit following damage of the posterior parietal cortex (Pisella, Binkofski, Lasek, Toni, & Rossetti, 2006; Pisella et al., 2009). Allograft recipients are certainly not to be compared to brain-damaged patients, yet both amputation and graft markedly impact the cerebral sensory-motor re-organization. As recalled in the introduction, after limb amputation sensory and motor cortical maps undergo plastic modifications induced by the deprivation of their sensory afferent inputs and motor effectors. The impact of these widespread cortical reorganization on the planning of visuomotor control remains unclear: in chronic upper limb amputees, for example, Philip and Frey (Philip & Frey, 2011) have reported preserved abilities in a grip selection task, whereas Metzger and colleagues (Metzger et al., 2010) have found abnormal feedforward control strategies in a reaching task. If one assumes that visuomotor planning could be at least partially altered by amputation induced cortical reorganization, then its recovery after allografts may require some further plastic adaptations. Moreover, the new hand does not have morphological and biomechanical features identical to those of the native hand, and the re-innervation is a relatively random process, which does not reproduce the original pattern. Thereby, hand allotransplantation is not the mere return in place of a terminal effector, but the implementation of a new one, which might require at least partial adaptation of motor planning, instead of the simple reactivation of the pre-amputation one. Taken together, the massive reorganization undergone by motor and sensory networks could well result in a less efficient motor planning, and possibly contribute to the delayed reaction time and global movement lengthening reported here. However, additional arguments may imply alternative explanations that we consider in the following section.

4.3 Persistent invalidating peripheral limitations

The lengthened reaction time and movement duration could alternatively reflect execution difficulties due to peripheral neuro-orthopedic sequelae. The task being more demanding in terms of execution for patients than for controls, patients could indeed need more time to plan it (Fitts & Peterson, 1964; Harrington & Haaland, 1987; Sternberg et al., 1978) (Fitts and Peterson, 1964; Sternberg et al., 1978; Harrington and Haaland, 1987). As reported in Table 1B, all patients experienced mild somatosensory and motor deficits due to the imperfect reinnervation after the graft, as well as important limitations of the active range of motion of the wrist and the fingers, secondary to tendinous lesions (adhesions, ruptures), muscular and articular lesions (fatty degenerations, fibrosis, retractions) and surgical constraints (strong tension imposed to the flexor system in order to preserve grip strength, tendinous grouping resulting in no or few dissociation of the digits movement, etc.). This “peripheral” account for patients’ difficulties is supported by several arguments when considering patients’ grasping kinematics.

Despite a preserved grip scaling, patients showed difficulties in pre-shaping their hands, particularly while approaching the largest object. Interestingly, following central lesions, patients with visuo-motor transformation deficit most typically show the opposite problem, consisting in an over extended grip aperture (Jeannerod, Decety, & Michel, 1994; Milner et al., 2001). Indeed, increase in movement safety requirements (and/or accuracy), e.g. because of absence of visual or somatosensory feedback, generally leads to increase of the maximal grip aperture to provide a greater safety margin (Chieffi & Gentilucci, 1993; Gentilucci, Toni, Chieffi, & Pavesi, 1994; Gentilucci, Toni, Daprati, & Gangitano, 1997; Jakobson & Goodale, 1991). Here, the general slowness exhibited by patients, their significantly higher intra-subject variability for numerous kinematic patterns, would have predicted an increased grip aperture. In sharp contrast, not only their maximum grip size was noticeably reduced (though delayed) as compared to healthy controls, but also their intra-subject variability of the very same parameter was significantly reduced. Patients were seemingly constrained to take smaller security margins than controls when grasping objects. In healthy subjects, object size may have no effect on movement time (Jeannerod, 1981, 1984; Paulignan, Frak, Toni, & Jeannerod, 1997), as we report here. Alternatively, increasing object size may result in shortening of the movement time, as the accuracy demands decrease due to a larger surface area available for contact (Chieffi & Gentilucci, 1993). Distinctively, allografted patients exhibited the opposite effect, the larger object giving rise to longer movement times. Wider fingers opening was accompanied by longer opening duration in patients: the 2 cm increase in diameter between the medium and the large object delayed the grip aperture latency by 100 ms, without having cost in controls. Overall, these findings are in favor of a role played by mechanical constraints: patients almost reached their widest possible grip aperture when required to grasp the larger object. In sum, patients exhibited a marked difficulty in opening their grip to ensure the safety margin while grasping the large object, which was hence the more demanding condition. In addition, grasping an object located in the contralateral or ipsi-lateral side does not require the same articular displacements, both at proximal and distal levels. Side-induced modulations of the transport parameters were present in controls, but absent in hand allotransplantation recipients where the biomechanical constraints may be responsible for their altered pattern of prehension.

The peripheral hypothesis finds additional support in the pattern of correlations. The correlations between the different kinematic parameters highlight a profound alteration of the coupling between the Transport and the Grasp components (e.g. the latencies of the transport were loosely correlated with the MGA latency in patients due to a delayed MGA, whereas the controls’ movements display a tighter link). This perturbed synchronization may largely depend upon the patients’ reduced grasping skills, as also emphasized by the reduced temporal coupling of the grasp parameters among themselves. As long as the transport parameters are considered, patients and controls displayed no difference in correlation parameters, though both the wrist and the hand were part of the allotransplantation in all patients. Moreover, the lack of correlations between the movement time and the earlier temporal parameters in patients highlight the major role played by the final phase of the movement in the patients’ difficulties. This phase needs higher precision and online control as it corresponds to the closing and fine adjusting of the fingers configuration to the object in order to achieve a stable grip, and was strikingly prolonged in patients (see Fig. 3). This alteration of the final stage of the movement, important enough to provoke a partial temporal disorganization of the movement, may be explained by a lack of dexterity that could follow from tactile sensitivity deficit, impaired motor capabilities and reduced range of motion of the grafted hands, which perturbed the fingersbiomechanics.

The correlations between the clinical parameters and movement kinematics are also coherent with peripheral limitations. Indeed, we did not find correlations between the kinematics and the time elapsed between amputation and allograft, or the time since allograft, clinical parameters that index the time available for cortical reorganization to take place. On the contrary, our results suggest an association between the time of ischemia, a major source of tissue damage, and the movement performances.

4.4 Limitations

Our study is somewhat intrinsically limited by the small number of patients and their heterogeneity. The sample size was overall constrained by the absolute number of bi-transplanted patients: at the time of the testing, only 25 patients worldwide have received a bilateral hand transplantation, including six French patients. So despite the fact that this study includes almost all of the French bilateral hand transplanted patients, the number in absolute terms remain small. Despite the corollary limitations of a potentially higher variability, it should be underlined that patients exhibited a quite similar pattern of results when considering most of the key kinematic variables (see Fig. S1 in supplementary materials). Furthermore the statistics were specially designed to ensure the best possible power and sensitivity. However, our correlation analysis between clinical and kinematic parameters was also limited by the small sample size. Future studies including more recently transplanted patients will help elucidating the links between prehension performance and clinical and anamnestic data.