Multidimensional evaluation of a two-level-of-compliance quasi-passive prosthetic wrist: functional, biomechanical, cognitive and user-experience outcomes in trans-radial limb loss

Abstract

Background

Advances in upper limb prosthetics have expanded functional capabilities for individuals with trans-radial limb loss, yet compensatory movements remain common due to mobility limitations of current prosthetic devices. This study evaluated the functional performance of a quasi-passive prosthetic wrist with two degrees of freedom and adjustable compliance operating according to the Synchronous Hand-Wrist Operation principle (compliant during reaching and automatically stiffening at grasp onset), focusing on clinical function, compensatory strategies, cognitive workload and user experience.

Methods

Six male prosthetic users (53 ± 13 years) completed two experimental sessions on consecutive days. Session one included familiarization and clinical evaluation using the Southampton Hand Assessment Procedure (SHAP). In session two, compensatory strategies were assessed during six standardized manual tasks using motion capture (Vicon, UK) and compared against age- and sex-matched healthy controls. Prosthetic users performed all procedures with the device in compliant-while-reaching (ON) and full-stiff modes (OFF) in randomized order. Clinical, spatio-temporal and kinematic outcomes were analyzed using linear mixed-effect models. Subjective evaluation included the NASA-Task Load Index (NASA-TLX), System Usability Scale (SUS) and a custom questionnaire, assessing the overall experience with the device in its intended compliant operating mode.

Results

In the SHAP, the ON condition yielded significantly lower mean hand functionality scores than OFF (55.0 vs. 58.3, ~ 5% decrease; p = 0.034) and longer completion times in 4/26 items (+ 1.4 s on average, + 18–44%; p < 0.05). Biomechanical analysis showed increased reaching phase duration in ON across all tasks, with minimal increases in total hand trajectory. Task-specific kinematic adaptations were evident: when wrist deviation during grasping occurred in ON, reduced compensations were observed at more proximal joints, bringing joint ranges of motion closer to healthy controls. Additional motor strategies were required, particularly in tasks involving surface interaction or complex hand-object alignment. Subjective evaluations indicated moderate cognitive workload (NASA-TLX: 48/100), above-average usability (SUS: 74/100) and overall positive acceptance.

Conclusions

A multi-DoF quasi-passive prosthetic wrist can reduce proximal compensations, although introducing longer execution times and novel motor demands. Functional benefits are strongly task-dependent. These findings stress the importance of balancing added mechanical complexity with usability to optimize clinical outcomes and user satisfaction.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12984-026-02007-5.

Background

Individuals with trans-radial limb loss face a health condition of complete absence of hand and wrist movements, significantly compromising their ability to independently perform daily activities involving the upper limb, especially those requiring fine motor control. Although advances in motor control and functionality of upper limb prostheses in recent decades have significantly expanded the range of actions that users can perform [1–3], prosthesis users often rely on compensatory motor strategies to adapt to the functional and physical limitations of current prostheses [4, 5]. In particular, the restricted distal mobility, typical of prostheses with fixed wrists, alters upper limb kinematics during manual tasks, leading to increased involvement of proximal joints (shoulder, elbow, and spine) compared to physiological movement patterns [4, 6–9]. As a result, additional work and mechanical load place a strain on the musculoskeletal structure of the upper limb, likely resulting in an increased risk of injury, pain and disuse in the long term [10, 11].

To address these limitations, research and development in the field of upper limb prosthesis components has progressively introduced solutions to specifically restore part of the lost wrist mobility [2, 12, 13]. In terms of design approaches, prosthetic wrists can be broadly classified into active and passive systems [14]. Active devices use motorized actuation to control hand orientation, whereas passive mechanisms rely on manual repositioning and locking. Most existing prosthetic wrist devices are directly controlled by the user and typically provide pronation–supination, flexion–extension or a combination of these movements. These degrees of freedom (DoF) are essential, as they have been shown to significantly enhance performance in activities of daily living [15], improve overall usability of the prosthesis and reduce compensatory movements [4, 16]. An alternative concept is represented by compliant wrists, which passively respond to environmental interactions without requiring additional control input from the user, enabling indirect adjustment of hand orientation [13]. In this framework, recent studies have highlighted that the effectiveness of prosthetic wrist compliance is task-dependent, as it varies across different phases of movement [17]. Specifically, compliant wrists have been shown to perform better during the reaching phase, while stiffer configurations result in more natural arm and trunk kinematics during object manipulation. These findings have driven the exploration of prosthetic solutions with adjustable or switchable compliance, aiming to balance optimal wrist orientation with minimal user effort.

Despite these technological advancements, the body of scientific literature assessing the functional impact of mobile wrist prostheses, particularly in individuals with trans-radial limb loss, remains limited. Notably, most of the studies have been conducted using prototype devices during varying manual tasks and primarily involved healthy participants wearing simulated prostheses [12, 17–19], which hinders the comparison of findings between studies and may not accurately reflect the experiences and movement patterns of actual prosthesis users. When studies include individuals with trans-radial limb loss, the findings are mixed. For instance, Deijs et al. [20] found that, while users tend to prefer flexible passive wrists, there was no clear evidence of compensatory motion reduction. In contrast, a recent study by Olsen et al. [12] reported a significant reduction in compensatory thorax forward leaning and lateral bending when participants used a powered, actively controlled wrist during a standardized clinical test. These discrepancies likely stem from differences in study design, sample size, task complexity, and prosthesis type. To address these inconsistencies and gain a comprehensive understanding of prosthetic wrist function, adopting a multi-factorial framework in the evaluation of new prosthetic technologies is fundamental. Such an approach should integrate diverse assessment domains, including usability, cognitive workload, ecological and laboratory-based use, clinical outcome measures as well as task-specific performance metrics. This holistic characterization is essential as a preliminary step before recommending extended home use, ensuring the device is both functionally effective and aligned with users’ daily needs.

Overall, while the capacity of mobile prosthetic wrists to reduce proximal compensations and improve the naturalness of the gesture remains to be clarified, their increased mechanical complexity raises additional important concerns regarding usability, cost-effectiveness, maintenance and repair, all factors that are particularly relevant for everyday prosthesis users [21–23]. To balance the need for increased wrist mobility with the goals of simplicity and low cost, a prototype of a quasi-passive flexible prosthetic wrist with two levels of compliance was recently developed to offer both stability and adaptability during manual tasks [24]. The device is built around a 2-DoF spherical joint coupled with a compression spring and a stepper-motor-based locking mechanism. This architecture enables two discrete operating modes. In compliant mode, the joint allows passive, multi-directional wrist movement (i.e., flexion/extension and radial/ulnar deviation) when the device interacts with a solid surface, facilitating natural wrist positioning prior to grasp. In the stiff mode, the locking mechanism limits wrist motion to avoid unwanted deviations induced by the weight of the grasped object or the prosthetic hand during lifting tasks. Critically, no additional control input from the user is required to switch between modes: transitions occur automatically and synchronously with the prosthetic hand opening and closing commands, following the Synchronous Hand-Wrist Operation principle [24]. Specifically, when the hand initiates a closing movement (grasp onset), the wrist is driven into stiff mode; conversely, when the hand opens, the wrist returns to compliant mode. In this way, the wrist is compliant during the reaching phase, when passive deflection against environmental constraints can facilitate optimal hand orientation, and stiff during the manipulation phase, ensuring safe and predictable handling of objects (Fig. 1). This design aims to enhance wrist functionality while maintaining mechanical simplicity and affordability, making it a promising solution for everyday prosthetic use.

Fig. 1.

Schematic representation of the prosthetic wrist functioning via the Synchronous Hand-Wrist Operation principle

In this regard, usability and cognitive workload are particularly relevant outcome dimensions in the context of multi-DoF prosthetic systems, whose added mechanical complexity may increase the attentional and cognitive demands placed on the user [25, 26]. Indeed, controlling an upper limb prosthesis is inherently cognitively demanding, and mental workload has been recognized as a critical criterion for device acceptance since the earliest myoelectric systems [25]. When excessive, this load may prevent full exploitation of the device’s capabilities and ultimately lead to abandonment [25, 26], a risk that has been reported to be linked to discomfort, limited function and difficulty of use [27]. The Synchronous Hand-Wrist Operation principle adopted within the prototype development was explicitly designed to mitigate this concern by eliminating any additional control requirement for wrist switching; nevertheless, empirical evaluation of cognitive load remains essential to verify that this design goal is achieved in practice.

The primary aim of the present study is to evaluate the functional performance of the newly developed prosthetic wrist in individuals with trans-radial limb loss. To this end, prosthetic users were assessed under two experimental conditions: the device operating according to its design intent (compliant during reaching and automatically switching to stiff at grasp onset, i.e., ON condition) and the device locked permanently in stiff mode throughout all task phases, serving as a rigid baseline (OFF condition). Performance in both prosthetic conditions was additionally contrasted against a group of age- and sex-matched healthy controls to provide a reference for normative movement patterns. Specifically, the study assesses the wrist’s impact on upper limb function using both standardized clinical tests and selected manual tasks, focusing on joint motion, the presence and extent of compensatory motor strategies during task execution, perceived cognitive load, and overall user experience. By investigating these parameters, the study seeks to provide a comprehensive understanding of the prosthetic wrist’s usability, effectiveness, and potential benefits in real-world scenarios.

Methods

Study design and setting

This cross-sectional study was conducted at the Bioengineering and Human Performance Laboratory of the Italian Defense Institute of Biomedical Sciences. Each participant underwent two experimental sessions on two consecutive days. The first session focused on familiarization and clinical assessments with prosthetic devices, while the second session aimed to evaluate compensatory motor strategies and coordination during instrumented functional tasks. At the end of the first session, an additional real-world scenario test was proposed to let the participant experience the prosthesis functioning in a less structured and constrained context. The workload, device usability, and user experience were assessed through standard and ad-hoc questionnaires at the end of the protocol, with participants being instructed to refer to their experience with the prosthetic system, with particular attention to its compliant operating mode, which represents the device’s intended configuration.

Participants

Seven prosthetic users with unilateral trans-radial amputation were recruited for the study between May and October 2024. Participants were recruited through the Italian Defence Veterans Centre and other relevant institutions via flyers and advertisements. Inclusion criteria required unilateral upper limb amputation at the trans-radial level, with proficiency in using a socket myoelectric prosthesis for at least one year. Exclusion criteria included the presence of concurrent general or localized comorbidities or disabilities that could interfere with the study; stump skin lesions affecting proper prosthesis use; the use of medications or substances that could impair sensory or motor functions; and any declared psychological or psychiatric comorbidities. In line with this, one of the seven individuals initially recruited was excluded after the clinical evaluation, which confirmed non-compliance with the eligibility criteria. Seven sex- and age-matched (± 5 years) healthy controls (HC) without any upper limb disabilities or impairments were recruited to provide reference data. All participants provided written informed consent for the experiments. Experimental procedures were approved by the institutional review board of Scuola Superiore Sant’Anna of Pisa (Approval No.10/2023).

Prosthetic wrist prototype

In each session, all participants were fitted with a prosthetic system consisting of a custom socket [28], the prosthetic wrist prototype, a commercial multi-fingered myoelectric hand (Mia Hand, Prensilia S.r.l., Italy), and a battery. The prosthetic wrist consists of one proximal and one distal frame linked through an external compression spring wrapping a spherical joint. This spherical joint allows a range of motion of ± 30° in wrist flexion/extension and radial/ulnar deviation, with a stepper-motor-based locking mechanism providing progressive locking positions at steps of 5°. The prototype has a total mass of 115 g, which is lighter than about 63% of currently available wrist units from 1 to 3 DoFs [13]. The prosthetic wrist was equipped with a commercial quick-disconnect flange (Quick Disconnect Wrist, Steeper, United Kingdom) for connecting with the multi-fingered myoelectric hand. Two surface EMG sensors (MyoBock 13E200, Ottobock, Germany) were positioned over the forearm muscles (wrist flexors and extensors) and used to control the opening and closing of the hand, as well as to switch the wrist mode from compliant to stiff. Specifically, the multi-fingered myoelectric hand processed signals from the EMG sensors using a state-of-the-art two-state amplitude-modulated control [29], which also governed wrist functionality following the Synchronous Hand-Wrist Operation principle. Under this principle, initiating a hand-closing movement automatically sets the wrist to stiff mode, while initiating a hand-opening movement switches the wrist to compliant mode. It is important to note that, when in compliant mode and without holding any object, the prosthetic wrist can only be bent through interaction with a solid surface, such as the table, prior to hand closing. This is because the spring stiffness is specifically calibrated to counteract the gravitational load of the prosthetic hand under typical arm postures, preventing unintended passive deflection during the reaching phase [24]. When the grasping phase begins and the hand starts to close, the wrist switches to stiff mode, and the locking pin mechanism engages to prevent further wrist deviations caused by the object’s weight during lifting. If the wrist is bent, the pin progressively engages with the peripheral steps until it reaches the central plughole, corresponding to the neutral wrist position.

Experimental protocol

The first experimental session (day one) focused on familiarization with the prosthetic apparatus and clinical assessment. After fitting the prosthetic system, the participants performed several object-picking and manipulation tasks, familiarizing themselves with different hand grip types (e.g., power, spherical, lateral grips) and wrist modalities. Then, a standardized clinical assessment, the Southampton Hand Assessment Procedure (SHAP) [30], was performed in the two experimental conditions ON and OFF in randomized order. Participants were seated with their elbows flexed at 90° on the surface of a height-adjusted table, according to the standard test procedures. Participants were then asked to undertake a real-world scenario test by assembling a chest of drawers (Trotten, IKEA, Sweden) to provide an unstandardized user experience. The assembly test was completed in both OFF (first trial) and ON (second trial) wrist conditions, with only the frame and a single drawer being assembled. Required tools and components were pre-arranged on the table, and instructions were clarified before starting each step as per the manufacturer’s instructions, with researchers intervening only if needed to complete a specific step.

The second experimental session (day two) focused on the instrumental evaluation of compensatory motor strategies during the execution of functional tasks. Participants performed six manual goal-oriented tasks selected from the SHAP at a self-selected pace while standing in front of a standard-height Table (80 cm) with their arms resting on it. To ensure consistency, the standing position was standardized across participants and trials by marking the feet position on the ground with tape. These tasks were selected to cover the most common daily grasps and procedures were slightly modified from the standard SHAP protocol to account for both wrist and multi-fingered hand functionality (Table 1). Participants performed five trials per task in one wrist modality before repeating them in the other, with modality order being randomized across participants. HC participants performed the same tasks with their dominant hand.

Table 1.

Description of manual tasks and grips

Task	Procedure	Requested grip type
PWR	Participants moved the heavyweight cylindrical object from a rear slot of the board to a front slot, both lined up in front of the person.	Power
SPH	Participants moved the heavyweight spherical object from a rear slot of the board to a front slot, both lined up in front of the person.	Spherical
GLASS	Participants grasped the glass jar containing 100 ml of water from a rear slot of the board positioned in front of the evaluated side of the person. They poured the water into the carafe located in a rear central slot and then returned the jar to initial position.	Power
CARD	Participants moved a thick metal card (i.e., the heavyweight extension object of SHAP) from a rear slot of the board, positioned directly in front of the person, to a front slot located centrally.	Lateral*
COIN	Participants picked up one coin placed at the edge of the table in front of the evaluated side and placed it into the glass jar placed centrally on the table.	Lateral
PAGE	Participants picked up a piece of cardboard positioned at the edge of the table on the opposite side of the evaluated arm** and flipped it as if turning the page of a book, placing it on the same side of evaluated arm.	Lateral**

Board and objects are part of the SHAP tools. * two participants performed the task with an extension grip due to difficulties in grasping the object; and for three participants the SHAP heavyweight extension object was positioned closer to the evaluated arm (~ 10 cm) due to difficulties experienced in securing a safe object grip for transfer. ** cardboard was positioned more centrally (~ 15 cm closer to the evaluated arm) in two participants due to difficulties in securing an optimal object grip for completing the flipping task

At the end of the instrumented session, mental workload, device usability and user experience were respectively evaluated through the self-reported questionnaire NASA Task Load Index (NASA-TLX), System Usability Scale (SUS) and a custom-designed questionnaire (QUEST). The NASA-TLX is a six-item scale widely used to evaluate the mental workload according to six types of cognitive dimensions (i.e., mental, physical, perceived performance, temporal, effort and frustration) during the use of assisting wearable devices [25, 31]. Each item is graded as a bipolar scale with a score between 0 and 100 (min to max level of cognitive demand), with the raw overall workload index being calculated as the arithmetic mean of the 6 items. The SUS is a ten-item Likert scale that provides a global measure of system satisfaction and sub-scales of usability and learnability [32]. To compute the total score, responses are first converted (i.e., subtracting 1 from scores of positively worded odd-numbered items and subtracting scores of negatively worded even-numbered items from 5), summed and then multiplied by 2.5, yielding a final score between 0 and 100, where higher values reflect better usability. The QUEST consisted of 20 items rated on Likert scales (1 = disagree; 5 = agree), including both positively and negatively worded statements addressing various aspects of the prosthetic wrist usability (Table 2). Furthermore, certain items included open-ended response options to capture more detailed user feedback (items 18–20).

Table 2.

List of the custom-designed questionnaire (QUES) items

Item n.	Question
1	It was difficult to control the opening and closing of the prosthesis during activities.
2	I feel I can make the most of the device.
3	I had no coordination problems in tasks that require both hands.
4	It was difficult to grasp objects.
5	I think I can use the device in various everyday contexts.
6	I had to think of different strategies to perform tasks without involving the prosthesis.
7	It was easy to manipulate objects.
8	I hesitated to use the device out of concern about mistakes I might make.
9	I frequently used the prosthesis to grasp objects without having to rely on my hand.
10	I managed to coordinate both hands well when using them together.
11	I felt as if the prosthesis were part of me, incorporated into my body rather than as a tool.
12	I could immediately tell whether an object was reachable and manipulable with the prosthesis.
13	It seemed as though the prosthesis moved more autonomously than under my voluntary control.
14	It was difficult to learn how to use the wrist.
15	The wrist facilitated the use of the prosthesis.
16	In flexible mode, I found the wrist too rigid.
17	I would like to have this type of wrist integrated into my prosthesis.
18	I believe the device presents usage risks regardless of the user.
19	I believe the device can be improved in many ways.
20	I see many more advantages than disadvantages in using this prosthetic wrist.

Biomechanical assessment

Upper body kinematics during instrumented manual tasks was captured using a 12-camera optoelectronic system (Vero cameras, Vicon Motion Systems Ltd., UK, 100 Hz sampling frequency) with 2 synchronized digital video cameras (Vue, Vicon Motion Systems Ltd., UK, 100 Hz sampling frequency). A biomechanical upper limb model was developed to accurately represent the kinematics of the relevant segments, particularly accounting for the specific movement and design characteristics of the prosthetic wrist [33]. This model is applied to the trajectories of 24 retroreflective passive markers placed on the skin and on the surface of the prosthetic wrist to obtain the kinematics of the thorax, clavicle, shoulder, elbow and hand segments. The reference frame of each segment was defined according to the International Society of Biomechanics recommendations for the upper limb [34]. The glenohumeral joint centre was estimated through a functional approach [35]. The following DoFs were extracted:

• Thorax: flexion/extension (FE), lateral bending (LB), and axial rotation (AR).

• Shoulder: FE, adduction/abduction (AA), and internal/external rotation (IER).

• Clavicle elevation.

• Elbow: FE.

• Wrist: FE and radial/ulnar deviation (RUD).

Data analysis

To evaluate clinical performance between the two wrist modalities, both the SHAP Index of Functionality (S-IOF) and the Prosthesis Index of Functionality (P-IOF) were calculated [30, 36]. While the S-IOF is based on the completion times of all 26 items, the P-IOF is an adapted version validated for use in upper limb prosthetic users that is calculated using the completion times of only 16 items of the SHAP procedure [36]. Both scores range between 0 and 100 points, with higher scores indicating better hand function. Additionally, individual item completion times from the full SHAP procedure were analyzed separately.

Regarding the biomechanical assessment, marker trajectories were filtered prior to analysis using a Woltring filter (smoothing factor = 10, Mean Square Error mode), as typically applied for noise reduction in kinematic data [37]. The data analysis in this study focused exclusively on the reaching phase, as this is when the wrist operates when in compliant mode (ON), whereas during the object transfer phase, it consistently reverts to stiff mode regardless of the selected wrist modality. To this aim, the movement start and object grip events were automatically identified in each trial by a double-threshold approach based on both the magnitude (15 mm/s) and direction of the vertical linear velocity of the hand marker. This approach was implemented through custom algorithms developed in Vicon ProCalc (version 1.6.1, Vicon Motion Systems Ltd., UK) and executed within the Vicon Nexus software environment (version 2.14, Vicon Motion Systems Ltd., UK). All detected events were further verified through visual inspection of synchronized video recordings.

To evaluate and describe the overall motor performance during the instrumented tasks, the reaching phase, defined as the movement phase between the beginning of the movement and the grasp event, was identified. Phase duration and the total path travelled by the hand during it were computed. Specifically, total hand path during the reaching phase was computed as a measure of the overall spatial efficiency of the reaching movement, that is whether the additional action of pressing the hand against an external surface to achieve wrist deviation resulted in a measurable increase in the total trajectory traveled by the hand as the end-effector of the upper limb.

To quantify the functional contribution of the wrist during object grasp, wrist FE and RUD deviation angles were extracted at the timepoint immediately preceding the grasp, providing information on the deviation of the wrist prosthesis prior to hand closure and object contact. Specifically, local minima in the vertical coordinate of two markers (one calibrated marker of the wrist ulna, and the other on the index finger’s dorsal surface mounted on the prosthetic shell) were automatically detected and the one closest to the final object grip event was selected. This procedure allowed to reduce false positives (i.e., repeated wrist movements during grasp attempts preceding the final object grip) that occurred in some cases due to precise hand placement requirements of the task, particularly in ON condition. The detection process proved to be robust across trials (success rate: 100% in HC, 97% in OFF and 83% in ON), with instant detection being finally validated through visual inspection of video recordings. Instantaneous wrist FE and RUD angles as well as angular position of the remaining DoFs were then extracted at this timepoint to characterize the adopted postural strategy for the functional adoption of the wrist prosthesis. It is specified that the prosthetic wrist has a neutral position of ~ 12° in wrist extension and ~ 4° in radial deviation due to the configuration of the robotic hand connection, which are retained in the biomechanical model and, thus, in the presented data.

The compensatory movements within the upper limb kinematic chain were evaluated in each task through the range of motion (ROM) for each considered DoF, defined as the difference between the maximum and the minimum value in the reaching phase.

Statistical analysis

A linear mixed effect model (LMEM) approach was used to assess differences in S-IOF, P-IOF, individual SHAP task completion times, spatial-temporal parameters and ROMs between the two experimental conditions (OFF, ON), with wrist condition as fixed factor and subject as random intercept. This approach was selected for its robustness with small sample size [38]. In addition, separate LMEMs were run to compare each condition against the HC group. For these comparisons, the random intercept was excluded, and the fixed factor was defined as group (HC vs. prosthetic users), with significant results being discussed only when significant differences emerged between the experimental prosthetic conditions. To verify the assumption of normality prior to model application, residuals were extracted and evaluated in terms of skewness and kurtosis. The extracted values were converted to z-scores and compared to the threshold of ± 1.96, corresponding to a significance level of 0.05. All the values fell within this range, thus indicating a normal distribution of the data. Model convergence was also tested for each comparison. For each LMEM, the estimated mean of the dependent variable was derived for each level of the fixed factor, while the fixed effect coefficient (β) quantified the magnitude and direction of the difference between conditions. Effect size was also estimated using the standardized fixed effect (SFE), which accounts for multiple sources of variance, including within- and between-subject variability, making it more suitable for linear mixed-effects models with complex variance structures. Effect sizes were considered small for |SFE| ≥ 0.2, medium for |SFE| ≥ 0.5, large for |SFE| ≥ 0.8 and very large for |SFE| ≥ 1.2 [39]. Statistical significance was set at 0.05 level and only significant differences greater than 5° are reported, in line with the minimal detectable change and standard error of measurement previously reported for optical 3D motion capture systems in upper limb kinematic assessment [40, 41]. All the analyses were conducted using Matlab software (version R2024a, The Mathworks Inc., MA, USA).

Results

Detailed information on demographics of prosthetic users and HC groups, as well as amputation characteristics, is provided in Table 3.

Table 4.

Mean (standard deviation) completion time of SHAP items

SHAP item/score	OFF	ON	p-value	SFE
Sphere (light) (s)	4.5 (1.2)	5.6 (1.1)	0.010	0.9
Tripod (light) (s)	5.0 (1.1)	5.9 (1.6)	0.048	0.6
Power (light) (s)	3.9 (0.6)	5.6 (1.1)	0.011	1.3
Lateral (light) (s)	5.9 (2.0)	5.5 (1.2)	0.366	0.2
Tip (light) (s)	5.7 (1.9)	6.2 (1.8)	0.441	0.3
Extension (light) (s)	5.0 (1.5)	6.8 (1.6)	0.012	1.1
Sphere (heavy) (s)	5.1 (2.3)	5.4 (0.8)	0.674	0.1
Tripod (heavy) (s)	5.4 (1.1)	6.4 (2.1)	0.240	0.5
Power (heavy) (s)	4.3 (0.9)	5.7 (2.0)	0.062	0.8
Lateral (heavy) (s)	5.4 (0.9)	6.0 (1.4)	0.137	0.5
Tip (heavy) (s)	5.6 (0.5)	5.7 (1.5)	0.827	0.1
Extension (heavy) (s)	5.5 (1.7)	7.4 (1.6)	0.088	0.8
Pick Up Coins (s)	53.6 (33.8)	44.9 (27.4)	0.282	0.2
Button Board (s)	48.4 (32.7)	44.6 (21.1)	0.797	0.1
Simulated Food Cutting (s)	22.8 (13.5)	17.9 (5.6)	0.279	0.1
Page Turning (s)	10.2 (9.3)	7.2 (3.2)	0.398	0.2
Jar Lid (s)	5.1 (1.3)	5.7 (2.2)	0.295	0.3
Glass Jug Pouring (s)	15.0 (2.5)	15.2 (2.3)	0.875	0.1
Carton Pouring (s)	16.2 (3.3)	15.3 (1.8)	0.632	0.2
Lifting a Heavy Object (s)	5.5 (1.0)	5.9 (1.6)	0.513	0.2
Lifting a Light Object (s)	4.6 (0.9)	4.9 (1.6)	0.632	0.2
Lifting a Tray (s)	7.1 (1.0)	7.2 (1.2)	0.956	< 0.1
Rotate Key (s)	8.3 (3.8)	6.9 (2.5)	0.096	< 0.1
Open/Close Zip (s)	10.2 (4.0)	10.1 (2.4)	0.928	< 0.1
Rotate a Screw (s)	13.9 (5.1)	15.7 (9.0)	0.493	< 0.1
Door Handle (s)	4.7 (1.4)	4.8 (1.1)	0.942	< 0.1
S-IOF	58.3 (9.4)	55.0 (7.8)	0.034	0.3
P-IOF	74.2 (4.7)	71.9 (4.1)	0.020	0.5

Significant results are reported in bold. S-IOF: SHAP Index of Functionality, P-IOF: Prosthetic Index of Functionality

Table 3.

Demographics and characteristics of prosthetic users and healthy controls (HC)

Group	Prosthetic users	HC
n	6	7
Age (yr, mean(SD))	53.2 (13.1)	54.4 (9.5)
Sex (n Male,%)	6 (100%)	7 (100%)
Etiology (%)	Acquired (100%)	-
Time from Amputation (yr, mean(SD))	27.8 (12.5)	-
Prosthesis/Dominant Side (n Right, %)	6 (100%)	6 (100%)
Stump Length (cm, mean(SD))	16.8 (3.6)	-
Myoelectric Prosthesis Use (yr, mean(min-max))	24.8 (5–41)	-

Clinical assessment

Prosthetic users completed 97% of the SHAP tasks in both the ON and OFF wrist conditions. One participant was unable to perform the “Rotate Key”, “Rotate a Screw” and “Simulated Food Cutting” tasks, the latter being failed by a second participant also. Table 4 shows the completion times for individual SHAP items, the S-IOF and P-IOF scores, and the related statistical results of between-condition comparisons obtained via LMEM. Significant differences with effect sizes ranging from medium to very large were reported in a few items (4/26), with ON showing greater completion times compared to OFF (average β: +1.4 s; p < 0.05). For overall hand functioning, both S-IOF and P-IOF showed a significant decrease in ON with respect to OFF (Table 4).

Biomechanical assessment

The functional use of the prosthetic wrist and the associated compensatory movements varied across tasks, showing both common patterns and task-specific differences. Figure 2 summarizes, for each task, the significant differences in ROM between OFF and ON conditions (magnitude, direction and effect size), the adopted wrist deviation strategy for object grasping (FE, RUD or both), whether these differences brought ON values closer to HC, and whether significant differences in reaching phase duration and hand path were observed. Detailed task-specific results for both spatio-temporal and kinematic outcomes are reported in Additional file 1.

Fig. 2.

Summary of the biomechanical assessment of compensatory movements. For each task, the table reports: 1) average significant estimated mean differences (β) in ranges of motion (ROM) of the thorax, shoulder and elbow when comparing OFF and ON conditions, with effect sizes represented by color coding and negative values indicating reduced ROM from OFF to ON; 2) the wrist deviation strategy adopted for object grasping in ON relative to OFF (flexion/extension, FE, and/or radial/ulnar deviation, RUD); 3) average significant estimated mean differences (β) in reaching phase time and total hand path between OFF and ON (similarly to ROM reporting); and 4) whether these changes brought ON values closer to HC (small HC icon; green = closer, red = farer)

Across all six tasks, reaching phase duration was consistently and significantly longer in ON than in OFF (β range: +0.85 to + 1.30 s; p < 0.012), with both prosthetic conditions showing greater durations than HC (β range: +0.75 to + 3.16 s; p < 0.01). Total hand path showed only task-specific differences between ON and OFF, being significantly longer in ON only during COIN (β = +3.2 cm, p = 0.015) and GLASS (β = +3.4 cm, p = 0.045), suggesting that the additional time required to exploit the compliant wrist in ON was generally not accompanied by a proportional increase in hand trajectory length.

Regarding the postural strategy adopted at the moment of wrist prosthesis deviation before grasping, the ON condition was systematically associated with greater wrist deviation than OFF across most tasks, involving radial deviation alone (COIN: β = +10°; PAGE: β = +11°; both p = 0.001) or wrist extension and radial deviation simultaneously (GLASS: β = −10° FE, β = +19° RUD; PWR: β = −12° FE, β = +23° RUD; SPH: β = −17° FE, β = +17° RUD; all p ≤ 0.001) (wrist strategies summarized in Fig. 2). This functional adoption of the wrist in ON was consistently accompanied by reduced flexion angles at more proximal joints at object grasp instant. Compared to OFF, ON showed lower elbow flexion in PAGE (β = −14°, p < 0.001), CARD (β = −10°, p = 0.045), COIN (β = −9°, p = 0.026) and GLASS (β = −17°, p < 0.001), and reduced shoulder and/or increased thorax flexion in GLASS (shoulder FE: β = −12°, p < 0.001; thorax FE: β = +9°, p = 0.002), CARD (shoulder FE: β = −13°, p = 0.009), PWR (shoulder FE: β = −16°, p < 0.001; thorax FE: β = +12°, p < 0.001) and SPH (shoulder FE: β = −11°, p = 0.022). When contrasted with HC, OFF consistently showed larger and more numerous significant postural instantaneous deviations than ON (see Additional file 1 for task-specific results). The most prominent example was PWR, where OFF differed from HC by β = -19° in thorax FE, β = +27° in shoulder FE, and β = +23° in elbow FE, whereas ON differed from HC only in shoulder FE (β = +11°, p = 0.044). CARD presented an intermediate and noteworthy pattern: despite no significant wrist deviation being detected in ON, reduced shoulder and elbow flexion postural angles were still observed compared to OFF.

Regarding the ROM during the overall reaching phase, ON showed consistent reductions compared to OFF in shoulder flexion/extension (GLASS: β = −12°; CARD: β = −13°; PWR: β = −16°; SPH: β = −14°, p ≤ 0.006) and in shoulder internal/external rotation (PWR: β = −10°; SPH: β = −13°, p ≤ 0.007), all with medium-to-large effect sizes (Fig. 2). In tasks where ROM reductions occurred, ON values were consistently intermediate between OFF and HC, with OFF showing the largest deviations. By contrast, PAGE and COIN showed no significant ROM reductions in ON relative to OFF; in particular, COIN showed a greater elbow FE ROM in ON than OFF (β = +6°, p = 0.005), indicating that the strategy required to lever the thumb against the table edge for wrist radial deviation introduced additional proximal motion rather than reducing it.

Self-reported outcomes

Figure 3 presents the results of the self-reported questionnaires. For overall mental workload assessed by the NASA-TLX, the group’s median raw score was 48 (95% CI: 25–70) out of a scale’s maximum of 100. Among the individual subscales, median raw scores ranged from low to moderate in the physical, temporal, performance, effort, and frustration dimensions (Fig. 3), with the highest value observed in the mental demand subscale (65; 95% CI: 36–94). It is important to note that considerable variability in NASA-TLX scores was identified across all subscales (Fig. 3); however, an exploratory qualitative analysis indicated no clear association between these scores (and those from other scales) and individual compensatory movement outcomes. For device usability, the SUS responses showed a group median total scale score of 73.8, with individual values ranging from 47.5 to 87.5 (maximum scale score: 100).

Fig. 3.

Results of self-reported outcomes on mental workload, system usability and user experience. Box plots of scores obtained from NASA-Task Load Index (NASA-TLX), System Usability Scale (SUS) and the 20-item custom questionnaire (QUEST). Discrete points represent the individual responses of each prosthetic user (color code similar to that in figures included within the Additional file 1)

Responses to the 20-item QUEST indicated generally positive user experiences, with items promoting device use receiving moderate to high scores (i.e., 3–5 points for positive items, 1–3 for negative items) across prosthetic users (Fig. 3). While certain items, particularly those concerning perceived grasping performance (e.g., “Difficulty grasping object”, “Used alternative strategies” and “Device could be improved”), elicited more diverse responses, several others were rated consistently across participants. Five participants provided open-ended responses related to the items “Device could be improved” and “More pros than cons of wrist”, highlighting the usefulness of wrist mobility and the need for improvements in object grasping and prosthesis size/weight, alongside a shared request for extended usage time to provide more informed evaluations.

Discussion

The aim of this study was to explore the specific contribution of a prosthetic wrist with two degrees of freedom and two levels of compliance to upper limb performance in individuals with trans-radial limb loss. By comparing the same participant under the compliant-while-reaching (ON) and full-stiff (OFF) modes wrist configuration, the study examined how wrist mobility influences compensatory movement strategies during the execution of standardized tasks, considering also the perceived cognitive load and user-reported experience, thereby providing further insights into the wrist’s functional role and added value in task performance. The main findings of the study indicated that the tested prototype could reduce compensatory motions in specific manual tasks, with an overall low mental load and positive user experience. However, when clinically assessed, the device demonstrated worsening performance when enabled, requiring additional time to perform manual actions.

In the clinical assessment using the SHAP, prosthetic users completed the same number of items under both wrist conditions, indicating that changes in the prosthetic wrist compliance do not hinder the ability to perform manual activities. Nonetheless, significantly lower functionality scores and longer task completion times were observed when enabling the compliant-while-reaching mode of the prosthesis (ON condition). This result is in contrast with Deijs et al. [20], who did not find a significant performance difference in the SHAP when comparing flexible and static wrist units after a two-week use period. The authors hypothesized that this finding was attributable to the limited sensitivity of the clinical measures (i.e., SHAP and Box and Blocks test) to detect changes in wrist functionality when movement speed is the primary outcome metric. In the present study, the lower performance in wrist compliant mode is likely due to the novelty of the system, which also required additional movement within specific tasks and therefore required a longer training period for its use to be efficiently integrated within functional tasks. As a result, the time required to complete the SHAP tasks was increased in the ON condition. However, due to the lack of established minimum clinically important difference thresholds for SHAP in prosthetic users, it remains unclear whether the observed changes are clinically meaningful. Notably, functionality scores and completion times in both wrist conditions remain within the performance ranges previously reported for individuals with trans-radial limb loss [20, 36, 42].

The biomechanical assessment of the selected manual tasks revealed that additional motor strategies are required to use the prosthetic wrist, which have distinct features according to the specific demands of each task. These adaptations were predominantly influenced by the object position on the table (edge vs. center), the hand alignment required to achieve the grasp and the portion of the prosthetic hand exploited as a lever to induce wrist deviation (e.g., the thumb in PAGE and COIN, or the ulnar edge of the prosthetic hand in other tasks). Overall, the adoption of the flexible prosthetic wrist (ON) increased the time required to reach and grasp the object regardless of the task, but with small-to-no changes in the total trajectory traveled by the hand. The motor strategy adopted in ON to induce wrist deviation mainly involved extending the elbow, as indicated by the reduced flexion elbow angle in most tasks (except SPH) compared to OFF. This additional movement was accompanied by more proximal angular changes that varied with the specific task demands, particularly in clavicle elevation (PAGE), shoulder motion (flexion: GLASS, CARD, PWR and SPH; internal rotation: COIN, PWR and SPH; adduction: GLASS and PWR) and thorax flexion (GLASS and PWR). Despite the additional movement required to exploit the prosthetic wrist’s flexible functionality, the analysis of ranges of motion suggests an overall reduction in compensatory movements, particularly in shoulder flexion/extension and when the prosthetic wrist is deviated in both its two DoFs, reporting ROM values closer to those of healthy controls (Fig. 2). Nevertheless, no clear reductions in compensatory movements were observed in all tasks; for example, PAGE and COIN, both involving objects placed at the edge of the table and the thumb as a lever, did not benefit from the ON configuration. Furthermore, compensatory changes were also evident in tasks without clear differences in wrist deviation, such as the decreased shoulder flexion in CARD. This finding may reflect a perceptual adaptation, whereby prosthetic users, knowing that additional degrees of freedom are available, reorganize their reaching strategy even without fully exploiting the wrist device.

Caution is needed when comparing our findings with previous reports on the effectiveness of flexible, multi-DoF prosthetic wrists in reducing compensatory movements among trans-radial prosthetic users, given the heterogeneity in experimental protocols, wrist designs and the frequent use of healthy participants with simulated prostheses. For instance, Deijs et al. [20] found no clear reduction of compensatory movements at shoulder joint during daily activities using commercial flexible wrists, whereas Olsen et al. [12] reported that a powered 2-DoF wrist (pronation/supination and flexion/extension) reduced trunk involvement. Within this context, our data show that a flexible wrist can decrease compensatory motion at proximal joints but requires the adoption of additional motor strategies for effective use. By analyzing multiple tasks, we further show that these benefits are task-dependent: in some tasks (e.g., PAGE, COIN), while users still use the functionality of the prosthetic wrist, no clear reduction in compensatory movements is observed. This suggests that while flexible wrists can provide functional advantages, their impact varies across task demands, supporting the need for context-specific evaluation. Moreover, our findings corroborate prior evidence that the functionality of the prosthetic wrist shapes users’ decision-making in approaching an object and ultimately determines whole-body posture [43].

The subjective assessment provided important insights into the users’ perceived workload, usability and satisfaction with the prosthetic system. The results from NASA-TLX indicated a moderate level of overall cognitive workload, with mental demand emerging among the subscales. This suggests that users had to allocate additional cognitive resources to adapt their movement pattern to the additional compliance of the prosthetic wrist. Interestingly, the reported overall cognitive workload was lower than that observed in recent evaluations of more complex myoelectric wrist systems, supporting the idea that increased system complexity tends to elevate cognitive demand during prosthesis use [12]. Despite the elevated mental demand, users reported an above-average usability score on the System Usability Scale (median 74/100), suggesting that the system was generally perceived as intuitive and learnable, aspects which advocate for potential reduction of mental demands with long-term adoption. The custom questionnaire responses reinforced these findings, showing overall satisfaction with the device and consistent ratings across items. However, greater variability emerged in aspects related to grasping performance, reliance on compensatory strategies and areas for improvement. Qualitative feedback underscored both the value of added wrist mobility and concerns regarding the prosthesis’ weight, size, and reliability during grasping. The perceived heaviness might have been exacerbated by the use of a temporary socket, which could have negatively influenced comfort and load distribution. A potential contributor to the inconsistency related to reliability during grasping may be the increased grip uncertainty (and required hand placement control) arising from unfamiliar or unstable wrist configurations, which are known to affect the mental demand during fine motor tasks [44]. Taken together, these findings underline a core challenge in the early-stage design of prosthetic wrists: enhancing functional versatility through additional degrees of freedom must be carefully balanced with usability and cognitive load. This trade-off is particularly critical in trans-radial limb loss, where the absence of wrist and hand proprioception makes users rely on visual feedback to guide motor actions [45–47]. In this context, increasing mechanical complexity without adequate sensory support may exacerbate cognitive burden and compromise functional outcomes, or user satisfaction [48], ultimately leading to prosthesis abandonment. Future research should therefore explore the integration of proprioceptive feedback substitution strategies, such as haptic or vibrotactile feedback, and investigate how these modalities interact with the control and usability of multi-DoF wrist prostheses [49].

This study has several limitations that should be acknowledged when interpreting the findings. First, the sample size limits the generalizability of results to the wider population of prosthesis users, although it is larger than those of previous studies evaluating prosthetic wrist prototypes in individuals with trans-radial limb loss [12, 19]. To partly overcome this limitation, we employed a statistical approach specifically designed to account for individual-level variability in small-sample research [38]. Furthermore, we also acknowledge that having only male participants is a limitation of the current study. While this gender imbalance is not uncommon in research involving prosthetic users [3, 10, 50], it reduces the generalizability of findings to the broader upper limb prosthesis-user population. Future studies should actively recruit female participants to address this gap. As a second limitation, participants used the prosthetic wrist only during two experimental sessions over two consecutive days, without daily use over an extended period. While this limited exposure constrains conclusions about long-term usability and adaptability, our study represents an initial step in the prosthesis evaluation process, as outlined in the Introduction. Notably, participants’ frequent requests for longer trial periods support the importance of incorporating extended usage in future studies. Third, while the SHAP test remains a widely used clinical tool for assessing upper limb function, its sensitivity to wrist-specific contribution has been questioned [20], thereby advocating for the development of ad hoc assessment procedures that are more responsive to changes in wrist mobility. Finally, all participants used the same model of prosthetic hand (Mia Hand, Prensilia S.r.l., Italy) to control for variability in terminal device characteristics. Although this standardized approach minimized confounding effects due to differing prosthetic hand technologies, it may have introduced unfamiliarity-related workload. Nevertheless, participants reported that the Mia Hand operated similarly to their own prostheses and required minimal adaptation. Similarly, the custom sockets used in this study differed from participants’ habitual clinical sockets in terms of geometry and electrode placement configuration, which may have introduced additional learning demands and variability in electrode-skin contact quality. However, participants reported no major difficulties related to socket fit or electrode performance, suggesting that socket-related confounds were limited in practice. Methods for replicating clinically-made sockets in a research setting have recently been proposed [51] and could improve the ecological fidelity of future studies by further reducing socket-related variability.

Conclusions

In conclusion, this study provides a comprehensive evaluation of a two-DoF prosthetic wrist with adjustable compliance in individuals with trans-radial limb loss, highlighting both its functional potential and practical limitations. While the flexible wrist configuration reduced compensatory movements at proximal joints and was generally well-received by users, it also required additional motor strategies and longer task completion times, particularly during early-stage adoption. These findings underscore that the benefits of increased wrist mobility are task-dependent and influenced by the novelty of the system as well as the specific demands of manual actions (i.e., object position, surface interaction and hand-object alignment). Although the users can adapt to the additional functionality, variability in grasping performance and concerns regarding device weight and stability remain. Overall, the results support the functional value of multi-DoF prosthetic wrists, while emphasizing the need for extended training, task-specific evaluation, and potential integration of sensory feedback to optimize both performance and user experience. These insights provide a critical foundation for the continued development and clinical assessment of advanced prosthetic wrists in trans-radial limb loss.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

AA

Adduction/abduction

AR

Axial rotation

CARD

Heavyweight extension object task

CI

Confidence intervals

COIN

Pick up coin task

DoF

Degree of freedom

EMG

Electromyography

FE

Flexion/extension

GLASS

Glass jar task

HC

Healthy controls

IER

Internal/external rotation

LB

Lateral bending

LMEM

Linear mixed effect model

NASA-TLX

NASA task load index

OFF

Prosthetic wrist in full-stiff mode

ON

Prosthetic wrist in compliant-while-reaching mode

P-IOF

Prosthesis index of functionality

PAGE

Cardboard task

PWR

Heavyweight cylindrical object task

QUEST

Custom-made questionnaire

S-IOF

SHAP index of functionality

SFE

Standardized fixed effect

SHAP

Southampton hand assessment procedure

SPH

Heavyweight spherical object task

SUS

System usability scale

ROM

Range of motion

RUD

Radial/ulnar deviation

Author contributions

Conceptualization: LR, DA, MG and UDC. Methodology: EB, LR, DA, EZ and UDC. Software: EB, EZ and MN. Formal analysis: EB, LR and DA. Investigation: EB, LR, DA, EZ, MN. Resources: AL. Data curation: EB and LR. Writing – original draft: EB, LR and DA. Writing – review & editing: EZ, MG, AL and UDC. Visualization: EB and LR. Supervision: LR, AL and UDC. Funding acquisition: UDC.

Funding

This work was funded by the Secretariat General of Defense of the Italian Ministry of Defense under the program “Ricerca Sanitaria Interforze“ within the “BIOVET - Bionics for Veterans” project (prot. n. M_D ABBE6E3 REG2022 0002861).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the institutional review board of Scuola Superiore Sant’Anna of Pisa (Approval No.10/2023). All procedures followed were in accordance with the ethical standards of the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Consent for publication

Written informed consent for publication was obtained from all participants.

Competing interests

The authors declare no competing interests.