The appropriate fit and comfort when wearing prosthetic sockets are critical factors that contribute to the successful use of upper-limb prostheses.1 A well-designed and appropriately fitted socket allows for extended periods of wear without significant discomfort as well as a reduced rate of injury due to tissue breakdown associated with localized pressure and shear.2 An uncomfortable socket often results in reduced use of the prosthesis by the patients to avoid potential tissue injury.2–4 Objective metrics for quality-of-fit assessment during the socket-making process are virtually nonexistent. The prevailing metric in current clinical practice is self-reported patient discomfort or pain. This highly subjective and largely person-dependent method has many shortcomings. First, the tolerance threshold for discomfort or perception of pain varies by individual and can differ widely among a population.5 Second, the lack of pinpoint location specificity on the part of the patient can cause costly and time-consuming mistakes in socket revisions.5 Third, some patients may have insensate portions of their residual limb that would make them unable to tell the fit condition of their socket and therefore are unable to provide the critical feedback to guide the socket iteration process. Over time, failure to address these issues at fitting can result in potential patient injury due to skin shear or excessive pressure, which is a substantial concern for patients with issues of sensitivity.
A pilot study by Neumann5 tested the ability of three lower-limb amputees to detect increases in pressure and discomfort on their residual limb. The study found that the inexperienced prosthetic user (first time being fitted) was better at detecting global pressure changes, perceiving deep pressure changes, and associating discomfort to increased pressure. The two experienced prosthetic users (≥32 years of prosthetic use) had difficulty in detecting changes in pressure (increments of 10 mm Hg) until 40 mm Hg (0.77 psi) was reached. In the deep-pressure perception to discomfort comparison, one of the experienced users did not report any discomfort during the testing, whereas the other reported an increase in discomfort at the second-to-last pressure increase. Hence, the discomfort perception and pressure perception are likely physiologically distinct, but both may be problematic in detecting socket pressures that could lead to skin breakdown. A more reliable and objective metric is needed to improve socket fitting success rate.
Previous efforts to measure socket-skin interface pressures have been applied to lower residual limbs.6,7 Polliack et al.7 found that a pressure-mapping sensor system, Rinceo SFS system, had errors of 24.7% on flat surfaces and errors of 32.9% on the positive model of a transtibial residual limb. The more accurate system, Tekscan’s F-socket system model 9811, had errors of 8.5% on flat surfaces and errors of 11.2% on the positive model.7 The authors concluded that “these systems are adequate in indicating areas of high pressure at the stump socket interface for clinical purposes.”7 After an extensive literature review, no study was found that attempted to measure pressure at the socket-skin interface of upper-limb amputees and relate it to patient discomfort. Characterization of the pressures in the upper-limb sockets could provide insight into current pressures that are tolerable to upper-limb amputees who have different socket forces and pressure distributions due to their different anatomy (more complex surface geometry) and the substantially greater range of motion. Because the sensitivity of the skin (measured as spatial acuity) also differs between the upper and lower residual limbs,8 an investigation of pressure and discomfort correlations of the upper-limb amputee may also help determine the reliability of discomfort to detect higher relative pressures within the socket. Therefore, socket-skin interface pressure measurements need to be evaluated in upper-limb applications.
The objectives of this study were 1) to develop a method for quantifying upper-limb socket fit condition by measuring the socket-skin interface pressure distribution and 2) to detect any potential correlation between this pressure and patient discomfort in the residual limb. Because pressure on the skin is perceived to be the primary cause for discomfort in socket fitting of patients with prosthetic limb, it was expected and therefore hypothesized that socket pressure would correlate with the level of discomfort in patients with an upper-limb prosthesis if any confounding factors are well controlled. Patient perception of discomfort was chosen for quantification over perception of pressure because perceived discomfort of the patient is current practice for assessing the quality of the socket fit.
This study was approved by the Western Institutional Review Board and conducted on upper-limb amputees using pressure-mapping sensors fitted inside of the prosthetic sockets (Figure 1). The subjects were asked to position their arms in one of the prescribed positions with handheld weights while the pressure conditions were recorded at the socket-skin interface. The subjects were required to mark their discomfort level on an analog scale. This discomfort score was then converted into a digital number between 0.0 and 10.0. The maximum pressure from each test was extracted from the pressure map data. A statistical analysis of correlation between maximum pressure and discomfort score was performed. The details of the methodology are presented as follows.
Three research subject volunteers from each of the upper-limb amputee groups, transradial (TR), transhumeral (TH), and shoulder disarticulation (SHD), were recruited for this study. The subjects’ demographics are presented in Table 1. Subject inclusion criteria were as follows: 1) age between 18 and 80 years, 2) ability to use an upper-limb prosthesis intermittently for up to 3 hrs, 3) having a reasonably comfortable socket (as determined by patient interview), and 4) able to hold at least the weight of 1 gallon of liquid with their prosthetic limb in any position (most chose to hold the gallon of liquid straight down by their side). Subject exclusion criteria were as follows: 1) pregnancy, 2) less than 1 year since upper-limb amputation, 3) loss of feeling in the skin when wearing prosthetic socket that could interfere with the safe use of the limb or the protective feedback of discomfort (i.e., anyone with complete loss of sensation regardless of surface area was excluded), 4) current mental health problems that could lead to misunderstanding of protocol (assessed by a prosthetist), and 5) inability to read and understand English. Although specific limb length was not recorded, the volunteers with TR or TH amputation had at least 50% remaining of their residual limb after their elbow for TR or shoulder for TH. All subjects had been using upper-limb prostheses for at least 10 years.
This study used Tekscan pressure-mapping sensor system (Versatek 2-port Cuff System; Tekscan, Inc, South Boston, MA, USA). Two model 9811 sensor sheets (6 × 16 sensor cells each) were used for the TR and TH sockets, whereas two model 9833 sensor sheets (15 × 16 sensor cells each) were used for the SHD sockets. This force-sensitive resistor pressure sensor system was selected for its thin profile (0.18 mm) and bending compliance that allowed instrumentation of the subject without need for modification of the socket.
Calibration of the pressure-mapping sensors (in arrays of thin film sheets) was performed before their use according to manufacturer instructions. Each sensor array sheet was sandwiched between two metal plates with a pneumatic bladder. A uniform pressure was applied to every sensor by inflating the pneumatic bladder with controlled pressure increments and measured by a pressure gauge (Pennwalt 1500). The electrical signals from the individual sensor cells were recorded and converted to pressure using the manufacturer-supplied software program. The calibration was performed at 2 and 5 psi to fit the linear response curve of the sensor cells for the expected pressure application range of 0 to 20 psi as recommended by the manufacturer.
The accuracy of the pressure sensors was assessed before and after each subject testing using the same procedure as the sensor system calibration. The pressure on the plate was increased to induce an evenly distributed pressure on the individual sensors from 0.0 to 5.0 psi in an increment of 1.0 psi while the pressure readings from the sensors were recorded. The mean difference between the prescribed pressure and sensor-measured pressure indicated the error in the sensor. The same process was repeated before and after testing of each subject to determine the pretest and posttest error of the pressure sensor system. Comparison of the error size before and after the test would indicate any sensor sensitivity drift during the course of the test day.
HUMAN SUBJECT TESTING
The subjects were tested in their existing prosthetic socket that had no known problems. The sensor sheets were first contained between two layers of a thin fabric sleeve to prevent them from wrinkling and to prevent malalignment when the subject pushed the residual limb into the socket. The sensors were then fitted between the residual limb or the liner and socket. The subjects wore their prostheses in their usual manner (Figure 1). The three-dimensional (3D) coordinates of the individual sensor locations in situ were digitized using a Microscribe G2X™ coordinate measuring device (Revware Inc) to obtain the general shape of the residual limb and the 3D locations of the pressure sensors relative to the limb anatomy. The subject’s residual limb was restrained during digitization to not allow any relative motion.
A test matrix to detect the effect of arm position and holding weight on socket skin pressure was developed using a design-of-experiment statistical tool supplemented by manual adjustment. The test variables included shoulder and elbow angle as well as holding weight. The weight held by the prosthetic arm varied from 0, 1, and 2 kg. The arm angles included combinations of shoulder flexion, shoulder transverse abduction, elbow flexion, and shoulder internal rotation at the values of 0°, 45°, and 90°. Because every combination of joint angle and weight would have resulted in 243 test combinations, the total number of tests needed to be reduced to minimize test subject fatigue, an initial test matrix was generated using a design-of-experiment statistical tool to determine a minimum set of position-weight combinations. Physiologically redundant elbow-shoulder configurations were eliminated from the test matrix. Clinically critical positions were added to the test matrix such as arms in position for pushing a lawn mower, pushing a shopping cart, or holding a drink. However, most of the positions were not common physiological positions but provided systematic changes to arm joint angles (i.e., lifting the arm straight out to the side in the coronal plane—shoulder abduction, 90°; elbow flexion, 0°; shoulder internal rotation, 0°). The number of trials was reduced to 68 test configurations for the TR subjects and 70 test configurations for the TH and SHD subjects. The test order of the position-weight configurations was then randomized for each subject and each amputation type to arrive at the final test matrices.
To assess repeatability of the pressure and discomfort measures, the centered position-weight test configuration (central test configuration) was replicated 10 times randomly throughout the subject testing procedure. The centered configuration involved 45° shoulder flexion, 45° shoulder transverse abduction, 45° elbow flexion, and 45° shoulder internal rotation of the prosthetic arm while a 1.0-kg weight was held by the prosthetic hand. Any change in the measured pressure or discomfort score in this configuration at different times of the testing would provide a measure of the results’ repeatability throughout the entire test sequence. Discomfort scores could be influenced by physical conditions (i.e., fatigue), individual physiological change in the course of a test day (i.e., increased residual limb volume), and/or experience in performing the test configurations. The selection of the central configuration for this statistical test was chosen on the assumption that it is the midway of all other test configurations and therefore would have the most relevance to all experimental test conditions. The SD of mean value (measured pressure or subject-indicated discomfort score) for each subject from the 10 replicated central test configurations and coefficient of variation (CV) were calculated to determine the measure repeatability.
For each of the test configurations, the socket pressure map and patient discomfort score were recorded. The subject was instructed to use his/her prosthetic hand to hold a prescribed weight and position his/her prosthesis in a predetermined orientation with the assistance of a researcher. Another researcher monitored the pressure sensor responses in real-time mode. After maintaining the arm configuration for approximately 3 seconds, the socket pressure data were recorded digitally. The subject then indicated his/her level of discomfort from “no discomfort” to “most discomfort imaginable” using a recorded value of 1 to 10 on a visual analog scale (VAS).9 The values of the recorded pressures in each test configuration were color-coded and laid over the digitized 3D sensor location coordinates (Figures 2, 3) during posttest data processing to facilitate visualization of the pressure values relative to the anatomic locations on the patient.
DATA AND STATISTICAL ANALYSIS
Statistical analysis of the subject test data was conducted to determine the correlation between the maximum pressure and the perceived discomfort. The discomfort score, as perceived and quantified by the patient in each of the test configurations, was normalized by dividing each discomfort score by the mean discomfort scores from that particular subject to minimize individual discomfort perception or threshold bias. The pressure data from the sensors at the edge of the TR and TH sockets were removed from this analysis to minimize artificially inflated pressure readings due to bending of the sensors (bending of a sensor of this technology results in an artificially saturated signal). The maximum pressure for the subject in each test configuration was extracted from the pressure map data. The maximum pressure from each test condition was not normalized because the pressure values are an objective measure and independent of personal perception or threshold. Linear regression analysis between the normalized discomfort score and maximum pressure was performed for each subject using the least square method. The association between those two response variables was assessed on the basis of the resulting correlation coefficient (r) or coefficient of determination (r2).
The results of the sensor accuracy assessments are compiled in Table 2. For each test subject, the mean percentage error from the test of 1-, 2-, 3-, 4-, and 5-psi pressure levels are summarized for the pretest and posttest conditions as well as their differences and mean scores. The sensor errors were also averaged for each sensor array (model 9811 used in the TR and TH groups and model 9833 used in the SHD group). The pretest sensor error ranged from 4.56% to 16.8%, whereas posttest errors ranged from 7.06% to 34.9%. The mean (SD) sensor sensitivity drift was 7% (4%) through the entire test procedure (Table 2).
The results from pressure and discomfort measure repeatability assessment are summarized in Table 3. A CV of zero would indicate that the outcome measure (pressure or discomfort) gave the same result when arm configurations were repeated. Pressure was most reliable in the TH group, and discomfort was most reliable in the SHD group. The TH group had nearly the same reliability as that of the SHD group. When compared with the CV of all arm configurations, the repeated central test arm configurations had less variability in pressure and discomfort by 33% (5%) and 48% (7%), respectively.
The socket pressure distribution pattern and maximum magnitude varied by individual subject and arm-weight configuration. The pressure magnitude ranged between as low as 0 and just greater than 12 psi (83 kPa) overall. The pressure difference within each test configuration ranged from 3 to 12 psi depending on the test configuration and test subject. The maximum pressure within each test configuration had a similar range.
The subject test results showed statistical correlation between the maximum pressure and subject discomfort score in the TR and TH groups but not in the SHD group (Table 4).10 The data show that the discomfort score increased significantly (p < 0.05) with increasing maximum pressure magnitude for all of the three TR test subjects, but their correlation was rather weak (r2 < 0.21). The correlation between the measured pressure and perceived discomfort was stronger in two of the three TH test subjects (r2 0.28 and 0.60 with p < 0.0001). No statistical correlation between pressure and discomfort was found from any of the three SHD test subjects (p > 0.05). The presence of a liner or sock in the TR and TH groups did not consistently influence the correlation between discomfort and pressure. An increase in weight was found to correlate with an increase in discomfort and pressure for both TR and TH subjects (p < 0.05) but not the SHD subjects.
This research attempted to develop a more objective method to assess socket fit by measuring the pressure distribution at the socket-skin interface of the upper-limb socket. Current standard practice relies on self-reported patient discomfort or pain, which is a highly subjective and person-dependent process. Objective and reliable methods for measuring socket fit would provide prosthetists with information needed to minimize discomfort and injury due to tissue breakdown associated with localized pressure and shear.2 Locations where the socket presents as too tight would be revealed by exceedingly high pressure. Points with no contact would be indicated by zero pressure at the socket-skin interface. This study investigated the feasibility of using the Tekscan pressure-mapping sensor system in providing more objective and reliable information on socket fit.
This new method had the most success in the TH test group, in which pressure predicted 28% to 60% of the perceived discomfort, as was indicated by their medium to strong correlation (Table 4). This method had less success when applied to the TR test group, in which pressure had a weaker correlation to the perceived discomfort level (Table 4). The possible reasons for this lesser correlation include less repeatable measures in pressure and discomfort score in the TR test group (Table 3).
Possible reasons that the TR test group had less repeatability in pressure and discomfort values than the TH test group could be differences in the amount of soft tissue around the residual limb and a residual limb lever arm length (distance between the most distal joint and the distal end of the residual limb). The subjects in the TR group had more soft tissue (fat and muscle) near the distal end of their residual limb and longer lever arms than the subjects in the TH group. Soft tissue pads of the residual limb and a longer residual limb lever arm allow greater surface area for distributing the forces, which could result in improved overall comfort of a prosthetic socket. Indeed, the subjects in the TR group reported a maximum discomfort score of 3.0 (1.3), whereas the subjects in the TH group reported a maximum discomfort score of 6.3 (3.2). Therefore, the increased thickness of soft tissue may have made the pressures feel more comfortable in the TR group. The increased residual limb lever arm length may have reduced discomfort by reducing peak pressure: the TR group had mean pressures that were 1.2 psi less than that of the TH group. In addition, the pressure sensor accuracy may have been reduced in the TR group if the soft tissue caused additional curving of the pressure sensors relative to the TH group. Other studies have previously found that curvature of the sensors could result in artificially high pressure readings when the sensors were applied to the residual limbs.5,7 Additional study is needed to better understand the factors that affect repeatability of pressure and discomfort measurement of persons with an upper-residual limb.
The pressure-mapping method had no success predicting discomfort levels in the three SHD cases (Table 4). Because the repeatability of the discomfort scores was relatively good, this application failure is most likely due to the limitation of the pressure sensor technology in which more complex geometry of the disarticulated shoulder surface caused more severe bending and/or folding of the pressure sensor sheets such that high sensor readings were registered even without compression between the socket and the skin. Neumann5 reported on a similar problem with a pressure sensor for measuring socket pressures, but he did not report on the type of sensor or the data collected with the sensor because of its reliability issues. Polliack et al.7 found a 22% increase in accuracy error between testing on flat surfaces versus curved surfaces.7 Therefore, the sensor error range in the subject testing in this study was likely greater than those indicated by testing on a flat surface (Table 2). It is expected, however, that pressure-sensor technologies such as those based on form-fitting polymers11,12 could alleviate such issues and improve sensor accuracy in such applications.
In light of these limitations, this preliminary study found that the Versatek 2-port Cuff System’s accuracy error is similar to previous reported errors in the lower-limb application, whereas sensor drift had greater error.7 However, the subject trials took approximately an hour, which gave the sensors a longer period to experience sensor drift. Polliack et al.7 limited the study of sensor drift to 20 minutes to reflect a length of time commonly used to evaluate patient socket fit. However, more time in this study was necessary to characterize pressure and discomfort in relation to arm position and prosthetic weight.
A unique aspect of this study was the testing of socket pressure and discomfort reliability in upper-limb amputees. The results demonstrated a wide variety of reliability of both discomfort scores and peak pressures when repeating the same arm position. The variability of the pressure sensors, although overall lower than discomfort scores, was still higher than expected. Although some of the variability in both the pressure sensors and discomfort scores may have been due to an increase in blood flow during subject testing, and therefore tighter socket fit, there was no trend toward greater pressure over time. This suggests that the variability in the pressure sensors was likely due to the inaccuracies of the sensors in this application.
In general, the subjects with a TR amputation were least reliable in describing discomfort, and the subjects with the SHD were the most reliable. This result indicated that some patients may have difficulty determining whether an alteration to their socket improved or worsened its fit, thus indicating that fit as a measure of promoting or preserving health of the residual limb is not necessarily the same as fit as a measure of patient comfort. Neumann’s5 study on lower-limb amputees also found that discomfort was an insensitive measure of pressure, with some participants being completely unable to associate the pressure ranges tested with an increase in discomfort. Therefore, these data suggest that more accurate and reliable pressure sensors would be a highly useful tool for evaluating and optimizing socket fit rather than relying solely on the discomfort of the patient. However, the pressure sensors used in this study do not seem adequate in providing the necessary reliability. With more reliable pressure sensors, future studies could better evaluate the relationship between patient’s discomfort, actual pressure, and pressures that lead to adverse outcomes.
Despite the limitations of the pressure sensor accuracy, low to moderate correlations to discomfort were observed for some subjects. An increase in weight was found to correlate with an increase in discomfort and pressure for both TR and TH subjects. This supports minimizing weight when fitting a socket to a patient or designing novel upper-limb prostheses. Previously, reduced weight of the prosthesis was the highest priority design concern of 242 people with an upper-limb amputation.13 Therefore, future studies that examine weight of upper-limb prostheses and limb length would be beneficial to determine clinically meaningful changes in prosthetic weight and pressure distribution. Upper-limb joint angles of the elbow and the shoulder also correlated with discomfort and pressure for some of the subjects. Therefore, a variety of joint angles should be tested when fitting a socket to a patient.
Increased socket pressure contributed from none to 60% of the discomfort felt by the subjects. When considering weight and various joint angles, weight was the primary factor that led to increased discomfort and pressure. The lack of reliability in patients describing their discomfort due to pressure suggests that more accurate methods of measuring socket pressure and better-correlated pressure metrics could improve socket fit prediction capability. Future studies that would improve the discomfort predictability and accuracy would include the following: 1) use larger sample sizes to improve statistical power; 2) explore other pressure-based metrics, such as pressure gradient and location-/anatomic-specific peak pressure, to better correlate fit condition and discomfort perception; and 3) use novel pressure sensor technologies such as those based on form-fitting polymers,11,12 which would also improve prediction accuracy because there is evidence that they maintain accuracy when curved around complex geometries.
1. Lake C. The evolution of upper limb prosthetic socket design. J Prosthet Orthot 2008; 20 (3): 85–92.
2. Sanders J. Stump-socket interface conditions. In: Bader DL, CVCB, Colin D, Oomens C, eds. Pressure Ulcer Research: Current and Future Perspectives. Berlin, Germany: Springer; 2005: 129–148.
3. Biddiss EA, Chau TT. Upper limb prosthesis use and abandonment: a survey of the last 25 years. Prosthet Orthot Int 2007; 31 (3): 236–257.
4. Biddiss E, Chau T. Upper-limb prosthetics: critical factors in device abandonment. Am J Phys Med Rehabil 2007; 86 (12): 977–987.
5. Neumann ES. Measurement of socket discomfort—part I: pressure sensation. J Prosthet Orthot 2001; 13 (4): 99–110.
6. Dumbleton T, Buis AW, McFadyen A, et al. Dynamic interface pressure distributions of two transtibial prosthetic socket concepts. J Rehabil Res Dev 2009; 46 (3): 405–415.
7. Polliack AA, Sieh RC, Craig DD, et al. Scientific validation of two commercial pressure sensor systems for prosthetic socket fit. Prosthet Orthot Int 2000; 24 (1): 63–73.
8. Myles K, Binseel MS. The Tactile Modality: A Review of Tactile Sensitivity and Human Tactile Interfaces. Aberdeen Proving Ground: US Army Research Laboratory; 2007.
9. Laylay GB. The visual analog scale: its use in pain measurement. Rheumatol Int 1985; 5 (4): 148–149.
10. Hinkle DE, Wiersma W, Jurs SG. Applied Statistics for the Behavioral Sciences. Boston, MA: Houghton Mifflin Co; 1998.
11. Kanellos GT, Papaioannou G, Tsiokos D, et al. Two dimensional polymer-embedded quasi-distributed FBG pressure sensor for biomedical applications. Opt Express 2010; 18 (1): 179–186.
12. Vandeparre H, Watson D, Lacour SP. Extremely robust and conformable capacitive pressure sensors based on flexible polyurethane foams and stretchable metallization. Appl Phys Lett 2013; 103: 4.
13. Biddiss E, Beaton D, Chau T. Consumer design priorities for upper limb prosthetics. Disabil Rehabil Assist Technol 2007; 2 (6): 346–357.