The feasibility and validity of a wearable sensor system to assess the stability of high-functioning lower-limb prosthesis users

Abstract

Introduction:

Lower-limb prosthesis users (LLPUs) experience increased fall risk due to gait and balance impairments. Clinical outcome measures are useful for measuring balance impairment and fall risk screening but suffer from limited resolution and ceiling effects. Recent advances in wearable sensors that can measure different components of gait stability may address these limitations. This study assessed feasibility and construct validity of a wearable sensor system (APDM Mobility Lab) to measure postural control and gait stability.

Materials and Methods:

Lower-limb prosthesis users (n=22) and able-bodied controls (n=24) completed an Instrumented Stand-and-Walk Test (ISAW) while wearing the wearable sensors. Known-groups analysis (prosthesis versus controls) and convergence analysis (Prosthetic Limb Users Survey of Mobility [PLUS-M] and Activity-specific Balance Confidence [ABC] Scale) were performed on 20 stability-related measures.

Results:

The system was applied without complications; however missing anticipatory postural adjustment data points for nine subjects affected the analysis. Of the 20 analyzed measures output by the sensors, only three significantly differed (p≤.05) between cohorts, and two demonstrated statistically significant correlations with the self-report measures.

Conclusions:

The results of this study suggest the clinical feasibility but only partial construct validity of the wearable sensor system in conjunction with the ISAW test to measure LLPU stability and balance. The sample consisted of high-functioning LLPUs, so further research should evaluate a more representative sample with additional outcome measures and tasks.

1.0. Background

Over half of community-living individuals with lower-limb loss experience at least one fall within a given year^1,2, and those that experience a fall are at risk of chronic reduced quality of life^3,4. Increased fall risk likely results from compromised stability due to lost sensory feedback and active joint control^5-9. Lower-limb prosthesis users (LLPUs) show increased body sway during quiet standing compared to able-bodied individuals^10,11 and compensatory proactive strategies for enhancing walking stability reveal deficits in postural and body center-of-mass control (BCoM)^12-18. Although falls in LLPUs follow a variety of patterns and result from multiple factors, they often occur during walking on level ground². Accordingly, to effectively and efficiently screen LLPUs for greater fall risk, it would be advantageous for clinicians to capture stability-relevant information of the patient while walking.

Two characteristics limit current clinically accessible, performance-based methods to assess stability. First, many valid clinical outcome measures for LLPUs are useful for evaluating overall functional balance, but do not target gait stability (e.g., Berg Balance Scale^19,20, Four Square Step Test^21,22). Second, while some outcome measures address postural control during gait (e.g., Timed Up-and-Go (TUG)^21,23,24, L-Test²⁵, Narrow Beam Walking Test²⁶), they lack resolution to evaluate the specific components of gait stability (e.g., standing balance, anticipatory postural adjustments [APAs], trunk motion, gait regularity) and often suffer from ceiling effects²⁷. Although such components of gait stability can be accurately measured with motion capture systems, this requires dedicated facilities with sophisticated, expensive instrumentation and trained personnel not readily available in clinical environments. A more accessible and accurate means to collect stability-relevant measures in clinical settings, especially those of rehabilitation professionals such as therapists or prosthetists, would help address these limitations.

Recent advancements in wearable sensors (accelerometers, gyroscopes) potentially offer a clinically relevant solution to collecting human movement measures^28-30 with particular applications to stability assessment and rehabilitation monitoring^30-33. In particular, a new system developed for clinical implementation, the Mobility Lab (APDM, Portland, OR, USA), captures components of human movement³⁴. Specifically, the system collects standing balance and gait information during performance tasks commonly administered in clinical practice (e.g., TUG, stand-and-walk test). This system has demonstrated acceptable levels of validity and reliability for capturing spatiotemporal and kinematic gait measures in able-bodied individuals^35-37 and has been applied to quantify balance and gait characteristics in multiple patient groups prone to instability: children with neurological conditions³⁸, Parkinson’s disease ³⁹, multiple sclerosis^40-42, traumatic brain injury⁴³, older adults^44-46, and vestibular disorders ⁴⁷. Several measurement variables have been validated in comparison to criterion standard techniques including force plate measurement⁴⁸ and motion capture⁴⁹. However, to date this system has not been evaluated for LLPUs, a patient group which could benefit from a robust evaluation of fall risk via a clinically accessible means to quantify measures of stability during standing and ambulation.

An assessment on the feasibility and validity of the Mobility Lab system (combined equipment and post-processed measures) is necessary prior to advocating its use. The purpose of this study was to evaluate 1) the feasibility of performing a balance and gait assessment on LLPUs using the Mobility Lab system in a clinic-like environment, and 2) the construct validity (via known-groups and convergence analyses) of specific kinematic measures relevant to stability. Results from this study will contribute to the psychometric evidence of the Mobility Lab system and inform clinicians of its utility for capturing information relevant to stability, fall risk, and rehabilitation progress of LLPUs.

2.0. Methods

2.1. Participants

The study was approved by the Northwestern University Institutional Review Board, and all participants provided written informed consent. A convenience sample was recruited from attendees of a 3-day national orthotic and prosthetic conference based on the following inclusion criteria:

1. Lower-limb loss at or above the ankle;

2. At least 18 years of age;

3. Able to stand independently (without assistive device) for 30 seconds; and

4. Able to walk independently (without assistive device) for 14 meters.

Exclusion criteria included the following:

1. Unable to understand the protocol instructions in English;

2. Known pathologies that would affect gait or balance; and

3. Taking medications that affect gait or balance.

These criteria also applied to the recruitment of able-bodied control participants, excluding the first inclusion criteria related to amputation.

2.2. Outcome Measures

The APDM Mobility Lab used in this study consists of six triaxial inertial measurement sensors and accompanying data collection and analysis software. Each sensor (48.5x36.5x13.5 mm, 22 g) was fixed to the wrists and feet on both sides, upper chest, and lower back using straps (Figure 1). The sensor manufacturer determined this sensor placement for optimal capture of anatomical positions and spatial relationships during the stand and walk test. The system collected information during a trial that was started and stopped remotely through a laptop computer. The information was processed through a proprietary algorithm to output the data of interest, with basic definitions and descriptions of this process available in the user manual⁵⁰. In addition to gait speed (m/s) to aid with results interpretation, 20 available software-processed measures with relevance to standing balance and gait stability were selected for analysis:

Figure 1.

Sensor placement diagram (Image adopted from APDM user manual⁵⁰). When applied, all sensors are oriented with hardware ports facing down.

1. Postural sway given its relevance to standing balance control^10,11,51 (coronal and sagittal sway centroidal frequency [Hz], coronal and sagittal sway mean velocity [m/s], coronal and sagittal sway area radius [degrees]);

2. APA given its relevance to preparation for gait initiation and postural control^39,52-58 (APA duration [sec], APA first step range-of-motion (ROM) [degrees], APA anterior-posterior amplitude [m/s²], APA medial-lateral amplitude [m/s²]);

3. Trunk dynamics given their role in gait compensatory mechanisms and effect on postural control through BCoM positioning relative to the walking base-of-support (BoS)^59-62 (trunk coronal/sagittal/transverse ROM [degrees]);

4. Temporal variability given its relationship with gait regularity and falls^9,63,64 (sound and prosthetic limb step time standard deviation [s], and sound and prosthetic limb swing time standard deviation [%Gait Cycle]); and

5. 180° turning strategies given their relevance to negotiating obstacles and controlling medial-lateral balance^65-67 (turning duration [s], turn velocity [degrees/s], number of steps in turn).

Perceived mobility and balance confidence of LLPUs were measured with two self-report measures: the Prosthetic Limb Users Survey of Mobility (PLUS-M) 12-item Short Form version 1.2^68,69 and Activities-specific Balance Confidence (ABC) Scale⁷⁰, respectively. The PLUS-M asks LLPUs to rate their mobility capability during twelve ambulatory scenarios. Each item is scored with a 5-point ordinal scale anchored by “unable to do” (1) and “without any difficulty” (5). The sum across all items is converted to a T-score for subsequent analysis. The ABC Scale asks individuals to rate their perceived confidence during 16 ambulatory scenarios. Each item is scored along a percentage scale anchored by “No confidence” (0%) and “Completely confident” (100%), and the average across items is used for subsequent analysis. Both measures have acceptable validity and reliability for use with LLPUs^68,69,71 and higher scores indicate greater mobility (PLUS-M) or balance confidence (ABC Scale). These measures were used for assessing convergence validity given their ability to quantify aspects of the construct functional balance and selected specifically because of their established relationship to functional mobility and balance, clinical applicability, and level of psychometric properties.

2.3. Protocol

All components of the protocol were administered in a well-lit, carpeted room. After obtaining written informed consent, participant characteristics (age, height, weight, sex, limb dominance, shoe type) were collected via a standardized intake form. For LLPUs, additional collected information included amputation characteristics (side, level, unilateral or bilateral, etiology, date), prosthetic components, and prosthesis use in days/week and hours/day. Participants’ height and weight were measured using a scale and wall-mounted tape measure, respectively. Participants then completed the PLUS-M and ABC Scale in random order (randomized prior to data collection).

After participants completed the self-report measures, an investigator strapped the sensors to the participant and provided verbal instructions for the performance measure. All participants were afforded the opportunity to ask questions. Participants completed two trials of the Instrumented Stand-and-Walk (ISAW) test. The ISAW protocol was as follows:

1. Participants stood while their foot position was standardized to 10 cm between heels and 30° toe-out using a wooden template provided with the sensor kit (removed prior to starting data collection)⁷²;

2. After a first audible tone delivered from the laptop, participants stood naturally with their arms to their side and looking straight ahead for 30 seconds;

3. After a second audible tone after 30 seconds elapsed, participants walked straight ahead for 7 m at a “natural and comfortable pace,” leading with the prosthetic limb (LLPU cohort) or non-dominant limb (control cohort), turned 180° at a floor marker, and walked back along the same 7 m path. The trial ended when the participant crossed the start line.

As no instruction was provided on the turn direction in the first trial, participants were instructed to turn in the opposite direction for the second trial. The unilateral LLPUs were instructed to lead with the prosthetic limb for standardization; this has been reported as the most commonly preferred lead limb⁷³. Accordingly, able-bodied participants led with the non-dominant limb for consistency. Rest breaks between trials were provided as requested. An investigator followed participants from behind and to the side to ensure safety without pacing. LLPUs were not permitted to adjust any prosthesis settings throughout the protocol to maintain consistency between trials.

2.4. Data Analysis

The study assessed feasibility of using the Mobility Lab system to assess LLPUs by recording and evaluating any issues with system setup and data collection. Construct validity of the system’s preselected parameters was assessed through two analyses: known-groups and convergence. The known-groups analysis draws from an independent-groups comparison of the predetermined output parameters’ data between the LLPU and able-bodied cohorts. The convergence analysis examines bivariate correlation of the same parameters from the known-groups analysis and PLUS-M (T-score) and ABC Scale (average score) results for the LLPU cohort.

Before analysis, data normality was assessed using the Shapiro-Wilk test. If normality was violated, the Mann-Whitney U test was used for the known-groups analysis; otherwise a Student t-test was used with an adjustment if the assumption of equal variance was violated based on Levene test. These comparison tests were also performed on age and body mass index. The strength and significance of all correlations were evaluated through estimation of the Spearman ρ, to examine monotonic relationships rather than linear correlations to account for ordinal data of the self-report measures. Validity was suggested if the independent-group comparison analyses or correlation analyses were significant using a critical α of 0.05. The critical α was not adjusted to account for the multiple independent statistical tests to allow for identification of parameters with potentially meaningful differences, but all statistical results are provided to aid with interpretation. We anticipated that LLPUs would demonstrate reduced stability or postural control compared with controls, and that relevant Mobility Lab parameters would demonstrate at least moderate strength (ρ≥0.3) relationships⁷⁴ with both the PLUS-M and ABC Scale. Although higher-functioning prosthesis users may demonstrate relatively higher functional performance, this analysis assessed if the system is sensitive to determining effects of disrupted postural control due to lower limb loss with some level of statistical confidence. Further, according to the background information provided in the Outcome Measures section, we also expected a negative relationship between these self-report measures and sway, APA, trunk ROM, temporal variability, and turning parameters as an increase in these measures would suggest greater postural control demands.

3.0. Results

Twenty-two LLPUs (20 male/2 female, 45.0±15.5 years, 90.9±15.1 kg, 178.5±6.6 cm, body mass index=28.4±3.4 kg/m²) and twenty-four able-bodied controls (16 male/8 female, 32.4±12.7 years, 78.9±16.8 kg, 176.0±9.7 cm, body mass index=25.3±4.1 kg/m²) participated in this study. The LLPU cohort were on average older (p=0.004) and of higher body mass index (p=0.008). The LLPU cohort averaged 24.0±15.4 years since limb loss and included dysvascular (n=2) and traumatic/congenital (n=20) etiologies, unilateral (n=21) and bilateral limb loss (n=1); and transtibial (n=11), transfemoral (n=8), and knee disarticulation (n=2) level loss, with one individual with proximal femoral focal deficiency (n=1). The median (first quartile, third quartile) PLUS-M T-score and ABC Scale score were 64.5 (60.0, 71.4) and 97.8 (93.8, 100.0), respectively. LLPUs used a variety of prosthetic joint components. Those who wore prosthetic feet only used either passive dynamic (n=9) or a microprocessor-controlled articulated feet (n=2). Those who wore prosthetic feet and knees used either passive dynamic (n=9) or a microprocessor-controlled articulated feet (n=2), and either polycentric mechanical knees (n=1) or microprocessor-controlled knee (n=10).

3.1. Feasibility

All participants completed the protocol without issue and investigators did not encounter any notable challenges when applying the Mobility Lab sensors to LLPUs. However, the post-processing software was unable to calculate APA data for nine of the LLPU participants and one (4%) able-bodied participant. For LLPUs that APA data were not captured, all used a prosthetic knee joint (n=6 transfemoral, n=2 knee disarticulation, and n=1 proximal femoral focal deficiency). Consequently, these missing APA data were not included in the validity analyses. The entire protocol including equipment setup and data collection required an average of 15 minutes per participant.

3.2. Validity

There was no statistically significant (p=0.940) difference in mean gait speed between the able-bodied (1.3±0.1 m/s) and LLPU (1.3±0.2 m/s) cohorts. Only 3 of the 20 parameters demonstrated statistically significant differences between both cohorts (Table 1): APA first step ROM, trunk coronal ROM, and trunk transverse ROM; the ROM always greater for LLPUs compared to controls.

Table 1.

Known-groups Analysis Statistical Results

Parameter	Able- Bodied mean (SD) /median (IQR)	LLPU mean (SD) /median (IQR)	t or U value	p-value
Coronal sway centroidal frequency (Hz)	1.69 (0.38)	1.58 (0.36)	t=1.081	0.285
Sagittal sway centroidal frequency (Hz)	0.81 (0.18)	0.87 (0.25)	t=0.837	0.407
Coronal sway mean velocity (m/s)	0.02 (0.01)	0.03 (0.04)	U=320.0	0.204
Sagittal sway mean velocity (m/s)	0.1 (0.09)	0.11 (0.08)	U=260.0	0.930
Coronal sway area radius (deg)	0.26 (0.23)	0.29 (0.21)	U=323.5	0.190
Sagittal sway area radius (deg)	0.89 (0.37)	1.0 (0.6)	U=288.5	0.590
APA Duration (s)	0.49 (0.15)	0.52 (0.18)	t=0.459	0.649
APA first step ROM* (deg)	34.0 (6.9)	45.1 (4.7)	t=5.118	<0.001
APA anterior-posterior amplitude (m/s2)	0.45 (0.31)	0.32 (0.33)	U=144.5	0.871
APA medial-lateral amplitude (m/s2)	0.47 (0.21)	0.48 (0.18)	t=0.078	0.938
Trunk coronal ROM* (deg)	5.4 (2.0)	7.5 (2.8)	t=2.901	0.006
Trunk sagittal ROM (deg)	5.1 (1.2)	5.0 (1.2)	t=0.217	0.829
Trunk transverse ROM* (deg)	7.8 (1.4)	10.8 (2.1)	t=5.759	<0.001
Sound limb step time SD (s)	0.02 (0.01)	0.02 (0.01)	U=267.0	0.946
Prosthetic limb step time SD (s)	0.02 (0.01)	0.02 (0.01)	U=254.5	0.831
Sound limb swing time SD (%Gait cycle)	0.8 (0.2)	0.65 (0.36)	U=203.5	0.183
Prosthetic limb swing time SD (%Gait cycle)	0.8 (0.2)	0.71 (0.63)	U=258.5	0.904
Turn duration (s)	2.1 (0.27)	2.2 (0.41)	t=1.497	0.141
Turn velocity (m/s)	191.5 (28.0)	200.0 (39.7)	t=0.835	0.408
Number of steps in turn	3.5 (1.0)	4.0 (0.75)	U=291.0	0.539

SD=standard deviation; IQR=interquartile range; APA=anticipatory postural adjustment; ROM=range of motion. Parameters with significant differences (p≤.05) denoted by asterisks and bolded. Mean and median values presented for t or U tests, respectively.

Only 2 of the 20 parameters demonstrated statistically significant correlations with the PLUS-M and ABC Scale scores (Table 2): coronal sway centroidal frequency (moderate strength) and sagittal sway area radius (strong). Greater coronal sway centroidal frequency was associated with lower mobility and balance confidence (Figure 2), while greater sagittal sway area radius was associated with greater mobility and balance confidence (Figure 3).

Table 2.

Convergence Analysis Results

Parameter		PLUS-M	ABC Scale
Coronal sway centroidal frequency*	ρ	−.586	−.509
	p-value	.004	.016
Sagittal sway centroidal frequency	ρ	−.245	−.228
	p-value	.273	.307
Coronal sway mean velocity	ρ	.277	.283
	p-value	.212	.202
Sagittal sway mean velocity	ρ	.126	.104
	p-value	.577	.646
Coronal sway area radius	ρ	.306	.308
	p-value	.166	.163
Sagittal sway area radius*	ρ	.451	.422
	p-value	.035	.051
APA Duration	ρ	.182	−.072
	p-value	.552	.814
APA first step ROM	ρ	−.499	−.426
	p-value	.083	.146
APA anterior-posterior amplitude	ρ	−.146	−.047
	p-value	.635	.878
APA medial-lateral amplitude	ρ	−.025	−.112
	p-value	.934	.716
Trunk coronal ROM	ρ	.200	.219
	p-value	.373	.328
Trunk sagittal ROM	ρ	−.054	−.071
	p-value	.813	.755
Trunk transverse ROM	ρ	−.047	−.086
	p-value	.835	.703
Sound limb step time SD	ρ	−.064	.116
	p-value	.776	.606
Prosthetic limb step time SD	ρ	−.219	−.370
	p-value	.326	.090
Sound limb swing time SD	ρ	−.338	−.288
	p-value	.123	.193
Prosthetic limb swing time SD	ρ	−.092	−.199
	p-value	.684	.375
Turn duration	ρ	.411	.165
	p-value	.058	.464
Turn velocity	ρ	.040	.219
	p-value	.859	.328
Number of steps in turn	ρ	.358	.177
	p-value	.101	.432

PLUS-M=Prosthetic Limb Users Survey of Mobility; ABC=Activity-specific Balance Confidence; APA=anticipatory postural adjustment; ROM=range of motion; SD=standard deviation. Parameters with significant correlations (p≤.05) denoted by asterisks and bolded.

Figure 2.

Scatter plots of Coronal Postural Sway Centroidal Frequency versus Plus-M T-Score (left) and Activities-Specific Balance Confidence Scale score (right).

Figure 3.

Scatter plots of Sagittal Postural Sway Area Radius versus Plus-M T-Score (left) and Activities-Specific Balance Confidence Scale score (right).

4.0. Discussion

This study assessed the feasibility and construct validity of the Mobility Lab system for capturing stability-related measures of LLPUs during a performance-based task (ISAW).

4.1. Feasibility

Results of this study suggest that use of the Mobility Lab and associated protocol were largely feasible, appropriate for clinical application when assessing components of stability in conjunction with the ISAW task, and worthy of further exploration. Apart from the compact Mobility Lab system, few resources and little physical space were required to perform the assessment. A 15-minute protocol may be difficult to integrate into a primary care setting; however, rehabilitation professionals such as physical therapists or prosthetists can perform these assessments in their office during initial or follow-up evaluations. A noteworthy limitation is that the system failed to produce APA data for 41% of the LLPU cohort. All of the participants with missing APA data walked with a prosthetic knee, suggesting that use of this component might affect the system’s ability to capture APAs. This issue limited APA data and analysis to only two participants who used a prosthetic knee. Given the post-processing software, this limitation may relate to problems in the algorithm’s use of the lower back sensor data to identify possible APA periods, as has been reported for persons with Parkinson’s disease⁵⁶. In this study, two trunk measures demonstrated significantly greater ROM in the LLPU cohort, thus suggesting that greater trunk ROM may disproportionately affect the lower back sensor’s ability to detect APA periods in LLPUs compared to controls. Consequently, application in the LLPU population should recognize this limitation in the observed state and version of the Mobility Lab system, especially at more-proximal levels of amputation.

4.2. Known-Groups Assessment

Three measures demonstrated significant differences between LLPUs and controls, all of which described greater kinematic ROM of either the stepping limb at gait initiation (APA first step ROM serving as angular representation of initial step length) or trunk during walking (trunk coronal and transverse ROM) of LLPUs. These differences are reasonable given general compensatory methods LLPUs use to initiate gait and facilitate forward ambulation.

In agreement with observed APA first step ROM differences, a previous study observed that unilateral LLPUs more often lead with the prosthetic limb (a condition required in this study) and display longer leading limb step lengths than able-bodied individuals when initiating gait⁵⁷. These differences likely relate to postural control strategies to compensate for impaired propulsive ability. Evidence suggests that unilateral LLPUs spend more time in single limb stance on the sound limb during gait initiation when leading with the prosthetic limb to achieve larger stance-limb propulsive forces^57,73, and that a shorter APA phase with a longer step execution phase (i.e., time to complete the first step of gait initiation) facilitates adequate progression velocity at the end of the first step⁵³. Furthermore, a longer step length may be reflective of impaired volitional control of the BCoM to prepare for gait initiation but aid restoration of balance after the step execution phase⁵⁸.

The significantly greater trunk coronal and transverse ROM of LLPUs compared to controls agrees with the literature suggesting that this mechanism aids with advancing the limbs for forward ambulation to compensate for reduced lower-limb muscle strength and terminal stance power generation in which the trunk generates momentum and facilitates prosthetic limb heel rise and lift off^59,61,75. However, control of the trunk mass is critical to maintain gait stability as evidenced by its influence on regulating the walking BoS⁶². Exaggerated lateral trunk motion in LLPUs may also have the negative consequence of projecting the extrapolated BCoM (a velocity-weighted BCoM position)⁷⁶ closer to or beyond the lateral BoS⁶¹, increasing the risk of falls.

Overall, though only a few metrics significantly differed between the two cohorts, the direction of differences aligns with the literature and reflects impaired balance control of LLPUs relative to able-bodied controls. Consequently, these known-groups comparison results suggest partial validation of the Mobility Lab system’s ability to measure stability and balance evidenced by expected differences between data for LLPUs and controls.

4.3. Convergence Assessment

Two measures demonstrated a significant correlation with the self-report measures (PLUS-M and ABC Scale) in the LLPU cohort, both related to standing balance control (coronal sway centroidal frequency and sagittal sway area radius). To note, sagittal sway area radius related to both self-report measures with moderate strength, however although the PLUS-M (p=0.035) relationship demonstrated significance, the ABC Scale relationship was marginally significant (p=0.051).

Coronal sway centroidal frequency offers a sensitive, valid, and reliable metric of the frequency at which the medial-lateral postural fluctuation spectral mass is concentrated^48,77. The inverse relationships between this metric and the self-report measures suggest that greater frequency in LLPUs is associated with less mobility capability and balance confidence. Evidence suggests that overall centroidal frequency increases with loss of visual feedback, a pertinent component of mobility and balance^78,79. However, contrasting evidence shows that higher medial-lateral frequencies are associated with better balance control given its strong inverse relationship with age⁸⁰ and weak inverse relationship with Parkinson’s disease disability severity³⁹. The results from this study align with evidence suggesting that higher-functioning LLPUs with greater perceived balance and mobility may demonstrate less rapid corrective adjustments in the coronal-plane due to increased capabilities to compensate for lost sensory feedback as concluded by Ku et al. (2014)¹⁰. Additionally, LLPUs tend to have better coronal-plane standing balance control than in the sagittal plane due to limitations of actively controlling the prosthetic ankle^10,81. This reasoning may also explain the observed positive relationship between sagittal plane sway and mobility capability/balance confidence, as higher functioning LLPUs may more willingly explore their standing ROM, especially with a prosthesis that lacks active control⁶. Although this relationship was opposite to what was expected, this result would also be supported by findings that transtibial prosthesis users demonstrate a positive relationship between exploration of their forward-backward limits of stability and time since discharge from rehabilitation⁸², and similar relationship directions have been observed in postural sway amplitude and balance confidence in older women⁸³. Typically, LLPUs with more impaired balance control either from more proximal level amputation, reduced muscle strength, or reduced sensation display greater sagittal plane ROM^10,11. However, this relationship may be different for high-functioning LLPUs as assessed in this study and reflected by the absence of any difference between this cohort and the able-bodied controls to suggest high levels of postural control. Overall, the relationships between these variables warrant further exploration to more confidently suggest convergence validation of the Mobility Lab system to assess stability and balance of LLPUs.

None of the measured parameters demonstrated both significant difference in the known-groups assessment and significant correlation with self-report measures in the convergence assessment. Therefore, the findings of this study do not suggest an explicit parameter for unequivocal clinical utilization in measuring stability and balance in high-functioning LLPUs.

4.4. Limitations

The convenience sample used for analysis limits generalizability of these results. The sample includes a relatively high-functioning cohort of LLPUs based on the PLUS-M and ABC scores in the context of existing literature^19,68. The LLPU cohort was also predominantly male (91%) and of traumatic/congenital etiology (91%). Future studies should perform a similar feasibility and validity assessment with a more generalizable cohort. Furthermore, the analysis is limited by the heterogenous sample of prosthesis users in terms of device use and limb loss level, and this limitation should be considered when interpreting results. In addition, although this analysis considered a large selection of stability-relevant outcome measures, this selection is not exhaustive of the Mobility Lab’s available parameters and performance task protocols. Future studies should consider additional parameters and tasks associated with this technology. Finally, the critical α was not adjusted to account for the experiment type I error rate despite the multiple independent statistical analyses. This decision allowed for identification of results that may support this first validation analysis when considering the limited sample size. However, all p-values are presented to aid results interpretation. Future studies should perform similar analyses with larger and more representative samples while accounting for possible false-positives.

Conclusion

The results of this study suggest the clinical feasibility but only partial construct validity of the parameters collected using the Mobility Lab system in conjunction with the ISAW test to measure LLPU stability and balance. Although simply and quickly applied to LLPUs, the system was unable to estimate APA parameters for 41% of this cohort, all of whom walked with a prosthetic knee joint. The known-groups assessment demonstrated expected differences between cohorts in three measures to suggest validity, however results of the relationships between mobility capability/balance confidence and two standing balance measures were less clear. Interpretation of these results must also consider the relatively high-functioning LLPU cohort. Studies on an LLPU sample consisting of a wider range of mobility levels and with additional outcome parameters are needed to fully investigate the construct validity of this system for LLPUs prior to its use in clinical or research settings.