Adaptations from the prosthetic and intact limb during standing on a sway-referenced support surface for transtibial prosthesis users

Abstract

Purpose: To investigate the bilateral postural adaptations as a result of standing on an increasingly unstable sway-referenced support surface with both the intact and prosthetic limb for transtibial prosthesis users (TPUs).

Method: TPUs (n = 14) and matched controls (n = 14) stood quietly in multiple foot placement conditions (intact foot, prosthetic foot and both feet) on a sway-referenced support surface which matched surface rotation to the movement of the centre of pressure (CoP). Force and motion data were collected and used to analyse CoP mean position, displacement integral and force components under intact and prosthetic limbs.

Results: Significant differences were found between prosthesis users and controls in CoP mean position in anteroposterior (1.5 (95% CI, 1.2–1.8) cm) and mediolateral directions (3.1 (95% CI, 0.5–5.7) cm. CoP displacement integrals were significantly different greater for prosthesis user group in the anteroposterior direction. Force components differences were found in all planes (anteroposterior: 0.6 (95% CI, 0.4–0.8 N); mediolateral: 0.1 (95% CI, 0.0–0.2 N & 0.3 (95% CI, 0.2–0.4) N, inferosuperior: 2.2 (95% CI, 1.4–3.0) N).

Conclusions: TPUs have bilateral static and dynamic postural adaptations when standing on a sway-referenced support surface that is different to controls, and between prosthetic and intact sides. Results further support evidence highlighting importance of the intact limb in maintenance of postural control in prosthesis users. Differences indicate clinical treatment should be directed towards improving outcomes on the intact side.

Implications for rehabilitation

• Prosthesis users have bilateral adaptations when standing on a sway referenced support surface

• These adaptations are different to controls, and between prosthetic and intact sides.

• The intact limb is the major contributor to maintenance of postural control in prosthesis users.

• Clinical treatment should account for this when interventions are designed.

Keywords:

• Balance
• postural control
• transtibial
• amputee
• prosthesis

Introduction

The ability to maintain postural stability is a prerequisite in prevention of falls and fall injury. When a person lacks sensation from – or control of – the foot and ankle, postural stability becomes more difficult and risk of falling increases [1]. Transtibial prosthesis users (TPUs) have a complete removal of the foot/ankle complex and all related sensorimotor contributions associated. Whilst a causal relationship has not been established, we do know that these individuals have increased fear of falling [2, 3], fall incidence [4, 5] and that increased balance confidence and balance ability are strong indicators of future prosthetic use [6].

Efforts have been made to identify the individual biomechanical role the prosthesis and intact limbs play in postural control and postural stability [7–18]. Often, these methods rely on extracting information about postural stability via displacement of the centre of pressure (CoP) [19]. Movement of the CoP as a model for movement of centre of mass (CoM) in TPUs has been validated in quiet standing [20] and during rapid movements in TPUs [21]. It has been shown that TPUs have larger adaptations following support surface perturbations resulting in larger displacements of CoP on the intact limb and controls when compared to the prosthetic side [9]. This increased deviation of CoP under intact foot when compared controls in the anteroposterior direction has been identified in multiple postural control studies [10, 12, 16, 22–24] and highlight a significant challenge that TPUs have in maintaining postural control in the anteroposterior direction when compared to controls. Some evidence exists suggesting a positive association between increased prosthetic foot stiffness and increased measures of postural control, particularly in the anteroposterior direction during challenging postural tasks including stability indices and dynamic balance control [13–15, 25]. Though, other authors have suggested this may not be the case for at least small fluctuations of stiffness in the anteroposterior direction [17]. A limitation of investigations into associations of foot stiffness and postural control is that it is difficult to separate the contribution of sensory changes as a consequence of the prosthesis from changes in motor control due to biomechanical constraints. Despite this difficulty, efforts have also been made to understand the role of sensory deficits due to prosthesis use in postural control. Decreased sensory acuity is a strong indicator of previous falls and increased AP excursion of CoP in TPUs [26]. We know these individuals have a postural reorganization resulting in increased demand on the intact limb [27, 28] and that the prosthetic limb is thought to contribute by providing spatial information about the position of CoM [29, 30]. A method of delivering inaccurate sensory information about support-surface position is called sway referencing [31]. Sway-referenced support surface refers to a surface rotation matched to movement of a persons CoP. As the CoP moves in the anteroposterior direction, support surface rotations are elicited in same magnitude and direction, thus removing sensory input and biomechanical constraints which allow for corrective moments from ankle. This method creates a possibility to systematically control extent to which sensory information from the support surface is available to a person – and to quantify effects of this on measures of postural stability and control. In the case of the side with the prosthesis, there is no active ankle moment. Though, the intersection of the prosthetic foot and support surface is still the mechanical interface for forces transmitted to and from the residual limb/socket interface. Thus, sway referencing would still be matched to corrective movements of the CoP resulting from active control from more proximal joints. This method has not previously been used to investigate sensory role of intact and prosthetic limbs in maintenance of postural stability in TPUs. The aim of this study is to investigate what extent TPUs adapt to maintain postural stability while standing a sway-referenced support surface with their prosthesis, and intact limb, by investigating displacement of CoP and ground reaction force (GRF) magnitude. The experimental hypotheses are that, when compared to controls, TPUs will have: postural adaptations as evidenced by (1) altered mean position of the CoP; (2) mean GRF magnitude; and (3) increased reactive control as evidenced by increased CoP displacement integral.

Methods

Participants

An experimental group of TPUs (n = 14) was recruited on basis that they; had a unilateral transtibial amputation with no concomitant health issues, no current issues regarding fit or function of prosthesis including wounds, blisters or skin breakdown and had been a regular prosthetic user for at least one year and subjectively responded as being able to comfortably stand for the length of time required to complete the test protocol. A control group (CON) (n = 14) was matched based on age, sex, height and weight (Table 1). All participants gave written, informed consent to the study which was approved by the Regional Ethical Review Board in Linköping, Sweden.

Table 1. Participant characteristics.

Participant	Sex	Height	Weight	Age	YSA	Cause	Amp	Dom	Liner	Susp	Foot	Control	Sex	Height	Weight	Age
1	M	182	82	63	18	Trauma	L	R	Gel	Vacuum	ESAR	C1	M	177	76	62
2	M	173	94	67	24	Trauma	R	R	Gel	Vacuum	ESAR	C2	M	172	103	65
3	M	181	66	30	13	Trauma	L	R	Gel	Pin	ESAR	C3	M	178	70	32
4	M	181	84	42	17	Trauma	L	R	Gel	Seal-in	ESAR	C4	M	183	86	47
5	M	180	65	59	21	Vascular	L	R	Gel	Pin	ESAR	C5	M	177	68	59
6	F	172	68	50	33	Infektion	L	R	Gel	Sleeve	ESAR	C6	F	172	76	55
7	M	190	78	44	8	Trauma	L	L	Pelite + gel	Pin	ESAR	C7	M	192	87	48
8	M	178	86	50	8	Trauma	R	R	Gel	Vacuum	ESAR	C8	M	183	85	50
9	M	176	92	80	10	Trauma	R	A	Gel	Pin	ESAR	C9	M	175	77	75
10	M	183	88	36	7	Trauma	R	R	Gel	Vacuum	ESAR	C10	M	186	85	37
11	M	182	72	48	6	Trauma	L	R	Gel	Pump	ESAR	C11	M	185	73	49
12	M	183	77	55	5	Trauma	R	R	Gel	Vacuum	ESAR	C12	M	178	74	51
13	M	183	82	67	7	Trauma	R	R	Gel	Vacuum	ESAR	C13	M	181	78	66
14	M	183	78	33	7	Trauma	L	R	Gel	Vacuum	ESAR	C14	M	180	84	35
Mean (SD)	M = 13; F = 1	180.5 (4.6)	79.4 (9.2)	51.7 (14.4)	13.1 (8.3)	–	–	–	–	–	–	–	M = 13; F = 1	179.9 (5.6)	80.1 (9.0)	52.2 (12.5)

Participants 1–14 from TPU with matched CON. Sex (M: male; F: female), Height (m), Weight (kg), Age (years), YSA: years since amputation (years); cause of amputation, Amp: amputation side; (L: left; R: right); Dom: dominant side; (L: left; R: right; A: ambidextrous); Liner: material interface between residual limb and socket (Gel: gel liners such as silicone, Pelite: pelite foam), Susp: suspension form (Vacuum: non-active, negative pressure interface (one-way valve), Pin: distal pin, Seal-in: Seal-in liner, Pump: active electrical pump), Foot: energy storing and return foot (ESAR). Summary statistics (means and standard deviation (SD)) presented where appropriate.

Experimental protocol

Participants were first fit with a body harness in case they experienced a fall. Participants in a number of foot positions (intact foot on platform, prosthetic foot on platform and both feet on platform) (Figure 1) on one of two force platforms: a forceplate capable of rotations of up to eight degrees in the anteroposterior direction (Pro Balance Master, NeuroCom International Inc., Pleasanton, CA) and a force structure mounted on one of two stationary embedded foceplates (BP400600, AMTI Inc.; Watertown, NY). During a number of 20 s standing trials, force data was collected via Qualisys Track Manager (QTM) (Qualisys AB, Gothenburg, Sweden) and exported for later analysis. Data capture frequency was 100 Hz. Data processing was conducted following export in Visual 3D (C-Motion Inc., Germantown, MD). Following a calibration trial where total body mass was collected on the AMTI forceplate, subjects were instructed to stand quietly with arms at their sides while looking forward at a computer monitor mounted at eye-level at a distance of approximately 1.2 m. Subjects received real-time feedback of weight-distribution via a custom Labview version 12.0 (National Instruments Corporation Inc., Austin, TX) virtual instrument from unfiltered vertical force (z-component) from the AMTI forceplate. Subjects were instructed to maintain a 50:50 weight distribution between left and right legs during testing. It is well established that weight-distribution is unequal for both prosthesis users [27, 28] and able-bodied individuals [32]. The decision to normalize weight-distribution was made in order to assure any postural adaptations were due control of the independent variable (support surface stability) and not due to the inherent inequality in weight-distribution between the groups.

Figure 1. The three individual foot position conditions (1–3). Condition 1 – the participant’s intact limb (black) is on the embedded forceplate (e) in the Neurocom Pro Balance Master (b), the individual’s prosthetic limb is standing on the force structure (d) which is resting on the embedded AMTI forceplate (c) in the floor (a). In condition 1, data from two forceplates are being collected: one set from the AMTI forceplate (STAT) (c) (prosthetic limb), and one set from the Neurocom forceplate (SWAY) (e) (intact limb). Condition 2 – the experimental conditions are the same but with reverse order of the intact and prosthetic limb (intact on the AMTI forceplate (c) and prosthetic on the Neurocom forceplate (e). In condition 2, data from two forceplates are being collected: one set from the AMTI forceplate (STAT) (c) (intact limb), and one set from the Neurocom forceplate (SWAY) (e) (prosthetic limb). Condition 3 – both limbs were on the Neurocom forceplate (e). In condition 2, data from one forceplate is being collected: the Neurocom forceplate (MIDDLE) (e) (resultant CoP from both limbs).

During standing trials, TPU participants stood in a number of foot positions (intact foot on platform (prosthesis on stable support surface), prosthetic foot on platform (intact foot on stable support surface), both feet on platform) (Figure 1). In each of these foot positions, three sway referencing magnitudes were controlled (stable – no sway referencing, equal – 1:1 ratio of sway referencing to postural sway, increased – 2:1 ratio of sway referencing to postural sway) (Figure 2). Sway-referencing calculation is based on sway-angle [33–36], which is formed by participants’ theoretical CoM and centre of force platform (start position for each trial) and A ratio of 1:1 for sway-referencing means, for every degree participants shift their CoM in anteroposterior direction, force platform rotates in same angular direction with same magnitude of shift using actuators responding with a 200 ms delay. By rotating support-surface in response to participants’ anteroposterior shift, it is possible to remove proprioceptive sensory information available in order to maintain postural stability [31]. Each position/sway-referencing magnitude was repeated three times in random order, resulting in 27 individual 20 s trials. Participants received a rest period after 9th and 18th trial as it was necessary to move equipment and recalibrate between foot position conditions (intact, prosthesis and both). Though participants were advised that additional rest periods were available, no participants requested additional rests. Total data collection time was approximately 30 min.

Figure 2. Data collection protocol. Each participant is tested in three-foot position conditions relating to which feet are on the Neurocom platform (both, intact and prosthetic). Coinciding with data collected from the Neurocom, was data collected from the AMTI at the same time. Whilst data from the intact foot on the Neurocom was collected (INTACT_SWAY), data was also collected from the prosthetic limb on the AMTI forceplate (PROS_STAT). For each of these foot position conditions, a randomized order of sway reference gains was applied to the support surface (stable, unstable, very unstable). Sway reference gains applied (1:1, 2:1) refer to the amount of angular sway referencing applied per angular deviation of the CoG in the anteroposterior plane [31].

Data analysis

Using CoP data from the AMTI forceplate a coordinate transformation was conducted from the forceplate CoP position to the force structure support surface (Figure 1(d)). Transformed CoP data was then exported and utilized in further analysis. CoP data from the Neurocom system was exported in original form for analysis.

Dependent variables which were analysed were calculated based on displacement of CoP and GRF. For this reason, forceplatform signals were processed to extract CoP signals required. Both CoP and force data were low-pass filtered using second order Butterworth filter with a cutoff frequency of 3 Hz. Prefixes are utilized in text to differentiate which forceplate is being used (AMTI forceplates have _STAT prefix, sway-referenced Neurocom forceplate have prefix _SWAY accompanying).

Outcome variables utilized in analyses were:

1. Mean position of anteroposterior and mediolateral CoP from both forceplates. Referred to as (_STATCoP_AP, _STATCoP_ML, _SWAYCoP_AP and _SWAYCoP_ML). Position units in meters (m).

2. Mean magnitude of GRF components from AMTI forceplates (GRF_AP, GRF_ML and GRF_Z). GRF units in Newton (N).

3. Integral of mediolateral and anteroposterior CoP displacement from both forceplates (_STAT∫_AP, _STAT∫_ML, _SWAY∫_AP and _SWAY∫_ML). Calculated using trapezoidal method on full-wave rectified signals of mediolateral and anteroposterior first derivative time series captured at 100 Hz. Integral units in arbitrary units (a.u.).

Statistical analysis

To assess the aims of the study three-way ANOVAs were used to determine effects of group (TPU and CON), position (PROS_STAT, PROS_SWAY, INTACT_STAT, INTACT_SWAY and MIDDLE) and surface (stable, unstable and very unstable) on each of the outcome variables (_SWAYCoP_AP, _SWAYCoP_ML, _STATCoP_AP, _SWAY∫_AP, _SWAY∫_ML, _STAT∫_AP, _STAT∫_ML, GRF_AP and GRF_ML). A two-way ANOVA was used to determine effects of group (TPU and CON) and surface (stable, unstable and very unstable) on _STATCoP_ML. Where three-way interaction effects were not statistically significant, two-way interaction effects were evaluated. Where two-way interaction effects were present, simple main group effects and pairwise comparisons were evaluated to establish if significant differences existed between TPU and CON groups. Statistical significance was determined using a critical alpha level of α = 0.05 for all tests. Data are mean ± standard deviation unless otherwise stated. All pairwise comparisons in post-hoc analyses were conducted using a Bonferroni adjustment. With 95% confidence intervals of difference in pairwise comparisons were calculated for each group (TPU – CON), GRFZ and weightbearing distribution. For all group comparisons PROS_STAT/INTACT_SWAY are compared with left position of CON group (Figure 1(1)) and INTACT_STAT/PROS_SWAY are compared with right position of CON group (Figure 1(2)).

Results

Weightbearing distribution

Pairwise analysis of weightbearing distribution between the left and right sides of the CON group, and the intact and prosthetic side of the TPU groups indicated no statistically significant differences for any of the support surface conditions (stable, unstable and very unstable). These differences are presented in Table 2.

Table 2. Mean difference and 95% CI of the difference for left-right (CON-Group) and intact-prosthetic side (TPU-Group) for each of the support surface conditions.

			95% confidence interval of the difference
Group	Condition	Mean difference (N)	Lower	Upper
TPU	Stable	−0.2	−1.0	0.6
	Unstable	1.2	−0.3	2.9
	Very Unstable	0.7	−2.1	3.6
CON	Stable	0.2	−0.4	0.8
	Unstable	0.3	−0.3	1.0
	Very Unstable	0.8	0.0	1.6

All units are in Newtons (N).

Mean anteroposterior and mediolateral CoP position

_STATCoP_AP: There was no statistically significant three-way interaction between group, position and surface, F(4,486) = 0.071, p = .931. There was a statistically significant group*position interaction, F(1,486) = 67.775, p = .000. Simple main effect of group on _STATCoP_AP in PROS_STAT condition was statistically significant (F(2,486) = 113.195, p = .000). _STATCoP_AP in PROS_STAT condition was 0.540 ± 0.002 m for TPU group and 0.513 ± 0.002 m for CON group, a statistically significant difference of 0.027 (95% CI, 0.022–0.032) m, p = .000. 95% CI for _STATCoP_AP given in Figure 3(A).

Figure 3. 95% CI of the anteroposterior CoP position for the STAT conditions (left – A) and the SWAY conditions (right – B) separated by support surface condition (stable, unstable, very unstable). Transtibial prosthesis users (TPU) group: Black lines indicate prosthetic limb, grey lines indicate intact limb, broken lines indicate middle position (for SWAY conditions (B) only). Control (CON) group: Black lines indicate left limb, grey lines indicate right limb, broken lines indicate middle position (for SWAY conditions (B) only). All units in meters (m).

_SWAYCoP_AP: There was no statistically significant three-way interaction between group, position and surface, F(4,729) = 0.583, p = .675. There was a statistically significant group*position interaction, F(2,729) = 18.815, p = .000. Simple main effect of group on _SWAYCoP_AP in both INTACT_SWAY and PROS_SWAY conditions were statistically significant (F(2,729) = 102.461, p = .000 and F(2,729) = 36.219, p = .000). _SWAYCoP_AP in INTACT_SWAY condition was 0.075 ± 0.002 m for TPU group and 0.048 ± 0.002 m for CON group, a statistically significant difference of 0.027 (95% CI, 0.022–0.032) m, p = .000. _SWAYCoP_AP in PROS_SWAY condition was 0.064 ± 0.002 m for TPU group and 0.048 ± 0.002 m for CON group, a statistically significant difference of 0.016 (95% CI, 0.011–0.021) m, p = .000. There was no statistically significant difference in _SWAYCoP_AP between groups for MIDDLE condition, p = .180. 95% CI for _SWAYCoP_AP given in Figure 3(B).

_STATCoP_ML: There was no statistically significant two-way interaction between group and surface in left (F(2,243) = 0.004, p = .996) or right condition (F(2,243) = 0.014, p = .986)). Simple main effect of group on _STATCoP_ML in both left and right conditions were statistically significant (F(1,243) = 31.066, p = .000) and (F(1,243) = 6.172, p = .014). _STATCoP_ML in left condition was 0.585 ± 0.001 m for TPU group and 0.576 ± 0.001 m for CON group, a statistically significant difference of 0.010 (95% CI, 0.006–0.013) m, p = .000. _STATCoP_ML in right condition was −0.164 ± 0.002 m for TPU group and −0.158 ± 0.002 m for CON group, a statistically significant difference of −0.006 (95% CI, −0.011 to −0.001) m, p = .014.

_SWAYCoP_ML: There was no statistically significant three-way interaction between group, position and surface, F(4,729) = 0.603, p = .661. There was a statistically significant group*position interaction, F(2,729) = 5.520, p = .004. Simple main effect of group on _SWAYCoP_ML in both MIDDLE and PROS_SWAY conditions were statistically significant (F(2,729) = 5.333, p = .021 and F(2,729) = 5.366, p = .021). _SWAYCoP_ML in MIDDLE condition was −0.042 ± 0.009 m for TPU group and −0.012 ± 0.010 m for CON group, a statistically significant difference of 0.031 (95% CI, 0.005–0.057) m, p = .021. _SWAYCoP_ML in PROS_SWAY condition was 0.013 ± 0.009 m for TPU group and −0.017 ± 0.009 m for CON group, a statistically significant difference of 0.030 (95% CI, 0.005–0.056) m, p = .021. There was no statistically significant difference in _SWAYCoP_ML between groups for INTACT_SWAY condition, p = .489.

GRF component magnitudes

GRF_AP: There was no statistically significant three-way interaction between group, position and surface, F(2,486) = 0.181, p = .834. There was a statistically significant group*position interaction, F(1,486) = 36.547, p = .000. Simple main effect of group on GRF_AP in INTACT_STAT condition was statistically significant (F(2,486) = 48.772, p = .000). GRF_AP in INTACT_STAT condition was 1.791 ± 0.061 N for TPU group and 1.185 ± 0.061 N for CON group, a statistically significant difference of 0.607 (95% CI, 0.436–0.777 N, p = .000. GRF_AP given in Figure 4(A).

Figure 4. 95% CI of the mean GRF amplitude for the STAT conditions in the anteroposterior direction (left – A) and mediolateral direction (right – B) separated by support surface condition (stable, unstable, very unstable). Transtibial prosthesis users (TPU) group: Black lines indicate prosthetic limb and grey lines indicate intact limb. Control (CON) group: Black lines indicate left limb and grey lines indicate right limb. All units in newtons (N).

GRF_ML: There was no statistically significant three-way interaction between group, position and surface, F(2,486) = 0.255, p = .775. There was a statistically significant group*position interaction, F(1,486) = 6.327, p = .012. Simple main effect of group on GRF_ML in both PROS_STAT and INTACT_STAT conditions were statistically significant (F(2,486) = 4.820, p = .029 and F(2,489) = 32.727, p = .000). GRF_ML in PROS_STAT condition was 0.688 ± 0.039 N for TPU group and 0.567 ± 0.039 N for CON group, a statistically significant difference of 0.121 (95% CI, 0.013–0.230 N, p = .029. GRF_ML in INTACT_STAT condition was 0.846 ± 0.040 N for TPU group and 0.527 ± 0.039 N for CON group, a statistically significant difference of 0.319 (95% CI, 0.209–0.428) N, p = .000. 95% CI for GRF_ML given in Figure 4(B).

GRF_Z: 95% CI of mean GRF_Z separated by condition (INTACT_STAT and PROS_STAT) and surface (stable, unstable and very unstable) are presented in Figure 5.

Figure 5. 95% CI of the mean GRF_Z amplitude for the STAT conditions separated by support surface condition (stable, unstable, very unstable). Transtibial prosthesis users (TPU) group: Black lines indicate prosthetic limb and grey lines indicate intact limb. Control (CON) group: Black lines indicate left limb and grey lines indicate right limb. All units in newtons (N).

CoP displacement integral

_STAT∫_AP: There was no statistically significant three-way interaction between group, position and surface, F(4,486) = 0.586, p = .557. There was a statistically significant group*position interaction, F(1,486) = 79.804, p = .000. Simple main effect of group on _STAT∫_AP in both PROS_STAT and INTACT_STAT conditions were statistically significant (F(2,486) = 30.069, p = .000 and F(2,489) = 51.028, p = .000). _STAT∫_AP in PROS_STAT condition was 0.187 ± 0.014 a.u. for TPU group and 0.294 ± 0.014 a.u. for CON group, a statistically significant difference of −0.107 (95% CI, −0.145 to −0.069) a.u., p = .000. _STAT∫_AP in INTACT_STAT condition was 0.436 ± 0.014 a.u. for TPU group and 0.296 ± 0.014 a.u. for CON group, a statistically significant difference of 0.140 (95% CI, 0.102–0.179) a.u., p = .000. 95% CI for _STAT∫_AP given in Figure 6(A).

Figure 6. 95% CI of the anteroposterior CoP integral for the STAT conditions (left – A) and the SWAY conditions (right – B) separated by support surface condition (stable, unstable and very unstable). Transtibial prosthesis users (TPU) group: Black lines indicate prosthetic limb, grey lines indicate intact limb, broken lines indicate middle position (for SWAY conditions (B) only). Control (CON) group: Black lines indicate left limb, grey lines indicate right limb, broken lines indicate middle position (for SWAY conditions (B) only). All units in arbitrary units (a.u.).

_SWAY∫_AP: There was no statistically significant three-way interaction between group, position and surface, F(4,729) = 1.267, p = .282. There was a statistically significant group*position interaction, F(2,729) = 3.591, p = .028. Simple main effect of group on _SWAY∫_AP was statistically significant in MIDDLE condition (F(2,729) = 10.771, p = .001). _SWAY∫_AP in MIDDLE condition was 0.329 ± 0.018 a.u. for TPU group and 0.242 ± 0.019 a.u. for CON group, a statistically significant difference of 0.087 (95% CI, 0.035–0.139) a.u., p = .001. 95% CI for _SWAY∫_AP given in Figure 6(B).

_STAT∫_ML: There was no statistically significant three-way interaction between group, position and surface, F(2,486) = 0.459, p = .632. There was no statistically significant two-way interactions for group*position, F(1,486) = 0.3.583, p = .059, group*surface, F(2,486) = 0.155, p = .857, position*surface, F(2,486) = 1.351, p = .260.

_SWAY∫_ML: There was no statistically significant three-way interaction between group, position and surface, F(4,729) = 1.525, p = .193. There was no statistically significant two-way interactions for group*position, F(2,729) = 0.397, p = .672, group*surface, F(2,729) = 0.050, p = .951, position*surface, F(4,729) = 0.876, p = 0.478.

Discussion

The aim of this study was to investigate what extent TPUs utilize sensory information from their prosthesis, and intact limb, in maintenance of postural stability by investigating displacement of CoP while participants stand on a sway-referenced support surface. Results suggest that TPUs have postural adaptations and compensations which differ from control participants when standing on a sway-referenced support surface. These differences were of larger amplitude in the anteroposterior direction, and amount of sway-referencing (within range utilized in this study) did not have a significant effect on these adaptations.

The first hypothesis was partly confirmed. Results indicate that in the anteriorposterior direction there are two compensation strategies utilized by the TPU group when compared to the CON group as evidenced by significant differences in both _SWAYCoP_AP and _STATCoP_AP conditions (Figure 3(A,B)). TPUs shifted the CoP under the prosthetic limb anteriorly on the stable support surface but not under the intact limb during same conditions. These results support previous research showing TPUs have CoP position approximately 5 cm anterior to that of the intact limb (no able-bodied control group) during quiet standing [37]. The current results indicate an anterior shift under the prosthetic side of 2.7 (95% CI, 2.2–3.2) cm which are of similar magnitude, regardless of amount of sway referencing (Figure 3). When considering compensations on the unstable platform there are compensations that are in contrast to that of the stable platform. CoP position under intact limb was in a more anterior position when compared to the prosthetic limb (approximately 1.1 cm difference) and this difference between TPU group and CON group was 1.5 (95% CI 1.2–1.8) cm. Although methods differ, in investigations utilizing dynamic postural tasks similar results have been presented. Mouchnino et al. [27] showed that TPUs had a greater mean (SD) magnitude of anterior CoP shift during initiation of a “thrust” movement during a lateral shift of CoM in controls 98 mm (SD 21) when compared to prosthesis users’ intact (85 mm (SD 30)) and prosthetic side (80 mm (SD 34)). In anterior shifts of support surface, Bolger et al. [9] showed that reactive balance responses, indicated by peak CoP excursion, was also greater in the intact limb (∼10 cm) than both prosthetic side (∼5 cm) and controls (∼8 cm). Though, these authors found prosthetic limb had the least CoP excursion (∼5 cm), and not controls, as was found in this study. Although these previous authors used external support surface perturbations to elicit a physiological response from participants, current results provide additional evidence that multiple types of dynamic postural tasks (for instance support surface translation vs. rotation) elicit larger adaptations from intact side than that of prosthetic side.

The second hypothesis was also partly confirmed. Consistent with previous research there was a significant increase in magnitude of anteroposterior GRF for intact limb of the TPU group compared to the CON group [9,38] (Figure 4(A)). This difference showed a positive increase for TPU group, which indicates an anteriorly directed force component on the intact side. To maintain equilibrium, this requires that the contralateral limb (prosthesis on sway referenced surface) has a posteriorly directed force. When viewed in combination with mediolateral GRF results (Figure 3(B)) there appears to be a compensation where TPUs are utilizing hip control mechanisms for stability in both AP and ML directions. TPUs are directing the prosthetic limb anteriorly and laterally, with contralateral forces (intact limb) being directed posteriorly and laterally. Despite using different methods, this group difference in mediolateral direction is in contrast to Bolger et al. [9], who found no significant difference in mediolateral peak GRF between TPUs and control following platform perturbations.

Similar to the two previous hypotheses, the third hypothesis was also partly confirmed. There is evidence that there is increased demand, or reactive control coming from intact foot when compared to prosthetic side and controls. This is evident with significant differences found in both _STAT∫_AP and _SWAY∫_AP conditions (Figure 4(A,B)). These differences were limited to anteroposterior direction, with no differences seen in any conditions in the mediolateral direction. These results in _STAT∫_AP condition are in agreement with previous research indicating greater deviations of CoP in the anteroposterior direction on the intact side when compared to the prosthetic side [9, 10, 22, 26–28, 39] and control participants [7,9,27,28,39] (Figure 4(A)). What is also interesting is pattern of increasing magnitude of anteroposterior integral as gain of sway referencing was increased. Though not statistically significant, results suggest that as support surface becomes less stable, magnitude of integral increased for both sides (intact and prosthetic) and groups, an effect not seen in CoP position data. This is the first study, to the authors knowledge to show specifically an effect of surface stability on anteroposterior force integral. This is in contrast to previous research which concluded this compensation on the intact side was not altered when a balance task went from quiet standing to standing on a moving platform [38]. This previously named research utilized sinusoidal movements of the support surface in the anteroposterior direction and not a real-time sway-referenced support surface and these contrasting results are likely related to these methodological differences. When considering _SWAY∫_AP variable, this same effect of increased sway reference gain on increased integral magnitude is seen for the TPU group compared to the CON group. There was a statistically significant difference between groups when both feet were on the sway-referenced support surface. Though, magnitude of difference is smaller than that in static conditions. This is likely due to inability of _SWAY∫_AP conditions to separate roles of individual limbs as this outcome is a measure of resultant CoP integral from both limbs. In this sense, it is possible that increased magnitude of CoP displacement under the intact limb is attenuated by reduced CoP displacement associated prosthetic side while the resultant CoP is coming from both limbs.

CoP results for mediolateral direction are quite singular in differences between TPU and CON groups. In both _STATCoP_ML and _SWAYCoP_ML conditions, CoP was shifted to a more lateral position. In _STATCoP_ML conditions, regardless of position, TPUs shifted the CoP laterally by approximately 1 cm (left: 1.0 (95% CI, 0.6–1.3) cm; right: −0.6 (95% CI, −1.1 to −0. 1) cm). In _SWAYCoP_ML conditions this shift was approximately 3 cm (3.0 (95% CI, 0. 5–5.6) cm). Rougier and Bergeau [10] also found similar results with similar magnitudes of lateral shift (approximately 1 cm).

Although there are differences in ML position one must acknowledge that ML position of resultant _MLCoP (whole body) is systematically shifted to dominant limb side [32]. The size of systematic shift is less than one centimeter (range 0.17–0.59 cm), which is of the same approximate magnitude found in this study. When considered against any effect caused by dominant limb loading, results do suggest a lateral shift of CoP of significant magnitude.

Considered in conjunction with GRF_Z results, CoP results are of interest. When compared to controls, TPUs combined a strategy of increased loading on the static support surface with a lateral shift of CoP on the sway referenced surface. This would accomplish two tasks which alleviate unstable conditions: increased loading on the stable surface and greater base-of-support in the mediolateral direction. When both feet were on the sway referenced surface TPUs shifted their CoP towards their intact foot resulting in a statistically significant difference (3.1 (95% CI, 0.5–5.7) cm) when compared to the CON group in MIDDLE condition. This shift is also in agreement with previous studies showing postural shifts towards the intact limb for prosthesis users [10, 18, 20, 22, 24, 40, 41]. What is interesting is that, although participants received real-time feedback about weight-distribution, as the sway reference gain was increased they did not maintain this 50:50 distribution and shifted their weight to the stable support surface. Whether this shift was voluntary or a required shift in order to maintain postural control is unknown. Despite this shift, results do not suggest that TPUs adapt in a fashion linking instability of support surface and transfer of body mass towards the intact limb, irregardless of how unstable support surface is on prosthetic side (Table 2). Using experimental conditions which also gradated difficulty of postural tasks previous research has showed no differences in transfer of loading in mediolateral direction due to altered vision [22, 24] and dual-task conditions [24]. Though methodologically different to graded conditions of sway referencing used in this study, results support previous evidence that increasing difficulty of the balance task does not lead to additional load distribution inequalities in the mediolateral plane irrespective of whether the weightbearing distribution is freely chosen by participants or mandated as in this research.

Certain limitations must be discussed. Due to analysis form in the _STATCoP_ML variable, sample size was cut in half for this specific analysis. Of the total participants in TPU group, six used a prosthesis on the right side and eight on the left, with seven participants having amputation on their dominant side and seven on non-dominant side. Although the groups appear balanced, interpretation should be made with this limitation in mind for this particular variable. The group was quite homogenous in some characteristics and caution should be exercised in generalization of these results to all prosthetic users, or other specific groups of prosthetic users (such as those amputated due to vascular disease, persons having recently become prosthesis users, and young prosthesis users). In addition, the choice of including the weightbearing distribution should be considered. This was chosen to remove the documented differences in weightbearing between the groups [27, 29, 32] as a confounder. As feedback regarding weightbearing distribution is known to affect TPUs and able-bodied controls equally during postural assessment [42], any effect should be equal on the results for both groups. Though, the potential effect of this task on the results should be acknowledged. The variability of some outcomes in the TPU group was greater than the CON group (GRF_AP on intact side and GRF_Z) (Figures 5 and 6). Although this does not challenge the validity of the results given the utilized analysis methods, it cannot provide insight into how individual participants responded. In future research, it would be of sound utility to investigate how individual TPUs respond to varying instability of the support surface.

Despite some limitations, this study is the first to utilize real-time, gradated, support surface sway referencing to investigate bilateral effects on postural control in TPUs. These results provide new evidence that TPUs have postural adaptations due to the level of instability of the support surface in dynamic sitations which differ between intact and prosthetic sides, and compared to able-bodied control participants. The magnitude of these compensations in anteroposterior and mediolateral direction may be directly linked to the difficulty of the postural task. The clinical implication of this research highlight the importance of treatment interventions which focus on adaptations made on the intact side. Armed with the knowledge that TPUs can maintain weightbearing distributions equally as well, yet that the adaptations that are made are different to able-bodied individuals, allows clinicians to focus treatment on relevant outcomes.

Conclusions

TPUs have postural adaptations due to standing on a sway referenced support surface which are affected by the degree of stability of the support surface. These adaptations are different between prosthetic and intact sides of prosthesis users, and different to able-bodied controls. These differences in postural adaptation presented as altered mean position of CoPs in both anteroposterior direction, and in reactive control in anteroposterior direction.

Acknowledgements

The author wishes to thank the following individuals for their assistance in completion of this project: Johan Qvick, Paul Forsling, Alen Cavka, Dan Karlsson, Kjell-Åke Nilsson, Jessica Crafoord and all participants.

Disclosure statement

The author was fully involved in study and preparation of manuscript and material within has not been and will not be submitted for publication elsewhere.

Additional information

Funding

This study was completed using financial contributions made by the Promobilia Foundation (Grant# 12066).