Abstract
Background and Purpose Roughly 50% of individuals with lower limb amputation report a fear of falling and fall at least once a year. Perturbation-based gait training and the use of virtual environments have been shown independently to be effective at improving walking stability in patient populations. An intervention was developed combining the strengths of the 2 paradigms utilizing continuous, walking surface angle oscillations within a virtual environment. This case report describes walking function and mediolateral stability outcomes of an individual with a unilateral transfemoral amputation following a novel perturbation-based gait training intervention in a virtual environment.
Case Description The patient was a 43-year-old male veteran who underwent a right transfemoral amputation 7+ years previously as a result of a traumatic blast injury. He used a microprocessor-controlled knee and an energy storage and return foot.
Outcomes Following the intervention, multiple measures indicated improved function and stability, including faster self-selected walking speed and reduced functional stepping time, mean step width, and step width variability. These changes were seen during normal level walking and mediolateral visual field or platform perturbations. In addition, benefits were retained at least 5 weeks after the final training session.
Discussion The perturbation-based gait training program in the virtual environment resulted in the patient's improved walking function and mediolateral stability. Although the patient had completed intensive rehabilitation following injury and was fully independent, the intervention still induced notable improvements to mediolateral stability. Thus, perturbation-based gait training in challenging simulated environments shows promise for improving walking stability and may be beneficial when integrated into a rehabilitation program.
More than 75,000 lower limb amputations are performed in the United States each year.1,2 Although the majority of people with lower limb amputations regain the ability to walk functional distances,3,4 many demonstrate gait deficits, including impaired walking stability. A return to independent community ambulation presents a challenge for individuals with lower limb amputations, as the community environment requires navigating diverse terrains, including varied surface angles. Thus, there is little surprise that 49% of people with lower limb amputations express a fear of falling, and 52% report falling at least once a year.2 Therefore, it is important to design and evaluate interventions to improve stability in individuals with lower limb amputations.
There are many intervention approaches that have been used to improve stability in patient populations. Perturbation-based gait training programs have been particularly effective at improving stability responses to destabilizing events and environments.5–8 Clinical researchers have successfully used virtual environments to assist in quantifying stability9–11 and assessing the efficacy of perturbation-based gait training interventions.6,12 Virtual environments also have been shown to be an effective rehabilitation tool offering extensive control, repeatability, and adaptability.13 Kaufman et al6 demonstrated that perturbation-based gait training could be combined with assessments in a virtual environment to improve tripping responses in a group of individuals with transtibial amputation. They used rapid treadmill belt accelerations to train proper tripping responses. Using a similar task in a virtual environment, the authors6 found that participants were able to improve their tripping responses. Although the intervention was successful, that specific tripping perturbation is only one of many types of destabilizing environments that individuals encounter in the community. This type of simulated trip using treadmill belt accelerations prompts discrete corrective stepping reactions. On the other hand, providing continuous physical perturbations requires sustained effort toward the maintenance of stability and may improve learning through the repeated exposure to a wide range of perturbations. An intervention that combines the strengths of training and evaluation in a virtual environment with the well-established benefits of a perturbation-based gait training framework has the potential to be highly effective at improving stability in people with amputations. This case report describes the effect of gait training using walking surface angle perturbations in a virtual environment on improving mediolateral walking stability of an individual with a unilateral transfemoral amputation (TFA).
Case Description: Patient History and Systems Review
At the time of the first visit, the patient was a 43-year-old male veteran (height=1.83 m, weight=106.6 kg) who had undergone a right TFA more than 7 years previously. The mechanism of injury was a vehicle-born improvised explosive device encountered during Operation Iraqi Freedom. The patient underwent several limb-salvage surgeries over 2 months before amputation. His residual limb length from greater trochanter to the terminal portion of soft tissue was 48 cm. Comorbidities included one-third pack per day smoking, shrapnel in his left leg and right forearm, posttraumatic stress disorder, and osteoarthritis in his left knee and both shoulders. His medications included citalopram, bupropion, lisinopril, and amlodipine. He participated in intensive physical therapy, occupational therapy, and recreational therapy at the Center for the Intrepid following injury. He had been an independent prosthetic ambulator for approximately 7 years at the time of his initial visit, but stated he was generally sedentary between the initial episode of care and the current intervention. Due to the intensive physical therapy that he received and the length of time as a community ambulator since the amputation, he was expected to have reached his rehabilitation plateau.
At the time of the patient's initial visit, primary efforts to stay active included neighborhood walks and yard work. He used a carbon-fiber ischial containment socket with a one-way expulsion valve to allow passive suction. He wore an Össur Seal-In silicone liner (Össur, Reykjavik, Iceland) and an oversocket neoprene suspension sleeve. His prosthesis consisted of an Ottobock X3 microprocessor-controlled knee (MCK) (Otto Bock Healthcare LP, Austin, Texas) and an Ottobock Triton foot. The patient complained of groin pain (3/10) when kneeling due to socket rubbing and central low back pain (5/10) described as muscle soreness after prolonged activity. He continued to attend prosthetic follow-up to obtain new gel liners and perform the occasional adjustments to the Ottobock X3 knee. He also complained of decreased cardiovascular endurance, walking efficiency, socket tolerance, and stability on paved and grassy slopes. These deficits limited his full participation in yard work and sports such as golf due to fear of instability and falling. The patient reported a history of falls and stumbles, especially on sloped terrain, but could not quantify their frequency. His goals included a return to sports (ie, stand-up basketball and golf), increased walking efficiency with decreased metabolic costs, and increased stability on slopes.
Individuals with a TFA are at significantly greater risk for falls than those with a transtibial amputation.1,2 Individuals with TFA must rely on the prosthetic knee joint to provide stance stability in the absence of knee extensors.14 During able-bodied gait, knee extensors and flexors coactivate to control knee flexion and prevent buckling during a rapid weight shift onto the stance limb at loading response.15 Many MCK prostheses can now provide stance knee flexion resistance during the loading response to prevent knee buckling.14,16,17 The use of MCK prostheses by individuals with a TFA has been reported to reduce stumbles and falls18–21 and improve gait stability on sloped surfaces.16,18,22 Although gait stability appears to improve with MCK prosthesis use, the devices require a particular gait pattern to function properly. Thus, the device and its user may both have particular difficulty adapting to continuous walking surface perturbations.
Examination
Prior to the intervention training, we evaluated the patient's functional status and lateral stability to identify deficits and establish a baseline. As a general measure of function and agility, we measured the time it took the patient to step with both feet into 4 quadrants in a horseshoe pattern (similar to the Four Square Step Test23). Following a practice trial, the patient completed the test 2 times, and the faster time was recorded. A faster time indicates better function and greater stability and agility.23 In addition to the functional stepping test, we measured the patient's self-selected walking speed. This measure has been used as a global measure of functional ability, with faster walking speeds associated with greater function.24 The patient was instructed to walk comfortably on a level walkway. We calculated the average self-selected walking speed over 8 passes using motion capture data from the middle, steady-state portion of the pass.
Theoretical modeling and experimental studies25,26 have shown that able-bodied individuals are most unstable in the mediolateral direction. Those with lower limb amputations are especially destabilized by laterally directed physical perturbations during walking.27–30 Therefore, we also evaluated the patient's lateral walking stability within a virtual environment (Computer-Assisted Rehabilitation Environment, Motek Medical, Amsterdam, the Netherlands), which consists of a 1.8- × 2.8-m treadmill on a 6-degree-of-freedom platform within a 300-degree dome. The patient walked at 1.27 m/s for two 3-minute trials through a scene consisting of a forest path with evenly spaced fence posts under each of 3 conditions: unperturbed walking (NOP); continuous, pseudorandom mediolateral visual scene oscillations (VIS); and continuous, pseudorandom mediolateral platform oscillations (PLAT).31 Kinematic data were collected at 60 Hz using a 24-camera Vicon motion capture system (Vicon Motion Systems, Oxford, United Kingdom). The cameras tracked the trajectories of 57 markers affixed to the patient's body segments.32 The marker position and digitized locations of joint centers were combined to create a 13-segment model using Visual 3D analysis software (C-Motion Inc, Germantown, Maryland).33
From the kinematic data, we extracted metrics related to lateral stability. Step width and step width variability have been correlated to fall risk and are particularly relevant to lateral stability.34 Step width was defined as the mediolateral distance between the heel markers at subsequent heel-strikes (Fig. 1). We quantified the step width variability as the standard deviation of the step width across all steps in a given condition.
Diagram illustrating the measurement of step width and margin of stability (MOS). The MOS is calculated as the mediolateral distance between the lateral edge of the foot and an adjusted position of the center of mass (COM) that also takes into account the COM velocity (COMvel) to compute an “extrapolated” center of mass (XCOM).
The dynamic margin of stability (MOS) quantifies the relationship between the movement of the center of mass and the base of support (Fig. 1). This measure is calculated as the minimum distance between the “extrapolated center of mass,” which accounts for the motion of the center of mass, and the edge of the base of support, defined as the lateral portion of the stance foot, during the stance phase and is used to determine the mechanical stability of a step. A negative value indicates that the center of mass is moving outside of the base of support, denoting an unstable step.27,35–37 The MOS was used in several ways to quantify the patient's walking stability. We used the percentage of unstable (negative MOS) steps per condition as a measure of how stable the patient was across each task. We also created a first-return plot to visualize how the stability of the patient changed from step to step. The first-return plot graphs the MOS on one step (step n) versus the MOS on the next step (step n+1). This plot allowed us to define each step as “stable” (a positive MOS followed by a positive MOS), “destabilizing” (a positive MOS followed by a negative MOS), “unstable” (a negative MOS followed by a negative MOS), or “recovering” (a negative MOS followed by a positive MOS).36 Lastly, we calculated the variability of the MOS for each condition as a global measure of relative stability control.
These metrics were compared with normative values obtained from studies previously conducted in the same facility with participants who were healthy and able-bodied.32,36,37–39 We considered a value outside of the mean (±1 standard deviation) of the normative data to indicate a functional or stability deficit.
To track the effect of the intervention training, we completed identical evaluations at a mid-training time point halfway through the intervention and posttraining at the conclusion of the intervention. We also conducted a follow-up evaluation 5 weeks after the posttraining test to assess retention. In addition to comparisons with normative values, when available, we used published minimal detectable change (MDC) values to identify clinically meaningful improvements.11,39 After the follow-up evaluation, the patient completed an exit survey regarding perceived stability and balance confidence.
Clinical Impressions
During the initial examination, we identified several functional and stability deficits. The self-selected walking speed on the level walkway was 1.21 m/s, which is below the normal threshold and even below the fifth percentile for male military service members39 (Fig. 2B), indicating decreased functional ability. In relation to stability measures, the patient's step width was above the normal threshold for all conditions (Fig. 3A). His step width variability also was greater than normative values during the VIS and NOP conditions (Fig. 3B). During the PLAT condition, the patient's step width variability was inside, but still on the high end of the normative threshold. These wider and more variable steps likely indicate the patient's lack of confidence in his own balance control and suggest that this individual had a greater risk of falling.34,40
(A) Functional stepping test times and (B) self-selected walking speed values at the 4 time points (examination, mid-training, posttraining, and follow-up). Improved locomotor function is indicated by decreased stepping times and increased walking speed.
(A) Step width mean and (B) variability at the 4 time points (examination, mid-training, posttraining, and follow-up). Improved lateral stability is indicated by decreased mean step width and step width variability. The dashed lines indicate the normal threshold calculated as the mean (±1 standard deviation) of normative data previously collected in the laboratory. A value above the line indicates a functional or stability deficit. VIS=continuous, pseudorandom mediolateral visual scene oscillations; PLAT=continuous, pseudorandom mediolateral platform oscillations; and NOP=unperturbed walking.
The patient had the greatest difficulty with the VIS condition, taking unstable steps over 7% of the time (Fig. 4A), greater than observed in able-bodied individuals.36 For the NOP and PLAT conditions, the patient's percent of unstable steps were within the reference values. Although the MOS variability during the VIS and NOP conditions was greater than typically observed in individuals without impairments, all initial testing values were within normal ranges (Fig. 4B).
(A) Number of unstable steps and (B) margin of stability (MOS) variability at the 4 time points (examination, mid-training, posttraining, and follow-up). Steps were considered unstable if the MOS was negative. Improved lateral stability is indicated by fewer unstable steps and decreased MOS variability. The dashed lines indicate the normal threshold calculated as the mean (±1 standard deviation) of normative data previously collected in the laboratory. A value above the line indicates a functional or stability deficit. VIS=continuous, pseudorandom mediolateral visual scene oscillations; PLAT=continuous, pseudorandom mediolateral platform oscillations; and NOP=unperturbed walking.
Temporal-spatial parameters from the initial examination also indicated that the patient was most challenged by the VIS condition. Furthermore, step width, step width variability, and percent of unstable steps were all well above the normative ranges during at least one condition. These findings indicated that the patient had stability deficits and would likely benefit from a gait training intervention that promoted effective stability responses to perturbations.
As the intervention used a combination of virtual environments and perturbation-based gait training with a novel perturbation scheme (ie, continuous pitch and roll oscillations), it was important to determine the patient's tolerance prior to completing a full-scale trial. The patient's specific injury made the intervention highly relevant, and his extensive experience as a community ambulator made him an ideal candidate to evaluate the intervention.
Intervention
To improve lateral walking stability, the patient walked in a progressively destabilizing virtual environment. The intervention consisted of 2 training sessions a week for 4 weeks for a total of 8 training sessions. Each training session included a 3-minute warm-up walking on the level treadmill followed by eight 3-minute variable terrain trials with increasing difficulty and a 3-minute cool-down, again walking on the level treadmill. This procedure resulted in a total of 30 minutes of walking for each training session.
Unlike typical perturbation-based gait training, which uses discrete perturbations, usually of treadmill belt speed, we used continuous walking surface angle perturbations to destabilize the patient in order to encourage him to develop strategies that maintained stability. To specifically avoid “training to the test,” we took advantage of our ability to precisely control and customize perturbations by pitching and rolling the walking surface to create a constantly changing virtual environment. Based on feedback from previous patients in the facility and the situations where they had the most difficulty, we placed the intervention in a real-world context similar to the challenging variable terrain encountered while walking outdoors, hiking, or playing golf, which was a specific goal of the patient.
While the patient walked on the treadmill through a canyon scene, the platform went through a seemingly random sequence of perturbations with a prescribed maximum magnitude (Fig. 5A; videos 1 and 2). Under the direction of a physical therapist, the difficulty of the task was systematically increased, based on the patient's tolerance and performance, by adjusting walking speed and maximum pitch and roll magnitude across trials within each training session and across the 8 training sessions (Fig. 5B). By the end of the intervention, the patient was walking at 1.27 m/s with up to 9.3 degrees of pitch and roll (see videos 1 and 2).
Illustration of the training paradigm, including (A) training environment and platform state profiles and (B) training progression. The training progressions show the maximum pitch and roll of the platform for each of the 8 trials within each of the 8 training sessions.
During the training sessions, the patient received no explicit instructions on how to modify his gait. The goal of the intervention was to promote appropriate stability responses through repeated exposures to a wide range of continuous, random physical (walking surface) perturbations. In this context, the patient could explore and develop strategies to successfully adapt and complete the walking task. This approach was important, as each patient is unique and has unique capabilities. Thus, there may not necessarily be one “correct” strategy to be pursued.
Outcomes
Throughout the course of the intervention, the patient improved his functional and mediolateral stability measures. By the posttraining evaluation, the patient had decreased his functional stepping test time by 13% from the pretraining evaluation (Fig. 2A). He also increased his self-selected walking speed by 0.09 m/s, from 1.21 m/s to 1.30 m/s, a change greater than the 0.05 m/s MDC,39 which brought it above the fifth percentile and nearly within normative ranges (Fig. 2B). Step width decreased by more than the condition-specific MDC11 for all conditions. The postintervention step width values reduced to within normative ranges for the VIS and PLAT conditions (Fig. 3A). Although the final values of the patient's step width were still above those of able-bodied individuals, their baseline decreased by more than the MDC of 0.58 cm during the NOP condition.11 The patient's step width variability also was reduced by more than the condition-specific MDC to within normal ranges for the VIS and NOP conditions (Fig. 3B). During the PLAT condition, the step width variability remained within normal ranges, with no notable difference between posttraining and pretraining values.
The number of unstable steps for the PLAT condition remained around the preintervention level of below 2%, similar to the NOP condition, which had no unstable steps except for one during the pretraining evaluation (Fig. 4A). However, the patient was able to nearly eliminate all of the unstable steps during the VIS condition following the training. The MOS first-return plot for the VIS condition (Fig. 6) clearly shows the progression from many inconsistent, destabilizing, and even unstable steps at the pretraining evaluation to a controlled grouping of stable steps at the posttraining evaluation. Although the MOS variability started within normal ranges for all conditions, the patient did decrease the variability compared with baseline during the NOP and VIS conditions (Fig. 4B). The MOS variability during the PLAT condition was less consistent, showing no change on the intact side and a slight increase on the prosthetic side.
Margin of stability (MOS) first-return plot displaying the MOS for each step n versus the next step n+1. The quadrants categorize the sequence of steps as stable, recovering, unstable, and destabilizing.
The improvements observed at the end of training were retained or further enhanced at the 5-week follow-up evaluation (Figs. 2, 3, and 4). These included improvements in functional stepping times, self-selected walking speed, step width, step width variability, percent stable steps, and MOS variability.
In the exit survey after the follow-up evaluation, the patient noted that he felt that, even after years of walking with the prosthesis, walking became easier. He attributed this improvement to his stride feeling more “efficient.” He also commented that he walked more upright following the intervention. The patient attributed much of the benefits of the intervention to challenging him to learn how to focus on his surroundings and make the necessary gait and postural changes without the aid of anticipation.
Discussion
There were notable improvements in measures of function and mediolateral walking stability after the intervention. The benefits included improvements in global functional ability, as indicated by decreased functional stepping test times and increased self-selected walking speed. Additionally, benefits were observed in walking stability, as shown by reduced step width, step width variability, and MOS variability. These improvements in lateral stability were seen across all 3 conditions, although less evident in the PLAT condition. Although the patient was most challenged by the VIS condition (had the largest pretraining values), he also showed the greatest improvements during that condition, effectively eliminating unstable steps and bringing all measures to within normal ranges (Figs. 2, 3, 4, and 6). This improved response to the visual perturbation was achieved even though the intervention did not manipulate the visual scene, only the platform state.
Step width variability and MOS variability did not differ greatly between pretraining and posttraining evaluations. These findings were likely due to the fact that the PLAT condition induced variability of movement as a result of the random physical perturbations. This also would explain why the patient was within the normal range for both variability measures, as walking surface perturbations require more reactionary changes to maintain stability. There are limited strategies that can be adopted to respond to the physical perturbations of the walking surface, thus it is not surprising that there would be little change in the variability measures. Although there was no notable improvement in step width and MOS variability, the patient did reduce his mean step width during the PLAT condition, showing improvement in stability following the intervention.
Although mean step width was used as a measure of walking stability and fall risk, the reduced step width following the intervention likely provided additional benefits. Walking with a wide step width increases an individual's base of support, making the person more stable.40,41 However, an exceptionally large step width is associated with a cautious gait pattern and is likely more related to balance confidence and fear of falling.34 Walking with a wide step width increases the base of support, but it also increases the metabolic cost of walking.40,42 Thus, during the intervention, the patient likely developed effective balance strategies and became more confident in his walking. These improvements allowed him to reduce his step width, which likely had the secondary benefit of reducing the metabolic cost of walking.
The quantitative measures in the evaluations aligned well with the patient's qualitative responses to the exit survey. The patient's comment that he felt like he walked more efficiently can likely be attributed to his reduced mean step width. By walking more upright, as noted in the survey, and with narrower steps, he likely reduced his center of mass sway, which was seen in the reduced MOS variability during NOP and VIS.
These improvements were likely derived primarily from exposure to the destabilizing perturbations and the patient's self-exploration, as the therapist guided the progression of the training but did not provide any direct instructions to the patient on how to modify his gait or specific strategies to use. In addition, the improvements transferred from training to testing, even though the nature of the perturbations (pitch and roll versus mediolateral translations) and the visual scenes in the virtual environment were different. This finding suggests that benefits can be gained by exposing patients to unstable walking conditions in a safe, controlled environment without the need for direct feedback instructions. In the exit survey, the patient attributed the success of the intervention to challenging his stability and forcing him to make gait modifications without relying on anticipation.
This type of training is important for a number of reasons. Although anticipatory strategies can be effective, unanticipated disturbances pose the greatest fall risk. Thus, patients need to develop the skills and strategies required to effectively respond to these unexpected challenges. In addition, every patient is unique with respect to his or her sensory and motor deficits. Therefore, the best strategy for any given circumstance may differ, making it important for the patient to explore many different options to determine what works best for him or her.
Although some more elaborate systems can be large and expensive, virtual environments with movable walking surfaces are highly beneficial for this type of gait training because they can provide a rich and diverse range of experiences and challenges with enhanced safety and control. In addition, unlike standard treadmill-based perturbation gait training, a virtual environment provides immersive visual scenes with appropriate optic flow. Thus, patients are free to find their own optimal strategies to a diverse set of challenges through self-exploration without fear of injury.
All of the measured benefits were shown to be robust and to remain even 5 weeks after the last training session. This finding is of special importance because the impact of an intervention relies not only on the ability to improve performance but also on the ability to retain those improvements. The results of this intervention were especially promising because the patient was many years out from his injury and was believed to have hit his recovery plateau. The fact that we saw such marked improvements in this well-established individual strongly suggests that larger studies are warranted to determine whether similar improvements to walking function and lateral stability could be achieved in a larger pool of similar patients, or possibly by other patients in varying states of their rehabilitation.
Overall, although larger-scale studies are still needed, the perturbation-based gait training intervention in a virtual environment demonstrated promise in improving mediolateral walking stability measures in an individual with a unilateral TFA. This case report further adds to the existing literature supporting the use of both virtual environments6,9–12 and perturbation-based gait training5–8 as effective rehabilitation tools. It also extends the work of Kaufman et al6 to include a different level or injury and different means of perturbation training and evaluation within a virtual environment. Although a therapist was present to ensure safety and guide progression, the benefits were derived through exposure and self-exploration without direct therapist instruction regarding strategy. This intervention resulted in the benefits being retained at least 5 weeks following the conclusion of the intervention. However, the intervention occurred many years after this patient had completed his initial rehabilitation. We did not have data from his discharge from rehabilitation, nor from any point during the intervening years, until this case report. However, it is possible this individual may have slowly regressed over time after completing the initial rehabilitation. Further work is needed to determine whether a maintenance program is needed to retain the training benefits beyond the 5 weeks tested in this rehabilitation program. However, it also is possible that the increased activity following a successful intervention would provide a sufficient exposure to destabilizing situations that their continued activity may be enough to maintain the initial benefits and slow any regression.
Although exposure to destabilizing conditions is a key aspect of many rehabilitation programs, the ability to systematically progress the difficulty is a significant capability of virtual reality systems and a strength of this intervention. This fact supports the promise of integrating perturbation-based gait training with virtual reality to promote gait stability in rich, challenging environments and provides the foundation for larger studies in the future to confirm and extend the results.
Footnotes
Dr Rylander, Dr Dingwell, and Dr Wilken provided concept/idea/project design. Dr Sheehan, Dr Rábago, Dr Dingwell, and Dr Wilken provided writing. Dr Sheehan, Dr Rábago, and Dr Rylander provided data collection. Dr Sheehan, Dr Rylander, and Dr Dingwell provided data analysis. Dr Sheehan, Dr Rylander, Dr Dingwell, and Dr Wilken provided project management. Dr Dingwell and Dr Wilken provided fund procurement and institutional liaisons. Dr Wilken provided facilities/equipment. All authors provided consultation (including review of manuscript before submission). The authors thank Michael Vernon for designing and operating the perturbation-based gait training program in the virtual reality environment.
This work was supported by National Institutes of Health grant 1-R01-HD059844 and DoD/CDMRP/BADER Consortium W81XWH-11-2-0222 (to J.B.D. and J.M.W.).
The views expressed herein are those of the authors and do not reflect the official policy or position of Brooke Army Medical Center, the US Army Medical Department, the US Army Office of the Surgeon General, the Department of the Army and Department of Defense, or the US Government.
- Received October 19, 2015.
- Accepted May 30, 2016.
- © 2016 American Physical Therapy Association