Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Feb 1.
Published in final edited form as: Clin Biomech (Bristol). 2020 Dec 23;82:105249. doi: 10.1016/j.clinbiomech.2020.105249

Posterior Fall-Recovery Training Applied to Individuals with Chronic Stroke: A Single-Group Intervention Study

Jamie Pigman 1,6, Darcy S Reisman 2, Ryan T Pohlig 3, John J Jeka 1, Tamara R Wright 2, Benjamin C Conner 1,4, Drew A Petersen 1,5, Michael S Christensen 1, Jeremy R Crenshaw 1,*
PMCID: PMC7940569  NIHMSID: NIHMS1660479  PMID: 33421756

Abstract

Background:

To assess the effects of the initial stepping limb on posterior fall recovery in individuals with chronic stroke, as well as to determine the benefits of fall-recovery training on these outcomes.

Methods:

This was a single-group intervention study of 13 individuals with chronic stroke. Participants performed up to six training sessions, each including progressively challenging, treadmill-induced perturbations from a standing position. Progressions focused on initial steps with the paretic or non-paretic limb. The highest perturbation level achieved, the proportion of successful recoveries, step and trunk kinematics, as well as stance-limb muscle activation about the ankle were compared between the initial stepping limbs in the first session. Limb-specific outcomes were also compared between the first and last training sessions.

Findings:

In the first session, initial steps with the non-paretic limb were associated with a higher proportion of success and larger perturbations than steps with the paretic limb (p=0.02, Cohen’s d=0.8). Paretic-limb steps were wider relative to the center of mass (CoM; p=0.01, d=1.3), likely due to an initial standing position with the CoM closer to the non-paretic limb (p=0.01, d=1.4). In the last training session, participants recovered from a higher proportion of perturbations and advanced to larger perturbations (p<0.05, d>0.6). There were no notable changes in kinematic or electromyography variables with training (p>0.07, d<0.5).

Interpretation:

The skill of posterior stepping in response to a perturbation can be improved with practice in those with chronic stroke, we were not able to identify consistent underlying kinematic mechanisms behind this adaptation.

Keywords: Stroke, Balance, Falls, Stability, Rehabilitation, Perturbation Training

1. Introduction

Chronic stroke is the leading cause of long-term disability in the United States (CDC, 2009). Those with chronic stroke have a fall risk that is twice that of age- and sex-matched peers (Jorgensen and Jacosen, 2002). As many as 75% of those with stroke fall annually (Batchelor et al., 2012; Forster and Young, 1995; Gordon and Morris, 2008), and 84% of fractures in this population are due to a fall (Ramnemark et al., 1998). Trips and slips cause one-third of post-stroke falls (Schmid et al., 2013a). Therefore, interventions that improve reactions to these common causes are a promising means to prevent falls and injuries.

Individuals living with chronic stroke have an impaired response to a standing, posterior loss of balance. Compared to people with no stroke history, stroke survivors had posterior recovery steps that were shorter and closer to the whole-body center of mass (CoM), resulting in a 54–71% failure rate compared to 0% in controls (Patel and Bhatt, 2018; Salot et al., 2016). Survivors of stroke have lower posterior multiple-stepping thresholds, defined as the smallest standing perturbations that elicit more than one step (de Kam et al., 2017; Handelzalts et al., 2019b), as well as fall-thresholds, or the perturbations that elicit a fall into a harness (Handelzalts et al., 2019b). In another study of standing perturbations, those with stroke who fell into a harness had delayed, paretic-limb muscle activity (Marigold and Eng, 2006). Such deficiencies were associated with more trunk rotation and a CoM positioned further outside the base of support. From these studies, it is apparent that stroke-related impairment alters fall-recovery kinematics, reducing the capacity to recover.

The asymmetry of stroke-related function should, in theory, influence posterior balance reactions. Individuals with stroke prefer to step with their non-paretic limb (Handelzalts et al., 2019a, 2019b; Lakhani et al., 2011; Patel and Bhatt, 2015a), often avoiding a paretic step (Kajrolkar and Bhatt, 2016). These observations were made from standing, laboratory-induced perturbations—a context in which preference dictates the stepping limb. Perturbations in the free-living environment, such as those from a slip, often occur while walking (Schmid et al., 2013b). In this context, the stepping limb is not only influenced by preference, but circumstances such as the limb that is perturbed. Initial, non-paretic steps may not be feasible, necessitating steps with the less-preferred, paretic limb. The effects of the stepping limb on balance reactions, however, have been inconsistent across studies. Against our expectations, slip-recovery success rates were not affected by the stepping limb (i.e. paretic or non-paretic) (Kajrolkar and Bhatt, 2016; Patel and Bhatt, 2015a). In two separate studies, step-initiation times were not different between limbs (Inness et al., 2016; Kajrolkar and Bhatt, 2016). In response to standing perturbations, those with stroke have demonstrated shorter, less-stable posterior steps with the paretic limb (Patel and Bhatt, 2015a). However, a second study that delivered perturbations during walking observed that paretic, slip-recovery steps were placed in a more stable position relative to the CoM (Kajrolkar and Bhatt, 2016). The greater stability of paretic-limb steps was attributed to better control of the non-paretic stance limb. During a feet-in-place response to a standing perturbation, the paretic limb had a diminished, delayed muscle response characterized by co-contraction (Badke et al., 1987; Di Fabio et al., 1986; Richard P. Di Fabio, 1987; Di Fabio and Badke, 1988; Ruth Dickstein et al., 1989; Dickstein et al., 1994; R Dickstein et al., 1989; Diener et al., 1985; Dietz and Berger, 1984; Kirker et al., 2000; Marigold et al., 2004). Muscle activity of the stance limb during posterior stepping, however, has not been evaluated.

Posterior balance reactions of stroke survivors can be improved with practice (Bhatt et al., 2019; Handelzalts et al., 2019a; Kajrolkar et al., 2014; Punt et al., 2019; Schinkel-Ivy et al., 2019; Van Duijnhoven et al., 2018). In a study of repeated walking slips targeting the non-paretic limb, individuals with chronic stroke were able to modify responses on the second trial (Kajrolkar et al., 2014). Within-session adaptations included a more stable position of the whole-body CoM at foot-off, as well as longer steps. Perturbation-based training has improved standing reactive balance by reducing the number of steps taken, although direction-specific results and kinematic adaptations were not detailed (Schinkel-Ivy et al., 2019). Those with sub-acute stroke improved standing, posterior multiple-stepping thresholds with perturbation-based training, but kinematic mechanisms underlying this improvement were not explored (Handelzalts et al., 2019a). In another study, individuals with chronic stroke improved posterior reactive step kinematics with practice, placing the step further beyond the pelvis (Van Duijnhoven et al., 2018). Additionally, those with chronic stroke were able to increase posterior step lengths and improve stability in response to a standing perturbation after undergoing a single session of perturbation training (Bhatt et al., 2019). These studies suggest that perturbation training can improve reactive stepping in stroke survivors. None of these studies, however, assessed adaptations of both paretic and non-paretic steps. It is likely that asymmetric function associated with chronic stroke not only alters the stepping response, but it’s capacity for and specific mechanisms of adaptation.

The purpose of this study was to assess the effects of the initial stepping limb (i.e. paretic or non-paretic) on posterior fall-recovery performance and kinematics, as well as to determine the limb-specific benefits of fall-recovery training on those outcomes. We hypothesized that compensatory steps with the paretic limb would be associated with less effective fall-recovery performance and kinematics. We also hypothesized that fall recovery would improve with training. Performance was quantified as the proportion of successful recoveries within a series of perturbations, as well as the largest perturbation successfully met within that series. "Less effective" kinematic features included shorter and wider steps relative to the CoM, more posterior trunk rotation, and greater lowering in pelvic height (Cham and Redfern, 2001; Grabiner et al., 2008; Patel and Bhatt, 2015a; Redfern et al., 2001; Troy et al., 2008; Yang et al., 2012). To explore the neuromuscular response, we evaluated surface EMG of the non-stepping, stance limb plantar- and dorsi-flexors. We expected the paretic limb to be characterized by a delayed response with less dorsiflexor activity and more co-contraction between plantar- and dorsiflexor muscles. We also expected that training would result in less delay and greater magnitudes of dorsiflexor activity, as well as less co-contraction.

2. Methods

2.1. Participants

From the University of Delaware’s Stroke Studies Registry, we recruited 18 participants. Exclusion criteria included more than one stroke, a body mass greater than 136 kg (as to not interfere with the treadmill acceleration), any joint replacement within a year before participation, dementia, Parkinson’s disease, a history of back or neck injury, surgery, or pain in the month before participation, bone fracture within a year before participation, open lesions, insulin-dependent diabetes, use of a pacemaker, use of an ostomy pouch, pregnancy, or a doctor’s recommendation to avoid moderate physical activity. Participants had a self-reported ability to walk a city block (≈ 100 m) without assistance. Those who were 50 years of age or older underwent a Dual-energy X-ray absorptiometry (DXA) screening to ensure that they were not osteoporotic (total hip or femoral neck bone mineral density t-score < −2.5) (Kanis et al., 2009). This screening criterion, which we have used previously (Crenshaw et al., 2018), was in place to reduce the risk of fractures from stepping or falling into a harness. This study was approved by the University of Delaware’s Institutional Review Board, and all participants provided written informed consent. Participants and staff that appear in photographs and videos gave written consent.

2.2. Training Protocol

Our training was garnered from a previous protocol applied to those with no stroke impairment (Yang et al., 2013). The standing perturbations delivered within our training, which consisted of anterior treadmill belt translations (ActiveStep®, Simbex, Lebanon, NH; Figure 1), were designed to necessitate rapid posterior steps similar to that of slip-recovery (Patel and Bhatt, 2015b; Yang et al., 2013). Participants attempted to complete six training sessions, a number that aligns with previous applications to those with lower-extremity amputations (Crenshaw et al., 2013), as well as to those with sub-acute stroke (Mansfield et al., 2011). Sessions consisted of two limb-specific progressions, or sets, of perturbations (Figure 2). The first progression focused on initial steps with the non-paretic limb, and the second progression focused on initial steps with the paretic limb. Progressions were limited to either 10 minutes or 18 perturbations, whichever occurred first, with rest periods lasting approximately 5 minutes. Prior to these posterior progressions, two progressions focused on standing, anterior perturbations were delivered, the results of which have been reported previously (Pigman et al., 2019). Six sessions occurred over about three weeks.

Figure 1:

Figure 1:

Open in a new tab

The top series (red) shows a posterior fall during the first training session stepping with the non-paretic limb. The bottom series (green) shows a successful fall-recovery at the same initial belt acceleration of 5.0 m/s2 on the sixth training session stepping with the non-paretic limb.

Figure 2.

Figure 2.

Open in a new tab

Training progressions were based on participant responses to each perturbation. Within each of six training sessions, separate progressions were conducted for initial steps with the paretic or non-paretic limb. The first perturbation within each progression had an initial acceleration of 0.5 m/s2.

For each laboratory visit, participants wore their own well-cushioned, closed-toe shoes. Two participants wore prescribed articulating ankle-foot orthosis as worn daily. Participants were outfitted with a safety harness (Delta™, Capital Safety, Bloomington, MN) attached to an overhead rail. Support straps were adjusted so that hands and knees could not contact the treadmill. The harness was instrumented with a transducer (Dillon, Fairmont, MN) that recorded peak forces.

When awaiting a perturbation, participants stood on the treadmill, feet at a comfortable width and toes evenly positioned anteroposteriorly (Figure 1). The perturbation velocity followed a triangular waveform, consisting of 200 ms acceleration and deceleration phases. Participants were instructed to "try to recover in one step" with the non-paretic or paretic limb. This step-number constraint was established because slip recovery, an applicable target skill, is dictated by the first-step features (Troy et al., 2008). We did not want to encourage abbreviated, initial steps with the paretic limb, done with the intent of quickly arresting the fall with a second, non-paretic step (Patel and Bhatt, 2016). Also, the single-step constraint directly addressed the previous observation that those with stroke take more steps compared to those with no stroke (de Kam et al., 2017; Handelzalts et al., 2019b; Patel and Bhatt, 2016; Salot et al., 2016).

The first perturbation of each progression had an initial acceleration of 0.5 m/s2, resulting in a peak velocity of 0.1 m/s and a total displacement of 0.02 m. After a successful recovery, the subsequent perturbation had an acceleration +0.5 m/s2 greater than the previous perturbation, resulting in +0.1 m/s peak velocity and +0.02 m displacement (Crenshaw et al., 2013). After a failed recovery, the subsequent trial acceleration was reduced by −0.5 m/s2 (a flow chart of protocol decisions is in Figure 2). Failures were defined as reactions in which the participant engaged the harness with >20% body weight (Cyr and Smeesters, 2009), stepped with the wrong limb, or took multiple steps. Non-stepping responses, which occurred in response to the smallest perturbations, were permissible. Available treadmill displacements were 0.01–0.64 m, with peak velocities of 0.1–3.2 m/s and initial accelerations of 0.5–16.0 m/s2. Each perturbation was preceded by a 1–5 s delay, and 5 cm anterior perturbations were introduced about once every six trials to limit anticipatory adjustments. Participants were to inform research staff if the training intensity was intolerable. In such cases, training continued at the largest perturbation tolerated for the remainder of the session. A video example of perturbations is available as an online supplement.

Three-dimensional motion and surface electromyography were recorded for the first and last sessions. Kinematics were recorded with a 12-camera system (Motion Analysis®, Santa Rosa, CA, replaced mid-study with Qualisys®, Göteborg, Sweden; 120 Hz). The positions of 35 passive-reflective markers facilitated the definition of 13 body segments, using a modified Helen Hayes marker set (Kadaba et al., 1990). Missing marker trajectories were filled using linear or polynomial interpolation, with gaps less than 10 frames in length. Marker trajectories were filtered with a low-pass Butterworth filter (4th order, 6 Hz cutoff).

Muscle activity was recorded using surface electromyography (Delsys, Natick, MA, 1200 Hz) of the bilateral medial gastrocnemius (MG) and tibialis anterior (TA) (Celinskis et al., 2018; Marigold and Eng, 2006, Hermens et al., 2000). Signals were demeaned, bandpass filtered (1st order polynomial, 10–300 Hz), rectified, and lowpass filtered (8th order Butterworth) at 50 Hz for muscle onset latency and 4 Hz for peak activation and co-contraction ratios (R. P. Di Fabio, 1987). To limit the visit duration, middle sessions were not instrumented with motion capture or electromyography.

2.3. Analysis

Fall-recovery performance was quantified from the proportion of successful recoveries and the largest perturbation magnitude achieved. To determine how lower-extremity impairment affected performance at baseline, we compared outcomes from stepping with the paretic or non-paretic limb within the first training session. To evaluate if these measures changed with training, we compared limb-specific outcomes on the first and last training sessions.

Custom LabView software (National Instruments, Austin, TX) was developed to calculate biomechanical variables. Our hypotheses were focused on foot placement (i.e. the heel marker) relative to the whole-body CoM, maximum trunk rotation angles and angular velocities, and maximum displacement in the height of the pelvic center. The whole-body COM was determined from the weighted average of segment COM locations, as per anthropometric estimates (Winter, 2009). To be comprehensive, we reported values at perturbation onset, foot off, and foot strike, and we reported step length, width, and timing. The EMG variables were tibialis anterior onset latency and peak activation, as well as the co-contraction ratio of the tibialis anterior and medial gastrocnemius. The technical definitions of all variables are detailed in supplemental material.

To determine if the stepping limb (paretic or non-paretic) altered kinematic or EMG variables, we compared single trials from the first training session. In addition, these outcomes were compared from single trials of the first and the last sessions to evaluate a training effect. With these comparisons, successful responses to the highest common perturbation magnitude across limb or session were evaluated, allowing us to evaluate the response at a relatively challenging level for each individual. Note that between-limb and between-session comparisons were not necessarily made at the same perturbation magnitudes.

Between-limb and between-session comparisons of performance, kinematics, and EMG were made with paired t-tests (SPSS 25, IBM, Armonk, NY, α=.05) and effect sizes (Cohen’s d; Lakens, 2013). We assumed that 15 participants would complete the study, providing adequate power (0.80) to detect a large effect (d=0.77) as significant. This threshold was sufficient to detect between-limb effects of the largest perturbation (d=0.80) and the step width relative to the CoM (d=1.16), as per preliminary results of eight participants. With 15 participants, we were adequately powered to detect preliminary training effects on performance variables (d=0.8–1.3).

3. Results

Thirteen (10 men, 3 women, Hemisphere of stroke: 11 right, 2 left) of 18 participants completed at least five of six sessions and were included in this analysis (Table 1). As part of protocol development, our first participant’s training consisted of posterior perturbations delivered while walking. Although our treadmill can deliver walking perturbations relative to gait events, it is not able to discriminate left and right steps. Therefore, we could not administer a limb-specific progression of perturbations. With the second participant, we changed the protocol to standing perturbations, the results of which are reported here. Two participants voluntarily withdrew from the study in the first session due to uneasiness not related to physical discomfort. A third participant withdrew after three sessions due to a seizure away from the laboratory. One participant did not perform posterior progressions on the first day, stopping due to fatigue after anterior perturbation training. This participant later performed all posterior training in the following sessions. There was one non-serious, anticipated event when a participant delayed training due to muscle soreness of their non-paretic hip after the second session. After three days, the soreness subsided, and training was resumed. The data of the three participants that withdrew from training were not used in the analysis. With a sample size of 13, we had 80% power to detect an effect size of 0.85 or larger. Group means and standard deviations are presented as tables, with individual data presented as supplemental material.

Table 1.

Demographic and clinical assessment data (n=13)

Measure Mean (SD), Range
Age (Years) 59 (12), 29 – 77
BMI (kg/m2) 28.9 (3.8), 22.0 – 33.9
Years after stroke 5.2 (3.7), 2 – 15
Fugl-Meyer LE 24 (6), 8 – 32
Activities Specific Balance Confidence Scale (ABC) 91 (8), 76 – 100
Functional Gait Assessment (FGA) 17 (6), 9 – 29
Berg Balance Scale (BBS) 51 (7), 36 – 56
Open in a new tab

Note: Prior to starting fall-recovery training, descriptive measures of the Fugl-Meyer Lower Extremity assessment, (Fugl Meyer et al., 1975) Activities-Specific Balance Confidence (ABC) scale, (Powell and Myers, 1995) Berg-Balance Scale, (Berg et al., 2009) and the Functional Gait Assessment(Wrisley et al., 2004) were used to characterize our participants balance and mobility.

3.1. Between-limb Comparisons at Baseline

In the first session, steps with the non-paretic limb were associated with about a 25% greater success rate (p=0.02, d=0.8) and a larger perturbation achieved (p=0.02, d=0.8; Table 2). Steps with the paretic limb were, on average, placed more than twice as far laterally from the CoM as steps with the non-paretic limb (p=0.01, d=1.3, Table 3). Steps with the paretic limb, however, were also associated with an initial CoM position closer to the stance limb (p=0.01, d=1.4). During non-paretic steps, the pelvis lowered about 2 cm more than during steps with the paretic limb (p=0.05, d=1.0, Table 3). At foot strike, however, no between-limb differences were observed in pelvic height (p=0.15, d=0.6). No between-limb differences were observed in the sagittal-plane trunk or step kinematic variables. The paretic limb in stance had more co-contraction and delays in tibialis anterior activity (p<0.01, d=5.7, Table 3). The between-limb comparison sample size was limited to seven for the kinematic variables and five for the EMG variables. The smaller samples were due to six participants being unable to execute paretic-limb steps at a perturbation size common to that of the non-paretic limb. In addition, the two participants wearing AFOs did not have EMG on their paretic limb.

Table 2.

Between-limb performance on the first session and changes with training between-sessions. (n=13)

Variable Initial Step Limb First Session p-value (Cohen’s d) Change w/training p-value (Cohen’s d)
% Successful Trials (%) Non-Paretic 82 (13) 0.02* (0.8) +8 (13) 0.05* (0.6)
Paretic 57 (27) +15 (19) 0.01* (0.8)
Largest Perturbation (m/s2) Non-Paretic 4.0 (1.1) 0.02* (0.8) +0.6 (0.6) 0.004* (1.0)
Paretic 2.8 (1.4) +0.4 (0.7) 0.04* (0.6)
Open in a new tab

Note: Data from the first session, as well as the change observed on the last training session, are displayed as mean (SD).

*

Significant (p < 0.05) between-session differences from the first session and the last sessions of training.

Table 3.

Between-limb comparisons of kinematic variables. First training session. (n=7)

Category Variable Stepping Limb p-value (Cohen’s d)
Non-Paretic Paretic
Step timing after perturbation Foot off time (ms) 263 (59) 301 (58) 0.11 (0.7)
Foot strike time (ms) 536 (121) 585 (89) 0.32 (0.4)
Posterior foot placement CoM to stance-limb heel distance at perturbation onset (cm)† 13.4 (2.6) 14.0 (0.9) 0.52 (0.3)
CoM to stance-limb heel distance at foot off (cm)† 3.9 (2.9) 4.3 (3.2) 0.79 (0.1)
CoM to stepping-limb heel distance at foot strike (cm)† 20.9 (3.7) 22.1 (10.0) 0.79 (0.1)
Step length (cm) 22.9 (8.1) 22.1 (14.0) 0.89 (0.1)
Lateral foot placement CoM to stance-limb heel distance at perturbation onset (cm)† 10.6 (3.0) 6.5 (3.4) 0.01* (1.5)
CoM to stance-limb heel distance at foot off (cm)† 10.7 (2.4) 6.5 (3.5) 0.01* (1.5)
CoM to stepping-limb heel distance at foot strike (cm)† 5.1 (2.9) 12.1 (3.5) 0.01* (1.4)
Step width (cm) 17.8 (5.8) 21.2 (7.6) 0.40 (0.3)
Posterior Trunk Rotation Angle Trunk angle at perturbation onset (deg) 58.2 (8.8) 56.2 (10.3) 0.24 (0.5)
Change in trunk angle at foot off (deg)‡ −1.7 (1.9) −0.2 (1.7) 0.21 (0.5)
Change in trunk angle at foot strike (deg)‡ −3.7 (5.9) −8.5 (8.7) 0.38 (0.4)
Max change in trunk angle (deg)‡ −8.7 (4.4) −11.3 (8.9) 0.47 (0.3)
Time of maximum trunk angle (ms) 554 (193) 675 (96) 0.11 (0.7)
Posterior Trunk Rotation Angular Velocity Trunk angular velocity at perturbation onset (deg/s) −0.7 (1.7) −1.3 (2.5) 0.60 (0.2)
Trunk angular velocity at foot off (deg/s)‡ −8.2 (51.6) −30.9 (45.5) 0.52 (0.3)
Trunk angular velocity at foot strike (deg/s)‡ −7.4 (33.5) −31.4 (36.1) 0.27 (0.5)
Max trunk angular velocity (deg/s)‡ −72.6 (25.6) −71.0 (37.4) 0.91 (0.0)
Time of maximum trunk angular velocity (ms) 357 (122) 458 (133) 0.19 (0.6)
Pelvic Lowering Pelvic lowering at foot off (cm) 0.3 (0.8) 0.4 (0.9) 0.90 (0.0)
Pelvic lowering at foot strike (cm) −2.6 (2.8) −0.8 (1.4) 0.18 (0.6)
Max pelvic lowering (cm) −3.0 (2.8) −0.7 (0.7) 0.05* (0.9)
Time of max pelvic lowering (ms) 388 (174) 243 (136) 0.10 (0.7)
Stance limb EMG Variables (n=5) Paretic Stance Limb Non-Paretic Stance Limb p-value (Cohen’s d)
Tibialis anterior onset latency (ms) 103 (22) 65 (23) <0.01* (5.7)
Tibialis anterior peak activation (units of baseline activity) 23.5 (24.1) 41.6 (33.6) 0.27 (0.6)
Co-contraction ratio 0.54 (0.5) 0.10 (0.1) 0.07 (1.1)
Open in a new tab

Note: Data are displayed as mean (SD). Between-limb kinematic comparisons were limited to 7 participants, as opposed to 13, because 6 participants did not have similar-magnitude perturbations between stepping limbs. In other words, they did not step with the paretic limb, or they only stepped with the paretic limb in response to small perturbations that did not elicit a step with the non-paretic limb. Between-limb EMG comparisons were further limited to 5 participants because 2 participants wore ankle foot orthosis that prevented EMG sensor placement.

*

Significant (p < 0.05) between-limb differences on the first session of training at a common perturbation magnitude between limbs.

†

For distances related to the CoM, positive values denote a CoM position that is either anterior or medial to the heel marker.

‡

Negative values denote posterior trunk rotation or rotational velocity.

3.2. Between-Session Comparisons: Non-Paretic Steps

From the first to last sessions, participants increased the proportion of successful recoveries (p=0.05, d=0.6) and progressed to larger perturbations (p=0.004, d=1.0, Table 2). In the first session, 1% of responses were failures due to stepping with the wrong limb, and 19% of responses were failures due to more than one step. In the last session, these failures were reduced to 0% and 11% of all trials, respectively. No falls into the safety harness were observed when stepping with the non-paretic limb. There were no significant or meaningful changes in kinematic or EMG variables (Table 4).

Table 4.

Between-session comparisons of kinematic variables for paretic and non-paretic initial steps.

Category Variable Non-Paretic Limb Step (n=13) Paretic Limb Step (n=7)
First Session Change w/training p-value (Cohen’s d) First Session Change w/training p-value (Cohen’s d)
Step timing after perturbation Foot off time (ms) 262 (67) +25 (106) 0.42 (0.2) 311 (84) +34 (68) 0.23 (0.5)
Foot strike time (ms) 492 (63) +62 (203) 0.29 (0.3) 607 (103) +27 (66) 0.32 (0.4)
Posterior foot placement CoM to stance-limb heel distance at perturbation onset (cm)† 11.0 (8.2) 1.7 (8.5) 0.49 (0.2) 14.2 (1.2) −0.4 (1.9) 0.61 (0.2)
CoM to stance-limb heel distance at foot off (cm)† −0.3 (4.2) 0.5 (4.6) 0.68 (0.1) 4.0 (4.3) −1.3 (2.7) 0.25 (0.5)
CoM to stepping-limb heel distance at foot strike (cm)† 19.0 (7.7) +1.5 (6.3) 0.40 (0.2) 22.8 (9.6) −3.7 (8.1) 0.27 (0.5)
Step length (cm) 23.1 (13.9) +1.7 (10.5) 0.56 (0.2) 23.8 (13.9) −3.8 (8.5) 0.29 (0.4)
Lateral foot placement CoM to stance-limb heel distance at perturbation onset (cm)† 12.0 (3.8) −1.2 (3.3) 0.21 (0.4) 6.2 (3.3) +1.5 (3.5) 0.30 (0.4)
CoM to stance-limb heel distance at foot off (cm)† 11.6 (3.1) −0.7 (2.9) 0.39 (0.2) 6.6 (3.4) +1.2 (3.9) 0.43 (0.3)
CoM to stepping-limb heel distance at foot strike (cm)† 5.8 (4.2) −1.3 (4.9) 0.36 (0.3) 11.5 (4.8) −0.8 (3.3) 0.53 (0.2)
Step width (cm) 20.7 (8.0) −2.5 (8.3) 0.30 (0.3) 20.7 (9.2) −0.3 (6.7) 0.90 (0.1)
Posterior Trunk Rotation Angle Trunk angle at perturbation onset (deg) 58.2 (8.8) +0.4 (6.1) 0.84 (0.1) 56.2 (10.6) +1.6 (8.2) 0.62 (0.2)
Change in trunk angle at foot off (deg)‡ −0.1 (3.4) +1.0 (2.5) 0.17 (0.4) −0.1 (16) +0.7 (1.7) 0.33 (0.4)
Change in trunk angle at foot strike (deg)‡ 2.4 (8.6) +1.8 (5.6) 0.28 (0.3) −7.9 (6.8) −1.7 (5.5) 0.44 (0.3)
Max change in trunk angle (deg)‡ −6.7 (5.5) +2.9 (5.3) 0.07 (0.5) −10.6 (7.2) −1.3 (6.1) 0.60 (0.2)
Time of maximum trunk angle (ms) 622 (304) −262 (352) 0.02 (0.7) 668 (94) +5 (135) 0.93 (0.0)
Posterior Trunk Rotation Angular Velocity Trunk angular velocity at perturbation onset (deg/s) −0.5 (1.7) −1.4 (3.9) 0.21 (0.4) 0.8 (2.0) +2.0 (18.8) 0.78 (0.1)
Trunk angular velocity at foot off (deg/s)‡ 13.1 (53.6) +4.5 (44.0) 0.72 (0.1) −29.3 (43.5) +16 (53.4) 0.45 (0.3)
Trunk angular velocity at foot strike (deg/s)‡ −14.1 (57.2) −6.1 (64.9) 0.74 (0.1) −21.5 (45.5) −12.1 (54.3) 0.58 (0.2)
Max trunk angular velocity (deg/s)‡ −70.3 (30.3) +8.8 (38.5) 0.43 (0.2) −83.7 (55.6) −15.6 (30.2) 0.22 (0.5)
Time of maximum trunk angular velocity (ms) 497 (141) −6 (214) 0.92 (0.0) 485 (149) −40 (108) 0.37 (0.4)
Pelvic Lowering Pelvic lowering at foot off (cm) 0.4 (0.9) −0.1 (0.4) 0.33 (0.3) 0.5 (0.9) −0.4 (0.7) 0.16 (0.6)
Pelvic lowering at foot strike (cm) −2.2 (2.5) −0.3 (1.2) 0.44 (0.2) −1.1 (1.8) +0.6 (1.2) 0.22 (0.5)
Max pelvic lowering (cm) −2.2 (2.4) +0.2 (1.0) 0.52 (0.2) −0.9 (1.4) −0.1 (0.8) 0.69 (0.2)
Time of max pelvic lowering (ms) 400 (136) +15 (105) 0.62 (0.1) 237 (172) 33 (44) 0.09 (0.8)
Stance Limb EMG Variables Paretic Stance Limb (n=8) Non-Paretic Stance Limb (n=7)
First Session Change w/training p-value (Cohen’s d) First Session Change w/training p-value (Cohen’s d)
Tibialis anterior onset latency (ms) 102 (59) +52 (174) 0.42 (0.3) 85 (23) +7 (47) 0.70 (0.2)
Tibialis anterior peak activation (units of baseline activity) 26.6 (22.3) −3.9 (16.5) 0.52 (0.2) 39.8 (38.2) −4.1 (25.8) 0.69 (0.2)
Co-contraction ratio 0.35 (0.25) −0.02 (0.36) 0.88 (0.1) 0.29 (0.5) −0.06 (0.3) 0.59 (0.2)
Open in a new tab

Note: Data are displayed as mean (SD). Between-session kinematic comparisons when initially stepping with the paretic limb were limited to 7 participants, as opposed to 13, because 6 participants did not have similar-magnitude perturbations between training sessions. Between-session EMG comparisons of the paretic stance limb were limited to 8 participants because 2 participants wore ankle foot orthosis that prevented EMG sensor placement and 3 participants did not have similar-magnitude perturbations between sessions.

*

Significant (p < 0.05) between-limb differences on the first session of training at a common perturbation magnitude between-limbs.

†

For distances related to the CoM, positive values denote a CoM position that is either anterior or medial to the heel marker.

‡

Negative values denote posterior trunk rotation or rotational velocity.

3.3. Between-Session Comparisons: Paretic Steps

From the first to last sessions, participants increased the proportion of successful recoveries (p=0.02, d=0.8) and progressed to larger perturbations (p=0.04, d=0.6, Table 2). Participants increased their proportion of successful recoveries primarily by reducing their tendencies to step with the wrong limb (non-paretic or paretic). In the first session, 35% of responses were failures due to stepping with the wrong limb, 10% of responses were failures due to more than one step, and 1% of responses were failures due to safety-harness support. In the last session, these failures were reduced to 22%, 8%, and 0%, respectively. We observed no significant or meaningful changes in kinematic or EMG variables (Table 4).

4. Discussion

The purpose of this study was to investigate the effect of the initial stepping limb (i.e. paretic or non-paretic) on posterior fall-recovery performance and kinematics, and then determine the benefits of fall-recovery training on those outcomes. We hypothesized that compensatory steps with the paretic limb would be associated with worse performance and kinematics, and that such aspects would improve with training. Our hypotheses were partially supported. At baseline, paretic steps had less frequent success, smaller perturbation limits, altered frontal plane kinematics, and less pelvic lowering. We observed notable performance-based improvements with training (Table 2), but these observations were not aligned with kinematic changes (Table 3).

Paretic steps were associated with a lower success rate (Table 2), primarily due to stepping with the wrong limb. This observation aligns with the previously observed, "aborted" steps of the paretic limb in response to a slip (Kajrolkar and Bhatt, 2016). We also detected between-limb differences in the largest successful perturbation response (Table 2). This is, to our knowledge, the first evidence that paretic-limb steps are associated with lower posterior multiple-stepping thresholds. Paretic-limb steps were placed wider relative to the CoM, which may be problematic for retaining stability. After a slip, a lateral foot placement decreases the likelihood of recovery (Troy et al., 2008). The between-limb differences in lateral step placement were linked to an initial CoM position closer to the non-paretic limb (Table 3). We did not detect or control for this standing asymmetry during data collections, and we do not know how results would be altered with a symmetrical initial posture. This asymmetry is not specific to our standing perturbations, as those with stroke walk with asymmetric COM positioning (Van Meulen et al., 2016). Perhaps the initial postures observed in our study are representative of those taken outside the laboratory.

Steps with the paretic limb were associated with less pelvic lowering (Table 3), perhaps due to more robust support from the non-paretic stance limb. The lowest pelvic position occurred before step completion, with earlier timing during paretic-limb steps. These results contradict previous reports of no between-limb differences in pelvic lowering, but a later timing with paretic-limb steps, typically after foot strike (Patel and Bhatt, 2015a). Between-study differences may be due to definitions of the pelvic height, as we estimated it from the center of the segment, and Patel and Bhatt estimated it from the midpoint of the anterior superior iliac spines. Our measure may be more influenced by posterior pelvic rotation. Additionally, Patel and Bhatt did not constrain responses to single steps, resulting in step lengths of about 9% body height. Our study had step lengths of about 12% body height, with the longer steps perhaps resulting in more pelvic lowering. Regardless, the between-limb differences in pelvic lowering were not only small (≈2 cm) but were negligible at step completion.

A lack of between-limb differences in sagittal-plane kinematics contrasts with previously observed differences in step length and the CoM position (Patel and Bhatt, 2015a). Discrepancies in methods may have led to different conclusions. Patel and Bhatt (2015) delivered three trials of a 20 cm perturbation, with fall rates exceeding 20% and the mean number of steps more than 1.5. Our protocol was progressively challenging and participant-specific, only evaluating successful, single-step recoveries at near-capacity levels. In the between-limb comparisons, we evaluated responses to perturbations that were an average (s.d.) of 11 (5) cm in displacement. In addition, participants had experienced numerous perturbations leading up to that analyzed trial, so a short-term learning effect (Kajrolkar et al., 2014) may have altered between-limb differences. In other words, between-limb differences may have existed had we assessed larger perturbations delivered at the beginning of the session.

Participants had delayed muscle responses and, based on effect size (d=1.1), more co-contraction in the paretic stance limb (Table 3). These results align with that of feet-in-place responses (Badke et al., 1987; Di Fabio et al., 1986; Richard P. Di Fabio, 1987; Di Fabio and Badke, 1988; Ruth Dickstein et al., 1989; Dickstein et al., 1994; R Dickstein et al., 1989; Diener et al., 1985; Dietz and Berger, 1984; Kirker et al., 2000; Marigold et al., 2004). However, the altered muscle response of the paretic limb in stance did not affect sagittal-plane kinematics. Perhaps there was compensation from proximal joints or the non-paretic limb.

With practice, participants recovered from a higher proportion of perturbations, and they recovered from larger perturbations. These results are similar to previous evidence of improved reactive-balance with practice (Bhatt et al., 2019; Handelzalts et al., 2019a; Kajrolkar et al., 2014; Punt et al., 2019; Schinkel-Ivy et al., 2019; Van Duijnhoven et al., 2018). In some cases, the participants in our study corrected the inability to initiate a paretic-limb step. This skill is relevant, as the inability to take paretic steps has been prospectively related to falls (Mansfield et al., 2015). Post hoc, between-limb comparisons of the proportion of successful recoveries at the end of training suggest that a between-limb difference persisted (p=0.008, d=1.40). So, our approach did not eliminate the between-limb disparity.

Despite performance-based improvements, we did not observe kinematic or muscle-activation changes. By only comparing successful responses, we may not have considered large enough perturbations to observe training-based benefits. However, we could not practically examine failed recoveries, as the latter included non-stepping responses or steps with the wrong limb. It may be that, with our single-step constraint, the potential for kinematic adaptations were limited. In a previous study, those with stroke shortened their initial step length with training, preferring multistep responses (Patel and Bhatt, 2016). Although a multistep response may be effective when no step constraints are given, it may have limited efficacy for responses that are dependent on first-step features, such as a slip (Troy et al., 2008). It may also be that our variables did not capture the mechanisms underlying performance-based improvements, such as potential alterations in stepping-limb kinetics after foot contact, preventing limb collapse (Wang et al., 2019). Alternatively, adaptations may not be consistent across participants. For example, we observed a moderate training effect on the posterior trunk angle with non-paretic steps (d=0.5). The three participants with the worst trunk angles (>13 degrees), however, had notable reductions (>10 degrees, see online supplement). Biomechanical analyses do not capture psychological benefits, such as altered confidence to withhold a second step (i.e. preventing mechanically unnecessary steps). It is difficult to quantify this form of confidence, although perturbation training has benefitted general balance confidence, as per the Activities-Specific Balance Confidence scale (Handelzalts et al., 2019a).

We did not conduct a controlled experiment. Therefore, changes were not necessarily due to the training itself. Potential confounders include increased familiarity with the research setting or the benefits of more activity associated with participation. Another limitation is a small sample for our biomechanical variables, reduced from enrollment due to participant dropout and the initial inability of participants to step with the paretic limb. For kinematic variables, we only failed to detect small-to-medium effects (d<0.7) as significant. For EMG variables, we did observe one large, non-significant, between-limb effect on the co-contraction ratio (d=1.1, p=0.07). Perhaps we would have observed more profound adaptations had we captured the first paretic-limb steps of many participants. These steps often occurred, however, in middle sessions with no instrumentation.

Further study is needed to determine how this protocol could be applied broadly in a rehabilitation setting. For some participants, our protocol was halted or postponed due to fatigue, soreness, and uneasiness. Our inclusion and exclusion criteria were crafted to limit injury risk. In doing so, we likely prevented participation from those with more severe impairments. The protocol could be altered to include those with lower physical capacity, for example, by limiting the perturbation magnitude and number. It may be, however, that such modifications limit training efficacy. Previous controlled trials of perturbation training in those with chronic stroke had less than a 6% withdrawal rate (Mansfield et al., 2018). With uncertainty about adherence, future trials may consider an intention-to-treat analysis to account for dropout.

5. Conclusions

We demonstrated that paretic-limb steps are associated with worse balance-reaction performance in those with chronic stroke. Performance improved with practice, but we did not observe consistent kinematic adaptations with training. We have also demonstrated improvements in anterior fall-recovery in this cohort (Pigman et al., 2019). There is a complex interplay with how adaptation to reactions in one direction may interfere with that in the opposite direction (Bhatt et al., 2013). So, these results must be interpreted with knowledge that anterior balance-reaction training was delivered concurrently, an approach that aligns with previous, multi-directional protocols (Mansfield et al., 2018, 2017; Schinkel-Ivy et al., 2019).

Supplementary Material

1
2
Download video file (8.7MB, mp4)

Highlights.

  • Posterior steps with the non-paretic limb were associated with better performance.

  • Paretic-limb steps were wider relative to the center of mass.

  • With training, subjects recovered from more perturbations and larger perturbations.

  • Step and trunk kinematics did not change with training.

Acknowledgements

We thank all of the participants that volunteered for this study. Subject recruitment and scheduling were made possible with resources provided by the Delaware Rehabilitation Institute/DRI. We thank our Undergraduate Research Assistant Rebecca Peck who assisted with data collections.

Funding

This project was supported by the University of Delaware Research Foundation, the American Society of Biomechanics Junior Faculty Research Award, the National Institutes of Health (P30 GM103333) and a grant from the National Institute of General Medical Sciences (2P20 GM103446) from the National Institutes of Health and the State of Delaware.

Footnotes

Declaration of Conflict of Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Badke MB, Duncan PW, Di Fabio RP, 1987. Influence of prior knowledge on automatic and voluntary postural adjustments in healthy and hemiplegic subjects. Phys. Ther. 67, 1495–500. [DOI] [PubMed] [Google Scholar]
  2. Batchelor FA, Mackintosh SF, Said CM, Hill KD, 2012. Falls after stroke. Int. J. Stroke 7, 482–490. doi: 10.1111/j.1747-4949.2012.00796.x [DOI] [PubMed] [Google Scholar]
  3. Berg T, Scale B, Bbs T, Score T, 2009. Berg Balance Scale. Arch. Phys. Med. Rehabil. 73, 2–5. doi: 10.1111/j.1532-950X.2008.00478.x [DOI] [Google Scholar]
  4. Bhatt T, Dusane S, Patel P, 2019. Does severity of motor impairment affect reactive adaptation and fall-risk in chronic stroke survivors? J. Neuroeng. Rehabil. 16, 1–13. doi: 10.1186/s12984-019-0510-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bhatt T, Wang TY, Yang F, Pai YC, 2013. Adaptation and generalization to opposing perturbations in walking. Neuroscience 246, 435–450. doi: 10.1016/j.neuroscience.2013.04.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. CDC, 2009. Prevalence and most common causes of disability among adults--United States, 2005. MMWR. Morb. Mortal. Wkly. Rep. 58, 421–6. doi:mm5816a2 [pii] [PubMed] [Google Scholar]
  7. Celinskis D, Grabiner MD, Honeycutt CF, 2018. Bilateral early activity in the hip flexors associated with falls in stroke survivors: Preliminary evidence from laboratory-induced falls. Clin. Neurophysiol. 129, 258–264. doi: 10.1016/j.clinph.2017.11.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Cham R, Redfern MS, 2001. Lower extremity corrective reactions to slip events. J. Biomech. 34, 1439–1445. doi: 10.1016/S0021-9290(01)00116-6 [DOI] [PubMed] [Google Scholar]
  9. Crenshaw JR, Bernhardt KA, Atkinson EJ, Khosla S, Kaufman KR, Amin S, 2018. The relationships between compensatory stepping thresholds and measures of gait, standing postural control, strength, and balance confidence in older women. Gait Posture 65, 74–80. doi: 10.1016/j.gaitpost.2018.06.117 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Crenshaw JR, Kaufman KR, Grabiner MD, 2013. Compensatory-step training of healthy, mobile people with unilateral, transfemoral or knee disarticulation amputations: A potential intervention for trip-related falls. Gait Posture 38, 500–506. doi: 10.1016/j.gaitpost.2013.01.023 [DOI] [PubMed] [Google Scholar]
  11. Cyr M-AA, Smeesters C, 2009. Maximum allowable force on a safety harness cable to discriminate a successful from a failed balance recovery. J. Biomech. 42, 1566–1569. doi: 10.1016/j.jbiomech.2009.04.003 [DOI] [PubMed] [Google Scholar]
  12. de Kam D, Roelofs JMB, Bruijnes AKBD, Geurts ACH, Weerdesteyn V, 2017. The Next Step in Understanding Impaired Reactive Balance Control in People With Stroke: The Role of Defective Early Automatic Postural Responses. Neurorehabil. Neural Repair 31, 708–716. doi: 10.1177/1545968317718267 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Fabio Di, Richard P., 1987. Lower extremity antagonist muscle response following standing perturbation in subjects with cerebrovascular disease. Brain Res. 406, 43–51. doi: 10.1016/0006-8993(87)90767-0 [DOI] [PubMed] [Google Scholar]
  14. Di Fabio RP, 1987. Reliability of computerized surface electromyography for determining the onset of muscle activity. Phys. Ther. 67, 43–48. doi: 10.1093/ptj/67.1.43 [DOI] [PubMed] [Google Scholar]
  15. Di Fabio RP, Badke MB, 1988. Influence of cerebrovascular accident on elongated and passively shortened muscle responses after forward sway. Phys. Ther. 68, 1215–1220. [PubMed] [Google Scholar]
  16. Di Fabio RP, Badke MB, Duncan PW, 1986. Adapting human postural reflexes following localized cerebrovascular lesion: Analysis of bilateral long latency responses. Brain Res. 363, 257–264. doi: 10.1016/0006-8993(86)91010-3 [DOI] [PubMed] [Google Scholar]
  17. Dickstein R, Dvir Z, Ben Jehosua E, Rois M, Pillar T, 1994. Automatic and voluntary lateral weight shifts in rehabilitation of hemiparetic patients. Clin. Rehabil. 8, 91–99. doi: 10.1177/026921559400800201 [DOI] [Google Scholar]
  18. Dickstein Ruth Hocherman S, Dannenbaum E., Pillar T., 1989. Responses of ankle musculature of healthy subjects and hemiplegic patients to sinusoidal anterior-posterior movements of the base of support. J. Mot. Behav. 21, 99–112. doi: 10.1080/00222895.1989.10735468 [DOI] [PubMed] [Google Scholar]
  19. Dickstein R, Pillar T, Shina N, Hocherman S, 1989. Electromyographic responses of distal ankle musculature of standing hemiplegic patients to continuous anterior-posterior perturbations during imposed weight transfer over the affected leg. Phys. Ther. 69, 484–91. [DOI] [PubMed] [Google Scholar]
  20. Diener HC, Ackermann H, Dichgans J, Guschlbauer B, 1985. Medium- and long-latency responses to displacements of the ankle joint in patients with spinal and central lesions. Electroencephalogr. Clin. Neurophysiol. 60, 407–416. doi: 10.1016/0013-4694(85)91014-4 [DOI] [PubMed] [Google Scholar]
  21. Dietz V, Berger W, 1984. Interlimb coordination of posture in patients with spastic paresis. Impaired function of spinal reflexes. Brain 107, 965–978. doi: 10.1093/brain/107.3.965 [DOI] [PubMed] [Google Scholar]
  22. Forster A, Young J, 1995. Incidence and consequences of falls due to stroke: a systematic inquiry. BMJ 311, 83–6. doi: 10.1136/bmj.311.6997.83 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Fugl Meyer AR, Jaasko L, Leyman I, 1975. The post stroke hemiplegic patient. I. A method for evaluation of physical performance. Scand. J. Rehabil. Med. 7, 13–31. [PubMed] [Google Scholar]
  24. Gordon A, Morris R, 2008. Falls and neurological disorders. C. J. Geriatr. Med. 10, 107–113. [Google Scholar]
  25. Grabiner MD, Donovan S, Bareither Marone J.R M Lou., Hamstra-Wright K., Gatts S., Troy KL., 2008. Trunk kinematics and fall risk of older adults: Translating biomechanical results to the clinic. J. Electromyogr. Kinesiol. 18, 197–204. doi: 10.1016/j.jelekin.2007.06.009 [DOI] [PubMed] [Google Scholar]
  26. Handelzalts S, Kenner-Furman M, Gray G, Soroker N, Shani G, Melzer I, 2019a. Effects of Perturbation-Based Balance Training in Subacute Persons With Stroke: A Randomized Controlled Trial. Neurorehabil. Neural Repair 33, 213–224. doi: 10.1177/1545968319829453 [DOI] [PubMed] [Google Scholar]
  27. Handelzalts S, Steinberg-Henn F, Levy S, Shani G, Soroker N, Melzer I, 2019b. Insufficient Balance Recovery Following Unannounced External Perturbations in Persons With Stroke. Neurorehabil. Neural Repair 33, 730–739. doi: 10.1177/1545968319862565 [DOI] [PubMed] [Google Scholar]
  28. Hermens HJ, Freriks B, Disselhorst-Klug C, Rau G, 2000. Development of recommendations for SEMG sensors and sensor placement procedures. J. Electromyogr. Kinesiol. 10, 361–374. doi: 10.1016/S1050-6411(00)00027-4 [DOI] [PubMed] [Google Scholar]
  29. Inness EL, Mansfield A, Bayley M, McIlroy WE, 2016. Reactive Stepping After Stroke: Determinants of Time to Foot Off in the Paretic and Nonparetic Limb. J. Neurol. Phys. Ther. 40, 196–202. doi: 10.1097/NPT.0000000000000132 [DOI] [PubMed] [Google Scholar]
  30. Jorgensen L, Jacosen BK, 2002. Higher Incidence of Falls in Long-Term Stroke Survivors Than in Populationn Controls. Stroke 33, 542–547. [DOI] [PubMed] [Google Scholar]
  31. Kadaba MP, Ramakrishnan HK, Wootten ME, 1990. Measurement of lower extremity kinematics during level walking. J. Orthop. Res. 8, 383–392. doi: 10.1002/jor.1100080310 [DOI] [PubMed] [Google Scholar]
  32. Kajrolkar T, Bhatt T, 2016. Falls-risk post-stroke: Examining contributions from paretic versus non paretic limbs to unexpected forward gait slips. J. Biomech. 49, 2702–2708. doi: 10.1016/j.jbiomech.2016.06.005 [DOI] [PubMed] [Google Scholar]
  33. Kajrolkar T, Yang F, Pai YC, Bhatt T, 2014. Dynamic stability and compensatory stepping responses during anterior gait-slip perturbations in people with chronic hemiparetic stroke. J. Biomech. 47, 2751–2758. doi: 10.1016/j.jbiomech.2014.04.051 [DOI] [PubMed] [Google Scholar]
  34. Kanis JA, Melton LJ, Christiansen C, Johnston CC, Khaltaev N, 2009. The diagnosis of osteoporosis. J. Bone Miner. Res. 9, 1137–1141. doi: 10.1002/jbmr.5650090802 [DOI] [PubMed] [Google Scholar]
  35. Kirker SGB, Simpson DS, Jenner JR, Wing AM, 2000. Stepping before standing: Hip muscle function in stepping and standing balance after stroke. J. Neurol. Neurosurg. Psychiatry 68, 458–464. doi: 10.1136/jnnp.68.4.458 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Lakens D, 2013. Calculating and reporting effect sizes to facilitate cumulative science: A practical primer for t-tests and ANOVAs. Front. Psychol. 26. doi: 10.3389/fpsyg.2013.00863 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Lakhani B, Mansfield A, Inness EL, McIlroy WE, 2011. Compensatory stepping responses in individuals with stroke: A pilot study. Physiother. Theory Pract. 27, 299–309. doi: 10.3109/09593985.2010.501848 [DOI] [PubMed] [Google Scholar]
  38. Mansfield A, Aqui A, Danells CJ, Knorr S, Centen A, DePaul VG, Schinkel-Ivy A, Brooks D, Inness EL, Mochizuki G, 2018. Does perturbation-based balance training prevent falls among individuals with chronic stroke? A randomised controlled trial. BMJ Open 8, e021510. doi: 10.1136/bmjopen-2018-021510 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Mansfield A, Inness EL, Komar J, Biasin L, Brunton K, Lakhani B, McIlroy WE, 2011. Training Rapid Stepping Responses in an Individual With Stroke. Phys. Ther. 91, 958–969. doi: 10.2522/ptj.20100212 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Mansfield A, Schinkel-Ivy A, Danells CJ, Aqui A, Aryan R, Biasin L, DePaul VG, Inness EL, 2017. Does Perturbation Training Prevent Falls after Discharge from Stroke Rehabilitation? A Prospective Cohort Study with Historical Control. J. Stroke Cerebrovasc. Dis. 26, 2174–2180. doi: 10.1016/j.jstrokecerebrovasdis.2017.04.041 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Mansfield A, Wong JS, McIlroy WE, Biasin L, Brunton K, Bayley M, Inness EL, 2015. Do measures of reactive balance control predict falls in people with stroke returning to the community? Physiother. (United Kingdom) 101, 373–380. doi: 10.1016/j.physio.2015.01.009 [DOI] [PubMed] [Google Scholar]
  42. Marigold DS, Eng JJ, 2006. Altered timing of postural reflexes contributes to falling in persons with chronic stroke. Exp. Brain Res. 171, 459–468. doi: 10.1007/s00221-005-02936 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Marigold DS, Eng JJ, Timothy Inglis J, 2004. Modulation of ankle muscle postural reflexes in stroke: influence of weight-bearing load. Clin. Neurophysiol. 115, 2789–2797. doi: 10.1016/j.clinph.2004.07.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Patel P, Bhatt T, 2015a. Modulation of reactive response to slip-like perturbations: effect of explicit cues on paretic versus non-paretic side stepping and fall-risk. Exp. Brain Res. 3047–3058. doi: 10.1007/s00221-015-4367-9 [DOI] [PubMed] [Google Scholar]
  45. Patel P, Bhatt T, 2015b. Adaptation to large-magnitude treadmill-based perturbations: improvements in reactive balance response. Physiol. Rep. 3, e12247–e12247. doi: 10.14814/phy2.12247 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Patel PJ, Bhatt T, 2018. Fall risk during opposing stance perturbations among healthy adults and chronic stroke survivors. Exp. Brain Res. 236, 1–10. doi: 10.1007/s00221-017-5138-6 [DOI] [PubMed] [Google Scholar]
  47. Patel PJ, Bhatt T, 2016. Does aging with a cortical lesion increase fall-risk: Examining effect of age versus stroke on intensity modulation of reactive balance responses from slip-like perturbations. Neuroscience 333, 252–263. doi: 10.1016/j.neuroscience.2016.06.044 [DOI] [PubMed] [Google Scholar]
  48. Pigman J, Reisman DS, Pohlig RT, Jeka JJ, Wright TR, Conner BC, Petersen DA, Crenshaw JR, 2019. Anterior fall-recovery training applied to individuals with chronic stroke. Clin. Biomech. 69, 205–214. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Powell LE, Myers AM, 1995. The Activities-specific Balance Confidence (ABC) Scale. journals Gerontol. A, Biol. Sci. Med. Sci. 50A, M28–34. [DOI] [PubMed] [Google Scholar]
  50. Punt M, Bruijn SM, van de Port IG, de Rooij IJM, Wittink H, van Dieën JH, 2019. Does a Perturbation Based Gait Intervention Enhance Gait Stability in Fall Prone Stroke Survivors? A Pilot Study. J. Appl. Biomech. 1–26. doi: 10.1123/jab.2017-0282 [DOI] [PubMed] [Google Scholar]
  51. Ramnemark A, Nyberg L, Borsse B, 1998. International Original Article Fractures after Stroke. Stroke 630, 92–95. [DOI] [PubMed] [Google Scholar]
  52. Redfern MS, Cham R, Gielo-Perczak K, Grönqvist R, Hirvonen M, Lanshammar H, Marpet M, Pai CY-C, Powers C, 2001. Biomechanics of slips. Ergonomics 44, 1138–1166. [DOI] [PubMed] [Google Scholar]
  53. Salot P, Patel P, Bhatt T, 2016. Reactive Balance in Individuals With Chronic Stroke: Biomechanical Factors Related to Perturbation-Induced Backward Falling. Phys. Ther. 96, 338–347. doi: 10.2522/ptj.20150197 [DOI] [PubMed] [Google Scholar]
  54. Schinkel-Ivy A, Huntley AH, Aqui A, Mansfield A, 2019. Does Perturbation-Based Balance Training Improve Control of Reactive Stepping in Individuals with Chronic Stroke? J. Stroke Cerebrovasc. Dis. 1–9. doi: 10.1016/j.jstrokecerebrovasdis.2018.12.011 [DOI] [PubMed] [Google Scholar]
  55. Schmid AA, Yaggi HK, Burrus N, McClain V, Austin C, Ferguson J, Fragoso C, Sico JJ, Miech EJ, Matthias MS, Williams LS, Bravata DM, 2013a. Circumstances and consequences of falls among people with chronic stroke. J. Rehabil. Res. Dev. 50, 1277–1286. doi: 10.1682/JRRD.2012.11.0215 [DOI] [PubMed] [Google Scholar]
  56. Schmid AA, Yaggi HK, Burrus N, McClain V, Austin C, Ferguson J, Fragoso C, Sico JJ, Miech EJ, Matthias MS, Williams LS, Bravata DM, Klar Yaggi H, Burrus N, McClain V, Austin C, Ferguson J, Fragoso C, Sico JJ, Miech EJ, Matthias MS, Williams LS, Roudebush RL, 2013b. Circumstances and consequences of falls among people with chronic stroke. J. Rehabil. Res. Dev. 50, 1277–86. doi: 10.1682/JRRD.2012.11.0215 [DOI] [PubMed] [Google Scholar]
  57. Troy KL, Donovan SJ, Marone JR, Bareither M. Lou, Grabiner MD., 2008. Modifiable performance domain risk-factors associated with slip-related falls. Gait Posture 28, 461–465. doi: 10.1016/j.gaitpost.2008.02.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Van Duijnhoven HJR, Roelofs JMB, Den Boer JJ, Lem FC, Hofman R, Van Bon GEA, Geurts ACH, Weerdesteyn V, 2018. Perturbation-based balance training to improve step quality in the chronic phase after stroke: A proof -of -concept study. Front. Neurol. 9, 1–12. doi: 10.3389/fneur.2018.00980 [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Van Meulen FB, Weenk D, Van Asseldonk EHF, Schepers HM, Veltink PH, Buurke JH, 2016. Analysis of balance during functional walking in stroke survivors. PLoS One. 11, e0166789. doi: 10.1371/journal.pone.0166789 [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Wang S, Liu X, Pai Y-C, 2019. Limb Collapse or Instability? Assessment on Cause of Falls. Ann. Biomed. Eng. 47, 767–777. doi: 10.1007/s10439-018-02195-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Winter DA, 2009. Anthropometry, in: Biomechanics and Motor Control of Human Movement. John Wiley & Sons, Inc., New Jersey, pp. 82–106. [Google Scholar]
  62. Wrisley DM, Marchetti GF, Kuharsky DK, Whitney SL, 2004. Reliability, internal consistency, and validity of data obtained with the functional gait assessment. Phys. Ther. 84, 906–918. [PubMed] [Google Scholar]
  63. Yang F, Bhatt T, Pai YC, 2013. Generalization of treadmill-slip training to prevent a fall following a sudden (novel) slip in over-ground walking. J. Biomech. 46, 63–69. doi: 10.1016/j.jbiomech.2012.10.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Yang F, Espy D, Bhatt T, Pai YC-C, 2012. Two types of slip-induced falls among community dwelling older adults. J. Biomech. 45, 1259–1264. doi: 10.1016/j.jbiomech.2012.01.036 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1660479-supplement-1.xlsx (86.3KB, xlsx)
2
Download video file (8.7MB, mp4)

RESOURCES