AN INVESTIGATION OF THE NEUROPLASTICITY UNDERLYING AN

IMPROVED REACTIVE BALANCE RESPONSE

by

Jennifer H Barnes

A dissertation submitted to the Faculty of the University of Delaware in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Biomechanics and Movement Science.

Summer 2020

© 2020 Jennifer H Barnes All Rights Reserved

AN INVESTIGATION OF THE NEUROPLASTICITY UNDERLYING AN

IMPROVED REACTIVE BALANCE RESPONSE

by

Jennifer H Barnes

Approved: ______John J. Jeka, Ph.D. Chair of the Department of Kinesiology and Applied Physiology

Approved: ______Kathleen S. Matt, Ph.D. Dean of the College of Health Sciences

Approved: ______Douglas J. Doren, Ph.D. Interim Vice Provost for Graduate and Professional Education and Dean of the Graduate College

I certify that I have read this dissertation and that in my opinion it meets the academic and professional standard required by the University as a dissertation for the degree of Doctor of Philosophy.

Signed: ______Jeremy R. Crenshaw, Ph.D. Professor in charge of dissertation

I certify that I have read this dissertation and that in my opinion it meets the academic and professional standard required by the University as a dissertation for the degree of Doctor of Philosophy.

Signed: ______Jonathan R. Wolpaw, M.D. Member of dissertation committee

I certify that I have read this dissertation and that in my opinion it meets the academic and professional standard required by the University as a dissertation for the degree of Doctor of Philosophy.

Signed: ______John J. Jeka, Ph.D. Member of dissertation committee

I certify that I have read this dissertation and that in my opinion it meets the academic and professional standard required by the University as a dissertation for the degree of Doctor of Philosophy.

Signed: ______Christopher A. Knight, Ph.D. Member of dissertation committee

ACKNOWLEDGMENTS

Special thanks go to the following individuals: Dr. Jeremy Crenshaw for serving as my academic advisor, for always being encouraging and supportive, and for providing insightful advice and direction. Dr. Jonathan Wolpaw for serving as my scientific advisor and for providing both the guidance and resources needed for conducting this research. Dr. Jonathan Carp for sharing his scientific expertise. Drs. John Jeka and Christopher Knight for taking the time to serve on my committee. Dr. Amir Eftekhar, Mr. Timothy Fake, and Mr. Steve Carmack for their technological support in hardware and software development. My incredible research assistants: Ms. Susan Heckman, Ms. Allegra Wu, Ms. Kelly Fitzpatrick, for help with data collection; and importantly Dr. Edward Greenberg for assistance with data collection and processing throughout the entire project – thank you Ed, for your steadfast commitment! This manuscript is dedicated to my family, especially my husband Joe and daughters Anja and Marissa. Thank you for your unwavering love, support, patience, and encouragement – you are the best!!

iv TABLE OF CONTENTS

LIST OF TABLES ...... viii LIST OF FIGURES ...... ix ABSTRACT ...... xi

Chapter

1 INTRODUCTION ...... 1

1.1 Overview ...... 1 1.2 Specific Aims and Hypotheses ...... 5 1.3 Significance and Innovation ...... 7

2 AIM 1: THE EFFECT OF TRAINING ON PERFORMANCE OF A NON- STEPPING RESPONSE TO POSTERIOR SURFACE TRANSLATIONS ... 11

2.1 Introduction ...... 11 2.2 Methods ...... 14

2.2.1 Participants ...... 14 2.2.2 Protocol ...... 14 2.2.3 Data Collection and Processing ...... 20

2.3 Results ...... 21 2.4 Additional Analyses ...... 24

2.4.1 Effect of Displacement Range on Outcome ...... 25 2.4.2 Effect of Baseline Stepping Threshold on Outcome ...... 26 2.4.3 Effect of Age on Stepping Threshold ...... 27

2.5 Discussion ...... 29 2.6 Additional Applications ...... 30

3 AIM 2: THE EFFECT OF TRAINING A NON-STEPPING RESPONSE TO POSTERIOR SURFACE TRANSLATIONS ON THE SOLEUS H- REFLEX ...... 32

3.1 Introduction ...... 32

v 3.1.1 Short-term and Long-term Adaptation of the Soleus H-reflex ... 34 3.1.2 Perturbation-based Reactive Balance Training and the Soleus H-reflex ...... 39

3.2 Methods ...... 42

3.2.1 Participants ...... 42 3.2.2 Protocol ...... 43 3.2.3 Data Collection and Processing ...... 46

3.3 Results ...... 48

3.3.1 Effect of Reactive Balance Training on the Soleus H-reflex ...... 48 3.3.2 Correlation between Change in Stepping Threshold and Change in Soleus H-reflex Behavior ...... 52

3.4 Additional Analyses ...... 54

3.4.1 Effect of Baseline H-max/M-max on Change in H-reflex Control Trial Amplitude ...... 55 3.4.2 Effect of Baseline H-max/M-max on Baseline Stepping Threshold ...... 56

3.5 Discussion ...... 58

3.5.1 Differences in the Direction of H-reflex Change ...... 59 3.5.2 Between-leg Differences in H-reflex Behavior ...... 64

3.6 Conclusions ...... 67 3.7 Limitations and Future Directions ...... 67

4 AIM 3: THE EFFECT OF TRAINING A NON-STEPPING RESPONSE TO POSTERIOR SURFACE TRANSLATIONS ON POSTURAL STEADINESS ...... 70

4.1 Introduction ...... 70 4.2 Methods ...... 74

4.2.1 Participants ...... 74 4.2.2 Protocol ...... 74 4.2.3 Data Collection and Processing ...... 76

4.3 Results ...... 78

4.3.1 Effect of Reactive Balance Training on Postural Steadiness ...... 78

vi 4.3.2 Correlation between Change in Postural Steadiness and Change in Soleus H-reflex Behavior ...... 80

4.4 Additional Analyses ...... 84 4.5 Discussion ...... 86 4.6 Conclusions ...... 89 4.7 Limitations and Future Directions ...... 90

5 SUMMARY ...... 93

5.1 Summary of Results ...... 93 5.2 Important Remaining Questions ...... 95 5.3 Unique Features of our Study ...... 96

REFERENCES ...... 99

Appendix

A VALIDATION OF TREADMILL PERTURBATION SIZE ...... 122 B IRB APPROVAL ...... 130

vii LIST OF TABLES

Table 1: Perturbation Parameters...... 19

Table 2: Participant Data: Primary outcome measures at baseline, and change in primary outcome measures, from baseline to final, for Aim 2 and Aim 1...... 51

Table 3: Participant Data: Primary outcome measures for Aim 3 and Aim 2. .... 79

Table 4: Correlation between Postural Sway Variables and Stepping Threshold. 85

Table 5: Correlation between Postural Sway Variables and Soleus H-reflex Behavior, at Baseline...... 86

Table 6: Perturbation Peak Velocity: Comparison of expected versus observed magnitude, by level...... 127

Table 7: Perturbation Displacement: Comparison of expected versus observed magnitude, by level...... 128

viii LIST OF FIGURES

Figure 1: Technological Innovation: Integration of hardware and software...... 9

Figure 2: Aim 1 Experimental Protocol...... 15

Figure 3: Treadmill Set-up and Participant Instructions...... 18

Figure 4: Perturbation Profile...... 19

Figure 5: Stepping Threshold at Baseline and Final...... 22

Figure 6: Stepping Threshold by Participant and Session...... 22

Figure 7: Stepping Threshold Percent Change from Baseline, Group Averages by Session...... 24

Figure 8: Stepping Threshold by Perturbation Velocity and Perturbation Displacement...... 25

Figure 9: Effect of Baseline Stepping Threshold on Outcome...... 27

Figure 10: Effect of Age on Stepping Threshold...... 28

Figure 11: Aim 2 Experimental Protocol...... 44

Figure 12: Soleus H-reflex Control Trial Amplitude at Baseline and Final Assessments...... 48

Figure 13: Soleus H-reflex Control Trial Data Grouped by Leg...... 50

Figure 14: Change in Stepping Threshold vs. Change in Soleus H-reflex Control Trial Amplitude, by Leg...... 53

Figure 15: Change in Soleus H-reflex CT Amplitude vs. Baseline Soleus H- max/M-max...... 56

Figure 16: Baseline Stepping Threshold vs. Baseline H-max/M-max, by Leg...... 57

Figure 17: Aim 3 Experimental Protocol...... 75

ix Figure 18: Postural Sway in the Anterior-Posterior Direction at Baseline and Final...... 78

Figure 19: Postural Sway Variables Grouped by Direction of Change in Soleus H-reflex Control Trial Amplitude of the Stepping Leg...... 81

Figure 20: Change in Center of Pressure RMS vs. Change in Soleus H-reflex CT Amplitude, by Leg...... 83

Figure 21: Change in Center of Pressure Mean Velocity vs. Change in Soleus H- reflex CT Amplitude, by Leg...... 84

Figure 22: Expected and Observed Size of a Level 10 Perturbation...... 123

Figure 23: Size of Treadmill Perturbations...... 124

Figure 24: Bland-Altman Plot, Perturbation Peak Velocity...... 125

Figure 25: Bland-Altman Plot, Perturbation Displacement...... 126

x ABSTRACT

Perturbation-based training protocols, in which participants repeatedly attempt to recover from rapid surface translations, seek to improve the reactive response needed to maintain upright balance following an external disturbance. While these protocols have reduced fall incidence in some at-risk populations, the neural adaptations underlying their effects are unknown. To address this gap in knowledge, we examined the neuroplasticity associated with changes in the ability of healthy adults (aged 25-55yrs) to respond to brief (i.e., 600-700ms) rapid surface translations, without needing to step. Our objective was to determine the impact of this training on lower leg spinal reflexes and to determine whether the training affects motor behavior during other postural tasks. The central hypothesis was that training would improve performance on the reactive balance task, and would change specific spinal reflexes (e.g., up- or down-regulation of the soleus Hoffman reflex (H- reflex)). These changes might, in turn, alter motor behavior during tasks such as postural steadiness during quiet stance. Fifteen participants completed six sessions of reactive balance training; during each session s/he stood on a computer-controlled treadmill and responded to ~70-85 perturbations delivered through custom software and hardware. All perturbations were directed posteriorly (i.e., induced anterior sway); participants were instructed to try not to step when responding to perturbations. The size (i.e., velocity and displacement) of each disturbance was adjusted trial by trial, based on the participant’s performance. Stepping Threshold, defined as the perturbation level (size) that elicited

xi a step on three consecutive trials, increased significantly, pre- to post-training, for the group. To determine the effect of our training protocol on spinal reflex behavior, soleus H-reflex control trials (i.e., H-reflex responses elicited while background muscle activity and M-wave size were kept constant) were obtained from both legs during standing. For six participants H-reflex amplitude increased bilaterally, pre-to post-training; for four participants H-reflex amplitude decreased bilaterally; five participants showed a mixed response. Additionally, there was a significant positive correlation, of moderate effect size, between change in Stepping Threshold and change in the H-reflex amplitude of the Stepping leg (i.e., the leg used most often for stepping during unsuccessful perturbation trials). To determine if our protocol led to changes in postural steadiness, under-the- feet center-of-pressure data were collected during quiet bipedal stance, before and after training. We found a positive trend in the relationship between change in the H- reflex of the Stepping leg and changes in postural steadiness. Specifically, an increase in the amplitude of this reflex tended to correlate with a reduction in postural sway in the anterior-posterior direction. In sum, our data suggest that changes in reactive balance and postural steadiness during standing are mediated, in part, by changes in the reflex behavior of the soleus muscle. However, future analyses of associated kinematic and electromyographic data are needed to investigate why the direction of H-reflex change differed among participants. Additionally, the positive relationship between changes in postural steadiness and changes in soleus H-reflex amplitude warrants further investigation given its potential impact on fall prevention. That is, training that

xii specifically targets up-regulation of this reflex (e.g., operant conditioning protocols) may improve upright postural control in some populations at-risk for falls.

xiii Chapter 1

INTRODUCTION

1.1 Overview

Motor learning is a neurological process that occurs through practice or experience and leads to a relatively permanent change in the ability to perform a motor skill [1]. Motor learning induces plasticity from cortex to spinal cord [2]. This ubiquitous plasticity enables us to continually modify and expand our repertoire of motor skills throughout life. It is comprised of primary plasticity that underlies the acquisition of a new skill, compensatory plasticity that preserves key features of previously learned skills, and reactive plasticity that is due to the altered neuronal activity caused by the primary and compensatory plasticity [3-7]. Guiding beneficial plasticity is the ultimate goal of all rehabilitative interventions; success in achieving this goal depends on our ability to predict how the central nervous system (CNS) will respond to specific training protocols. Postural control, that is, control of the orientation and stability (i.e., equilibrium) of the body in space [8], is fundamental to all purposeful movement. Trauma, pathology, and aging can impair postural control and thereby increase the risk of injury due to falls [9-17] and reduce quality of life [18-21]. Thus, interventions that improve postural control can have widespread benefit. Perturbation-based training protocols, in which participants repeatedly attempt to recover equilibrium after a mechanical perturbation (e.g., a rapid surface translation), seek to improve the reactive

1 response needed to maintain upright posture during external disturbances (e.g., trips and slips) [22-41]. Previous studies have examined the ecological validity and efficacy of perturbation-based training protocols; three important findings from these studies are: 1) mechanically-induced perturbations can be used to identify the biomechanical features of effective and non-effective reactive balance responses [24,26,27,32,42-49]; 2) computer-controlled treadmills and moveable platforms can deliver perturbations that simulate trips and slips [24,28,30,50-55]; 3) perturbation-based training can improve the reactive response needed to arrest trips and slips, in some populations at- risk for falls [23,24,26-40]. Together these studies demonstrate the importance of perturbation-based protocols to fall prevention. However, the primary plasticity responsible for training-induced benefits, that is, changes to the neural substrate that underlie an improved reactive balance response, is not clearly understood. Knowledge of these changes could lead to improved therapeutic protocols, that is, to protocols that guide plasticity beneficial for postural control in the most effective and widespread manner. Motor learning, the neurological process underlying skill acquisition, induces a hierarchy of plasticity that progresses from the brain to the spinal cord [4-7,56].

Because the primary plasticity responsible for the acquisition on a new skill occurs at multiple sites within the CNS it can alter the performance of other motor skills (i.e., those not specifically targeted during training) if it occurs at sites shared by the neural substrates of each of these motor tasks [57-63]. If this primary plasticity improves performance of previously acquired motor skill(s), then no further plasticity is necessary; if it interferes with that performance, compensatory plasticity may be

2 needed to preserve key features of these other motor skills [64]. Compensatory plasticity can, in turn, affect the performance of yet other motor tasks. Thus, training- induced plasticity can change multiple behaviors concurrently [5-7]. Because training can lead to ubiquitous plasticity, the identification of site- specific changes to the neural substrate that underlie an improved motor response could help inform the development of protocols that generate widespread beneficial plasticity (i.e., plasticity that improves performance of the skill specifically trained as well as that of other skills not directly targeted by the training). This study explores the neuroplasticity associated with training-induced changes in the ability of healthy younger adults to perform a standing, feet-in-place, response to rapid posterior translations of a support surface (e.g., treadmill belts). The purpose is to determine the impact of this training on lower leg spinal reflexes and to determine whether the training affects other motor behaviors related to postural control. Spinal reflexes contribute to both simple and complex motor behaviors [2-7], are modulated based on movement requirements [65-81] and can change in response to training [82-103]. Changes in reflex behavior (i.e., spinal cord plasticity) can be reliably measured through the Hoffman reflex (H-reflex), a response that is elicited through low-level electrical stimulation of a mixed peripheral nerve [104-110]. This stimulation, when applied at the appropriate level, generates two electromyographic (EMG) responses in the target muscle: stimulation of the motor axons produces a direct muscle response (M-wave); and stimulation of the sensory nerve fibers produces an H-reflex (i.e., a muscle response generated mainly by Ia afferents synapsing on spinal motor neurons). A change in the amplitude of the H-reflex with respect to M- wave amplitude and to the background EMG (bEMG) level, indicates a change in

3 spinal reflex behavior [107,108]. Training-induced change in the H-reflex amplitude provides a window to the activity-dependent plasticity of the CNS underlying skill acquisition [5]. Previous studies indicate that muscle activation at the ankle occurs early in the motor response to perturbations of upright posture [111-114]. The early activation of muscles at this joint persists even if a stepping response is pre-planned and/or occurs [115]. Furthermore, when responding to posteriorly-directed perturbations (i.e., perturbations that induce anterior sway) plantar flexor muscles have been shown to contribute to the reactive response, whether these perturbations are delivered at the level of the torso (i.e., waist pulls), or at the level of the foot (i.e., surface translations)

[114,116-118]. The soleus muscle in particular is important also to the maintenance of equilibrium during unperturbed (steady-state) standing [12,119-122]. Therefore, training that induces plasticity in the soleus reflex pathway could be beneficial for the performance of multiple motor tasks that require control of upright posture. However, it is not known if perturbation-based training changes the reflex response of this muscle. To address this unknown we developed custom software for an instrumented commercially-available treadmill and we designed a training protocol that challenges healthy adults (aged 25-55yrs) to improve their non-stepping response to rapid posterior surface translations. By limiting our protocol to posteriorly directed translations (i.e., perturbations that induce anterior sway), and by constraining the reactive response to a feet-in-place (i.e., non-stepping) strategy [123], our training provides a robust approach for examining adaptations specific to the soleus muscle. By investigating this in younger, unimpaired, individuals we remove potential

4 confounds of age and morbidity, and can more easily identify training-specific adaptations. The central hypothesis of this study is that training a non-stepping response to posterior surface translations will improve performance on the task, and will change specific spinal reflexes (e.g., up- or down-regulation of the soleus H-reflex). These changes may, in turn, alter motor behavior of related tasks such as postural steadiness during quiet standing. The results of this study should increase our understanding of the neural substrates responsible for control of upright posture. This understanding could aid in the development of more effective rehabilitative interventions for postural control, that is, interventions that target specific neural sites in order to generate widespread beneficial plasticity.

1.2 Specific Aims and Hypotheses This study has three specific aims. Aim 1 is to determine the extent to which healthy individuals, aged 25-55yrs, can improve their non-stepping (i.e., feet-in- place) response to rapid posterior translations of the support surface. To address this aim, participants will undergo six sessions of reactive balance training, each comprising 70-85 perturbations (disturbances) to upright posture. Perturbations, induced with a computer-controlled treadmill and custom software, will consist of brief (i.e., 600-700ms) surface translations delivered posteriorly during quiet standing. Perturbation size (i.e., speed and displacement of the treadmill belts) will change according to a progressive practice approach, based on the participant’s response (i.e., stepping or non-stepping). A successful response is defined as maintaining balance without stepping. Improvement will be evaluated by comparing the maximum perturbation size the participant successfully responds to before training with that

5 achieved after the six training sessions. The ability to maintain upright stance after a perturbation is a motor skill and as such should improve with practice. Thus, the hypothesis for Aim 1 is that, with practice, healthy younger individuals will learn to respond, without stepping, to faster and larger posterior perturbations. Aim 2 of this study is to determine the effect of training a non-stepping response to rapid posterior perturbations on the soleus H-reflex, bilaterally. To address this aim, participants will undergo reactive balance training as described in Aim 1. Before and after the six-session training protocol soleus H-reflex/M-wave recruitment curves and soleus H-reflex control trials (i.e., H-reflex responses elicited while background EMG (bEMG) and M-wave size are kept constant) will be obtained from both legs during standing. Posterior surface translations induce anterior sway; the muscular response to this type of perturbation has been shown to begin at the ankle joint and involve the soleus muscle [112,113]. Thus, the hypothesis for Aim 2 is that soleus H-reflex behavior will change in response to the training of a non-stepping balance reaction to rapid posterior perturbations. Aim 3 is to determine the effect of reactive balance training on postural steadiness during quiet standing (i.e., during unperturbed bipedal stance). To address this aim, participants will undergo reactive balance training as described in

Aim 1. Before and after the six-session training protocol, postural steadiness will be measured during unperturbed standing. Specifically, movement of the body’s center- of-mass in the anterior-posterior direction will be estimated based on center-of- pressure data collected from side-by-side force plates located under the participant’s feet [12]. Motor learning induces plasticity throughout the CNS, thus the performance of motor skills not specifically targeted during training can be affected by new motor

6 learning [57-63]. That is, the primary plasticity underlying a new motor skill can affect multiple motor behaviors simultaneously if it occurs at sites shared by their neural substrates [4-7]. While postural balance during standing refers to one’s ability to stay upright or to recover equilibrium after a perturbation, postural steadiness refers to the ability to stand as still as possible [124,125]. Neural sites and synapses underlying each of these tasks may be shared across their substrates such that changes in behavior due to the training of one task, may affect performance of the other task. Thus, the hypothesis for Aim 3 is that the training of a non-stepping balance reaction to rapid posterior perturbations will lead to changes in postural steadiness during bipedal stance.

1.3 Significance and Innovation Postural control is essential to all goal-directed movement; losses in postural control greatly reduce quality of life and can lead to serious injury due to falls [9-21]. In the United States falls are the leading cause of accidental, nonfatal injuries [126]. People at-risk for falls include older adults and those with Parkinson’s disease, hemiparesis due to stroke, lower limb amputation, and traumatic brain injury [9-11,13- 16,18]. An estimated one in three adults, aged 65 and older, falls annually [127]; these falls often result in bone fractures and loss of ability to live independently [128-130]. Fear of falling or a history of falls, can inhibit participation in many activities including those with social interaction and/or physical health benefits [19-21,32]. Thus, protocols that improve postural control can have widespread benefit. This study is significant in that our results could help guide development of such protocols and thereby reduce fall incidence.

7 Training-induced plasticity occurs at many sites within the CNS, including the spinal cord [4-7]. How this plasticity is conducted across the CNS such that new motor behaviors are incorporated while previously learned ones are functionally maintained is not clear [5-7]. By investigating changes in the performance of motor tasks that are both targeted and non-targeted during training, and then correlating them with spinal cord plasticity measured through the H-reflex, we can begin to elucidate how spinal cord sites and properties are negotiated by multiple behaviors [7]. This knowledge should help us determine which sites within the CNS to target during motor rehabilitation in order to guide the most beneficial plasticity. In this study we will test the effect of training on the performance of a non- stepping reactive balance response to rapid posterior surface translations in healthy adults (Aim 1). The motivation for training this response is to determine if perturbation-based training changes lower leg spinal reflexes. While other studies have examined kinematic changes associated with an improved reactive balance response [24,26,27,32,42-50], little has been done to identify training-induced changes in the neural substrate. Our study is innovative in that it is designed to help fill this research gap; knowledge gained from this study will add to our understanding of the role of spinal reflexes in postural control and could help guide the development of new interventions targeting fall prevention. Specifically, we will examine the effect of our reactive balance training protocol on the soleus H-reflex, bilaterally (Aim 2). While the soleus muscle is important to the maintenance of upright posture, it is not known if perturbation-based training changes the reflex response in this muscle. It is also not known how this type of training impacts the performance of other postural tasks. To begin addressing this

8 gap in knowledge, we will examine the effect of our training protocol on postural steadiness during quiet bipedal stance (Aim 3). Finally, though it is now widely appreciated that the CNS changes throughout life, we lack a clear understanding of how the properties of shared neurons and synapses within the spinal cord are negotiated (i.e., determined) by both old and new behaviors [5-7]. By investigating the effect of our perturbation-based training protocol on the soleus H-reflex, a spinal pathway relevant to both a non-stepping balance reaction and postural steadiness during upright stance, we hope to expand our overall understanding of how the CNS conducts its plasticity to ensure that all desired motor outcomes are achieved.

Figure 1: Technological Innovation: Integration of hardware and software. (A.) A participant stands on a Bertec® split-belt treadmill; for safety s/he wears a harness that is attached to an overhead support structure. Custom software controlled by the investigator delivers treadmill perturbations, triggers motion capture equipment, and displays time-synced instructions on the monitor to the participant’s front. (B.) Custom tachometer: Electronics of a disassembled optical mouse are connected to an Arduino® Uno microcontroller platform. Treadmill movement data are relayed via USB to the software that delivers treadmill perturbations.

9 To conduct this study, we developed custom hardware and software to control a Bertec® (Columbus OH, USA) split-belt treadmill. Though this treadmill is commercially-available and common to many research facilities, to our knowledge we are the first laboratory to successfully produce reliable perturbations of this type using a Bertec® split-belt treadmill. Our technological innovation includes a custom-built tachometer, and custom software through which the investigator dictates the speed and time course of perturbations, triggers motion capture equipment, and provides the participant with instructions (Figure 1). These instructions are time-synced to the perturbation trials and displayed on a computer monitor located to the front of the participant. Our automation of instructions results in greater standardization of training, reduces the potential confound of participant-investigator interaction, and leads to more consistent inter-trial intervals. Our software was designed to also allow for the incorporation of other technologies in the future. Future versions of our software could include time-locked EEG data collection, automated peripheral nerve stimulation, and a wide selection of perturbation profiles.

10 Chapter 2

AIM 1: THE EFFECT OF TRAINING ON PERFORMANCE OF A NON- STEPPING RESPONSE TO POSTERIOR SURFACE TRANSLATIONS

2.1 Introduction Human upright posture is inherently unstable due to our high center of mass and small base of support (i.e., the area circumscribed by our feet during standing) [12,120,121]. Control of upright posture is a complex motor skill in which motor output is continually adjusted based on visual, vestibular and somatosensory input [12,131-135]. This skill develops during childhood and the mechanisms underlying it are modified throughout life based on changes in the body or environment, and in response to new task requirements [10,18,136-144]. Loss in postural control, due to trauma, pathology, or aging, hinders performance in many activities of daily living including mobility, and can increase the risk of injury due to falls [9-11,13-21]. Training protocols that perturb upright posture seek to reduce fall incidence by improving the reactive response necessary for maintaining balance during external (i.e., environment-based) disturbances such as contact with obstacles, collisions with other moving objects, or slippery or moving surfaces [22-41,135]. When a standing person’s balance is perturbed, stability (equilibrium) can be regained by taking a step, or by reaching to touch or grasp a fixed object (i.e., by a change-in-support reaction) [17,145,146]. Balance can also be regained by generating the appropriate muscle torques needed to return the center of mass to its original position over the base of support without stepping or reaching (i.e., by a fixed-support reaction)

11 [17,112,113,146]. Computer-controlled treadmills and movable platforms that rapidly translate the support surface have been used in perturbation-based training to improve stepping responses to external disturbances [22,24-26,28-30,32,38-40,50-55]. These devices may also be useful for training a non-stepping (fixed-support) response. For this study we developed custom software to control a commercially- available treadmill for the purpose of training a non-stepping response to rapid posterior surface translations. The impetus for this study was to address the current gap in knowledge regarding changes in the neural substrate that underlie training- induced improvement of reactive balance responses. Our goal was to determine if reactive balance training changes lower leg spinal reflexes. Pursuant to this goal, we designed a perturbation-based protocol that would target activation of the soleus muscle. We then tested the ability of our protocol to change motor performance in younger healthy adults, a requirement necessary for investigating training-induced reflex change without potential confounds of age and morbidity. That is, for Aim 1 of this study, we tested the extent to which healthy younger adults (i.e., those aged 25- 55yrs) could improve their non-stepping (i.e., feet-in-place) response to brief rapid surface translations induced posteriorly with a computer-controlled treadmill. The ability to maintain upright posture after an external disturbance is a motor skill and as such, should improve with practice. Previous perturbation-based training studies have shown that practice can improve reactive stepping responses; this study tested whether non-stepping responses could be improved through similar methods. Our hypothesis was that, with practice, younger healthy adults would learn to respond to faster and larger posterior perturbations, without stepping.

12 While our motive for developing this protocol was to investigate the effect of perturbation-based training on lower leg spinal reflexes, we believe our protocol may be useful beyond this purpose. Additional applications for our reactive balance training, and for the technology we developed to conduct this training, are presented at the end of this chapter (Section 2.6).

13 2.2 Methods

2.2.1 Participants

This study was approved by the University of Delaware Institutional Review Board and the New York State Department of Health Institutional Review Board, with regulatory oversight granted to the latter. The study was conducted at the National Center for Adaptive Neurotechnologies (a Biomedical Technology Resource Center supported by the National Institutes of Health, US Department of Health and Human Services), located within the Wadsworth Center, New York Department of Health, in Albany NY. Eighteen participants (6 men, 12 women), aged 26-53yrs, were recruited from communities in the vicinity of Albany, New York. All had normal or corrected- to-normal vision and none were taking medication(s) that affect balance. Exclusion criteria included history of neurological or neuromuscular disorder, and any current medical condition limiting use of or sensation in the neck, back, arms, or legs, or interfering with standing or walking. All participants provided written informed consent prior to participation; all were compensated for their time. Three of the 18 participants (all women) withdrew prior to completing the study: two withdrew for unrelated medical reasons; the third withdrew when she was no longer able to commit the time required for participation. Each of the remaining 15 participants (6 men, 9 women) completed the study without incidence.

2.2.2 Protocol Our experimental protocol (Figure 2A) consisted of eight sessions: one baseline assessment session, six training sessions [27,32], and one final assessment session. All participants completed the protocol within 3-4 weeks. Participation averaged 2-3 sessions per week [147,148].

14

Figure 2: Aim 1 Experimental Protocol. (A.) This protocol consisted of one baseline assessment session (BL), six training sessions (T#1-T#6), and one final assessment session (F). Stepping Threshold, defined as the level where the size of the perturbation was large enough to evoke stepping on three consecutive trials, was determined at the beginning of each of these eight sessions. At T#1-T#6 Stepping Threshold determination was followed by 60 trials of progressive practice. (B.) This figure depicts a typical training session. Perturbation levels ranged from 1-20 with higher values indicating faster and larger perturbations. Trials 1-13 depict Stepping Threshold determination; here the participant begins at perturbation level 1 (the lowest level of difficulty) and moves up one level at a time until s/he steps during three consecutive trials. Trials 14- 73 depict progressive practice; here the participant begins one level below Stepping Threshold, progresses to the next higher level after completing three successful (non-stepping) trials, and moves down one level if s/he steps.

During each session the participant stood on a Bertec® (Columbus OH, USA) split-belt treadmill while perturbations (rapid surface translations) were triggered at

15 random times by custom software (Figure 3A). During all trials the treadmill belts were tied (i.e., the two belts moved together) and traveled backwards (i.e., posteriorly); these posterior perturbations caused the participant to sway forward. For safety, the participant wore a harness (Guardian Fall Protection®, Kent WA, USA) that was attached to an overhead support structure. This harness was attached in a manner that would not interfere with a participant’s reactive response, unless interference was needed to prevent knee or hand contact with the treadmill. During each trial, instructions were displayed on a 40-inch computer monitor located 1.5m to the participant’s front, at eye level (Figure 3B); these instructions were time-synced with the treadmill perturbations through custom software. Before each perturbation the participant was instructed to stand quietly with his or her attention focused on the monitor (screen #1). When the investigator perceived the participant as ready, s/he initiated the perturbation trial. At trial initiation a 10-second motion capture trial was triggered, and screen #2 appeared; this screen instructed the participant to “try not to step” [149]. Three seconds later a green circle appeared (screen #3); participants were pre-instructed to focus on the white crosshairs located in the middle of this circle. The appearance of the green circle informed the participant that the treadmill belts would move within the next 2-4 seconds; the pre-perturbation time period was randomized through our software. When the perturbation was complete a red circle appeared (screen #4); this circle indicated that the perturbation was over. After a 3-second delay, the participant was instructed to reposition his/her feet (screen #5) at the trial start-line marked on the treadmill frame (this was to ensure the participant was located directly under the overhead support system at the start of each trial). Meanwhile, the investigator recorded the outcome of the trial as stepping

16 or non-stepping (for stepping trials the leg used for stepping (right or left) was also recorded), set the software to the perturbation level appropriate for the next trial (as described below and shown in Figure 2B), and reset the display on the instruction monitor to screen #1. When both the investigator and participant were ready, the next trial was initiated; time between trials was ~20s. Perturbation characteristics, as programmed, are shown in Table 1 and Figure 4. Perturbations followed a trapezoid-shaped velocity profile with each perturbation lasting ~600-700ms. Perturbation difficulty levels were defined by acceleration rate and peak velocity. Acceleration and deceleration rates were equal in absolute value; their rates and time periods (200ms each) were controlled through custom software, as was perturbation peak velocity. However, due to limitations in treadmill capabilities (system operating at 10Hz), the time at peak velocity could only be controlled within a 200-300ms range. Variability in time at peak velocity produced variability in treadmill belt displacement, at each level. That is, at each perturbation level a range of displacements was possible; these displacement ranges overlapped across levels, especially when perturbations were at the higher speeds. This overlap across levels challenged participants to respond to disturbances that varied differentially (i.e., those equal in velocity could vary in displacement and vice versa).

17

Figure 3: Treadmill Set-up and Participant Instructions. (A.) A Bertec® split-belt treadmill controlled by custom software was used to deliver posterior perturbations to a standing participant; for safety, the participant wore a harness attached to an overhead support. A Qualisys® motion capture system was used to record the movement of reflective markers (attached to the treadmill belts along the outer edge) during each perturbation trial. Instructions, time-synced to the perturbations through custom software, were displayed on the monitor located to the participant’s front. (B.) Participant instructions are shown sequentially from top to bottom. Before each trial the participant was instructed to stand quietly (Screen #1). At the start of each trial the participant was instructed to try not to step (Screen #2). After a 3s delay, a green circle appeared (Screen #3); participants were pre-instructed to focus on the crosshairs in the middle of this circle. Treadmill perturbation began after a randomly-generated 2- 4s delay. When the perturbation was complete a red circle appeared (Screen #4). Participants were then instructed to reposition their feet at the start-line marked on the treadmill frame (Screen #5). After repositioning, Screen #1 reappeared in preparation for the next trial.

18 Table 1: Perturbation Parameters.

Acceleration/ Peak Displacement Perturbation Deceleration Velocity Range* Level Rate (m/s2) (cm/s) (cm) 1 0.25 5 2-2.5 2 0.5 10 4-5 3 0.75 15 6-7.5 4 1 20 8-10 5 1.25 25 10-12.5 6 1.5 30 12-15 7 1.75 35 14-17.5 8 2 40 16-20 9 2.25 45 18-22.5 10 2.5 50 20-25 11 2.75 55 22-27.5 12 3 60 24-30 13 3.25 65 26-32.5 14 3.5 70 28-35 15 3.75 75 30-37.5 16 4 80 32-40 17 4.25 85 34-42.5 18 4.5 90 36-45 19 4.75 95 38-47.5 20 5 100 40-50 *Due to variability in time at peak velocity, each perturbation level had a range of displacements. The first value in each range is the displacement expected for a perturbation that was at peak velocity for 200ms; the second value is the displacement expected for a perturbation that was at peak velocity for 300ms.

Figure 4: Perturbation Profile. Perturbations followed a trapezoid-shaped velocity profile. Acceleration and deceleration periods were 200ms each. Time at peak velocity ranged from 200-300ms.

19 During each session, perturbations started at the lowest level and were increased one level at a time until the participant stepped during three consecutive trials (i.e., until Stepping Threshold was reached) [149]. Each participant was then challenged to raise his or her threshold through progressive practice [150] (Figure 2B). Specifically, s/he started one level below threshold, progressed to the next higher level when three successful (i.e., non-stepping) responses were achieved, and transitioned down a level after an observed step. As can be seen in the figure (2B), this up-down ratio ensured that each participant completed more non-stepping than stepping trials, and thus was successful on most of the trials. All participants completed two training runs per session; each run included 30 perturbations. A 5-minute break was provided between runs; during this time the participant was unhooked from the treadmill support system and given the opportunity to walk around or sit down, as s/he preferred. Completion of a Stepping Threshold assessment followed by two runs of progressive practice resulted in each participant experiencing ~70-85 perturbations per training session.

2.2.3 Data Collection and Processing An 11-camera Qualisys® (Göteborg, Sweden) motion capture system was used to record the movement of reflective markers attached directly to the treadmill belts. The purpose for this was to determine the displacement of each perturbation, and to verify its peak velocity. Reflective markers were also attached to participants’ shoes at the second toe and mid-heel. These shoe markers were used to verify stepping during unsuccessful trials. Data were collected at 125Hz. Custom-written MATLAB® codes (The MathWorks Inc., Natick MA, USA) were used to compute the displacement and peak velocity of the observed

20 perturbations. These values were compared with the values expected based on software commands. Results of this comparison, reported in Appendix A, confirm that the observed size of each perturbation was consistent with its commanded (i.e., programmed) parameters. Statistical analyses were completed with JMP®14 (SAS Institute Inc., Cary NC, USA) software. The primary outcome measure for Aim 1 was Stepping Threshold, defined as the level where the magnitude of treadmill-induced perturbations was enough to evoke a stepping response on three consecutive trials. A two-tailed paired samples t-test was used to test the effect of training on Stepping Threshold. Statistical tests used for additional analyses are reported with their respective results. Statistical significance for all analyses was set at α=0.05.

2.3 Results Individual Stepping Thresholds are shown in Figures 5 and 6. Figure 5 shows the overall effect of training on Stepping Threshold. All participants except one had a higher Stepping Threshold at final assessment than at baseline. For the group, the difference in Stepping Threshold from baseline to final was highly significant (t(14) = 5.932, p < .0001, d = 1.531). Figure 6 shows the session by session progress of each participant. The lowest baseline Stepping Threshold across all participants was Level 7 (P16); the highest was Level 15 (P26). Four participants (P17, P19, P20, P24) showed an initial decline in Stepping Threshold after baseline assessment followed by an overall increase in this measure at final assessment. The average increase in Stepping Threshold from baseline to final was 3.5 levels; the largest increase was 7 levels (P11, P26). Over the

21 course of the experiment no participant successfully responded to a Level 20 perturbation.

Stepping Threshold 20 P11 18 P12 16 P14 P16 14 P17 12 P19 10 P20 P21 8 P23 Perturbation Level 6 P24

4 P26 P27 2 P28 0 P29 Baseline Final

Figure 5: Stepping Threshold at Baseline and Final. Fifteen participants (P) completed reactive balance training. Stepping Threshold was defined as the perturbation level that evoked a stepping response on three consecutive trials. There was a significant difference in Stepping Threshold from baseline to final assessment, for the group (p < .0001).

Figure 6: Stepping Threshold by Participant and Session. For each participant (P), Stepping Threshold was determined at the beginning of each session. Stepping Threshold was defined as the perturbation level (L) that evoked a stepping response on three consecutive trials. Sessions included one baseline assessment (BL), six training sessions (T#1-T#6), and one final assessment (F). L-values in the lower right corner of each plot indicate the participant’s Stepping Threshold at BL and F, respectively.

22

23 Group averages for percent change in Stepping Threshold from baseline are shown, by session, in Figure 7. On average the greatest session to session changes were seen at the beginning of trainings #2-4. Change in Stepping Threshold appeared to level off after training #5.

Figure 7: Stepping Threshold Percent Change from Baseline, Group Averages by Session. Group averages were based on fifteen participants. Sessions were baseline assessment (BL), training (T#1-T#6), and final assessment (F). Error bars indicate SEM.

2.4 Additional Analyses While the primary purpose of Aim 1 was to determine if our reactive balance training protocol would lead to changes in motor performance as measured by Stepping Threshold, several additional analyses were conducted. The purpose of these additional analyses was to assess the impact of other variables on our outcome measure.

24 2.4.1 Effect of Displacement Range on Outcome

Figure 8: Stepping Threshold by Perturbation Velocity and Perturbation Displacement. The peak velocity and displacement of treadmill belts, computed from motion capture data, were averaged across the three trials that constituted Stepping Threshold. Averaged values were compared, baseline to final, for each variable. In each case the difference in Stepping Threshold was statistically significant (both p < .0001).

Due to limitations in treadmill capabilities (system operating at 10Hz), a range of displacements was possible at each perturbation level. These ranges overlapped across levels, especially when perturbations were at the higher speeds. To assess the impact of this overlap on our results, Stepping Thresholds were computed based on

25 both the peak velocity of the perturbations and on perturbation displacements. These perturbation variables were computed from treadmill movement data recorded through motion capture, and then averaged across the three unsuccessful (i.e. stepping) trials that constituted Stepping Threshold. As shown in Figure 8, for both variables, the difference in Stepping Threshold from baseline to final was highly significant (two- tailed paired samples t-test: for velocity, t(14) = 5.884, p < .0001; for displacement, t(14) = 5.561, p < .0001).

2.4.2 Effect of Baseline Stepping Threshold on Outcome It was possible that the degree to which a participant could improve Stepping Threshold depended on his or her baseline Stepping Threshold. To test for this possibility, we used a linear model to compare training outcome to baseline performance. Training outcome was defined two ways. First, it was defined as the change in Stepping Threshold from baseline to final assessment (i.e., computed as final Stepping Threshold - baseline Stepping Threshold). Next, it was defined as the percent change in Stepping Threshold from baseline to final (i.e., computed as ((final Stepping Threshold - baseline Stepping Threshold)/baseline Stepping Threshold) x100). As shown in Figure 9, training outcome was independent of baseline performance. That is, there was no significant correlation between change in Stepping Threshold and baseline Stepping Threshold (Figure 9A: r = 0.035, p = .901). There also was no significant correlation between percent change in Stepping Threshold and baseline Stepping Threshold (Figure 9B: r = -0.259, p = .352).

26

Figure 9: Effect of Baseline Stepping Threshold on Outcome. Based on a linear model, there was no effect of baseline Stepping Threshold on (A.) change in Stepping Threshold or on (B.) percent change in Stepping Threshold.

2.4.3 Effect of Age on Stepping Threshold Previous studies have shown that both age and morbidity affect one’s ability to respond to external disturbances of upright posture [151-158]. To avoid the potential confound of these factors to our results, we limited study participation to healthy adults aged 25-55yrs. However, even with these restrictions, it was possible that age-

27 related differences in performance occurred. Therefore, we used a linear model to test the effect of age on both baseline Stepping Threshold and change in Stepping Threshold (i.e., final Stepping Threshold minus baseline Stepping Threshold).

Figure 10: Effect of Age on Stepping Threshold. Based on a linear model, there was (A.) a significant negative effect of age on baseline Stepping Threshold but (B.) no significant effect of age on change in Stepping Threshold (i.e., final Stepping Threshold minus baseline Stepping Threshold).

28 Data shown in Figure 10 suggest a reduction in both of these variables after ~age 45. However, for our participants the effect of age was statistically significant only for baseline Stepping Threshold (Figure 10A: r = -0.526, p = .044; Figure 10B: r = -0.418, p = .120).

2.5 Discussion The primary purpose of this aim was to test the extent to which our training protocol could improve a non-stepping response to rapid posterior surface translations in healthy adults aged 25-55yrs. Younger healthy adults are not a population at-risk for falls. However, in order to investigate the neural adaptations underlying changes in the performance of this task, we had to establish that our training protocol could improve this balance reaction in these individuals. Our hypothesis was that with practice, healthy younger adults would learn to respond to faster and larger posterior perturbations, without stepping. As expected, most participants were able to learn this. That is, most participants were able to improve their feet-in-place response to posteriorly-directed surface translations as demonstrated by increases in Stepping Threshold from baseline to final assessment. Moreover, their ability to improve this response was not dependent on their baseline Stepping Threshold or on their age. Similar to other perturbation-based training studies we found that older participants tended to step in response to smaller perturbations [151-155,157], as evidenced by their lower baseline Stepping Threshold. Also, similar to other perturbation-based training studies, we found that motor improvement occurred quickly [40]. On average, the largest changes in Stepping Threshold were seen at the beginning of trainings #2-4. Change in Stepping Threshold appeared to level off after

29 training #5. This leveling off suggests that our protocol was of sufficient length to capture a participant’s capacity for training-induced change. In addition to demonstrating that our training protocol led to improved task performance, we also confirmed that the technology we developed led to treadmill perturbations that were of the size expected (Appendix A). A limitation to our technology was that the time at peak velocity could only be controlled within a 200- 300ms window. This led to a range of displacements at each perturbation level. However, this variability in displacement and the resultant overlap of displacement ranges across perturbation levels did not affect our overall results. Thus, in this chapter, we established that the protocol we developed can be used as a paradigm for investigating the neural adaptations underlying changes in motor performance due to perturbation-based training.

2.6 Additional Applications Our motivation for developing this protocol was to address the gap in knowledge regarding changes in the neural substrate that underlie an improved reactive balance response However, the protocol we developed may be useful beyond this purpose. Our training protocol is different from others in that it targets a non- stepping (fixed-support) response to external perturbations. Change-in-support reactions, such as stepping, have greater stabilization potential than do fixed-support (i.e., non-stepping) reactions, in that they quickly increase the size of the base of support [17,145]. However, factors such as space to step and/or a fixture to grasp are not always available (e.g., when someone is bumped while standing on a street curb). Training that targets a non-stepping response may be beneficial for restoring equilibrium under these circumstances.

30 Additionally, not all postural disturbances are external (i.e., environment- based) [135]. Upright balance can also be disturbed by movement generated internally, such as when turning, bending, or reaching [9,129,159]. Training that targets a non-stepping (fixed-support) response to external perturbations may help defend against falls caused by internally-generated disturbances. Finally, protocols that train stepping responses may not be appropriate for some individuals. The size of the perturbation needed to evoke large single- or multiple-step responses may be unsafe for individuals with low bone density or for those with neural, sensory, or musculoskeletal deficits that prevent rapid execution of complex limb movements [143]. Additionally, perturbations that are larger and faster than the ones we used may be intolerable for individuals who are highly anxious or fearful of falling. Our non-stepping protocol may be a more practical therapeutic approach for these individuals. The protocol we developed was well-received; participants expressed a desire to improve their Stepping Threshold and spontaneously displayed enthusiasm during training. No participants withdrew from the study due to fatigue or injury. Additionally, while perturbation magnitudes were enough to stimulate motor improvement, no participant reported training-related anxiety or fear of injury.

Finally, our protocol was challenging enough that no participant was able to respond successfully to the highest level of perturbation. Given these findings and observations we suggest that our reactive balance training protocol may be useful to other applications.

31 Chapter 3

AIM 2: THE EFFECT OF TRAINING A NON-STEPPING RESPONSE TO POSTERIOR SURFACE TRANSLATIONS ON THE SOLEUS H-REFLEX

3.1 Introduction The spinal cord plays an important role in motor control. It is the final point at which the central nervous system (CNS) can exert control over motor output and the first point in the system to receive sensory feedback regarding movement performance [160]. A spinal reflex (i.e., a motor response for which the neural circuitry responsible is contained entirely within the spinal cord) is the shortest pathway available for adjusting motor output based on sensory input; thus, spinal reflexes are important for both injury prevention and motor performance. Spinal reflexes are triggered by sensory receptors located in the skin, joints, and muscles [for review, 161]. Cutaneous reflexes include the flexion-withdrawal reflex in which flexor muscles contract in response to painful stimuli. Stretch reflexes occur in response to changes in muscle length detected by muscle spindles (i.e., sensory receptors located within the muscle tissue). A familiar example of this reflex is the knee-jerk reflex, in which the muscle stretch produced by tapping the tendon of the quadriceps femoris causes the homonymous muscle to contract, thereby extending the knee. Spinal stretch reflexes (referred to simply as spinal reflexes from here forward) contribute to the neural substrates underlying both simple and complex motor tasks [2- 6]. These spinal pathways are modulated based on task requirements [65-81,162] and

32 they can change during skill acquisition [82-103]. Activity-dependent change in spinal reflex behavior is likely the result of interactions between descending (supra- spinal) and peripheral inputs [4]. The importance of peripheral (i.e., sensory) inputs to spinal cord plasticity can be seen in studies of animals with transected spinal cords and humans with spinal cord injuries; in both these populations, treadmill training has been shown to modify the spinal cord and improve motor performance [163,164]. The importance of descending inputs can be seen in humans with damage to the brain (e.g., cerebral palsy, stroke, traumatic brain injury) and in animals with corticospinal tract ablations. For example, the loss of inputs from the brain can lead to changes in the spinal cord (e.g., hyperreflexia) that result in disordered movement or, as in the event of cerebral palsy, can result in a failure to develop proper reflex patterns [4,63]. Further, in studies with animals it has been shown that operant conditioning of spinal reflexes (i.e., change in reflex amplitude based on a reward contingency) only occurs when the corticospinal tract is intact, again highlighting the importance of supra-spinal inputs [165-167]. Spinal reflex behavior can be reliably measured through the Hoffman reflex (H- reflex) [104-106,109]. The H-reflex is a muscle response generated mainly by Ia (sensory) afferents synapsing on spinal motor neurons. This response is elicited through low-level percutaneous electrical stimulation of a mixed peripheral nerve. When stimulation is applied at the appropriate current level, it generates two electromyographic (EMG) responses in the target muscle: stimulation of the motor axons produces a direct muscle response (M-wave); stimulation of the sensory nerve fibers produces an H-reflex. Changes in H-reflex amplitude, with respect to M-wave

33 amplitude and to the background EMG (bEMG) level, provide a window to the CNS plasticity underlying short and long-term motor adaptations [107,108]. While the H-reflex can be elicited from any muscle containing muscle spindles [168], the H-reflex studied most extensively is that of the soleus muscle. Reasons for this selectivity include accessibility of the peripheral nerve innervating this muscle, as well as the predominance of slow motor units associated with this muscle [162]. The soleus muscle is also important to the control of human upright posture [12]. Studies comparing the magnitude of the soleus H-reflex between older and younger participants report that the size of this H-reflex appears to decrease with age [141], while studies investigating age-related changes in motor performance report that the risk of injury due to falls increases with age [10,18,143]. Perturbation-based training in which participants repeatedly attempt to recover from a mechanical perturbation has been shown to improve the reactive response needed to maintain upright balance (equilibrium) after an external disturbance, in older adults [26,28,29,33,34,36,38-40]. However, potential changes in soleus H-reflex behavior due to perturbation-based training have not been investigated. Thus, it is not known if changes in soleus H- reflex behavior contribute to these improved reactive responses. In this chapter (Aim 2) we begin to address this research gap by investigating the effect of training a non- stepping response to posterior surface translations on soleus H-reflex behavior, bilaterally.

3.1.1 Short-term and Long-term Adaptation of the Soleus H-reflex Modulation (i.e., short-term adaptation) of the soleus H-reflex has been demonstrated in several studies. These studies show that the size of the reflex differs depending on the motor task being performed. For example, the soleus H-reflex

34 down-regulates when position is changed from prone to standing [71,88]. Further, the amplitude of the H-reflex with respect to bEMG differs between standing, walking, and running. Here, the reflex is highest during standing, lower during walking, and lowest during running. It has been suggested that the larger reflex observed during standing is consistent with the soleus’s role in controlling ankle position during stance, while the reduction in reflex size (relative to bEMG) during running may be necessary to prevent saturation of motor output [65,66,70]. Thus, the CNS modulates soleus reflex behavior based on the functional requirements of the motor task [67]. Reflex size can also differ across different phases of a motor task. For example, during the gait cycle, the soleus H-reflex is largest during the stance phase and smallest during the swing phase [65,162]. Again, these modulations align with task requirements. Specifically, at the end of the stance phase the larger soleus reflex assists with foot push-off, while the absence of the soleus reflex during the swing phase allows the ankle to dorsiflex thereby preventing contact of the toes with the ground [67]. Context-dependent modulation of spinal reflexes has also been demonstrated during walking and standing. Specifically, soleus H-reflex amplitude is lower when walking on a narrow beam than when walking on a treadmill or over-ground [68,69].

It has also been found that this reflex down-regulates in response to increased postural anxiety [169] and postural threats such as when standing on the edge of a raised platform [170]. Down-regulation of the soleus reflex in response to threats to postural stability may represent an increase in cortical control (i.e., supra-spinal induced presynaptic inhibition) and/or co-contraction at the ankle joint for the purpose of increasing ankle stiffness [69,140,169,170].

35 Interestingly, reflex modulation in response to changes in environmental conditions during standing has been shown to differ between the young and the old. Specifically, when control of upright posture was complicated, first by standing on an unstable (foam) surface, and next by occlusion of vision (while standing on foam), young participants decreased reflex amplitude, in stepwise progression, as the task became more difficult. In contrast, when older participants stood on the unstable surface, reflex amplitude increased when vision was still available, and decreased when vision was occluded [76]. While the reasons for these differences are not clear, they may reflect age-related differences in strategies used to control upright posture. More permanent (long-term) change in soleus H-reflex behavior (i.e., spinal plasticity) has been associated with skill acquisition. For example, it was shown that modern dancers have smaller H-reflexes than do untrained controls [99]. Further, in a seminal study by Nielsen and colleagues the amplitude of the soleus H-reflex (measured while sitting at rest) was found to differ among participants based on the level and type of his or her physical activity; specifically, highly active individuals typically had a larger H-reflex than did sedentary individuals, however this reflex was smallest in ballet dancers even though their group was the most physically active of all the groups studied [85].

Other athletic populations that have been studied include swimmers, badminton players, power-trained athletes (e.g., sprinters, volleyball players), and endurance- trained athletes (e.g., triathletes, cross-country skiers) [82,83,87,91,98,102]. The overall conclusions drawn from these studies are that athletes trained for explosive movements (i.e., powerful anaerobic activities) have smaller H-reflexes than do non- trained controls, while athletes trained for endurance (i.e., aerobic) sports have larger

36 H-reflexes than do non-trained or power-trained individuals [82,83,91]. These differences in reflex behavior were identified by comparing soleus H-max/M-max ratios (i.e., the maximum size of the H-reflex with respect to the maximum M-wave, as determined from H-reflex and M-wave recruitment curves), across groups; differences in this ratio were the result of lower H-max values, not of higher M-max values [83]. The difference in H-max/M-max between these groups is considered to correspond with the functional properties of motor units; specifically, endurance athletes purportedly have a higher incidence of small, slow, fatigue-resistant motor units, while power athletes have either a higher incidence of, or preferentially recruit, large fast, easily-fatigued motor units.

The cross-sectional nature of the athlete studies referenced above makes it difficult to know if the between-group differences are a consequence of training, or if they are a function of genetic predisposition. An examination of two studies with swimmers provides support for the notion of training-specific influences on reflex behavior. First, in a cross-sectional study, experienced swimmers (i.e., those with >10yrs swimming experience) were shown to have a larger H-max/M-max ratio than non-trained controls [98]. Later, in a longitudinal study of swimmers, the effect of training on reflex behavior was tested at specific timepoints within a training regime; it was reported that soleus H-max/M-max ratios were reduced during periods of intense anaerobic training and rebounded during recovery periods (i.e., during periods when training was less intense) [87]. The authors interpretation of this result was that the H-reflex changes were related to training-induced changes in anaerobic power. Within-subject effects of training on H-reflex amplitude have been found in other studies as well. For example, twelve weeks of alpine skiing led to an increase in

37 soleus H-reflex amplitude in a group of elderly participants [100], while ten sessions of slackline training (i.e., standing or walking on tensioned ribbons of nylon webbing), conducted over the course of four weeks, led to a decrease in H-reflex amplitude in the participants who were trained [101]. Further, following a training protocol during which cyclists were instructed to maintain constant speed despite frequent changes in pedal resistance, it was found that performance improvement (i.e., improved ability to maintain constant speed) correlated with a reduction in soleus H-reflex amplitude [95]. Finally, strong support for training-induced change in reflex behavior comes from operant conditioning studies in which the size of an H-reflex (or spinal stretch reflex) was changed based on a reward contingency [171-178]. During operant conditioning, pathways descending from the brain influence spinal pathways either directly, or indirectly through spinal interneurons. When a change in behavior is rewarded, the descending influence is maintained; maintenance of this influence gradually changes the spinal cord itself. That is, during operant conditioning protocols, plasticity at the level of the brain gradually induces plasticity at the level of the spinal cord [for review, 179]. Operant conditioning protocols have been used to both up-regulate and down- regulate soleus H-reflexes (in animals and in humans) [173,177,178,180-184]. These behavioral changes were retained and led to improved motor function in both animals and humans with spinal cord injury. Specifically, up-conditioning (i.e., up-regulation through operant conditioning) of the right soleus H-reflex improved locomotor functioning in rats with midthoracic transection of the right lateral column of the spinal cord [57], while down-conditioning of the soleus H-reflex in people with

38 hyperreflexia due to incomplete spinal cord injury, led to improved walking speed and greater step symmetry [58,62]. In sum, these previous studies highlight the importance of spinal reflexes to motor control. They also demonstrate that spinal reflexes are plastic (i.e., spinal reflexes can change). Changes in spinal reflex behavior occur in response to context, activity, pathology, and aging. Understanding how training interventions affect spinal reflexes is important to the goal of understanding motor learning in general; this knowledge is also important to the more specific goal of developing training protocols that guide spinal plasticity in a beneficial manner.

3.1.2 Perturbation-based Reactive Balance Training and the Soleus H-reflex Perturbation-based training, in which a person stands on a moveable platform or computer-controlled treadmill and repeatedly attempts to recover from rapid surface translations, has been shown to improve the reactive response needed to maintain upright balance during external disturbances, in some populations at-risk for falls. These populations include older adults [26,28,29,33,34,38-40], those with Parkinson’s disease [23,35,36], and those with unilateral transfemoral or transtibial amputation [27,32]. Studies also show that the effects of perturbation-based training are retained [40,185,186], and that the improved reactive response leads to fewer reported falls [34]. The effectiveness of perturbation-based training has been attributed to its task- specificity [31]. Computer-controlled treadmills allow for the delivery of consistent and controlled perturbations. In a controlled environment, participants can safely practice generating the movement pattern necessary for producing a reactive response that arrests a fall. However, while perturbation-based studies demonstrate that

39 practice in responding to treadmill perturbations can lead to improved reactive balance responses, the neuroplasticity (i.e., changes in the brain and spinal cord) underlying acquisition of this new motor skill (i.e., acquisition of an improved balance reaction) is not clearly understood. For humans, maintaining upright balance (equilibrium) during postural disturbances is not a trivial task given our high center of mass and small base of support [12,120,121]. In addition, the larger the perturbation, the more rapid and coordinated a reactive balance response must be in order to prevent a fall. These responses can include the coordinated activation of leg, arm, and torso musculature [114,116-118,135,145,187-190]; as such the neuroplasticity underlying an improved reactive balance response likely occurs at multiple CNS sites including sites within the spinal cord. During standing, the muscular response to external perturbations typically begins at the ankle joint (ankle strategy) and extends proximally (hip strategy) when an ankle strategy alone is insufficient [111-113,191]. This early activation of ankle musculature persists even if a stepping response is pre-planned and/or occurs [115]. The early onset and persistence of this muscle response suggest that the contribution of lower leg spinal reflexes to reactive balance behaviors may be important to fall prevention. Soleus H-reflex behavior has been shown to differ between older and younger adults [71,73,76,77,141,192]. The increased fall risk in older persons may be associated with age-related differences in reflex behavior. To determine if treadmill-based perturbation training changes lower leg spinal reflexes, we developed a reactive balance training protocol that targets activation of the soleus muscle. We then used the H-reflex to determine if the reflex behavior of

40 this muscle changed pre- to post-training. In order to more easily identify adaptations specific to our training protocol, we limited our initial investigation to healthy younger adults. Healthy adults, aged 25-55yrs, completed our six-session training protocol over the course of 3-4 weeks. During each session the participant stood on a computer-controlled treadmill and responded to ~70-85 perturbations. All perturbations were directed posteriorly (i.e., induced anterior sway). Participants were instructed to try not to step when responding to perturbations. The size (i.e., velocity and displacement) of each perturbation was adjusted trial by trial, based on the participant’s performance. These trial by trial adjustments ensured that the participant was continually challenged to respond to perturbations of greater magnitude, while also ensuring that s/he was successful (i.e., able to recover standing equilibrium without stepping) on the majority of trials. Stepping Threshold, defined as the perturbation level (size) that elicited a step on three consecutive trials, was used to measure performance. (For additional information regarding our training protocol and training outcome, see Chapter 2 of this document.) By limiting our training protocol to non-stepping responses to posterior surface translations, we hoped to gain insight regarding spinal reflex adaptations specific to the soleus muscle. Given the previously established importance of the soleus muscle to anterior-posterior stability (equilibrium) during standing, we hypothesized that training a non-stepping response to rapid posterior surface translations would lead to changes in the amplitude of the soleus H-reflex (relative to M-wave amplitude and bEMG). We further hypothesized that these changes would correlate with an improved reactive balance response. That is, we expected a change in soleus H-reflex

41 behavior to correlate with a change in the size of the perturbation that a participant could successfully respond to, as measured by Stepping Threshold. Activity-dependent plasticity occurs at multiple sites within the CNS, including sites within the spinal cord [2-7]. By determining the effect of training a non-stepping response to rapid posterior perturbations on the soleus H-reflex, we begin to fill the gap in knowledge regarding the neural adaptations underlying perturbation-based reactive balance training. The insight gained from this study could provide information regarding specific sites within the CNS to target during training, when the goal of the intervention is to improve postural control and prevent falls.

3.2 Methods

3.2.1 Participants

This study was approved by the University of Delaware Institutional Review Board; regulatory review and oversight were granted to the New York State Department of Health Institutional Review Board. The study was conducted at the National Center for Adaptive Neurotechnologies, located within the Wadsworth Center, New York Department of Health, in Albany NY and was completed by the same participants (6 men and 9 women, aged 26-53yrs) as was Aim 1 (Chapter 2).

Each participant reported no history of neurological disease or injury, no current medical condition limiting use of or sensation in the neck, back, arms, or legs, or interfering with standing or walking, and no current use of medication(s) that affect balance. All participants had normal or corrected-to-normal vision and provided written informed consent prior to participation. All were compensated for their time.

42 3.2.2 Protocol Each participant completed reactive balance training as described in Aim 1 (Chapter 2). Before training began and after completion of the six training sessions, soleus H-reflex/M-wave recruitment curves and soleus H-reflex control trials (i.e., H- reflex responses elicited while bEMG and M-wave size were kept constant) were recorded bilaterally during standing, using Evoked Potential Operant Conditioning Software (EPOCS) [178] (Figure 11). Training-induced change in H-reflex control trial (CT) amplitude was determined by comparing baseline CT amplitude with that measured following completion of our reactive balance training protocol. H-reflex data were collected according to methodology previously established and reported by our group [59,61,178]. Specifically, optimal locations for both the percutaneous stimulation of the posterior tibial nerve (PTN), and the recording of soleus and tibialis anterior (TA) muscle activity, were identified during screening sessions and mapped with respect to physical landmarks (e.g., moles, scars, and veins). Templates, indicating these locations and landmarks, were made with clear 12- gauge vinyl (Jo-Ann Stores, LLC.; Hudson OH, USA); these templates were used during baseline and final assessments to ensure consistent placement of electrodes across sessions. Additionally, at screening, soleus and TA EMG levels during normal standing (i.e., bEMG levels) were determined for each participant. Soleus H-reflex/M-wave recruitment curves began below M-wave and H- reflex thresholds (i.e., at a stimulus level 1.0mA below that required to elicit an EMG response) and terminated at M-max (i.e., at the stimulus level required to elicit the maximum M-wave) (Figure 11C). Stimulus level was incremented in 1.0mA steps. For both the H-reflex and the M-wave, EMG responses were averaged across four trials per stimulus level.

43

Figure 11: Aim 2 Experimental Protocol. (A.) This protocol consisted of two screening sessions (S1, S2), two baseline assessment sessions (BL1, BL2), six reactive balance training sessions, and two final assessment sessions (F1, F2). Soleus H-reflex/M-wave recruitment curves (RCs) and soleus H-reflex control trials (CTs) were collected at S1, S2, BL1, BL2, F1, F2. (B.) During collection of H-reflex/M-wave RCs and H-reflex CTs, the participant used visual feedback to maintain background electromyographic (bEMG) activity of the soleus muscle (green bar) within a specific range; this range was determined at screening, based on the individual’s normal muscle activity during quiet standing, and visually displayed with hatch-marks. (C.) Soleus H-reflex/M-wave RCs were elicited through percutaneous electrical stimulation of the posterior tibial nerve. All RCs began below M-wave (black circle) and H-reflex (blue circle) thresholds (i.e., at a stimulus level 1.0mA below that required to elicit an EMG response) and terminated at M-max (i.e., at the stimulus level required to elicit the maximum M-wave); the stimulus level was incremented in 1.0mA steps. (D.) During soleus H-reflex CTs the size of the M-wave was kept constant. M-wave size was pre- determined for each participant individually, based on his/her recruitment curve; selection criterion was the lowest stimulus level that elicited a change in both the M-wave and the H-reflex. To determine if reactive balance training changed soleus reflex behavior, the amplitude of the H- reflex with respect to bEMG level was compared across CTs, pre- to post-training.

44 The M-wave value used for soleus H-reflex CTs (Figure 11D) was determined for each participant individually, based on his/her recruitment curves. M-wave selection criterion was the lowest stimulus level at which a change in stimulus level corresponded with a change in both the M-wave and the H-reflex. During CTs, M- wave size was monitored by the investigator; if needed small occasional adjustments were made to the stimulus level in order to maintain the M-wave at its pre-determined size. Participants received no feedback regarding H-reflex size during CTs. During the collection of both recruitment curves and CTs, participants used visual feedback to maintain soleus bEMG within his/her specified range as determined at screening (Figure 11B). Electrical stimulation occurred only after bEMG was within range for at least two seconds and only after a minimum 5-second inter- stimulus interval. To control for reciprocal inhibition, TA muscle activity was monitored; no electrical stimulation occurred if antagonist muscle activity was above the specified level. Participants typically completed two screening sessions (one session per leg); additional sessions were conducted if needed. The first three participants completed one baseline and one final assessment each, in accordance with the protocol of our initial pilot study; the next twelve participants completed two baseline and two final assessments each. Soleus H-reflex/M-wave recruitments curves and soleus H-reflex CTs were recorded separately for each leg at each assessment session. To control for diurnal variation in H-reflex size [193-196], all H-reflex sessions occurred at the same time of day for each participant. Each H-reflex session lasted 60-90 minutes; to avoid fatigue, participants were given regular breaks from

45 standing. Time between each of the two baseline sessions and each of the two final sessions was typically 1-2 days.

3.2.3 Data Collection and Processing During all H-reflex sessions soleus and TA EMG activity were collected bilaterally with a Bortec Biomedical® (Calgary AB, Canada) AMT 8 EMG Wire Telemetry System. These data were sampled at 3200Hz, amplified, and bandpass filtered at 10-1000Hz. A Digitimer® (Hertfordshire, UK) DS8R constant-current stimulator was used to stimulate the left and right PTNs percutaneously. For each muscle, a pair of self-adhesive surface Ag-AgCl electrodes (2.2 x 3.5cm; Vermed, Buffalo NY, USA) was placed on the skin, in vertical arrangement, with electrode centers ~3cm apart and the long dimension of the electrode pad perpendicular to the line between the electrode centers. The TA electrodes were placed over the muscle belly; soleus electrodes were placed below the gastrocnemii at the location where the largest M-wave could be recorded. Stimulation electrodes (self-adhesive surface Ag-AgCl; Vermed) were placed behind the knee, with the cathode (2.2 x 2.2cm) located in the popliteal fossa and the anode (2.2 x 3.5cm) placed ~7cm superiorly. Exact locations for all electrodes were determined through an iterative process at screening, such that the soleus H-reflex threshold was minimized, soleus H-max and M-max were maximized, and stimulation of other nerves was avoided. For all participants, a 1ms monophasic square-wave pulse was used during electrical stimulation. For safety, the stimulation level was capped at 50mA. In addition, each participant was instructed to notify the investigator if s/he did not want the stimulus level increased further. We were not able to get complete recruitment

46 curves for three participants: for two participants maximal M-wave was not reached; for one participant amplifier blocking occurred at the higher stimulus levels. The primary outcome measure for this aim was soleus H-reflex CT amplitude, defined as the magnitude of the soleus H-reflex normalized to its bEMG (averaged over a 50ms pre-stimulation window). This variable was calculated based on mean rectified EMG data as output from EPOCS. For each participant, CT amplitude was averaged across five trials per leg, per session. For participants with two baseline sessions and two final sessions, CT values were averaged across each of these two sessions (i.e., across the two baseline sessions and across the two final sessions). Peak-to-peak EMG values were used to determine soleus H-max and M-max amplitudes. Statistical analyses were completed with JMP®14 (SAS Institute Inc., Cary NC, USA) software. A two-tailed paired samples t-test was used to determine the effect of training on soleus H-reflex CT amplitude. A bivariate analysis based on a linear model was conducted to determine if changes in soleus H-reflex behavior correlated with changes in Stepping Threshold (the primary outcome measure for Aim 1, Chapter 2). Statistical significance for all analyses was set at α=0.05.

47 3.3 Results

3.3.1 Effect of Reactive Balance Training on the Soleus H-reflex

Figure 12: Soleus H-reflex Control Trial Amplitude at Baseline and Final Assessments. Soleus H-reflex control trial (CT) data were collected bilaterally from fifteen participants.

To determine the effect of reactive balance training on soleus H-reflex behavior, CT amplitude was compared baseline to final. As shown in Figure 12, soleus H-reflex CT amplitude changed differently across participants and legs. To

48 examine these differences more closely, data were separated according to the leg’s primary role during reactive balance training in Aim 1 (Chapter 2). Stepping leg was defined as the leg used most often for stepping during unsuccessful perturbation trials (i.e., during trials when the participant stepped in response to a perturbation). The other leg was defined as the Stance leg. For multi-step responses, Stepping and Stance leg assignment was based on the first step. Stepping and Stance legs were determined based on data collected across the entire Aim 1 protocol (i.e., as the cumulative result of all perturbation trials conducted during baseline assessment, reactive balance training sessions #1-6, and final assessment). For one participant (P21) Stepping and Stance legs could not be discerned as she stepped equally with each leg: 50 of 56 steps were taken with her right leg during baseline through training #3; 55 of 61 steps were taken with her left leg during training #4 through final assessment. Data from the remaining fourteen participants are shown in Figure 13 [template, 197]. For the group, there was no significant difference in CT amplitude, from baseline to final, for either leg (for Stepping leg: t(13) = 1.024, p = .324, d = 0.274; for Stance leg: t(13) = 0.471, p = 0.645, d = 0.126). As can be seen in Figure 13 and Table 2, the direction of change (i.e., up- or down-regulation of the soleus H-reflex) was not consistent across participants. In the Stepping leg, CT amplitude increased for nine participants and decreased for five participants. In the Stance leg, CT amplitude increased for seven participants and decreased for seven participants. The median difference in CT amplitude, from baseline to final, was 2.82 for the Stepping leg and -0.16 for the Stance leg.

49

Figure 13: Soleus H-reflex Control Trial Data Grouped by Leg. Bilateral H-reflex data from fourteen participants were separated according to the leg’s primary role during unsuccessful perturbation trials in Aim 1 (i.e., according the whether the leg was used for Stepping or for Stance during perturbation trials that led to a stepping response). Data were further categorized based on the direction of H-reflex change in the Stepping leg, from baseline to final. Specifically, if the participant showed an increase in Stepping leg control trial (CT) amplitude, data for both the Stepping leg and the Stance leg are shown in red, regardless of the direction of change in the Stance leg. Similarly, if the participant showed a decrease in Stepping leg CT amplitude, data are shown in blue for both legs, regardless of the direction of change in the Stance leg. For ten participants the direction of H-reflex change was the same for both legs; six participants showed a bilateral increase in CT amplitude and four participants showed a bilateral decrease. For three of the four remaining participants, CT amplitude increased in the Stepping leg and decreased in the Stance leg; for the fourth participant results were the reverse. For the group, there were no significant differences in CT amplitude, baseline to final, for either leg (for Stepping, p = .324; for Stance, p = .645). The median difference in CT amplitude (black bar) was 2.82 for the Stepping leg and -0.16 for the Stance leg.

50 To indicate whether the direction of H-reflex change was consistent within participants (i.e., if the direction change was the same across a participant’s two legs), data in Figure 13 were categorized (color-coded) based on the direction of change in the Stepping leg. Stepping leg was chosen as the criterion variable because, for the group, the absolute value of the median difference in CT amplitude was greater for this leg. Additionally, on average, participants with a mixed response (i.e., those for whom the direction of H-reflex change differed across legs) showed greater change in the Stepping leg.

Table 2: Participant Data: Primary outcome measures at baseline, and change in primary outcome measures, from baseline to final, for Aim 2 and Aim 1.

Change in % Change H-reflex CT H-reflex CT Baseline Change %Change Baseline Amplitude Amplitude H-Max/M-Max Stepping Stepping Stepping Participant Step Stance Step Stance Threshold Threshold Step Stance Threshold P11 23.01 7.83 73% 10% 7 58% 0.43 0.86 12 P12 2.18 -1.26 37% -12% 5 45% NA NA 11 P14 -10.36 -13.31 -14% -26% 6 67% 0.79 0.54 9 P16 19.23 23.69 42% 66% 7 100% 0.45 0.33 7 P17 4.62 -5.09 17% -11% 1 11% 0.27 0.49 9 P19 -8.90 -16.58 -6% -9% 1 9% 0.85 0.73 11 P20 -11.46 -13.93 -24% -19% 2 17% 0.72 0.94 12 P21* 10.52 -5.06 20% -11% 6 60% 0.58 0.58 10 P23 4.89 -2.59 3% -1% 3 30% 0.74 0.82 10 P24 3.25 34.81 60% 103% 3 30% NA NA 10 P26 6.88 8.48 9% 17% 4 27% 0.54 0.68 15 P27 -0.27 1.80 -3% 14% 2 18% 0.22 0.24 11 P28 2.57 1.14 13% 6% 3 30% 0.27 0.35 10 P29 -1.25 -0.72 -15% -12% 0 0% NA NA 8 P30 3.06 0.40 41% 12% 2 22% 0.26 0.22 9 *P21 Stepping leg was indeterminate: values in Stepping column are for the leg used for stepping at baseline and training sessions #1-3, values in Stance column are for the leg used stepping during training sessions #4-6 and at final assessment. NA: Not available. For three participants we were not able to obtain complete recruitment curves.

51 For ten participants the direction of H-reflex change was the same for both legs, with six participants (P11, P16, P24, P26, P28, P30) showing a bilateral increase in CT amplitude and four participants (P14, P19, P20, P29) showing a bilateral decrease. Five participants had a mixed response. For three of these five participants (P12, P17, P23) CT amplitude increased in the Stepping leg and decreased in the Stance leg; for the fourth participant (P27) results were the reverse. For the fifth participant (P21, data not shown in Figure 13) there was an increase in CT amplitude for the leg used for stepping during (Aim 1) baseline through training #3 (for H-reflex: baseline CT=54, final CT=64), and a decrease in CT amplitude for the leg used for stepping during (Aim 1) training #4 through final (for H-reflex: baseline CT=46, final

CT=41).

3.3.2 Correlation between Change in Stepping Threshold and Change in Soleus H-reflex Behavior

A linear model was used to determine if there was a relationship between change in soleus H-reflex behavior and change in Stepping Threshold (the primary outcome measure for Aim 1, Chapter 2), following perturbation-based reactive balance training. Change in soleus H-reflex behavior was defined as the difference in H-reflex CT amplitude from baseline to final. Stepping Threshold was defined (in

Aim 1) as the perturbation level that elicited a stepping response on three consecutive trials; change in Stepping Threshold was defined as the difference in Stepping Threshold from baseline to final.

52

Figure 14: Change in Stepping Threshold vs. Change in Soleus H-reflex Control Trial Amplitude, by Leg. Based on a linear model, there was (A.) a significant positive correlation between change in Stepping Threshold and change in H-reflex CT amplitude, for the Stepping leg, but (B.) no significant correlation between change in Stepping Threshold and change in H-reflex CT amplitude, for the Stance leg.

As shown in Figure 14, there was a statistically significant positive correlation between change in Stepping Threshold and change in H-reflex CT amplitude of the

53 Stepping leg (Figure 14A: r = 0.561, p = .037); in this leg, change in H-reflex CT amplitude accounted for ~32 percent of the variance in change in Stepping Threshold. There was no significant correlation between change in Stepping Threshold and change in H-reflex CT amplitude for the Stance leg (Figure 14B: r = 0.351, p = .220). The results of this analysis were based on data from fourteen participants. Stepping leg for P21 was indeterminate, therefore her data were not included in the by-leg analysis. For P21, Stepping Threshold increased six levels. For the leg used for stepping during (Aim 1) baseline through training #3, change in H-reflex CT amplitude was +10; for the leg used for stepping during (Aim 1) training #4 through final, change in H-reflex CT amplitude was -5.

3.4 Additional Analyses The primary purpose of this aim (Aim 2) was to determine the longitudinal effect of our reactive balance training protocol on the soleus H-reflex, bilaterally. The primary outcome measure for this aim was CT amplitude, defined as the mean rectified value of the H-reflex at a specific M-wave, divided by the bEMG. This outcome measure is different than that used in cross-sectional studies of the H-reflex. In cross-sectional studies, the measure typically used to identify between-group differences in H-reflex behavior is the H-max/M-max ratio (that is, the maximum size of the H-reflex with respect to the maximum M-wave, as determined from H-reflex and M-wave recruitment curves, and measured peak to peak). Previous studies have reported between-group difference in H-max/M-max, based on both age [71,73,141,192] and past athletic experience [82-85,91,98,103]. In our study, participation was limited to those aged 25-55yrs. However, our participants were not selected or grouped according to either past athletic experience or the size of

54 his/her H-reflex at baseline, as measured by H-max/M-max. As shown in Table 2, for our study, baseline H-max/M-max varied among participants and legs. To determine the effect of baseline H-max/M-max on our results, two additional analyses were conducted. First, we tested the effect of baseline H-max/M- max on change in H-reflex CT amplitude. Next, we tested the effect of baseline H- max/M-max on baseline Stepping Threshold, for each leg.

3.4.1 Effect of Baseline H-max/M-max on Change in H-reflex Control Trial Amplitude

A linear model was used to determine the effect of baseline H-max/M-max on training-induced change in soleus H-reflex behavior. Analyses were based on bilateral data from the twelve participants for whom we had complete recruitment curves; P21 was not included in the by-leg analyses as her Stepping leg was indeterminate. As shown in Figure 15, a lower baseline H-max/M-max tended to predict an increase in CT amplitude, while a higher baseline H-max/M-max tended to predict a decrease in CT amplitude. Effect size was greatest for the Stepping leg. The correlation between measures was statistically significant only for combined data (i.e., based on data from both limbs); sample size for the by-leg analyses was likely too small to detect an effect. For the Stepping leg: r = -0.465, p = .150; for the Stance leg: r = -0.370, p =

.263; for both legs combined: r = -0.414, p = .045.

55

Figure 15: Change in Soleus H-reflex CT Amplitude vs. Baseline Soleus H-max/M- max. Based on a linear model, there was a negative effect of baseline H- max/M-max on changes in H-reflex CT amplitude, for the soleus muscle. Effect size was largest for the Stepping leg. Analyses were based on bilateral data from the twelve participants for whom we had complete recruitment curves. Black dots indicate the participant for whom Stepping leg was indeterminate; this participant was not included in by- leg analyses.

3.4.2 Effect of Baseline H-max/M-max on Baseline Stepping Threshold A linear model was used to determine the effect of baseline H-max/M-max on baseline Stepping Threshold. A by-leg analysis was completed based on data from the twelve participants for whom we had complete recruitment curves. Similar to earlier analyses, limb data were separated based on the leg’s primary role during unsuccessful perturbation trials in Aim 1 (Chapter 2). Data from P21 (Stepping leg indeterminate) were included in these analyses, as her baseline H-max/M-max was the same bilaterally.

56

Figure 16: Baseline Stepping Threshold vs. Baseline H-max/M-max, by Leg. (A.) Stepping leg and (B.) Stance leg data are from the twelve participants for whom we had complete recruitment curves. Based on a linear model, baseline Stepping Threshold was more strongly correlated with the baseline H-max/M-max of the Stance leg than with that of the Stepping leg.

As shown in Figure 16, the correlation between baseline H-max/M-max and baseline Stepping Threshold was stronger for the Stance leg than it was for the Stepping leg. Correlations were not statistically significant for either leg (for

57 Stepping: r = 0.167, p = .601; for Stance: r = 0.531, p = .076). However, our data suggest that, at baseline, there may have been a moderate effect of Stance leg H- max/M-max on Stepping Threshold, that we were not sufficiently powered to detect.

3.5 Discussion Perturbation-based training has recently emerged as an effective method for improving reactive balance during upright posture [31,34,36,38-40]. Previous studies have used computer-controlled treadmills and moveable platforms to improve the reactive response needed to arrest trips and slips, in some patient populations [23,24,26-29,32-36,38-40]. These devices have also been used to identify biomechanical differences between successful versus unsuccessful reactive responses, to these types of perturbations [24,26,27,32,42,43,45-47,49]. However, changes in the neural substrate that underlie an improved reactive response to rapid surface translations, have not been identified. In this aim we sought to determine the effect of treadmill-based perturbation training on lower leg reflex behavior as measured by the H-reflex. We hypothesized that the training of a non-stepping response to rapid posterior surface translations would lead to changes in the amplitude of the soleus H-reflex, relative to M-wave amplitude and to bEMG. We further hypothesized that these changes would correlate with changes in Stepping Threshold (i.e., with changes in the size of the perturbation a participant could respond to without stepping). Our hypotheses were partially supported. First, while for the group there was no significant difference in CT amplitude, from baseline to final, the soleus H-reflex did change (>10%) in at least one leg, for most participants (see Table 2). However, the direction of this reflex change (i.e., up- or down-regulation of the soleus H-reflex)

58 differed across participants; and in some participants the direction of change differed across legs. Second, a significant positive correlation of moderate effect size was found between change in Stepping Threshold and change in H-reflex CT amplitude, for the Stepping leg. However, there was no significant correlation between change in Stepping Threshold and change in H-reflex CT amplitude, for the Stance leg. Two important questions arise from these primary results. The first question is, why did the direction of H-reflex change differ across participants (and in some cases across a participant’s two legs)? The second question is, why was the relationship between change in Stepping Threshold and change in H-reflex CT amplitude stronger for the Stepping leg than for the Stance leg? While the answers to these questions are not yet clear, we propose two interrelated factors that could have contributed to our results; these factors are 1) reactive response strategy and 2) baseline physiology.

3.5.1 Differences in the Direction of H-reflex Change In our study, participants were not given explicit instructions on how to respond to the perturbations; they were simply instructed to try not to step. Given the many degrees of freedom available to a multiple-joint system, more than one movement pattern could lead to a successful reactive response (i.e., a response that avoids a step). The (conscious or unconscious) choice of response strategy could be due to factors that include past experience and/or baseline physiology. Results of previous studies indicate that a disturbance of upright posture can generate motor responses across multiple body segments [116,198-201]. While the non-stepping protocol we developed was designed to target the soleus muscle, activation of the soleus muscle is only one part of a reactive response. Reactive

59 balance responses can include arm and torso musculature as well as that of the leg [114,116-118,135,145,187-190]. Thus, some participants could have used a response that was less reliant on ankle musculature (i.e., ankle strategy) and more reliant on counter-rotation strategies, either at the level of the pelvis (i.e., hip strategy), the level of the arms, or both. An ankle strategy consists of controlling center of mass position through torques generated primarily at the ankle and knee, while the trunk is maintained in vertical orientation; during a hip strategy the center of mass is repositioned quickly by bending the torso at the hips [113,202]. These strategies are not hard-wired but rather learned through experience; further, they are not mutually exclusive [113,135]. That is, people use a continuum of strategies. During reactive responses the ankle strategy tends to dominate during smaller perturbations, while a mixed (ankle and hip) strategy emerges as the magnitude of the perturbation increases. However, no response is purely reactive; responses are influenced by context, behavioral goals, and past experience [135]. For example, a person’s tripping response likely differs depending on whether or not s/he is carrying a cup of hot coffee. In our study participants knew a perturbation was coming (even though they did not know exactly when or what size the perturbation would be); as such, each participant could proactively strategize, based on his/her own unique past experiences [200]. Thus, the degree to which each participant used an ankle strategy likely differed across the group; this difference may have contributed to our differing results. An in-depth analysis of kinematic data (beyond the scope of this dissertation) should help clarify the influence of different strategies on our H-reflex results.

60 It is also possible that some participants may have co-activated ankle musculature in an effort to limit the magnitude of AP sway (i.e., postural oscillations). The ability to respond to a perturbation is influenced by the direction of sway at the time of the perturbation; a large forward oscillation at the start of a perturbation would make it more difficult to maintain a feet-in-place position. Differences in levels of co- contraction could also explain differences in the direction of H-reflex change among participants. As mentioned earlier, dancers tend to have smaller soleus H-reflexes than do untrained controls [85,88,99]. It has been suggested that the smaller H- reflexes found in ballet dancers are the result of years of training during which co- contraction of the soleus and TA muscles is required in order to maintain balance during classical ballet postures [85]. Additionally, it has been shown that, in humans, down-regulation of the soleus H-reflex occurs when walking along a narrow beam, and that this down-regulation is accompanied by co-contraction of the ankle muscles [69]. Thus, in our population, those who down-regulated the H-reflex may also have been co-contracting ankle musculature. An examination of the associated EMG (beyond the scope of this dissertation) should reveal whether differences in co- contraction correlate with differences in the direction of H-reflex change. The direction of H-reflex change (and the response strategy) could also have been a function of baseline reflex behavior. Studies have shown that even when different strategies (or combinations of strategies) are used, early activation of muscles at the ankle joint persists [115]; however, there is a physiological limit to reflex size [168]. If a participant’s reflex was either at or approaching this limit at baseline, the strategy s/he used to improve Stepping Threshold may have been different than that used by a participant who had a small baseline reflex (i.e., a participant who had room

61 to increase reflex size). Indeed, baseline H-max/M-max values for three of the four participants showing a bilateral H-reflex decrease, were higher than that of the other participants (Table 2). Specifically, the average baseline H-max/M-max across all twelve participants for whom we had recruitment curves, was .51 for the Stepping Leg and .57 for the Stance Leg. For those with a bilateral decrease in H-reflex amplitude, Stepping and Stance leg values were, respectively, .79 and .54 (P14), .85 and .73 (P19), and.72 and .94 (P20). For the fourth participant with a bilateral H-reflex decrease (P29) we were not able to get complete recruitment curves. An alternate, albeit speculative, interpretation of our results is one of optimization. As shown in Figure 15, higher baseline H-max/M-max values tended to predict decreases in CT amplitude, while lower baseline H-max/M-max values tended to predict increases in CT amplitude. This convergence towards a mid-point could represent a compromise between two competing requirements for our task. In order to produce a successful non-stepping response to a rapid posterior surface translation using an ankle strategy, muscle torques of sufficient magnitude must be generated quickly in order to return the center of mass to its original position within the base of support. For this requirement an increase in reflex size would intuitively seem more useful than a decrease. However, excessive reflex activity, prior to and following the initial perturbation response, could have destabilizing effects which would be counter-productive. Several earlier studies have shown that when self-movement during standing must be limited in order to maintain equilibrium within a small, fixed, base of support, the soleus H-reflex is down-regulated. For example, participants down-regulated this reflex when standing on the edge of a raised

62 platform [170], when standing on a compliant surface [72,76,203], and when standing with their feet in tandem [77]. In our study, participants were similarly instructed to stand within a small, fixed, base of support (i.e., with feet side-by-side, at approximately shoulder width apart), however they also had the additional task of responding to a large, rapid, external disturbance, while remaining in this position. Thus, a difference between our study and these earlier studies is that, in our study, the environmental conditions changed, and the participant needed to be able to respond to this change quickly and sufficiently. Finally, it has been shown that participants scale the magnitude of a response, to the magnitude of disequilibrium, based on a prediction of perturbation characteristics [204,205]. However, in order to ensure that participants were continually challenged, we adjusted the size of the perturbation trial by trial; therefore, the size of the perturbation could not be predicted. Additionally, there was a random 2-4 second delay between the start of a perturbation trial and the time the treadmill belts began to move; therefore, participants could not accurately predict the start of the perturbation. In an earlier study, when participants were instructed to walk backward on a treadmill, the size of the soleus H-reflex increased presumably due to the uncertainty of foot placement and possible loss of balance associated with acquisition of this new motor skill [94]. In our study, uncertainty regarding both the size (i.e., speed and displacement) and the timing of each perturbation may have produced similar effects. Thus, during our protocol, reflex size may have been adjusted in a manner that best satisfied competing requirements. Down-regulation of a large reflex would help

63 ensure that excessive reflex activity did not destabilize position, while up-regulation of a small reflex would ensure that a rapid reactive response could be produced when needed. Training may have led to an optimization of reflex size that addressed both these competing needs. To summarize, it is not yet clear why the direction of H-reflex change differed among participants. Differences in the direction of H-reflex change could have resulted from differences in the type of strategy used to respond to perturbations (e.g. co-contraction, ankle strategy, hip strategy). Differences in response strategy could have been the result of past experience and/or baseline reflex behavior. Differences in the direction of H-reflex change could also reflect a task-dependent optimization of reflex size. Future in-depth analyses of associated kinematic & EMG data should provide insight regarding these differences.

3.5.2 Between-leg Differences in H-reflex Behavior Our second interesting result was that of limb-specific differences in H-reflex behavior. A unique feature of our study was the collection of H-reflex data bilaterally. While the primary purpose of this study was not to investigate functional lateralization of motor behavior, our bilateral data did reveal between-leg differences in both baseline H-max/M-max and training-induced change in H-reflex (Table 2); further, these differences were associated with the primary role of the leg during unsuccessful perturbation trials (i.e., during trials that elicited a step). At baseline, Stepping Threshold appeared to correlate positively with the H- max/M-max of the soleus muscle in the Stance leg (Figure 16). That is, participants with larger Stance leg H-reflexes tended to have higher baseline Stepping Thresholds. There was no correlation between Stepping leg H-reflexes and Stepping Threshold, at

64 baseline. It was also observed that participants tended to step with the leg that had the smaller reflex (i.e., the leg with the lower baseline H-max/M-max ratio). At baseline, for seven of the twelve participants with complete recruitment curves, the Stepping leg H-max/M-max was lower than the Stance leg H-max/M-max; for 4 participants this was the reverse. For one participant baseline H-max/M-max was equal across both legs; interestingly, this is also the participant for whom Stepping leg was indeterminate. Of further interest is the observation that, for this participant, the H- reflex only increased in the leg used for stepping during the first half of the protocol; this observation suggests that when the participant switched her Stepping legs, she may have switched to the leg with the smaller reflex.

Training-induced changes in Stepping Threshold (which in Aim 1 were shown to be independent of baseline Stepping Threshold) tended to correlate with changes in the H-reflex of the Stepping Leg. For most participants, change in the Stepping leg H- reflex was positive (i.e., for nine of fourteen participants the soleus H-reflex of the Stepping leg up-regulated). For these (nine) participants, the average change in Stepping Threshold was 3.9 levels (median change was 3 levels); for the (five) participants whose Stepping leg H-reflex down-regulated, the average change in Stepping Threshold was 2.2 levels (median change was 2 levels). That is, the average change in Stepping Threshold was higher for those whose Stepping leg H-reflex up- regulated. For the (five) participants whose Stepping leg H-reflex down-regulated, this reflex tended to be larger at baseline than it was for those whose Stepping leg H- reflex up-regulated. Additionally, for four of these five participants (i.e., Stepping leg down-regulators) down-regulation was bilateral. Thus, in general, our perturbation- based reactive balance training protocol led to an increase in the size of the soleus H-

65 reflex of the Stepping leg; generally, the Stepping leg was the leg with the lower baseline H-max/M-max. Departures from these trends tended to be the participants who had the larger Stepping leg H-reflexes at baseline. In general, these participants tended to down-regulate the soleus H-reflex bilaterally, and to show less training- induced change in Stepping Threshold. With respect to lateralization, it is difficult to compare our results with that of other studies, as the soleus H-reflex is typically not tested bilaterally. While some studies report testing the H-reflex of the dominant leg [91,169], the idea of leg dominance is itself, controversial. In the upper extremities (i.e., the arms and hands) functional differentiation between the right and left sides is easily demonstrated through activities such as writing and throwing [206]; however, for the legs many of our daily activities depend on right-left symmetry (e.g., walking, running, swimming). Additionally, among those who support the idea of lower limb functional lateralization, there is little agreement on what defines leg dominance [207-211]. That said, for the upper extremities it has been suggested that lateralization reflects specialized but complimentary motor control of each limb. Specifically, the dominant hemisphere purportedly utilizes predictive mechanisms to control the dynamic properties of movement while the non-dominant side optimizes postural stability [212]. Similarly, functional differentiation of the lower limbs has been defined as precision versus support [213], and as mobilization versus stabilization [210]. For our study these terms would seem to align best with the roles of the Stepping leg and the Stance leg, respectively. Given the between leg differences found in our results, future studies should consider limbs distinctly in terms of their roles in the specific task.

66 3.6 Conclusions In conclusion, our reactive balance training protocol led to changes in both Stepping Threshold and reflex behavior of the soleus muscle, for most participants. However, there were within and between subject differences in the direction of H- reflex change. These differences in direction could have been due to differences in the strategy used to improve Stepping Threshold. Differences in strategy could have been due to past experience and/or the participant’s baseline physiology. These differences could also represent an optimization of reflex behavior based on competing postural control requirements. Future in-depth analyses of associated kinematic & EMG data should provide further insight regarding the within and between subject differences found in this study. Interestingly, we also found within-subject (between-leg) differences in participant baseline H-max/M-max ratios. While the concept of laterality, or limb dominance, is well recognized with respect to the arms (and hands), laterality with respect to the legs is controversial. Our bilateral data lends support to the idea of a functional directionality in motor behavior across the lower limbs. We found that higher baseline Stepping Thresholds tended to be associated with larger baseline Stance leg H-reflexes, and that participants tended to step with the leg that had the smaller H-reflex at baseline.

3.7 Limitations and Future Directions

A limitation to our study was that participants were not given explicit instructions on how to respond to the perturbations; they were simply instructed to try not to step. As such, the strategy used for responding may have differed among participants. Bilateral soleus and TA EMG data were collected during all perturbation

67 trials; additionally, full-body kinematic data were collected during perturbation trials at baseline and final assessments. Future in-depth analysis of these associated data should provide insight regarding possible differences in response strategies as well as their potential effect on our outcome measures. Additionally, future experiments in which elements of the reactive response (e.g., movement at the level of the pelvis or arms) are controlled, could be conducted to test the effect of response strategy on soleus H-reflex behavior. Interestingly, previous treadmill-based studies have reported that the biomechanical changes associated with an improved response to posterior surface translations include a decrease in both the angle and the velocity of trunk flexion, and an increase in step length [24,26,27,32]. As mentioned earlier, during an ankle strategy feet-in place response, trunk orientation remains vertical [135]; thus, these reported changes are more in line with that of an ankle strategy than that of a hip strategy. Additionally, studies related to gait indicate that the H-reflex is largest at the end of the stance phase, presumably to assist with push off of the foot [67]. Thus, perturbation-based training, in which the reactive response to posterior perturbations is limited to a feet- in-place ankle strategy, could lead to an increase in H-reflex behavior that proves beneficial to both non-stepping and stepping responses.

Another limitation to our study was that we were not able to obtain complete recruitment curves from all participants; as such, we were not able to fully evaluate the effect of baseline H-max/M-max on our results. Further, participants were not screened or grouped based on past experience and/or baseline reflex behavior. Differences in past experience and/or baseline reflex behavior could have, together or separately, contributed to the emergence of two groups (i.e., those whose H-reflex up-

68 regulated and those whose H-reflex down-regulated); we were not adequately powered for two different groups. In future studies participants could be grouped according to baseline reflex behavior (as measured during screening), based on an a priori criterion value; participants could be recruited such that the size of each group was equal, and both groups were sufficiently powered for statistical analyses. Finally, a limitation to our study was the use of younger (i.e., ages 25-55yrs), healthy participants. The results of this study do not inform us as to the effect of our training protocol on the soleus H-reflex in patient populations or in older adults. While previous studies indicate that the size of the soleus H-reflex decreases with age [73,76,141,192], it has also been shown that older participants can increase the size of this reflex with training [100]. Given that age-related decreases in the size of the soleus H-reflex have been associated with decreased stability [141], and that older participants can improve their stepping response to external disturbances through perturbation-based training [26,28,29,33,34,36,38-40], an important research question is whether these improved stepping responses (or, as in our protocol, improved non- stepping responses) are accompanied by an increase in soleus H-reflex amplitude, in older adults. Future studies should examine the effect of perturbation-based training on the soleus H-reflex in both older adults and patient populations at risk for falls.

69 Chapter 4

AIM 3: THE EFFECT OF TRAINING A NON-STEPPING RESPONSE TO POSTERIOR SURFACE TRANSLATIONS ON POSTURAL STEADINESS

4.1 Introduction The central nervous system (CNS) is highly adaptive; plasticity, defined here as changes in the brain and spinal cord that underlie changes in motor performance, occurs continually in response to growth and maturation, trauma or pathology, and skill acquisition [3,4]. Motor learning, the neurological process underlying skill acquisition, induces plasticity at multiple sites within the CNS, including the spinal cord [2-7]. Spinal reflexes (i.e., motor responses in which the neural circuitry responsible is contained entirely within the spinal cord) contribute to both simple and complex motor behaviors [2,5,6] and can change in response to training [82-103]. Because the spinal cord is shared by multiple behaviors, changes in a spinal pathway (e.g., a spinal reflex) that comprise the primary plasticity underlying a newly acquired skill can alter the performance of previously learned skills or trigger compensatory plasticity in their CNS substrates [2-7]. Thus, identifying the effect of specific training protocols on specific spinal reflexes, and then assessing the impact of these reflex changes on the performance of other motor tasks, are important to the development of rehabilitation protocols that induce and guide beneficial plasticity [2,4,5,58,59,61-63]. For example, operant conditioning protocols developed to down- regulate the soleus H-reflex during standing, have been used to improve locomotion in patients with spinal cord injury [58].

70 In Aim 1 of this study we trained standing participants to respond to rapid posterior surface translations without stepping; the primary purpose for this training was to investigate the neuroplasticity underlying improvement in this type of response. In Aim 2 we found that changes in Stepping Threshold correlated positively with changes in the amplitude of the soleus H-reflex of the Stepping leg (i.e., the leg used most often for stepping during unsuccessful perturbation trials). In this chapter (Aim 3) we examine the effect of both our training protocol and changes in H-reflex behavior on a second postural task, specifically postural steadiness during quiet bipedal stance. While postural balance during standing refers to one’s ability to stay upright or to recover equilibrium after a perturbation; postural steadiness refers to the ability to stand as still as possible [124,125]. Postural sway [136] (i.e., movement of the center of mass in the horizontal plane) is a measure of steadiness; it is typically analyzed based on center-of-pressure (CoP) data collected from force plates located under a person’s feet during quiet standing [124,214-220]. Increases in postural sway have been correlated with pathology, aging, and increased risk for falls [9,136,221-228]. Increased postural sway (i.e., unsteadiness) may also contribute to a fear of falling [229], and thereby deter individuals from participation in physical and social activities beneficial to their overall health [17,19-21]. While our reactive balance training protocol did not target postural steadiness directly, our non-stepping (i.e., feet-in-place) reactive balance task is functionally relevant to postural steadiness in that both skills require control of the center of mass within a fixed base of support. This control is achieved, in part, through activation of the soleus muscle. During standing, the position of the body’s center of mass is

71 located anterior to the ankle joint; to maintain upright posture, activation of the soleus muscle is necessary in order to counter the effects of gravity [12]. During a rapid posterior surface translation, the activation level of the soleus muscle increases in response to the anterior sway generated by the posterior perturbation [112,113,116]. Given the biomechanical congruence between these two tasks, it seems likely that certain sites and synapses within their neural substrates are shared. If neural sites and synapses are shared across two or more substrates, then training that improves the performance of one motor task, may also affect the performance of the other motor task(s) that share these sites and synapses [7]. To determine if changes in postural steadiness occurred in response to reactive balance training, standing CoP data were recorded during quiet bipedal stance, before and after completion of our six-session training protocol. These data were then used to quantify postural sway (a measure of steadiness) in the anterior-posterior (AP) direction. The postural sway variables that we computed and analyzed were the root mean square of CoP displacement (CoPrms) and the mean velocity of CoP displacement (CoPvel). These two variables were selected because they have been shown to be highly reliable, and to have good discriminant capability regarding fall risk [125,215,217]; higher sway values are considered to indicate greater instability

[216]. The primary hypothesis for this aim was that the training of a non-stepping response to rapid posterior surface translations would lead to changes in both AP

CoPrms and AP CoPvel. Further, due to similarities in the role of the soleus muscle for each task, our secondary hypothesis was that training-induced changes in postural steadiness would correlate with the changes found in soleus H-reflex behavior in Aim

72 2. That is, we expected changes in soleus H-reflex behavior to mediate changes in postural steadiness during quiet unperturbed stance. The relationship between reactive balance ability and postural steadiness has been tested in two earlier studies. Specifically, in the first study postural sway measures were compared with the ability to produce stepping responses following three different types of postural disturbances (release from a tethered forward lean position, recovery of gait following treadmill acceleration, and recovery following a trip ) [230]; later, sway measures were compared with the ability to respond to release from a tethered forward lean position without stepping (i.e., with a feet-in-place response) [231]. In both cases, no correlation was found between postural steadiness and the ability to respond to postural disturbances. However, in both of these studies, the results were based on cross-sectional samples. In contrast, the design of our study was longitudinal; that is, we tested whether training-induced changes in reactive balance correlated with changes in postural steadiness. Additionally, the participants of the two earlier studies were older adults (aged ≥ 70yrs), while the participants in our study were younger adults (aged 25-55yrs). Finally, H-reflex data were not collected in either of the earlier studies. While CoPrms and CoPvel have been shown to be reliable measures, there is no consensus as to the underlying postural control mechanism that an adjustment in these variables represents [216]; our comparison of training-induced changes in H-reflex and CoP data could provide insight regarding the postural control mechanism(s) associated with each of these variables.

73 4.2 Methods

4.2.1 Participants

This study was approved by the University of Delaware Institutional Review Board; regulatory review and oversight were granted to the New York State Department of Health Institutional Review Board. The study was conducted at the National Center for Adaptive Neurotechnologies, located within the Wadsworth Center, New York Department of Health, in Albany NY. Fifteen participants (6 men, 9 women), aged 26-53yrs, completed the study. These participants were the same participants as in Aim 1 (Chapter 2) and Aim 2 (Chapter 3). All participants provided written informed consent prior to participation and were compensated for their time. All participants had normal or corrected-to-normal vision; all had no history of neurological disease or injury, no current medical condition limiting use of or sensation in the neck, back, arms, or legs, or interfering with standing or walking, and no current use of medication(s) that affect balance.

4.2.2 Protocol Each participant completed six sessions of reactive balance training as described in Aim 1(Chapter 2). Before and after training, postural sway was measured during quiet standing (i.e., during bipedal stance) (Figure 17A). To measure sway,

CoP data were collected from side-by-side force plates located under the belts of a split-belt treadmill, while the participant stood with feet approximately shoulder width apart and one foot on each belt/plate. During postural sway trials each participant was instructed to “stand as still as possible” [219], with head facing forward, arms relaxed at the side, and eyes closed [219,232,233]. A blind fold was used to ensure vision remained occluded during each trial. Initial foot placement was traced onto a paper

74 template; this template was used to replicate standing position across trials and sessions [234] (Figure 17B).

Figure 17: Aim 3 Experimental Protocol. (A.) This protocol consisted of one baseline assessment session (BL), six reactive balance training sessions, and one final assessment session (F). Postural sway was measured at BL and F, while participants stood quietly with eyes closed. (B.) During postural sway trials participants stood on an instrumented split-belt treadmill; center of pressure (CoP) data were collected from side by side force plates located under the treadmill belts. Initial foot placement was traced onto a paper template; this template was used to replicate standing position across trials and sessions. (C.) Postural sway variables were computed based on movement of the CoP in the anterior-posterior (AP) direction.

75 Each participant completed three bipedal sway trials per assessment (i.e., at baseline and final). Rest breaks were provided between trials. For the first three participants each trial lasted 60 seconds; for the remaining twelve participants trials were 120 seconds each. For consistency, only data from the first 60 seconds of each trial were used for all participants. For each trial the CoPrms and CoPvel were computed based on movement of the CoP in the AP direction (Figure 17C).

4.2.3 Data Collection and Processing A Qualisys® (Göteborg, Sweden) motion capture system was used to record the CoP and force data for each trial. These data were collected from the two force plates embedded under the belts of an instrumented Bertec® (Columbus OH, USA) split-belt treadmill. Data, collected at 1750Hz (first thee participants) and 1250Hz (last twelve participants) were processed offline, with custom-written MATLAB® (The MathWorks Inc., Natick MA, USA) codes.

CoPnet, defined as the weighted sum of the time-varying position of the CoP collected from each force plate, was calculated for each trial as follows,

��, �(�) ��, �(�) ������(�) = ����(�) � + ����(�) � ��, �(�) + ��, �(�) ��, �(�) + ��, �(�) where CoPl and CoPr are the AP CoP coordinates under the left foot and right foot respectively, and where Fz,l and Fz,r are the vertical ground reaction forces under each

th foot, respectively [12,226]. CoPnet data were then filtered (4 order recursive Butterworth with a 5Hz cut-off), down-sampled to 125Hz, and demeaned. The primary outcome measures for this aim (Aim 3), CoPrms, and CoPvel, were calculated based on CoPnet.

76 For each outcome measure (i.e., CoPrms, and CoPvel), data were averaged across the three bipedal sway trials collected at each time point (i.e., at baseline and final assessments), for each participant. A two-tailed paired samples t-test was used to determine the effect of reactive balance training on each postural sway variable. Additionally, for each postural sway variable, bivariate analyses, based on a linear model, were conducted, by leg, to determine if changes in postural steadiness correlated with changes in soleus H-reflex behavior (the primary outcome measure for Aim 2, Chapter 3). For all analyses statistical significance was set at α=0.05. All statistical analyses were completed with JMP®14 (SAS Institute Inc., Cary NC, USA) software.

77 4.3 Results

4.3.1 Effect of Reactive Balance Training on Postural Steadiness

Figure 18: Postural Sway in the Anterior-Posterior Direction at Baseline and Final. For fifteen participants, postural sway variables were computed based on movement of the center of pressure (CoP) in the anterior-posterior direction. For the group, there was no significant difference in CoP root mean square (RMS) or CoP mean velocity, from baseline to final (for RMS, p = .994; for mean velocity, p = .444).

To determine the effect of reactive balance training on postural steadiness,

CoPrms and CoPvel were compared, baseline to final (Figure 18). No significant effect of training was found, overall (for CoPrms: t(14) = -0.007, p = .994, d = 0.002; for

78 CoPvel: t(14) = -0.788, p = .444, d = 0.203). As in Aim 2, individual data show that the direction of change in these variables differed across participants.

Table 3: Participant Data: Primary outcome measures for Aim 3 and Aim 2. Change and percent change values indicate the difference in outcome measures from baseline to final assessment. Stepping leg refers to the limb that was used most often for stepping during unsuccessful perturbation trials in Aim 1; the other limb is referred to as the Stance leg.

Center of Pressure Center of Pressure Stepping Leg Stance Leg Root Mean Square Mean Velocity H-reflex CT H-reflex CT (mm) (mm/s) Amplitude Amplitude Participant Change %Change Change %Change Change %Change Change %Change P11 -0.53 -12% -0.38 -5% 23.01 73% 7.83 10% P12 0.05 1% 0.37 5% 2.18 37% -1.26 -12% P14 3.35 62% 1.66 24% -10.36 -14% -13.31 -26% P16 -0.89 -12% -5.60 -31% 19.23 42% 23.69 66% P17 -0.34 -9% -1.15 -12% 4.62 17% -5.09 -11% P19 0.13 3% -0.48 -4% -8.90 -6% -16.58 -9% P20 0.51 17% 0.07 1% -11.46 -24% -13.93 -19% P21* 0.24 5% -0.79 -10% 10.52 20% -5.06 -11% P23 -1.00 -16% 0.07 1% 4.89 3% -2.59 -1% P24 -0.52 -10% 0.42 7% 3.25 60% 34.81 103% P26 -1.65 -19% 0.72 7% 6.88 9% 8.48 17% P27 0.25 7% 0.18 3% -0.27 -3% 1.80 14% P28 1.08 29% 0.30 3% 2.57 13% 1.14 6% P29 2.12 36% 2.37 32% -1.25 -15% -0.72 -12% P30 -2.85 -22% -3.81 -21% 3.06 41% 0.40 12% *P21 Stepping leg was indeterminate: values in Stepping column are for the leg used for stepping at baseline and training sessions #1-3, values in Stance column are for the leg used stepping during training sessions #4-6 and at final assessment.

As shown in Table 3, the absolute value of the change in CoPrms, from baseline to final, tended to be greater than that for CoPvel. For CoPrms, this change was ≥ 10% for ten participants; for CoPvel, only six participants showed a change ≥ 10%.

79 Additionally, the direction of change in CoP variables tended to be opposite of the direction of change in H-reflex CT amplitude (the primary outcome measure for Aim 2).

4.3.2 Correlation between Change in Postural Steadiness and Change in Soleus H-reflex Behavior

Figure 19 [template, 197] shows the relationship between the direction of change in CoP variables and the direction of change in soleus H-reflex CT amplitude as reported in Aim 2 (Chapter 3). In Aim 2 it was found that changes in Stepping Threshold were associated primarily with changes in the H-reflex behavior of the Stepping leg (i.e., the leg used most often for stepping during unsuccessful perturbation trials in Aim 1). Therefore, data in this figure were color-coded according to the direction of H-reflex change in the Stepping leg. As shown in Figure 19 and Table 3, there was an inverse relationship between the change in CoPrms and change in the H-reflex of the Stepping leg for 12 of 14 participants; specifically, CoPrms increased for all five participants who showed a post- training decrease in Stepping leg H-reflex CT amplitude (P14, P19, P20, P27, P29), while CoPrms decreased for seven of the nine participants who showed a post-training increase in Stepping leg H-reflex CT amplitude (P11, P16, P17, P23, P24, P26, P30).

This inverse relationship was not as strong for CoPvel; for this measure the direction of change was opposite of that of the H-reflex for only eight of fourteen participants (P11, P14, P16, P17, P20, P27, P29, P30). For P21 Stepping Leg was indeterminate, therefore her data are not included in Figure 19; for her, CoPrms increased and CoPvel decreased, baseline to final.

80

Figure 19: Postural Sway Variables Grouped by Direction of Change in Soleus H- reflex Control Trial Amplitude of the Stepping Leg. This figure shows the relationship between the direction of training-induced change in postural sway variables and the direction of training-induced change in H-reflex control trial amplitude (CT) of the leg used most often for stepping during reactive balance training (i.e., the Stepping leg in Aim 1). Data presented are from fourteen participants; sway variables were computed based on movement of the CoP in the anterior-posterior direction. For participants whose Stepping leg H-reflex increased (data shown in red), the center of pressure root mean square (CoPrms) tended to decrease; for participants whose Stepping leg H-reflex decreased (data shown in blue), the CoPrms tended to increase. For center of pressure mean velocity (CoPvel) there was a similar inverse relationship for participants whose Stepping leg H-reflex decreased, but not for those whose Stepping leg H-reflex increased. The median difference (black bar) in CoPrms from baseline to final was -0.15mm; for CoPvel the median difference was +0.12mm.

To determine the strength of the relationship between training-induced changes in postural steadiness and soleus H-reflex behavior, changes in both CoPrms and CoPvel

81 were compared with changes in the H-reflex CT amplitude of each leg. Again, the Stepping leg was the leg used most often for stepping during unsuccessful perturbation trials in Aim 1; the other leg was the Stance leg. Because Stepping leg was indeterminate for P21, her data were not included in the by-leg analyses.

Based on a linear model, for both CoPrms and CoPvel, the correlation between change in H-reflex and change in postural steadiness was stronger for the Stepping leg than for the Stance leg, however none of these correlations reached the statistical significance level of p < .05. The correlation between change in CoPrms and change in H-reflex CT amplitude is shown in Figure 20A for the Stepping leg (r = -0.505, p = .066), and Figure 20B for the Stance leg (r = -0.385, p = .175). Figures 21A and 21B show the correlation between the change in CoPvel and change in H-reflex CT amplitude for the Stepping leg (r = -0.489, p = .076) and the Stance leg (r = -0.315, p = .272), respectively.

Interestingly, data in Table 3 indicate that when CoP change was ≥ 10%, the direction of change in H-reflex tended to be the same bilaterally. Specifically, for five of the six participants who showed a ≥10% decrease in CoPrms, H-reflex CT amplitude increased bilaterally (P11, P16, P24, P26, P30); for three of the four participants who showed ≥10% increase in CoPrms, H-reflex CT amplitude decreased bilaterally (P14,

P20, P29). Similarly, for two of the four participants whose CoPvel decreased by ≥10%, H-reflex CT amplitude increased bilaterally (P16, P30); for the two participants whose CoPvel increased by ≥10%, H-reflex CT amplitude decreased bilaterally (P14, P29). This observation suggests that changes in postural steadiness depended on whether the direction of H-reflex change was the same bilaterally.

82

Figure 20: Change in Center of Pressure RMS vs. Change in Soleus H-reflex CT Amplitude, by Leg. Based on a linear model, the relationship between change in CoPrms and change in soleus H-reflex CT amplitude was stronger for the (A.) Stepping leg than for the (B.) Stance leg; however, the correlation between these two variables did not reach the statistical significance level of p < .05 for either leg.

83

Figure 21: Change in Center of Pressure Mean Velocity vs. Change in Soleus H- reflex CT Amplitude, by Leg. Based on a linear model, the relationship between change in CoPvel and change in soleus H-reflex CT amplitude was stronger for the (A.) Stepping leg than for the (B.) Stance leg; however, the correlation between these two variables did not reach the statistical significance level of p < .05 for either leg.

4.4 Additional Analyses In two earlier studies it was found that, in older women, the ability to recover from postural disturbances could not be predicted from measures of postural

84 steadiness [230,231]. To determine if this was also true for our (younger) population, we tested the baseline relationship between postural steadiness (i.e., CoPrms and

CoPvel) and the ability of a participant to respond successfully to a postural disturbance, as measured by Stepping Threshold in Aim 1. Similar to these earlier cross-sectional studies, we found no correlation between Stepping Threshold and postural steadiness, at baseline. Additionally, to examine the longitudinal effects of our reactive balance training protocol on this relationship, we compared changes in

Stepping Threshold with changes in CoPrms and changes in CoPvel. Again, based on a linear model, there was no correlation between change in CoP measures and change in Stepping Threshold. For results of these analyses see Table 4.

Table 4: Correlation between Postural Sway Variables and Stepping Threshold.

Baseline Baseline Change in Change in Center of Pressure Center of Pressure Center of Pressure Center of Pressure Root Mean Square Mean Velocity Root Mean Square Mean Velocity r p r p r p r p Baseline -0.124 .661 -0.327 .234 Stepping Threshold Change in -0.005 .984 -0.275 .321 Stepping Threshold

Finally, our data suggest that changes in postural steadiness were related to changes in the amplitude of the soleus H-reflex. That is, increases in H-reflex CT amplitude tended to be associated with increases in postural steadiness as demonstrated by reduced postural sway, and vice versa. To determine if a similar relationship existed at baseline, we compared baseline CoP measures to baseline H- max/M-max values for the twelve participants who had complete recruitment curves; results are shown in Table 5. The strongest relationship between these measures was

85 for CoPvel and the H-max/M-max of the Stance leg; however, none of these correlations met the criterion level for statistical significance.

Table 5: Correlation between Postural Sway Variables and Soleus H-reflex Behavior, at Baseline.

Baseline Baseline Center of Pressure Center of Pressure Root Mean Square Mean Velocity r p r p

Baseline H-max/M-max - 0.176 .584 -0.285 .369 Stepping Leg

Baseline H-max/M-max - 0.378 .226 -0.504 .095 Stance Leg

4.5 Discussion Motor learning induces plasticity throughout the CNS, thus the performance of motor skills not specifically targeted during training or rehabilitation can be affected by new motor learning. That is, the primary plasticity underlying a new motor skill can affect multiple motor behaviors simultaneously if it occurs at sites shared by their neural substrates [2,3,7]. In this aim we examined the effect of training a non-stepping (i.e., feet-in-place) response to rapid posterior surface translations on postural steadiness during quiet unperturbed standing. The motor skill we directly targeted during training was reactive balance; the motor skill not directly targeted, but tested for performance, was postural steadiness. Due to the biomechanical similarity of these two motor skills (i.e., upright control of the center of mass within a fixed base of support), it was hypothesized that our reactive balance training protocol would change performance not only of the targeted task but also of this non-targeted task. Given the importance of the soleus muscle to each of these postural tasks, it was further

86 hypothesized that changes in postural steadiness would correlate with changes in soleus H-reflex behavior. Though our hypotheses were not supported, changes in postural steadiness did occur. For twelve of the fifteen participants the absolute value of change was ≥ 10% for at least one postural sway variable; the measure for which more participants showed ≥10% change was CoPrms (Table 3). However, with respect to our primary hypothesis, we found no overall effect of training on postural steadiness for the group, based on either CoPrms, or CoPvel (Figure 18). As in Aim 2, the direction of change in our outcome measures (i.e., CoP variables) was not the same across participants. Similar to previous studies we found that reactive balance ability could not be predicted by measures of postural steadiness. At baseline, there was no correlation between Stepping Threshold and CoPrms, or between Stepping Threshold and CoPvel. There was also no correlation between change in Stepping Threshold and change in postural steadiness following reactive balance training. Thus, our results extend the findings of two earlier studies in which no relationship was found between postural steadiness and the ability of older adults to respond to postural disturbances156,157; the baseline results for our younger adults indicate the same. Further, we found no relationship between training-induced changes in reactive balance ability and postural steadiness. With respect to our secondary hypothesis, we observed trends suggesting that changes in postural sway were inversely related to changes in H-reflex CT amplitude; again, this finding was primarily for CoPrms (Figure 19). Our by-leg correlations indicate that the inverse relationship between changes in sway and changes in H-reflex was stronger for the Stepping leg than for the Stance leg (for both CoP variables this

87 tended to account for ~25% of the variance in the Stepping leg); however, none of these correlations reached the p < .05 level of significance (Figures 20 and 21). Based on data in Table 3, it was mainly the participants for whom the direction of change in H-reflex was the same bilaterally, who showed ≥ 10% change in postural sway; these participants included all six of the bilateral H-reflex up-regulators and three of the four bilateral H-reflex down-regulators.

For both of our postural sway variables (CoPrms and CoPvel), lower values are considered to indicate better postural control [216]. Therefore, the participants who showed improvement in postural steadiness were primarily those for whom the soleus H-reflex amplitude increased. This is an important finding, given that older adults tend to have both smaller reflexes [141,192] and greater postural sway [136,223], than do younger adults. That is, during standing, larger soleus H-reflexes have been associated with greater postural steadiness. If our reactive balance training protocol can increase soleus H-reflex amplitude in older adults, as it did in some of our younger participants, then our results suggest that this training should also improve their postural steadiness. A similar result was reported for a study that investigated the effect of twelve weeks of alpine skiing (skiing ~3.5 hours/day for ~30 days) on soleus H-reflex behavior in older adults (aged ~60-70yrs). Specifically, for the skiing group, soleus H-reflex amplitude increased while postural sway decreased; for the control group there was no change in either of these measures [100]. In our study, it is interesting that changes in sway tended to correlate with changes in H-reflex, given that we found no relationship between changes in sway and changes in Stepping Threshold. The reason for this may be that, while Stepping Threshold increased for all but one participant, the direction of change in H-reflex

88 associated with Stepping Threshold improvement differed across participants; as discussed in Aim 2 (Chapter 3), differences in the direction of H-reflex change may have resulted from differences in perturbation response strategy (e.g., ankle strategy versus hip strategy) and/or baseline physiology. Again, future in-depth analyses of associated kinematic and EMG data should help clarify the relationship between these outcome measures (i.e., H-reflex CT amplitude and postural sway) and the response strategy a participant used during perturbation trials. Finally, our baseline comparison of H-max/M-max to CoP measures hints at an inverse relationship between CoPvel and the H-reflex of the Stance leg (Table 5). Although the correlation between these two variables was not statistically significant, it is interesting to note that, in Aim 2, baseline correlation between H-max/M-max and Stepping Threshold tended to be stronger for the Stance leg, as well (Figure 16). Thus, overall, baseline H-reflex and sway measures tended to correlate more strongly with the Stance leg, while training-induced changes in these measures tended to correlate more strongly with the Stepping leg. That is, based on our outcome measures, the Stepping leg was more responsive to training. With respect to our two postural sway variables, pre- to post-training changes were greater for CoPrms than for

CoPvel (i.e., more participants showed a ≥10% change in this variable). This observation suggests a difference in postural control mechanisms underlying CoPrms and CoPvel; changes in H-reflex behavior appear to exert greater influence over mechanisms underlying CoPrms.

4.6 Conclusions In conclusion, we found no overall effect of training on postural steadiness, for the group; the direction of change in CoP variables differed across participants. In

89 general, the direction of change in these variables tended to be opposite of the direction of change in soleus H-reflex CT amplitude found in Aim 2; this inverse relationship was strongest for CoPrms. Additionally, the relationship between change in postural sway and change in H-reflex amplitude was stronger for the Stepping leg than for the Stance leg; however, it was primarily the participants for whom the direction of H-reflex change was the same bilaterally, that showed the greatest (i.e., ≥10%) change in postural sway. The most important finding of this aim was the inverse relationship between change in H-reflex CT amplitude and change in CoPrms. Stated differently, we found that reduced postural sway (i.e., improved postural steadiness) was associated with an increase in the size of the soleus H-reflex. Though our results were not statistically significant, this finding warrants further investigation due to the impact it could have on fall prevention. Older adults are at risk for falls; additionally, the soleus H-reflex tends to be smaller in older adults [141,192]. If increasing this reflex improves postural steadiness, then protocols that specifically target up-regulation of the soleus H-reflex (e.g., operant conditioning protocols) could reduce the risk of injury due to falls, in older adults.

4.7 Limitations and Future Directions The main limitation to this study is that analyses were based only on CoP time domain variables; analyses in the frequency domain or CoP structural analyses could provide additional insight regarding how changes in H-reflex behavior relate to changes in postural sway. The frequency content of a CoP signal can reveal preferential involvement of the various sensory inputs that contribute to postural control (i.e., inputs from the visual, vestibular, and somatosensory systems) [220].

90 Further, structural analyses that decompose the signal into subunits in order to characterize the dynamics of postural sway have been used to reveal underlying motor control processes [220,235,236]. For example, in a study examining the relationship between changes in postural sway and modulation of the soleus H-reflex while wearing prism googles, a detrended fluctuation analysis revealed differences in the dynamic characteristics of sway based on whether the H-reflex was up-regulated or down-regulated [236]; these differences were found even though there was no correlation between the H-reflex and the CoP trajectory based on a traditional time domain analysis of sway area. Further, it has been suggested that the balance skills of ballet dancers are not well captured by these traditional measures. While dancers reportedly have smaller soleus H-reflexes [85,88,99] no differences in postural sway were found between dancers and non-dancers [237], or between dancers and track athletes [235], based on time domain CoP measures; however, a recurrence quantification analysis revealed that the dynamic patterns of postural sway in dancers were different than that of the track athletes. Specifically, the dancers’ sway was found to be more stationary, less regular (lower recurrence), and less complex (lower entropy) [235]. Thus, for our study, future analyses of the frequency distribution and structure of the CoP trajectory, could provide insight regarding the directional differences found in our outcome measures. Finally, our Aim 1 data suggest a specialization of function across limbs (i.e., Stepping leg versus Stance leg); our Aim 2 and Aim 3 data indicate different levels of responsiveness to training based on this differentiation. However, Aim 3 outcome measures were computed based CoPnet (i.e., data combined from the two side by side force plates, each one located under a different limb). Further, the greatest levels of

91 change in these CoP variables were for those participants whose H-reflex change was in the same direction for both limbs. To investigate the by-limb relationship between the H-reflex and sway more closely, CoP variables could be computed based on the data from each force plate individually. Additionally, a by-limb comparison based on CoP data collected during unipedal stance, could be conducted. While unipedal stance requires greater mediolateral control than bipedal stance, the soleus muscle has been shown to also contribute significantly to control of upright balance during this task [203]. However, it is not known how changes in soleus H-reflex behavior affect motor performance during unipedal stance.

92 Chapter 5

SUMMARY

For this study, a computer-controlled treadmill and custom software were used to perturb upright balance during quiet stance for the purposes of improving reactive balance, identifying training-induced changes in spinal reflex behavior, and examining the impact of this training on another postural task. The theoretical basis for this study was the negotiated equilibrium model of spinal cord function [7]. According to this model, the primary plasticity underlying acquisition of a new motor skill (i.e., behavior) reflects a negotiation of spinal cord properties among all the skills that share the spinal neurons and synapses with the new skill; thus, when a new motor skill is acquired, changes in spinal cord properties can affect multiple skills concurrently. This spinal plasticity preserves and may even improve the performance of previously acquired skills.

5.1 Summary of Results

Aim 1 showed that our reactive balance training protocol could improve motor performance in healthy younger adults. For 14 of 15 participants (93%) the speed and displacement of the largest posterior surface translation to which they could respond successfully (i.e., without stepping) increased with training. Additionally, the technology we developed to conduct this study proved reliable. The system hardware and software performed as expected; perturbation magnitudes, observed with motion capture technology, matched those expected based on the software commands.

93 Aim 2 found a moderate positive correlation between changes in Stepping Threshold and changes in the H-reflex CT amplitude of the Stepping leg. However, in contrast to this overall result, three of the participants who increased Stepping Threshold also decreased H-reflex CT amplitude, bilaterally. It is not clear why the direction of H-reflex change differed for these participants; comprehensive analyses of the associated kinematic and EMG data may help to answer this question. Aim 3 did not find a correlation between changes in Stepping Threshold and changes in postural steadiness. However, it did reveal a non-significant trend in the relationship between changes in the H-reflex CT amplitude of the Stepping leg and changes in postural steadiness. These changes in postural steadiness were greatest in the participants for whom the direction of change in H-reflex CT amplitude was the same bilaterally. Taken together, these data suggest that changes in postural steadiness were mediated by changes in soleus H-reflex behavior. From a theoretical perspective our results were consistent with the negotiated equilibrium model of spinal cord function [7]. The spinal plasticity associated with the acquisition of the new motor skill appeared to change the performance of a previously acquired skill (i.e., postural steadiness). That is, changes in the soleus H- reflex that correlated with changes in Stepping Threshold also appeared to correlate with changes in postural steadiness. From an applied perspective our results suggest that training protocols designed to increase the amplitude of the soleus H-reflex during standing, might improve reactive balance and postural steadiness concurrently, provided that the reflex is not already large. However, this assertion is made with caution; not all of our

94 results were statistically significant. Future studies could test this hypothesis in populations at risk for falls.

5.2 Important Remaining Questions With respect to future research, several questions remain. Regarding our protocol in particular, the first question is, would the results of our study have been different if the perturbation response strategy used by participants had been specified by the investigator or constrained by the design (e.g. a bar that limited trunk flexion)? If further analyses of the kinematic and EMG data collected during this study suggest that differences in the direction of H-reflex change were related to differences in the biomechanics of the reactive response, future studies may want to investigate how limiting specific movements during this response affects reflex behavior. Before each training trial, the participants did not know what the size of the perturbation would be. They knew only that perturbation size would be near their current Stepping Threshold. Furthermore, they did not know, from trial to trial, whether their Stepping Threshold was improving. Thus, a second question regarding our protocol is whether the training outcome would be different if participants knew the size of the impending perturbation. For example, might this knowledge affect the strategy they adopt? Third, our participants were healthy younger adults; we do not know what the results of our training protocol would be in other populations. A key question is, who would be the most likely to benefit from this training? If age-related reductions in soleus H-reflex amplitude contribute to fall risk, then older adults seem promising candidates. Older adults have benefited from other types of perturbation-based training (e.g., training that improves stepping responses), although it is not known how

95 those protocols affect the soleus H-reflex. Further, previous studies have shown that older adults tend to exhibit more postural sway [136,221,238] and higher levels of soleus and TA co-activation, during standing, than do younger adults [239-241]. While co-contraction is often viewed as a mechanism for stabilizing upright posture, it also reduces the size of the soleus H-reflex through reciprocal inhibition [242,243]. If the smaller H-reflexes found in older adults contribute to fall risk, then co-contraction of the ankle muscles may actually be counterproductive for this population. For example, in two recent studies postural steadiness (as measured by CoP variables in the time domain) decreased in response to ankle muscle co-contraction [241,244]. An important question regarding older adults is whether co-activation of ankle musculature during standing correlates with their overall soleus H-reflex behavior (i.e., H-reflex amplitude and its task-dependent modulation). Finally, our results suggest that up-regulation of the soleus H-reflex may be beneficial for persons at risk for falls, if their baseline H-reflex is relatively small. This possibility might be evaluated with an operant conditioning protocol in which postural steadiness and reactive balance were measured before, during, and after up- conditioning of the soleus H-reflex. The aim would be to determine whether the H- reflex increase was associated with improved control of upright posture.

5.3 Unique Features of our Study The unique features of our study included the hardware and software development. Using a custom-built tachometer and custom software we were able to deliver reliable perturbations with a commercially-available Bertec® split-belt treadmill. While this treadmill is common to many research facilities, it has not to our knowledge ever been successfully programmed for this purpose. Our technology also

96 automated the participant instructions, synchronized them with the perturbation trials, and displayed them on a monitor in front of the participant. This design standardized training across participants and sessions, particularly with respect to inter-trial intervals and participant-investigator interactions. Additionally, while many studies have used computer-controlled treadmills to improve stepping responses to externally-imposed perturbations, we used this device to train a non-stepping response to such disturbances. To do this, our protocol reinforced non-stepping behavior; and as performance improved, challenged participants with larger and larger perturbations. By raising the difficulty level of the perturbation after three successful (non-stepping) trials, and then reducing this level after one unsuccessful (stepping) trial, the protocol maintained a 3:1 ratio between non-stepping and stepping responses. The protocol overlapped displacement magnitudes across perturbation levels. While the incremental change in acceleration and peak velocity between levels remained constant, the displacements associated with each level varied according to the length of time (i.e., 200-300ms) that a perturbation was at peak velocity; the range of displacements at each level overlapped with the ranges of the adjacent levels. This overlap was due to a limitation in treadmill capability. It was nevertheless beneficial, for it made the transition between levels less abrupt and thus enabled a more natural progression to the next perturbation level. Finally, in contrast to previous studies that have focused on the biomechanical characteristics of an improved reactive response, our study explored the changes in the neural substrate that underlie this improved response. In particular, we examined the effect of perturbation-based training on the soleus H-reflexes of both legs (the

97 Stepping leg and the Stance leg). Interestingly, soleus reflex behavior has not typically been examined bilaterally; it has been assumed that, in healthy adults, reflex behavior is similar across legs. In contrast to this assumption we found between-leg differences in soleus H-reflex behavior. We also found that training-induced changes in this behavior depended on how the limb was used during the motor task (i.e., when responding to perturbations).

98 REFERENCES

1. Salmoni AW, Schmidt RA, Walter CB: Knowledge of results and motor learning: a review and critical reappraisal. Psychological bulletin 1984, 95(3):355-386.

2. Wolpaw JR: What can the spinal cord teach us about learning and memory? The Neuroscientist: a review journal bringing neurobiology, neurology and psychiatry 2010, 16(5):532-549.

3. Wolpaw JR: The complex structure of a simple memory. Trends in neurosciences 1997, 20(12):588-594.

4. Wolpaw JR, Tennissen AM: Activity-dependent spinal cord plasticity in health and disease. Annual review of neuroscience 2001, 24:807-843.

5. Wolpaw JR: The education and re-education of the spinal cord. Progress in brain research 2006, 157:261-280.

6. Wolpaw JR: Spinal cord plasticity in acquisition and maintenance of motor skills. Acta physiologica (Oxford, England) 2007, 189(2):155-169.

7. Wolpaw JR: The negotiated equilibrium model of spinal cord function. The Journal of physiology 2018, 596(16):3469-3491.

8. Horak FB: Clinical measurement of postural control in adults. Physical therapy 1987, 67(12):1881-1885.

9. Overstall PW, Exton-Smith AN, Imms FJ, Johnson AL: Falls in the elderly related to postural imbalance. British medical journal 1977, 1(6056):261- 264.

10. Horak FB, Shupert CL, Mirka A: Components of postural dyscontrol in the elderly: a review. Neurobiology of aging 1989, 10(6):727-738.

11. Forster A, Young J: Incidence and consequences of falls due to stroke: a systematic inquiry. BMJ (Clinical research ed) 1995, 311(6997):83-86.

12. Winter D: Human balance and posture control during standing and walking. Gait & posture 1995, 3(4):193-214.

99 13. Kulkarni J, Toole C, Hirons R, Wright S, Morris J: Falls in Patients with Lower Limb Amputations: Prevalence and Contributing Factors. Physiotherapy 1996, 82(2):130-136.

14. Bloem BR, Grimbergen YA, Cramer M, Willemsen M, Zwinderman AH: Prospective assessment of falls in Parkinson's disease. Journal of neurology 2001, 248(11):950-958.

15. Miller WC, Speechley M, Deathe B: The prevalence and risk factors of falling and fear of falling among lower extremity amputees. Archives of physical medicine and rehabilitation 2001, 82(8):1031-1037.

16. Jorgensen L, Engstad T, Jacobsen BK: Higher incidence of falls in long- term stroke survivors than in population controls: depressive symptoms predict falls after stroke. Stroke 2002, 33(2):542-547.

17. Maki BE, McIlroy WE: Change-in-support balance reactions in older persons: an emerging research area of clinical importance. Neurologic clinics 2005, 23(3):751-783, vi-vii.

18. Maki BE, McIlroy WE: Postural control in the older adult. Clinics in geriatric medicine 1996, 12(4):635-658.

19. Tinetti ME, Williams CS: The effect of falls and fall injuries on functioning in community-dwelling older persons. The journals of gerontology Series A, Biological sciences and medical sciences 1998, 53(2):M112-119.

20. Miller WC, Deathe AB, Speechley M, Koval J: The influence of falling, fear of falling, and balance confidence on prosthetic mobility and social activity among individuals with a lower extremity amputation. Archives of physical medicine and rehabilitation 2001, 82(9):1238-1244.

21. Schmid AA, Van Puymbroeck M, Altenburger PA, Dierks TA, Miller KK, Damush TM, Williams LS: Balance and balance self-efficacy are associated with activity and participation after stroke: a cross-sectional study in people with chronic stroke. Archives of physical medicine and rehabilitation 2012, 93(6):1101-1107.

22. Shimada H, Obuchi S, Furuna T, Suzuki T: New intervention program for preventing falls among frail elderly people: the effects of perturbed walking exercise using a bilateral separated treadmill. American journal of physical medicine & rehabilitation 2004, 83(7):493-499.

100 23. Protas EJ, Mitchell K, Williams A, Qureshy H, Caroline K, Lai EC: Gait and step training to reduce falls in Parkinson's disease. NeuroRehabilitation 2005, 20(3):183-190.

24. Bieryla KA, Madigan ML, Nussbaum MA: Practicing recovery from a simulated trip improves recovery kinematics after an actual trip. Gait & posture 2007, 26(2):208-213.

25. Shapiro A, Melzer I: Balance perturbation system to improve balance compensatory responses during walking in old persons. Journal of neuroengineering and rehabilitation 2010, 7:32.

26. Grabiner MD, Bareither ML, Gatts S, Marone J, Troy KL: Task-specific training reduces trip-related fall risk in women. Medicine and science in sports and exercise 2012, 44(12):2410-2414.

27. Crenshaw JR, Kaufman KR, Grabiner MD: Compensatory-step training of healthy, mobile people with unilateral, transfemoral or knee disarticulation amputations: A potential intervention for trip-related falls. Gait & posture 2013, 38(3):500-506.

28. Lurie JD, Zagaria AB, Pidgeon DM, Forman JL, Spratt KF: Pilot comparative effectiveness study of surface perturbation treadmill training to prevent falls in older adults. BMC geriatrics 2013, 13:49.

29. Rosenblatt NJ, Marone J, Grabiner MD: Preventing trip-related falls by community-dwelling adults: a prospective study. Journal of the American Geriatrics Society 2013, 61(9):1629-1631.

30. Yang F, Bhatt T, Pai YC: Generalization of treadmill-slip training to prevent a fall following a sudden (novel) slip in over-ground walking. Journal of biomechanics 2013, 46(1):63-69.

31. Grabiner MD, Crenshaw JR, Hurt CP, Rosenblatt NJ, Troy KL: Exercise- based fall prevention: can you be a bit more specific? Exercise and sport sciences reviews 2014, 42(4):161-168.

32. Kaufman KR, Wyatt MP, Sessoms PH, Grabiner MD: Task-specific fall prevention training is effective for warfighters with transtibial amputations. Clinical orthopaedics and related research 2014, 472(10):3076-3084.

33. Dijkstra BW, Horak FB, Kamsma YP, Peterson DS: Older adults can improve compensatory stepping with repeated postural perturbations. Frontiers in aging neuroscience 2015, 7:201.

101 34. Mansfield A, Wong JS, Bryce J, Knorr S, Patterson KK: Does perturbation-based balance training prevent falls? Systematic review and meta-analysis of preliminary randomized controlled trials. Physical therapy 2015, 95(5):700-709.

35. Peterson DS, Dijkstra BW, Horak FB: Postural motor learning in people with Parkinson's disease. Journal of neurology 2016, 263(8):1518-1529.

36. Gerards MHG, McCrum C, Mansfield A, Meijer K: Perturbation-based balance training for falls reduction among older adults: Current evidence and implications for clinical practice. Geriatrics & gerontology international 2017, 17(12):2294-2303.

37. Mansfield A, Schinkel-Ivy A, Danells CJ, Aqui A, Aryan R, Biasin L, DePaul VG, Inness EL: Does Perturbation Training Prevent Falls after Discharge from Stroke Rehabilitation? A Prospective Cohort Study with Historical Control. Journal of stroke and cerebrovascular diseases: the official journal of National Stroke Association 2017, 26(10):2174-2180.

38. McCrum C, Gerards MHG, Karamanidis K, Zijlstra W, Meijer K: A systematic review of gait perturbation paradigms for improving reactive stepping responses and falls risk among healthy older adults. European review of aging and physical activity: official journal of the European Group for Research into Elderly and Physical Activity 2017, 14:3.

39. Okubo Y, Schoene D, Lord SR: Step training improves reaction time, gait and balance and reduces falls in older people: a systematic review and meta-analysis. British journal of sports medicine 2017, 51(7):586-593.

40. Papadimitriou A, Perry, M: A systematic review of the effects of perturbation training on preventing falls. New Zealand Journal of Physiotherapy 2017, 45(1):31-49.

41. Pigman J, Reisman DS, Pohlig RT, Wright TR, Crenshaw JR: The development and feasibility of treadmill-induced fall recovery training applied to individuals with chronic stroke. BMC neurology 2019, 19(1):102.

42. Grabiner MD, Koh TJ, Lundin TM, Jahnigen DW: Kinematics of recovery from a stumble. Journal of gerontology 1993, 48(3):M97-102.

43. Pavol MJ, Owings TM, Foley KT, Grabiner MD: Mechanisms leading to a fall from an induced trip in healthy older adults. The journals of gerontology Series A, Biological sciences and medical sciences 2001, 56(7):M428-437.

102 44. Grabiner MD, Donovan S, Bareither ML, Marone JR, Hamstra-Wright K, Gatts S, Troy KL: Trunk kinematics and fall risk of older adults: translating biomechanical results to the clinic. Journal of electromyography and kinesiology: official journal of the International Society of Electrophysiological Kinesiology 2008, 18(2):197-204.

45. Hurt CP, Rosenblatt N, Crenshaw JR, Grabiner MD: Variation in trunk kinematics influences variation in step width during treadmill walking by older and younger adults. Gait & posture 2010, 31(4):461-464.

46. Crenshaw JR, Rosenblatt NJ, Hurt CP, Grabiner MD: The discriminant capabilities of stability measures, trunk kinematics, and step kinematics in classifying successful and failed compensatory stepping responses by young adults. Journal of biomechanics 2012, 45(1):129-133.

47. Crenshaw JR, Kaufman KR, Grabiner MD: Trip recoveries of people with unilateral, transfemoral or knee disarticulation amputations: Initial findings. Gait & posture 2013, 38(3):534-536.

48. Patel PJ, Bhatt T: Fall risk during opposing stance perturbations among healthy adults and chronic stroke survivors. Experimental brain research 2018, 236(2):619-628.

49. Chu VWT, Hornby TG, Schmit BD: Stepping responses to treadmill perturbations vary with severity of motor deficits in human SCI. Journal of neurophysiology 2018, 120(2):497-508.

50. Owings TM, Pavol MJ, Grabiner MD: Mechanisms of failed recovery following postural perturbations on a motorized treadmill mimic those associated with an actual forward trip. Clinical biomechanics (Bristol, Avon) 2001, 16(9):813-819.

51. Sessoms PH, Wyatt M, Grabiner M, Collins JD, Kingsbury T, Thesing N, Kaufman K: Method for evoking a trip-like response using a treadmill- based perturbation during locomotion. Journal of biomechanics 2014, 47(1):277-280.

52. Lee A, Bhatt T, Pai YC: Generalization of treadmill perturbation to overground slip during gait: Effect of different perturbation distances on slip recovery. Journal of biomechanics 2016, 49(2):149-154.

53. Liu X, Bhatt T, Pai YC: Intensity and generalization of treadmill slip training: High or low, progressive increase or decrease? Journal of biomechanics 2016, 49(2):135-140.

103 54. Engel T, Mueller J, Kopinski S, Reschke A, Mueller S, Mayer F: Unexpected walking perturbations: Reliability and validity of a new treadmill protocol to provoke muscular reflex activities at lower extremities and the trunk. Journal of biomechanics 2017, 55:152-155.

55. Beom-Chan L, Martin BJ, Thrasher TA, Layne CS: A new fall-inducing technology platform: Development and assessment of a programmable split-belt treadmill. Conference proceedings: Annual International Conference of the IEEE Engineering in Medicine and Biology Society IEEE Engineering in Medicine and Biology Society Annual Conference, 2017:3777-3780.

56. Norton JJS, Wolpaw JR: Acquisition, maintenance, and therapeutic use of a simple motor skill. Current Opinion in Behavioral Sciences 2018, 20:138-144.

57. Chen Y, Chen XY, Jakeman LB, Chen L, Stokes BT, Wolpaw JR: Operant conditioning of H-reflex can correct a locomotor abnormality after spinal cord injury in rats. The Journal of neuroscience: the official journal of the Society for Neuroscience 2006, 26(48):12537-12543.

58. Thompson AK, Pomerantz FR, Wolpaw JR: Operant conditioning of a spinal reflex can improve locomotion after spinal cord injury in humans. The Journal of neuroscience: the official journal of the Society for Neuroscience 2013, 33(6):2365-2375.

59. Thompson AK, Wolpaw JR: Operant conditioning of spinal reflexes: from basic science to clinical therapy. Frontiers in integrative neuroscience 2014, 8:25.

60. Thompson AK, Wolpaw JR: The simplest motor skill: mechanisms and applications of reflex operant conditioning. Exercise and sport sciences reviews 2014, 42(2):82-90.

61. Thompson AK, Wolpaw JR: Targeted neuroplasticity for rehabilitation. Progress in brain research 2015, 218:157-172.

62. Thompson AK, Wolpaw JR: Restoring walking after spinal cord injury: operant conditioning of spinal reflexes can help. The Neuroscientist: a review journal bringing neurobiology, neurology and psychiatry 2015, 21(2):203-215.

104 63. Eftekhar A, Norton JJS, McDonough CM, Wolpaw JR: Retraining Reflexes: Clinical Translation of Spinal Reflex Operant Conditioning. Neurotherapeutics: the journal of the American Society for Experimental NeuroTherapeutics 2018, 15:669-683.

64. Chen Y, Chen L, Wang Y, Wolpaw JR, Chen XY: Operant conditioning of rat soleus H-reflex oppositely affects another H-reflex and changes locomotor kinematics. The Journal of neuroscience: the official journal of the Society for Neuroscience 2011, 31(31):11370-11375.

65. Capaday C, Stein RB: Amplitude modulation of the soleus H-reflex in the human during walking and standing. The Journal of neuroscience: the official journal of the Society for Neuroscience 1986, 6(5):1308-1313.

66. Capaday C, Stein RB: Difference in the amplitude of the human soleus H reflex during walking and running. The Journal of physiology 1987, 392:513-522.

67. Stein RB, Capaday C: The modulation of human reflexes during functional motor tasks. Trends in neurosciences 1988, 11(7):328-332.

68. Llewellyn M, Prochazka A, Yang JF: Human H-reflexes are reduced in difficult beam-walking compared to stance and normal walking. The Journal of physiology 1989, 418:104P.

69. Llewellyn M, Yang JF, Prochazka A: Human H-reflexes are smaller in difficult beam walking than in normal treadmill walking. Experimental brain research 1990, 83(1):22-28.

70. Edamura M, Yang JF, Stein RB: Factors that determine the magnitude and time course of human H-reflexes in locomotion. The Journal of neuroscience: the official journal of the Society for Neuroscience 1991, 11(2):420-427.

71. Koceja DM, Markus CA, Trimble MH: Postural modulation of the soleus H reflex in young and old subjects. Electroencephalography and clinical neurophysiology 1995, 97(6):387-393.

72. Hoffman MA, Koceja DM: The effects of vision and task complexity on Hoffmann reflex gain. Brain research 1995, 700(1-2):303-307.

73. Angulo-Kinzler RM, Mynark RG, Koceja DM: Soleus H-reflex gain in elderly and young adults: modulation due to body position. The journals of gerontology Series A, Biological sciences and medical sciences 1998, 53(2):M120-125.

105 74. Rietdyk S, Patla AE: Context-dependent reflex control: some insights into the role of balance. Experimental brain research 1998, 119(2):251-259.

75. Zehr EP, Stein RB: What functions do reflexes serve during human locomotion? Progress in neurobiology 1999, 58(2):185-205.

76. Earles DR, Koceja DM, Shively CW: Environmental changes in soleus H- reflex excitability in young and elderly subjects. The International journal of neuroscience 2000, 105(1-4):1-13.

77. Chalmers GR, Knutzen KM: Soleus H-reflex gain in healthy elderly and young adults when lying, standing, and balancing. The journals of gerontology Series A, Biological sciences and medical sciences 2002, 57(8): B321-329.

78. Larsen B, Voight M, Grey MJ: Changes in the soleus stretch reflex at different pedaling frequencies and crank loads during pedaling. Motor control 2006, 10(3):265-279.

79. Tokuno CD, Carpenter MG, Thorstensson A, Garland SJ, Cresswell AG: Control of the triceps surae during the postural sway of quiet standing. Acta physiologica 2007, 191(3):229-236.

80. Tokuno CD, Garland SJ, Carpenter MG, Thorstensson A, Cresswell AG: Sway-dependent modulation of the triceps surae H-reflex during standing. Journal of applied physiology 2008, 104(5):1359-1365.

81. Kawaishi Y, Domen K: The relationship between dynamic balancing ability and posture-related modulation of the soleus H-reflex. Journal of electromyography and kinesiology: official journal of the International Society of Electrophysiological Kinesiology 2016, 26:120-124.

82. Rochcongar P, Dassonville J, Le Bars R: [Modification of the Hoffmann reflex in function of athletic training (author's transl)]. European journal of applied physiology and occupational physiology 1979, 40(3):165-170.

83. Casabona A, Polizzi MC, Perciavalle V: Differences in H-reflex between athletes trained for explosive contractions and non-trained subjects. European journal of applied physiology and occupational physiology 1990, 61(1-2):26-32.

84. Koceja DM, Kamen G: Segmental reflex organization in endurance-trained athletes and untrained subjects. Medicine and science in sports and exercise 1992, 24(2):235-241.

106 85. Nielsen J, Crone C, Hultborn H: H-reflexes are smaller in dancers from The Royal Danish Ballet than in well-trained athletes. European journal of applied physiology and occupational physiology 1993, 66(2):116-121.

86. Trimble MH, Koceja DM: Modulation of the triceps surae H-reflex with training. The International journal of neuroscience 1994, 76(3-4):293-303.

87. Raglin JS, Koceja DM, Stager JM, Harms CA: Mood, neuromuscular function, and performance during training in female swimmers. Medicine and science in sports and exercise 1996, 28(3):372-377.

88. Mynark RG, Koceja DM: Comparison of soleus H-reflex gain from prone to standing in dancers and controls. Electroencephalography and clinical neurophysiology 1997, 105(2):135-140.

89. Patikas D, Bassa H, Kotzamanidis C, Koceja DM, Giatsis S, Ampatzidis G, Giannakos A: High frequency depression of the soleus H-reflex in trained swimmers and untrained subjects. Medical science research 1999, 27:579- 581.

90. Trimble MH, Koceja DM: Effect of a reduced base of support in standing and balance training on the soleus H-reflex. The International journal of neuroscience 2001, 106(1-2):1-20.

91. Maffiuletti NA, Martin A, Babault N, Pensini M, Lucas B, Schieppati M: Electrical and mechanical H(max)-to-M(max) ratio in power- and endurance-trained athletes. Journal of applied physiology 2001, 90(1):3-9.

92. Ross A, Leveritt M, Riek S: Neural influences on sprint running: training adaptations and acute responses. Sports medicine (Auckland, NZ) 2001, 31(6):409-425.

93. Mynark RG, Koceja DM: Down training of the elderly soleus H reflex with the use of a spinally induced balance perturbation. Journal of applied physiology 2002, 93(1):127-133.

94. Schneider C, Capaday C: Progressive adaptation of the soleus H-reflex with daily training at walking backward. Journal of neurophysiology 2003, 89(2):648-656.

95. Mazzocchio R, Kitago T, Liuzzi G, Wolpaw JR, Cohen LG: Plastic changes in the human H-reflex pathway at rest following skillful cycling training. Clinical neurophysiology: official journal of the International Federation of Clinical Neurophysiology 2006, 117(8):1682-1691.

107 96. Gruber M, Taube W, Gollhofer A, Beck S, Amtage F, Schubert M: Training-specific adaptations of H- and stretch reflexes in human soleus muscle. Journal of motor behavior 2007, 39(1):68-78.

97. Taube W, Kullmann N, Leukel C, Kurz O, Amtage F, Gollhofer A: Differential reflex adaptations following sensorimotor and strength training in young elite athletes. International journal of sports medicine 2007, 28(12):999-1005.

98. Ogawa T, Kim GH, Sekiguchi H, Akai M, Suzuki S, Nakazawa K: Enhanced stretch reflex excitability of the soleus muscle in experienced swimmers. European journal of applied physiology 2009, 105(2):199-205.

99. Ryder R, Kitano K, Koceja DM: Spinal reflex adaptation in dancers changes with body orientation and role of pre-synaptic inhibition. Journal of dance medicine & science: official publication of the International Association for Dance Medicine & Science 2010, 14(4):155-162.

100. Lauber B, Keller M, Gollhofer A, Muller E, Taube W: Spinal reflex plasticity in response to alpine skiing in the elderly. Scandinavian journal of medicine & science in sports 2011, 21 Suppl 1:62-68.

101. Keller M, Pfusterschmied J, Buchecker M, Muller E, Taube W: Improved postural control after slackline training is accompanied by reduced H- reflexes. Scandinavian journal of medicine & science in sports 2012, 22(4):471-477.

102. Masu Y, Muramatsu K: Soleus H-reflex modulation during receive stance in badminton players in the receive stance. Journal of physical therapy science 2015, 27(1):123-125.

103. Ceballos-Villegas ME, Saldana Mena JJ, Gutierrez Lozano AL, Sepulveda- Canamar FJ, Huidobro N, Manjarrez E, Lomeli J: The Complexity of H- wave Amplitude Fluctuations and Their Bilateral Cross-Covariance Are Modified According to the Previous Fitness History of Young Subjects under Track Training. Frontiers in human neuroscience 2017, 11:530.

104. Schieppati M: The Hoffmann reflex: a means of assessing spinal reflex excitability and its descending control in man. Progress in neurobiology 1987, 28(4):345-376.

105. Zehr EP: Considerations for use of the Hoffmann reflex in exercise studies. European journal of applied physiology 2002, 86(6):455-468.

108 106. Misiaszek JE: The H-reflex as a tool in neurophysiology: its limitations and uses in understanding nervous system function. Muscle & nerve 2003, 28(2):144-160.

107. Palmieri RM, Ingersoll CD, Hoffman MA: The Hoffmann reflex: methodologic considerations and applications for use in sports medicine and athletic training research. Journal of athletic training 2004, 39(3):268- 277.

108. Stein RB, Thompson AK: Muscle reflexes in motion: how, what, and why? Exercise and sport sciences reviews 2006, 34(4):145-153.

109. Knikou M: The H-reflex as a probe: pathways and pitfalls. Journal of neuroscience methods 2008, 171(1):1-12.

110. Gajewski J, Mazur-Rozycka J: The H-reflex as an important indicator in kinesiology. Human movement 2016, 17(2):64-71.

111. Nashner LM: Adapting reflexes controlling the human posture. Experimental brain research 1976, 26(1):59-72.

112. Nashner LM: Fixed patterns of rapid postural responses among leg muscles during stance. Experimental brain research 1977, 30(1):13-24.

113. Horak FB, Nashner LM: Central programming of postural movements: adaptation to altered support-surface configurations. Journal of neurophysiology 1986, 55(6):1369-1381.

114. Hwang S, Tae K, Sohn R, Kim J, Son J, Kim Y: The balance recovery mechanisms against unexpected forward perturbation. Annals of biomedical engineering 2009, 37(8):1629-1637.

115. McIlroy WE, Maki BE: Changes in early 'automatic' postural responses associated with the prior-planning and execution of a compensatory step. Brain research 1993, 631(2):203-211.

116. Henry SM, Fung J, Horak FB: Control of stance during lateral and anterior/posterior surface translations. IEEE transactions on rehabilitation engineering: a publication of the IEEE Engineering in Medicine and Biology Society 1998, 6(1):32-42.

117. Misiaszek JE, Stephens MJ, Yang JF, Pearson KG: Early corrective reactions of the leg to perturbations at the torso during walking in humans. Experimental brain research 2000, 131(4):511-523.

109 118. Misiaszek JE: Early activation of arm and leg muscles following pulls to the waist during walking. Experimental brain research 2003, 151(3):318- 329.

119. Winter DA, Prince F, Frank JS, Powell C, Zabjek KF: Unified theory regarding A/P and M/L balance in quiet stance. Journal of neurophysiology 1996, 75(6):2334-2343.

120. Gatev P, Thomas S, Kepple T, Hallett M: Feedforward ankle strategy of balance during quiet stance in adults. The Journal of physiology 1999, 514 (Pt 3):915-928.

121. Loram ID, Maganaris CN, Lakie M: Human postural sway results from frequent, ballistic bias impulses by soleus and gastrocnemius. The Journal of physiology 2005, 564(Pt 1):295-311.

122. Boyas S, Hajj M, Bilodeau M: Influence of ankle plantarflexor fatigue on postural sway, lower limb articular angles, and postural strategies during unipedal quiet standing. Gait & posture 2013, 37(4):547-551.

123. McIlroy WE, Maki BE: Task constraints on foot movement and the incidence of compensatory stepping following perturbation of upright stance. Brain research 1993, 616(1-2):30-38.

124. Goldie PA, Bach TM, Evans OM: Force platform measures for evaluating postural control: reliability and validity. Archives of physical medicine and rehabilitation 1989, 70(7):510-517.

125. Lafond D, Corriveau H, Hebert R, Prince F: Intrasession reliability of center of pressure measures of postural steadiness in healthy elderly people. Archives of physical medicine and rehabilitation 2004, 85(6):896- 901.

126. Centers for Disease Control and Prevention, N.C. for I.P. and C. 2018. Web-based Injury Statistics Query and Reporting System (WISQARS) [WWW Document]. https://www.cdc.gov/injury/wisqars/index.html (accessed 11.1.18).

127. Bergen G, Stevens MR, Burns ER: Falls and fall injuries among adults aged >/= 65 years – United States 2014. MMWR Morbidity and Mortality Weekly Report 2016, 65:993-998.

128. Tinetti ME, Williams CS: Falls, injuries due to falls, and the risk of admission to a nursing home. The New England journal of medicine 1997, 337(18):1279-1284.

110 129. Luukinen H, Herala M, Koski K, Honkanen R, Laippala P, Kivela SL: Fracture risk associated with a fall according to type of fall among the elderly. Osteoporosis international: a journal established as result of cooperation between the European Foundation for Osteoporosis and the National Osteoporosis Foundation of the USA 2000, 11(7):631-634.

130. Schwartz AV, Nevitt MC, Brown BW, Jr., Kelsey JL: Increased falling as a risk factor for fracture among older women: the study of osteoporotic fractures. American journal of epidemiology 2005, 161(2):180-185.

131. Berthoz A, Lacour M, Soechting JF, Vidal PP: The role of vision in the control of posture during linear motion. Progress in brain research 1979, 50:197-209.

132. Diener HC, Dichgans J, Guschlbauer B, Mau H: The significance of proprioception on postural stabilization as assessed by ischemia. Brain research 1984, 296(1):103-109.

133. Allum JH, Pfaltz CR: Visual and vestibular contributions to pitch sway stabilization in the ankle muscles of normals and patients with bilateral peripheral vestibular deficits. Experimental brain research 1985, 58(1):82- 94.

134. Horstmann GA, Dietz V: A basic posture control mechanism: the stabilization of the centre of gravity. Electroencephalography and clinical neurophysiology 1990, 76(2):165-176.

135. Horak FB, Henry SM, Shumway-Cook A: Postural perturbations: new insights for treatment of balance disorders. Physical therapy 1997, 77(5):517-533.

136. Sheldon JH: The effect of age on the control of sway. Gerontologia clinica 1963, 5:129-138.

137. Hansen PD, Woollacott MH, Debu B: Postural responses to changing task conditions. Experimental brain research 1988, 73(3):627-636.

138. Panzer VP, Bandinelli S, Hallett M: Biomechanical assessment of quiet standing and changes associated with aging. Archives of physical medicine and rehabilitation 1995, 76(2):151-157.

139. Breniere Y, Bril B: Development of postural control of gravity forces in children during the first 5 years of walking. Experimental brain research 1998, 121(3):255-262.

111 140. Carpenter MG, Frank JS, Silcher CP, Peysar GW: The influence of postural threat on the control of upright stance. Experimental brain research 2001, 138(2):210-218.

141. Mynark RG, Koceja DM: Effects of age on the spinal stretch reflex. Journal of applied biomechanics 2001, 17:188-203.

142. Mackey DC, Robinovitch SN: Mechanisms underlying age-related differences in ability to recover balance with the ankle strategy. Gait & posture 2006, 23(1):59-68.

143. Maki BE, McIlroy WE: Control of rapid limb movements for balance recovery: age-related changes and implications for fall prevention. Age and ageing 2006, 35 Suppl 2:ii12-ii18.

144. Simmons RW, Levy SS, Simmons NK: A Longitudinal Assessment Of Standing Balance In Healthy Adults. Experimental aging research 2017, 43(5):467-479.

145. Maki BE, McIlroy WE: The role of limb movements in maintaining upright stance: the "change-in-support" strategy. Physical therapy 1997, 77(5):488-507.

146. Pollock AS, Durward BR, Rowe PJ, Paul JP: What is balance? Clinical rehabilitation 2000, 14(4):402-406.

147. Lesinski M, Hortobagyi T, Muehlbauer T, Gollhofer A, Granacher U: Dose-response relationships of balance training in healthy young adults: a systematic review and meta-analysis. Sports medicine (Auckland, NZ) 2015, 45(4):557-576.

148. Lesinski M, Hortobagyi T, Muehlbauer T, Gollhofer A, Granacher U: Effects of Balance Training on Balance Performance in Healthy Older Adults: A Systematic Review and Meta-analysis. Sports medicine (Auckland, NZ) 2015, 45(12):1721-1738.

149. Crenshaw JR, Kaufman KR: The intra-rater reliability and agreement of compensatory stepping thresholds of healthy subjects. Gait & posture 2014, 39(2):810-815.

150. Christiansen L, Madsen MJ, Bojsen-Moller E, Thomas R, Nielsen JB, Lundbye-Jensen J: Progressive practice promotes motor learning and repeated transient increases in corticospinal excitability across multiple days. Brain stimulation 2018, 11(2):346-357.

112 151. Pai YC, Rogers MW, Patton J, Cain TD, Hanke TA: Static versus dynamic predictions of protective stepping following waist-pull perturbations in young and older adults. Journal of biomechanics 1998, 31(12):1111-1118.

152. Hall CD, Woollacott MH, Jensen JL: Age-related changes in rate and magnitude of ankle torque development: implications for balance control. The journals of gerontology Series A, Biological sciences and medical sciences 1999, 54(10):M507-513.

153. Jensen JL, Brown LA, Woollacott MH: Compensatory stepping: the biomechanics of a preferred response among older adults. Experimental aging research 2001, 27(4):361-376.

154. Mille ML, Rogers MW, Martinez K, Hedman LD, Johnson ME, Lord SR, Fitzpatrick RC: Thresholds for inducing protective stepping responses to external perturbations of human standing. Journal of neurophysiology 2003, 90(2):666-674.

155. Sturnieks DL, Menant J, Vanrenterghem J, Delbaere K, Fitzpatrick RC, Lord SR: Sensorimotor and neuropsychological correlates of force perturbations that induce stepping in older adults. Gait & posture 2012, 36(3):356-360.

156. Piirainen JM, Linnamo V, Cronin NJ, Avela J: Age-related neuromuscular function and dynamic balance control during slow and fast balance perturbations. Journal of neurophysiology 2013, 110(11):2557-2562.

157. Crenshaw JR, Grabiner MD: The influence of age on the thresholds of compensatory stepping and dynamic stability maintenance. Gait & posture 2014, 40(3):363-368.

158. Crenshaw JR, Petersen DA, Conner BC, Tracy JB, Pigman J, Wright HG, Miller F, Johnson CL, Modlesky CM: Anteroposterior balance reactions in children with spastic cerebral palsy. Developmental medicine and child neurology 2020, 62(6):700-708.

159. Topper AK, Maki BE, Holliday PJ: Are activity-based assessments of balance and gait in the elderly predictive of risk of falling and/or type of fall? Journal of the American Geriatrics Society 1993, 41(5):479-487.

160. Christiansen L, Lundbye-Jensen J, Perez MA, Nielsen JB: How plastic are human spinal cord motor circuitries? Experimental brain research 2017, 235(11):3243-3249.

113 161. Kandel ER, Schwartz JH, Jessell TM, Siegelbaum SA, Hudspeth AJ (eds): Principles of neural science, fifth edition. McGraw-Hill Companies 2013, 709-811.

162. Crenna P, Frigo C: Excitability of the soleus H-reflex arc during walking and stepping in man. Experimental brain research 1987, 66(1):49-60.

163. Shurrager PS, Dykman RA: Walking spinal carnivores. Journal of comparative and physiological psychology 1951, 44(3):252-262.

164. Dobkin BH: An Overview of Treadmill Locomotor Training with Partial Body Weight Support: A Neurophysiologically Sound Approach Whose Time Has Come for Randomized Clinical Trials. . Neurorehabilitation and Neural Repair 1999, 13(3):157-165.

165. Chen XY, Wolpaw JR: Dorsal column but not lateral column transection prevents down-conditioning of H reflex in rats. Journal of neurophysiology 1997, 78(3):1730-1734.

166. Chen XY, Carp JS, Chen L, Wolpaw JR: Corticospinal tract transection prevents operantly conditioned H-reflex increase in rats. Experimental brain research 2002, 144(1):88-94.

167. Chen XY, Chen Y, Chen L, Tennissen AM, Wolpaw JR: Corticospinal tract transection permanently abolishes H-reflex down-conditioning in rats. Journal of neurotrauma 2006, 23(11):1705-1712.

168. Burke D: Clinical uses of H reflexes of upper and lower limb muscles. Clinical neurophysiology practice 2016, 1:9-17.

169. Tanaka Y: Spinal Reflexes During Postural Control Under Psychological Pressure. Motor control 2015, 19(3):242-249.

170. Sibley KM, Carpenter MG, Perry JC, Frank JS: Effects of postural anxiety on the soleus H-reflex. Human movement science 2007, 26(1):103-112.

171. Wolpaw JR, Braitman DJ, Seegal RF: Adaptive plasticity in primate spinal stretch reflex: initial development. Journal of neurophysiology 1983, 50(6):1296-1311.

172. Wolpaw JR, O'Keefe JA: Adaptive plasticity in the primate spinal stretch reflex: evidence for a two-phase process. The Journal of neuroscience: the official journal of the Society for Neuroscience 1984, 4(11):2718-2724.

173. Wolpaw JR: Operant conditioning of primate spinal reflexes: the H-reflex. Journal of neurophysiology 1987, 57(2):443-459.

114 174. Evatt ML, Wolf SL, Segal RL: Modification of human spinal stretch reflexes: preliminary studies. Neuroscience letters 1989, 105(3):350-355.

175. Wolf SL, Segal RL: Reducing human biceps brachii spinal stretch reflex magnitude. Journal of neurophysiology 1996, 75(4):1637-1646.

176. Segal RL: Plasticity in the Central Nervous System: Operant Conditioning of the Spinal Stretch Reflex. Topics in stroke rehabilitation 1997, 3(4):76- 87.

177. Chen XY, Chen L, Wolpaw JR: Time course of H-reflex conditioning in the rat. Neuroscience letters 2001, 302(2-3):85-88.

178. Thompson AK, Chen XY, Wolpaw JR: Acquisition of a simple motor skill: task-dependent adaptation plus long-term change in the human soleus H- reflex. The Journal of neuroscience: the official journal of the Society for Neuroscience 2009, 29(18):5784-5792.

179. Wolpaw JR, Chen XY: Operant conditioning of reflexes. Encyclopedia of neuroscience 2009, 7:225-233.

180. Chen XY, Wolpaw JR: Operant conditioning of H-reflex in freely moving rats. Journal of neurophysiology 1995, 73(1):411-415.

181. Chen XY, Wolpaw JR, Jakeman LB, Stokes BT: Operant conditioning of H-reflex in spinal cord-injured rats. Journal of neurotrauma 1996, 13(12):755-766.

182. Chen XY, Wolpaw JR, Jakeman LB, Stokes BT: Operant conditioning of H-reflex increase in spinal cord--injured rats. Journal of neurotrauma 1999, 16(2):175-186.

183. Chen Y, Chen XY, Jakeman LB, Schalk G, Stokes BT, Wolpaw JR: The interaction of a new motor skill and an old one: H-reflex conditioning and locomotion in rats. The Journal of neuroscience: the official journal of the Society for Neuroscience 2005, 25(29):6898-6906.

184. Carp JS, Tennissen AM, Chen XY, Wolpaw JR: H-reflex operant conditioning in mice. Journal of neurophysiology 2006, 96(4):1718-1727.

185. Epro G, Mierau A, McCrum C, Leyendecker M, Bruggemann GP, Karamanidis K: Retention of gait stability improvements over 1.5 years in older adults: effects of perturbation exposure and triceps surae neuromuscular exercise. Journal of neurophysiology 2018, 119(6):2229- 2240.

115 186. Konig M, Epro G, Seeley J, Catala-Lehnen P, Potthast W, Karamanidis K: Retention of improvement in gait stability over 14 weeks due to trip- perturbation training is dependent on perturbation dose. Journal of biomechanics 2019, 84:243-246.

187. McIlroy WE, Maki BE: Early activation of arm muscles follows external perturbation of upright stance. Neuroscience letters 1995, 184(3):177-180.

188. Tang PF, Woollacott MH, Chong RK: Control of reactive balance adjustments in perturbed human walking: roles of proximal and distal postural muscle activity. Experimental brain research 1998, 119(2):141- 152.

189. Misiaszek JE, Krauss EM: Restricting arm use enhances compensatory reactions of leg muscles during walking. Experimental brain research 2005, 161(4):474-485.

190. Rapp van Roden EA, Petersen DA, Pigman J, Conner BC, Tyler Richardson R, Crenshaw JR: The contribution of counter-rotation movements during fall recovery: A validation study. Journal of biomechanics 2018, 78:102-108.

191. Dietz V, Horstmann GA, Berger W: Interlimb coordination of leg-muscle activation during perturbation of stance in humans. Journal of neurophysiology 1989, 62(3):680-693.

192. Kido A, Tanaka N, Stein RB: Spinal excitation and inhibition decrease as humans age. Canadian journal of physiology and pharmacology 2004, 82(4):238-248.

193. Wolpaw JR, Seegal RF: Diurnal rhythm in the spinal stretch reflex. Brain research 1982, 244(2):365-369.

194. Chen XY, Wolpaw JR: Circadian rhythm in rat H-reflex. Brain research 1994, 648(1):167-170.

195. Lagerquist O, Zehr EP, Baldwin ER, Klakowicz PM, Collins DF: Diurnal changes in the amplitude of the Hoffmann reflex in the human soleus but not in the flexor carpi radialis muscle. Experimental brain research 2006, 170(1):1-6.

196. Carp JS, Tennissen AM, Chen XY, Wolpaw JR: Diurnal H-reflex variation in mice. Experimental brain research 2006, 168(4):517-528.

116 197. Weissgerber TL, Milic NM, Winham SJ, Garovic VD: Beyond bar and line graphs: time for a new data presentation paradigm. PLoS biology 2015, 13(4):e1002128.

198. Eng JJ, Winter DA, Patla AE: Strategies for recovery from a trip in early and late swing during human walking. Experimental brain research 1994, 102(2):339-349.

199. Marigold DS, Bethune AJ, Patla AE: Role of the unperturbed limb and arms in the reactive recovery response to an unexpected slip during locomotion. Journal of neurophysiology 2003, 89(4):1727-1737.

200. Misiaszek JE: Neural control of walking balance: if falling then react else continue. Exercise and sport sciences reviews 2006, 34(3):128-134.

201. Marigold DS, Misiaszek JE: Whole-body responses: neural control and implications for rehabilitation and fall prevention. The Neuroscientist: a review journal bringing neurobiology, neurology and psychiatry 2009, 15(1):36-46.

202. Runge CF, Shupert CL, Horak FB, Zajac FE: Ankle and hip postural strategies defined by joint torques. Gait & posture 1999, 10(2):161-170.

203. Alizadehsaravi L, Bruijn SM, Maas H, van Dieen JH: Modulation of soleus muscle H-reflexes and ankle muscle co-contraction with surface compliance during unipedal balancing in young and older adults. Experimental brain research 2020, doi:10.1007/s00221-020-05784-0.

204. Diener HC, Horak FB, Nashner LM: Influence of stimulus parameters on human postural responses. Journal of neurophysiology 1988, 59(6):1888- 1905.

205. Horak FB, Diener HC, Nashner LM: Influence of central set on human postural responses. Journal of neurophysiology 1989, 62(4):841-853.

206. Oldfield RC: The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 1971, 9(1):97-113.

207. Chapman JP, Chapman LJ, Allen JJ: The measurement of foot preference. Neuropsychologia 1987, 25(3):579-584.

208. Coren S: The lateral preference inventory for measurement of handedness, footedness, eyedness, and earedness: Norms for young adults. Bulletin of the Psychonomic Society 1993, 31(1):1-3.

117 209. Elias LJ, Bryden MP, Bulman-Fleming MB: Footedness is a better predictor than is handedness of emotional lateralization. Neuropsychologia 1998, 36(1):37-43.

210. Grouios G, Hatzitaki V, Kollias N, Koidou I: Investigating the stabilising and mobilising features of footedness. Laterality 2009, 14(4):362-380.

211. van Melick N, Meddeler BM, Hoogeboom TJ, Nijhuis-van der Sanden MWG, van Cingel REH: How to determine leg dominance: The agreement between self-reported and observed performance in healthy adults. PloS one 2017, 12(12):e0189876.

212. Mutha PK, Haaland KY, Sainburg RL: Rethinking motor lateralization: specialized but complementary mechanisms for motor control of each arm. PloS one 2013, 8(3):e58582.

213. Castañer M, Andueza J, Hileno R, Puigarnau S, Prat Q, Camerino O: Profiles of Motor Laterality in Young Athletes' Performance of Complex Movements: Merging the MOTORLAT and PATHoops Tools. Frontiers in psychology 2018, 9:916.

214. Kapteyn TS, Bles W, Njiokiktjien CJ, Kodde L, Massen CH, Mol JM: Standardization in platform stabilometry being a part of posturography. Agressologie: revue internationale de physio-biologie et de pharmacologie appliquees aux effets de l'agression 1983, 24(7):321-326.

215. Prieto TE, Myklebust JB, Hoffmann RG, Lovett EG, Myklebust BM: Measures of postural steadiness: differences between healthy young and elderly adults. IEEE transactions on bio-medical engineering 1996, 43(9):956-966.

216. Palmieri RM, Ingersoll CD, Stone MB, Krause BA: Center-of-pressure parameters used in the assessment of postural control. Journal of sport rehabilitation 2002, 11:51-66.

217. Raymakers JA, Samson MM, Verhaar HJ: The assessment of body sway and the choice of the stability parameter(s). Gait & posture 2005, 21(1):48- 58.

218. Santos BR, Delisle A, Lariviere C, Plamondon A, Imbeau D: Reliability of centre of pressure summary measures of postural steadiness in healthy young adults. Gait & posture 2008, 27(3):408-415.

118 219. Ruhe A, Fejer R, Walker B: The test-retest reliability of centre of pressure measures in bipedal static task conditions--a systematic review of the literature. Gait & posture 2010, 32(4):436-445.

220. Paillard T, Noe F: Techniques and Methods for Testing the Postural Function in Healthy and Pathological Subjects. BioMed research international 2015, 2015:891390.

221. Hasselkus BR, Shambes GM: Aging and postural sway in women. Journal of gerontology 1975, 30(6):661-667.

222. Hufschmidt A, Dichgans J, Mauritz KH, Hufschmidt M: Some methods and parameters of body sway quantification and their neurological applications. Archiv fur Psychiatrie und Nervenkrankheiten 1980, 228(2):135-150.

223. Maki BE, Holliday PJ, Fernie GR: Aging and postural control. A comparison of spontaneous- and induced-sway balance tests. Journal of the American Geriatrics Society 1990, 38(1):1-9.

224. Ingersoll CD, Armstrong CW: The effects of closed-head injury on postural sway. Medicine and science in sports and exercise 1992, 24(7):739-743.

225. Geurts AC, Ribbers GM, Knoop JA, van Limbeek J: Identification of static and dynamic postural instability following traumatic brain injury. Archives of physical medicine and rehabilitation 1996, 77(7):639-644.

226. Lafond D, Corriveau H, Prince F: Postural control mechanisms during quiet standing in patients with diabetic sensory neuropathy. Diabetes care 2004, 27(1):173-178.

227. Ruhe A, Fejer R, Walker B: Center of pressure excursion as a measure of balance performance in patients with non-specific low back pain compared to healthy controls: a systematic review of the literature. European spine journal: official publication of the European Spine Society, the European Spinal Deformity Society, and the European Section of the Cervical Spine Research Society 2011, 20(3):358-368.

228. Degani AM, Santos MM, Leonard CT, Rau TF, Patel SA, Mohapatra S, Danna-dos-Santos A: The effects of mild traumatic brain injury on postural control. Brain injury 2017, 31(1):49-56.

229. Maki BE, Holliday PJ, Topper AK: Fear of falling and postural performance in the elderly. Journal of gerontology 1991, 46(4):M123-131.

119 230. Owings TM, Pavol MJ, Foley KT, Grabiner MD: Measures of postural stability are not predictors of recovery from large postural disturbances in healthy older adults. Journal of the American Geriatrics Society 2000, 48(1):42-50.

231. Mackey DC, Robinovitch SN: Postural steadiness during quiet stance does not associate with ability to recover balance in older women. Clinical biomechanics (Bristol, Avon) 2005, 20(8):776-783.

232. Carpenter MG, Frank JS, Winter DA, Peysar GW: Sampling duration effects on centre of pressure summary measures. Gait & posture 2001, 13(1):35-40.

233. van der Kooij H, Campbell AD, Carpenter MG: Sampling duration effects on centre of pressure descriptive measures. Gait & posture 2011, 34(1):19- 24.

234. Chiari L, Rocchi L, Cappello A: Stabilometric parameters are affected by anthropometry and foot placement. Clinical biomechanics (Bristol, Avon) 2002, 17(9-10):666-677.

235. Schmit JM, Regis DI, Riley MA: Dynamic patterns of postural sway in ballet dancers and track athletes. Experimental brain research 2005, 163(3):370-378.

236. Tahayori B, Port NL, Koceja DM: The inflow of sensory information for the control of standing is graded and bidirectional. Experimental brain research 2012, 218(1):111-118.

237. Simmons RW: Sensory organization determinants of postural stability in trained ballet dancers. The International journal of neuroscience 2005, 115(1):87-97.

238. Teasdale N, Stelmach GE, Breunig A: Postural sway characteristics of the elderly under normal and altered visual and support surface conditions. Journal of gerontology 1991, 46(6):B238-244.

239. Laughton CA, Slavin M, Katdare K, Nolan L, Bean JF, Kerrigan DC, Phillips E, Lipsitz LA, Collins JJ: Aging, muscle activity, and balance control: physiologic changes associated with balance impairment. Gait & posture 2003, 18(2):101-108.

120 240. Benjuya N, Melzer I, Kaplanski J: Aging-induced shifts from a reliance on sensory input to muscle cocontraction during balanced standing. The journals of gerontology Series A, Biological sciences and medical sciences 2004, 59(2):166-171.

241. Vette AH, Sayenko DG, Jones M, Abe MO, Nakazawa K, Masani K: Ankle muscle co-contractions during quiet standing are associated with decreased postural steadiness in the elderly. Gait & posture 2017, 55:31-36.

242. Nielsen J, Kagamihara Y: The regulation of presynaptic inhibition during co-contraction of antagonistic muscles in man. The Journal of physiology 1993, 464:575-593.

243. Perez MA, Lundbye-Jensen J, Nielsen JB: Task-specific depression of the soleus H-reflex after cocontraction training of antagonistic ankle muscles. Journal of neurophysiology 2007, 98(6):3677-3687.

244. Warnica MJ, Weaver TB, Prentice SD, Laing AC: The influence of ankle muscle activation on postural sway during quiet stance. Gait & posture 2014, 39(4):1115-1121.

245. Giavarina D: Understanding Bland Altman analysis. Biochemia medica 2015, 25(2):141-151.

121 Appendix A

VALIDATION OF TREADMILL PERTURBATION SIZE

The data presented in this appendix are reported for the purpose of validating that the sizes of our treadmill perturbations were consistent with that expected based on software commands. For this validation, reflective markers were attached directly to the treadmill belts; Qualisys® (Göteborg, Sweden) motion capture technology was then used to record marker trajectories at a rate of 125Hz, during each perturbation trial. Custom-written MATLAB® (The MathWorks Inc., Natick MA, USA) codes were used to determine the observed peak velocity and displacement of each perturbation based on marker data. For each trial, perturbation onset and termination were identified based on a criterion value (set just above the level of signal noise) for change in marker position in the anterior-posterior direction; these times points were then verified through visual inspection of the marker trajectory. Perturbation displacement was computed by subtracting marker position at onset from that at termination (Figure 22B). Perturbation velocity was computed by taking the first derivative of position with respect to time. Peak velocity (Vpeak) was computed by averaging perturbation velocity across the middle twelve data points of each trial (i.e., across a 96ms time window) (Figure 22A). Note: For all perturbation trials, the duration of time at peak velocity was between 200-300ms. For each trial, observed perturbation displacement and peak velocity values were averaged across data collected from the right and left treadmill belts.

122

Figure 22: Expected and Observed Size of a Level 10 Perturbation. Expected values were the perturbation magnitudes expected based on software commands; observed values were computed from motion capture data. Total time duration for all trials was expected to be 600-700ms; the average observed duration was 642ms. (A.) For all perturbations, expected acceleration and deceleration periods were 200ms each; time at peak velocity was expected to range from 200-300ms. Observed peak velocity (Vpeak) was computed by averaging the observed velocity across the middle 96ms of the trial. For perturbation Level 10 the expected peak velocity was 50.00cm/s; the average observed peak velocity (Vpeak) was 50.68cm/s. (B.) For each perturbation level there was a range of expected displacements that corresponded with the minimum to maximum time a perturbation was expected to be at peak velocity. For Level 10, expected perturbation displacement ranged from 20-25cm; the average observed displacement was 24.6cm.

Observed displacement and peak velocity were compared with that expected based on software commands. For each perturbation level the range of expected displacements was computed based on the minimum and maximum amount of time a perturbation was expected to be at peak velocity (i.e., 200-300ms). Data shown in the figures and tables below are that of Stepping Threshold trials collected at baseline and final assessments, and at the beginning of each training session, for all participants.

123

Figure 23: Size of Treadmill Perturbations. Each red line represents the perturbation size (i.e., displacement and peak velocity) expected for a particular perturbation level (L) based on software commands. Duration at peak velocity ranged from 200-300ms; thus, for each perturbation level there was a range of expected displacements. The size of the displacement range is shown by the length of the red bar; higher peak velocities resulted in larger displacement ranges. Each blue dot represents the observed size of a single perturbation based on movement of the treadmill belts as recorded through motion capture technology. Data include all trials used for analyses in this study (i.e., all participant Stepping Threshold trails).

124 As shown in Figure 23, there was a slight upward shift in the magnitude of the perturbations as observed through motion capture. This shift occurred for both peak velocity and displacement. However, the difference between these observed and expected values was similar across all perturbations and levels.

Figure 24: Bland-Altman Plot, Perturbation Peak Velocity. Each blue dot represents a single perturbation. Data include all trials used for analyses in this study (i.e., all participant Stepping Threshold trails). Data columns, from left to right, correspond to perturbation levels 7-20, respectively. Expected values were the velocities commanded by the computer software; observed velocities were computed based on motion capture data. The mean difference between observed and expected values was 0.97cm/s; the standard deviation (SD) of this difference was 0.57cm/s. Limits of agreement were set at 1.96SD; 93% of all trials were within these limits.

Bland-Altman plots [245], completed separately for perturbation peak velocity and perturbation displacement, are shown in Figures 24 and 25. For each plot, the difference between the observed and the expected magnitude was plotted against that

125 expected, for each trial. For the peak velocity plot, the value used as the expected velocity was the velocity commanded by the software; for the displacement plot the expected value used was the median value of the expected displacement range. For each plot, mean and standard deviation (SD) of the difference between observed and expected values were computed based on all the trials; limits of agreement were set a priori at 1.96SD. Data separated by perturbation level are reported in Tables 6 and 7, for peak velocity and displacement respectively.

Figure 25: Bland-Altman Plot, Perturbation Displacement. Each blue dot represents a single perturbation. Data include all trials used for analyses in this study (i.e., all participant Stepping Threshold trails). Data columns, from left to right, correspond to perturbation levels 7-20, respectively. Expected displacements were the median values computed for each displacement range based on programmed parameters; observed displacements were computed based on motion capture data. The mean difference between observed and expected values was 2.36cm; the standard deviation (SD) of this difference was 1.86cm. Limits of agreement were set at 1.96SD; 97% of all trials were within these limits.

126 As shown in Figure 24, the average difference between the observed peak velocity and the expected peak velocity, across all trials, was ~1cm/s; 93% of all trials were within the limits of agreement. As shown in Figure 25, the average difference between the observed displacement and the median value of the expected displacement range, across all trials, was ~2.4cm. The percentage of trials within the limits of agreement for displacement was 97%.

Table 6: Perturbation Peak Velocity: Comparison of expected versus observed magnitude, by level. Data shown in this table are that of the Stepping Threshold trials used for analyses in this study. Expected values were the velocities commanded by the computer software. For observed values, peak velocity, computed for each trial based on motion capture data, was averaged across all trials within the specified level. Velocity difference, was the observed velocity minus the expected velocity, computed separately for each trial and then averaged across all trials within the level. SD is the standard deviation for both the observed velocity and the velocity difference.

Number Expected Observed Velocity Perturbation of Velocity Velocity Difference Level Trials (cm/s) Mean (cm/s) Mean (cm/s) SD 7 5 35.00 35.47 0.47 0.27 8 15 40.00 40.74 0.74 0.31 9 27 45.00 45.62 0.62 0.38 10 62 50.00 50.68 0.68 0.51 11 48 55.00 55.89 0.89 0.55 12 54 60.00 61.07 1.07 0.67 13 36 65.00 66.07 1.07 0.50 14 36 70.00 71.07 1.07 0.52 15 21 75.00 76.27 1.27 0.64 16 18 80.00 81.24 1.24 0.31 17 3 85.00 86.64 1.64 0.32 18 9 90.00 91.50 1.50 0.33 19 12 95.00 96.30 1.30 0.46 20 3 100.00 102.01 2.01 0.20

127 Table 7: Perturbation Displacement: Comparison of expected versus observed magnitude, by level. Data shown in this table are that of the Stepping Threshold trials used for analyses in this study. The expected displacement range was computed based on software commands and time duration at peak velocity: the lowest value for each range was the displacement expected for a trial at peak velocity for 200ms; the highest value was that expected for a trial at peak velocity for 300ms. For observed values, perturbation displacement, computed for each trial based on motion capture data, was averaged across all trials within the specified level. To determine the difference between observed and expected displacement the median value of the range of expected displacements was used as the expected value. Displacement difference (i.e., observed minus expected displacement) was computed for each trial and then averaged across all trials within the perturbation level. SD is the standard deviation for both the observed displacement and the displacement difference.

Number Expected Observed Expected Observed Displacement Perturbation of Displacement Displacement Displacement Displacement Difference Level Trials Range (cm) Range (cm) Median (cm) Mean (cm) Mean (cm) SD 7 5 14.00 - 17.50 17.17 - 18.54 15.75 18.00 2.30 0.50 8 15 16.00 - 20.00 18.21 - 21.68 18.00 20.00 2.00 1.10 9 27 18.00 - 22.50 20.39 - 24.58 20.25 22.40 2.10 1.20 10 62 20.00 - 25.00 22.47 - 27.70 22.50 24.60 2.10 1.30 11 48 22.00 - 27.50 24.17 - 30.31 24.75 26.70 2.00 1.50 12 54 24.00 - 30.00 25.45 - 32.35 27.00 29.10 2.10 1.80 13 36 26.00 - 32.50 27.56 - 35.65 29.25 31.70 2.50 1.90 14 36 28.00 - 35.00 30.54 - 39.20 31.50 34.50 3.00 2.20 15 21 30.00 - 37.50 32.87 - 40.80 33.75 37.70 4.00 2.30 16 18 32.00 - 40.00 34.57 - 41.20 36.00 38.00 2.00 2.10 17 3 34.00 - 42.50 37.93 - 42.71 38.25 39.80 1.60 2.50 18 9 36.00 - 45.00 39.34 - 46.40 40.50 42.70 2.20 3.10 19 12 38.00 - 47.50 42.14 - 50.50 42.75 46.10 3.30 2.90 20 3 40.00 - 50.00 43.69 - 46.75 45.00 45.40 0.40 1.60

For both peak velocity and displacement, differences between observed and expected perturbation magnitudes were similar across perturbation levels. Consistent with expectations, the observed peak velocity increased ~5cm/s with each increase in

128 perturbation level (Table 6). Additionally, the mean difference between the observed displacement and the median value of the expected displacement range was stable across levels (Table 7). Further, the SD of the observed displacement increased with perturbation level; this stepwise increase was consistent with expectations based on the corresponding increase in the size of the displacement range. Finally, the total time expected for each perturbation ranged from 600-700ms; the average observed perturbation time, across all trials, was 642ms (data not shown). In sum, we conclude that the software and hardware used in this study delivered reliable perturbations. There was a slight shift in perturbation magnitude between the expected and observed data, for both peak velocity and displacement.

However, this shift was systematic (i.e., consistent across trials and perturbation levels) and, as such, it did not affect the overall results of our study.

129 Appendix B

IRB APPROVAL

130

131