DOES THE EFFECT OF WALKING BALANCE PERTURBATIONS GENERALIZE ACROSS CONTEXTS?

Abstract

Balance perturbations are used to study locomotor instability. However, these perturbations are designed to provoke a specific context of instability that may or may not generalize to a broader understanding of falls risk. The purpose of this study was to determine if the effect of balance perturbations on instability generalizes across contexts. 29 healthy younger adults and 28 older adults completed four experimental trials, including unperturbed walking and walking while responding to three perturbation contexts: mediolateral optical flow, treadmill-induced slips, and lateral waist-pulls. We quantified the effect of perturbations as an absolute change in margin of stability from unperturbed walking. We found significant changes in mediolateral and anteroposterior margin of stability for all perturbations compared to unperturbed walking in both cohorts (p<0.042). In older adults, the mediolateral effects of lateral waist-pulls significantly correlated with those of optical flow perturbations and treadmill-induced slips (r ≥0.398, p-values≤0.036). In younger adults but not in older adults, we found positive and significant correlations between the anteroposterior effect of waist-pull perturbations and optical flow perturbations, and the anteroposterior and mediolateral effect of treadmill-induced slips (r≥0.428, p-values≤0.021). We found no “goldilocks” perturbation paradigm to endorse that would support universal interpretations about locomotor instability. Building the most accurate patient profiles of instability likely requires a series of perturbation paradigms designed to emulate the variety of environmental contexts in which falls may occur.

1.). Introduction

Falls among older adults are detrimental to their health, independence, and quality of life, and most of these falls occur during locomotor activities such as walking. This public health challenge has for many decades motivated physicians, scientists, and engineers to investigate ways to inform improved diagnostics, treatment, and preventative measures to mitigate falls. Unfortunately, the factors leading to an increased risk of falls are incredibly complex, including declines in sensory and motor acuity (Melzer et al., 2004; Menz et al., 2005; Springer et al., 2006; Thelen et al., 1996). Even preclinical changes in these factors can influence balance and precipitate a first fall (Franz et al., 2015). Moreover, our group has shown that age effects on walking balance can go undetected using usual, steady-state walking alone (Franz et al., 2015). Indeed, observational measures from steady-state walking poorly predict vulnerability to balance challenges in older adults. Ultimately, this motivates the use of balance perturbations to elicit walking-related instability. However, balance perturbations, by their very nature, are designed to provoke a specific context of instability that may or may not generalize to a broader understanding of falls risk. For this work, we define generalization as a fundamental similarity in the response to different types of balance perturbation contexts.

In response to different contexts of balance perturbations (e.g., mechanical versus sensory challenges), walking balance exhibits direction-dependent responses to mitigate instability. Depending on the nature of the balance challenge, lateral instability is accommodated using disproportionately more active control whereas anteroposterior stability benefits at least in part from passive dynamics (Bauby & Kuo, 2000). Moreover, balance responses are orchestrated via complex interactions between sensory (i.e., vestibular, visual, and peripheral sensation) and motor (i.e., muscle strength/power, reaction time) factors while navigating everyday activities (Lord & Sturnieks, 2005). Accordingly, the instability elicited by lateral versus anteroposterior challenges or sensory versus mechanical threats to balance may differ fundamentally in requisite muscle demand, coordination, neural circuitry, etc. Indeed, we see evidence of this in the demographics of falls. In a review of self-reported causes of falls among older adults, a National Health Interview Survey reported 59% from trips, slips, or stumbling, 22% from a loss of balance, dizziness, fainting, or seizure, and 19% from other causes (Kochera, 2002). Motivated by neuromechanics of walking balance and the complex demographics of falls, individual perturbation paradigms have been designed and implemented in the research community to emulate various environmental challenges. Common examples include optical flow perturbations implemented in virtual reality, treadmill-induced slip or surface perturbations, and impulsive force perturbations. Optical flow perturbations emulate intrinsic problems in sensorimotor integration used to navigate visual feedback of our environment. Treadmill-induced slip perturbations emulate instability applied to the base of support such as slipping on ice. Lastly, impulsive force perturbations emulate instability applied to the center of mass such as being bumped into while walking.

Unfortunately, the ecological relevance of balance perturbation infrastructure comes at a cost. As a research community, we are continually progressing toward open-source and more cost-effective solutions to emulate balance challenges faced in our community (e.g., see (Guan Rong Tan, 2020)). However, financial, space, and time constraints can limit our selection of perturbation paradigms in the design of translational research programs and thus the clinical impact of our work. Indeed, it is generally unrealistic and logistically prohibitive to fully capture the instability elicited by various environmental contexts that could precipitate a fall. But, what if the effect of walking balance perturbations generalized across various contexts? Research and/or clinical environments could then leverage a single perturbation paradigm to make widespread inferences about instability and balance impairment. However, researchers and clinicians currently lack the evidence base to inform whether such a generalization exists.

Past literature has shown positive translational improvements from training on balance responses such as slips from a moveable platform to that on an oily surface or treadmill slip perturbations to novel overground slips. These positive improvements have also been shown between slip and trip perturbations (forces acting upon the base of support in the frontal place in different directions) or carry-over between limbs of the same perturbations. Yet, other literature has shown a lack of transfer in performance from training with treadmill perturbations acting on the base of support to a lean-and-release task that focuses on the center of mass control. Our line of research moves away from carry-over from training protocols and focuses on the diagnostics of determining the instability of a participant in the lab. For this research, we also approach the gap of using vastly different forms of perturbation contexts through the use of a frontal plane base of support perturbations, a sagittal plane center of mass perturbation, and a sagittal plane sensory perturbation.

Thus, the purpose of this preliminary study was to leverage a suite of walking balance perturbation paradigms in a cohort of younger and older adult participants as a first step to address this knowledge gap. As younger adults are often used in biomechanics research to make widespread inferences and act as controls against an aging population, we decided to include a younger adult cohort to compare how the effect of perturbations may generalize differently between them and older adults. Given the fundamental mechanical and sensory differences between common perturbation contexts – namely, optical flow perturbations, mechanical waist-pull perturbations, and treadmill-induced slip perturbations – we hypothesized that independent of age, balance perturbation effects will not generalize across different contexts. We evaluated this hypothesis by testing the prediction that perturbation effects (measured herein using absolute change in margin of stability from unperturbed walking [ΔMoS]) would not significantly correlate between the three walking perturbation contexts.

2.). Methods

2.1.). Participants:

Twenty-nine healthy younger adults (14 female, age: 22.4±3.0 yrs, height: 1.73±0.08 m, mass: 67.2±8.6 kg) and twenty-eight older adults (15 female, age: 73.0±5.94 yrs, height: 1.70±0.08 m, mass: 71.6±19.6 kg) participated. We included participants 18–35 years (younger adults) and over 65 years (older adults) with no leg injuries or prostheses, without neurological, musculoskeletal, or cardiopulmonary disease, and who could walk without the use of an assistive device. Methods and recruitment procedures for this study were approved by the University of North Carolina at Chapel Hill Institutional Review Board (20–0555). Participants provided written informed consent prior to participating.

2.2.). Equipment:

A 10-camera motion capture system (Motion Analysis Corporation, Santa Rosa, California, USA) recorded the positions of retroreflective markers placed on the anterior and posterior iliac spines, sacrum, lateral femoral condyles, lateral malleoli, posterior calcanei, first and fifth metatarsal heads, acromials, 7th cervical spine, 10th thoracic spine, sternum, and the sternal notch plus an additional 14 tracking markers placed in clusters on the lateral thighs and shanks at 100 Hz. This project was completed during a single session as part of a larger study on the neuromechanics of walking balance control. Participants walked on a dual-belt, instrumented treadmill (Bertec, Columbus, Ohio, USA) while wearing 8 bipolar recording electrodes on each leg (Delsys, Natick, Massachusetts, USA); however, force and electromyography data were not used in the current analysis or results reporting.

2.3.). Protocol:

The preferred walking speed for each participant (YA: 1.34±0.12 m/s, OA: 1.19±0.19 m/s) was obtained from the average of four 30-m overground walking trials timed using photocells (Bower Timing Systems, Draper, Utah, USA). Participants completed a three-minute warm-up walk before data collection to acclimate to treadmill walking. For data collection, participants completed four walking trials at their preferred walking speed in a randomized order while wearing an overhead harness. In one trial, participants walked for two minutes without exposure to balance perturbations. The participants then also experienced three novel and diverse perturbation paradigms designed to emulate balance challenges faced in their communities. One paradigm was continuous mediolateral optical flow perturbations (Fig. 1A). Here, participants walked on the treadmill while watching a speed-matched virtual hallway rear-projected on an immersive, semi-circular curved screen (2.75 m high, 2.25 m wide) and controlled using Simulink Desktop Real-Time^™ (MathWorks, Natick, Massachusetts, USA). The only instruction given was to walk naturally while looking down the hallway. The mediolateral optical flow perturbation consisted of the sum of three sinusoids with the full amplitude (0.35 m) applied at 0.250 Hz and half that amplitude applied at 0.125 Hz and 0.442 Hz. The optical flow perturbation parameters used in this research study are identical to those used in previous studies from our group (Allen & Franz, 2018; Franz et al., 2015; Qiao et al., 2018; Richards et al., 2019; Selgrade et al., 2020). In another perturbation paradigm, participants were exposed to treadmill-induced slips (Fig. 1B). Participants experienced sudden 6 m/s² of treadmill belt decelerations of 200 ms in duration applied to 5 randomly-selected heel strikes per leg, each separated by at least 10 steps using a custom MATLAB script (MathWorks, Natick, Massachusetts, USA) (Crenshaw & Grabiner, 2014; Lee et al., 2016; Liu et al., 2016). Following the 200 ms perturbation, the treadmill belt returned to the preferred walking speed at 6 m/s². Finally, participants responded, in two different block trials, to right or left side impulsive lateral waist-pull perturbations (Fig. 1C). Specifically, participants were exposed 5 times per side to an unexpected 5% bodyweight, 100 ms force applied at the instant of toe-off via a servomotor-driven cable attached to their waist controlled via custom software (National Instruments, Austin, TX, USA). These perturbations were also separated by at least 10 steps and the hardware details are described in detail by Conway et al. (Conway et al., 2018; Conway & Franz, 2020).

Figure 1:

Our study protocol included four randomized conditions, including walking normally without perturbations and walking while responding to: (A) mediolateral optical flow perturbations, (B) treadmill-induced slip perturbations, and (C) lateral waist-pull perturbations. Perturbation signal amplitudes accompany each paradigm and include the oscillation magnitude for the mediolateral optical flow perturbations, the velocity signal for the treadmill-induced slip perturbations, and the force signal for the lateral waist-pull perturbations.

2.4.). Measurements & Analysis:

Our primary outcome measure in this study was margin of stability (MoS). There are a variety of walking balance outcomes available including local dynamic stability (McAndrew et al., 2011; Qiao et al., 2018), gait variability (Brach et al., 2005; Collins & Kuo, 2013; Dean et al., 2007), and whole-body angular momentum (Nolasco et al., 2019; Silverman et al., 2012). However, the scope of our study required an outcome that is both meaningful but also rigorously applicable to both continuous and discrete perturbations. Some outcome measures (i.e., local dynamic instability) are unable to characterize instability elicited by discrete perturbations (McAndrew et al., 2011). Others (i.e., step width variability) are sensitive to differences in the number of samples analyzed between conditions. Conversely, MoS overcame many of these methodological limitations. We calculated MoS using previously described methods (Hof et al., 2005; McAndrew Young et al., 2012; Richards et al., 2019). First, we filtered marker position data using a 4^th order low-pass Butterworth filter with a cutoff frequency of 12 Hz. We calculated the center of the polygon created by the anterior and posterior iliac spine marker positions as an approximation for the center of mass (x)(de Jong et al., 2020; Hak et al., 2013; Peebles et al., 2017; Rosenblatt & Grabiner, 2010). We next calculated extrapolated center of mass (xCoM) using:

$$\mathit{\text{xCoM}}=x+\frac{\dot{x}+V_{\mathit{\text{treadmill}}}}{{\omega }_0}$$

where x is the CoM position, $\dot{x}$ is the CoM velocity, V_treaccdmill is the treadmill belt velocity, and ω₀ is the natural frequency of an inverted pendulum model of the stance phase:

$${{\omega }_0}^2=\frac{g}{L}$$

where g is acceleration due to gravity and L is participant leg length. We estimated leg length as the mean distance between the sacral marker and the heel marker at heel-strike. We calculated MoS as the distance between the boundary of the base of support (BoS), defined as the 5^th metatarsal marker for mediolateral MoS or as the 1^st metatarsal marker for anteroposterior MoS, and the xCoM projected to the treadmill belt.

We then extracted the MoS at heel strike of the left and right leg. Mediolateral (MoS_ML) and anteroposterior (MoS_AP) MoS were averaged across all strides for the optical flow perturbation trial and the normal walking trials. Following published guidelines for discrete perturbations, MoS outcomes were calculated at the instant of heel strike directly following perturbation onset (i.e., the recovery step) and then averaged across all perturbation occurrences within the trial (Golyski et al., 2022; Martelli et al., 2016). Calculated ML & AP MoS at the participant’s preferred walking speed, unperturbed, is consistent with that of previously published literature (Kao et al., 2014).

2.5.). Statistical Analysis:

A repeated measures analysis of variance (ANOVA) tested for significant effects of condition on MoS_ML and MoS_AP with an alpha level of 0.05. When a significant effect was found, post-hoc paired t-tests with a Holm adjustment identified pairwise differences. Perturbation effects on stability were specifically quantified as the actual change in MoS (ΔMoS) from unperturbed walking to response to perturbations and reported in cm. To test our primary hypothesis, we first calculated bivariate Pearson correlations between the ΔMoS of each perturbation separately in the mediolateral and anteroposterior directions (i.e., AP effect for slip perturbations vs AP effect for optical flow perturbations). We next calculated correlations for the directional effects within each perturbation (i.e., AP vs ML effect for slip perturbations). Finally, we repeated that analysis using ΔMoS in the direction acted upon by each perturbation (i.e., AP for slip perturbations and ML for optical flow and waist perturbations). In total, we assessed 11 correlations per age group. There were no multiple comparison corrections applied to the correlations. We assessed age effects on ΔMoS across perturbation contexts using t-tests with a Holm adjustment.

3.). Results

In both younger and older adults, we found significant main effects of condition on MoS_AP (YA: p<0.001, η_p² = 0.588 OA: p<0.001, η_p² = 0.646) and MoS_ML (YA: p<0.001, η_p² = 0.887 OA: p<0.001, η_p² = 0.587). Across perturbation contexts, MoS_AP decreased by an average of 1.2–9.4 cm in younger adults and 0.9–14.1 cm in older adults and MoS_ML increased by an average of 0.5–4.1 cm in younger adults and 0.8–2.8 cm in older adults compared to unperturbed walking (YA AP: p≤0.042, d≥0.40; OA AP: p≤0.023, d≥0.42; YA ML: p≤0.008, d≥0.59; OA ML: p<0.001, d≥0.59) (Fig. 2).

Figure 2:

Boxplots for margin of stability (MoS) across experimental conditions in the mediolateral (ML) and anteroposterior (AP) directions. In both directions, MoS changed significantly (increase in ML and decrease in AP directions) from unperturbed walking for each of the balance perturbations in both age groups. In YA, MoS_ML was significantly larger for waist-pull perturbations than for either optical flow or treadmill slip perturbations. For OA, MoS_ML was significantly larger for treadmill slip perturbations than for either optical flow or waist-pull perturbations. For MoS_AP, there was a significant decrease in MoS between the treadmill slip perturbations and the waist-pull and optical flow perturbations in both age groups. A positive MoS_AP denotes an extrapolated CoM that is posterior to the BoS and a positive MoS_ML denotes an extrapolated CoM that is medial to the BoS. Asterisks (*) indicate significant pairwise differences (p<0.05) between conditions.

3.1.). Perturbation-Specific Effects:

Pairwise comparisons in younger and older adults revealed significant pairwise differences between all perturbation contexts for both MoS_AP and MoS_ML. In younger and older adults, MoS_AP for treadmill slips (YA: −7.76 ± 7.91 cm, OA: −10.1 ± 10.4 cm) were significantly smaller than for optical flow perturbations (YA: −0.42 ± 1.10 cm, OA: 3.15 ± 6.66 cm) (YA: p<0.001, d=1.16, OA: p<0.001, d=1.29) or waist-pulls (YA: −4.80 ± 4.48 cm, OA: −3.41 ± 4.81 cm) (YA: p=0.042, d=0.397, OA: p<0.001, d=1.29)(Fig. 2A&C). MoS_AP for waist-pulls (YA: −4.80 ± 4.48 cm, OA: −3.41 ± 4.81 cm) were also significantly smaller than for optical flow perturbations (YA: −0.42 ± 1.10 cm, OA: 3.15 ± 6.66 cm) (YA: p<0.001, d=2.72, OA: p<0.001, d=2.15). In younger adults, MoS_ML for waist -pull perturbations(16.7 ± 2.46 cm) were significantly larger than for optical flow perturbations (13.0 ± 1.58 cm) (p<0.001, d=2.50) or treadmill slips (15.2 ± 1.97 cm) (p=0.001, d=0.95)(Fig. 2B). MoS_ML for treadmill slips (15.2 ± 1.97 cm) were also significantly larger than for optical flow perturbations (13.0 ± 1.58 cm) (YA; p<0.001, d=1.71). Conversely, older adult MoS_ML for treadmill slips (15.0 ± 1.72 cm) was significantly larger than for optical flow perturbations (13.0 ± 1.73 cm) (p<0.001, d≥1.61) or waist-pulls (13.8 ± 1.83 cm) (p=0.001, d=0.97)(Fig. 2D). MoS_ML for waist-pulls (13.8 ± 1.83 cm) were also significantly larger than optical flow perturbations (13.0 ± 1.73 cm) (OA; p=0.018, d=0.593).

3.2.). Correlations:

We found several significant correlations between margin of stability outcomes across perturbation contexts. In younger but not in older adults, we found positive and significant correlations between: (i) ΔMoS_AP due to lateral waist-pulls and oscillations in optical flow (ΔMoS_AP: r=0.428, p-value=0.021) (Fig. 3), and (ii) ΔMoS_AP and ΔMoS_ML of treadmill-induced slips (r=0.468, p-values=0.011) (Fig. 4). Conversely, in older adults, we found positive and significant correlations of ΔMoS_ML between waist-pull perturbations and optical flow perturbations and waist-pull perturbations and treadmill-induced slips (ΔMoS_ML: r≥0.398, p-values≤0.036)(Fig. 5).

Figure 3:

YA between-condition correlations of perturbation effects (ΔMoS in cm) in the ML and AP direction. Pertrubation effect only generalized in the AP direction between our two ML acting perturbations (i.e., optical flow and waist-pulls). Significant correlations defined using p<0.05.

Figure 4:

Within-condition correlations for treadmill slips (AP vs ML) in both YA and OA. ΔMoS_AP and ΔMoS_ML direction only correlated for YA. Significant correlations defined using p<0.05.

Figure 5:

OA between-condition correlations of perturbation effects (ΔMoS in cm) in the ML and AP direction. Perturbation effects generalized in the ML direction between lateral waist-pulls and: (i) treadmill-induced slips and (ii) optical flow perturbations. Significant correlations defined using p<0.05.

3.3.). Age Effects:

We found that older adults exhibited significantly larger AP responses than younger adults to treadmill-induced slips and lateral waist-pulls (p<0.041, d≥0.555) (Fig. 6). Conversely, younger adults exhibited larger ML responses than older adults to lateral waist-pulls (p<0.001, d=1.40) (Fig. 6). We found no significant between-group differences for optical flow in the AP (p=0.88) or ML (p=0.068) directions or for treadmill-induced slips in the ML direction (p=0.36).

Figure 6:

Boxplots for the response to perturbations, actual change in margin of stability (ΔMoS), in the anteroposterior (AP) and mediolateral (ML) directions from unperturbed walking to the three perturbation paradigms: mediolateral optical flow, treadmill-induced slips, and lateral waist-pulls. Older adults had a significantly larger AP response to treadmill-induced slips and lateral waist-pulls compared to younger adults. Younger adults had a significantly larger ML response to lateral waist-pulls compared to older adults. Asterisks (*) indicate significant pairwise differences (p<0.05) between cohorts.

4.). Discussion

Factors and scenarios with the potential to precipitate a fall are various and complex, including those arising from intrinsic and extrinsic sources. Balance perturbation paradigms designed to emulate those factors and scenarios have revolutionized our scientific understanding of walking balance control and inspired diagnostic approaches and mitigation strategies (e.g. (Mansfield et al., 2015; McCrum et al., 2022; Selgrade et al., 2020)). In addition to benchmarking age-related differences in the response to various walking balance perturbations, the central goal of this work was to investigate the extent to which a single perturbation paradigm can be deployed to make widespread inferences about instability and balance impairment. Generally, our suite of perturbation paradigms elicited instability consistent with that reported by previous studies (Golyski et al., 2022; Selgrade et al., 2020). In partial agreement with our hypothesis, we found that the effect of walking balance perturbations did not generalize across all contexts. However, this was not universally true. In older adults, ΔMoS_ML exhibited some generalization in ΔMoS_ML between lateral waist-pulls and: (i) treadmill-induced slips and (ii) optical flow perturbations. Conversely, in younger adults, ΔMoS_AP for optical flow and waist-pull perturbations (which both act in the ML direction) and ΔMoS_ML and ΔMoS_AP for treadmill-induced slip perturbations, exhibit generalization. Despite these findings, we found no “goldilocks” perturbation paradigm to endorse that would support universal interpretations about instability across younger and older adults. However, as we discuss in more detail below, that at least some generalization exists between very different perturbation contexts is scientifically and clinically important.

Contrary to our hypothesis, we found generalization for ΔMoS_ML between lateral waist-pulls and: (i) treadmill-induced slips and (ii) optical flow perturbations in older adults. Indeed, despite differences in the direction of the perturbation, the location of the force application (i.e., BoS vs. CoM), and their context (i.e., mechanical vs. sensory), older adult participants whose ML stability was more affected by lateral waist-pull perturbations were also those whose ML stability was more effected by the treadmill-induced slip and the optical flow perturbations. Maintaining ML stability is crucial to safely performing daily activities such as walking (Hilliard et al., 2008; Maki & McIlroy, 2006; Rogers & Mille, 2003) and, compared to AP stability, disproportionately requires active sensorimotor control during steady-state walking. We add here that the corrective motor adjustments necessary to regulate ML instability in response to both discrete mechanical perturbations and continuous sensory perturbations are at least in part independent of context. Our data suggest that a single perturbation paradigm may in these cases provide insight into ML instability that may generalize across contexts for older adults.

Conversely, though consistent with our hypothesis, we found that younger adults had no generalization of ΔMoS_ML between perturbation contexts. We interpret our findings to suggest that corrective motor adjustments necessary to regulate ML stability in younger adults are context specific. These context-specific responses suggest that younger adults exhibit a broader range of available perturbation responses than older adults. Indeed, compared to younger adults, the more generalized response of ΔMoS_ML in older adults suggests that age decreases the ability to deploy context-specific corrective motor adjustments to maintain walking balance in that direction. There are several possible explanations for this age-related loss of task-specific control. Older adults walk with greater antagonist leg muscle coactivation than younger adults during walking (Hortobagyi & Devita, 2006; Ortega & Farley, 2015; Thompson et al., 2018). Antagonist coactivation alludes to a more generalized control of muscle coordination in older adults. Additional support for this theory is our recent evidence suggesting that, compared to younger adults, older adults walk with a reduced peripheral motor repertoire (i.e., a smaller number of motor modules extracted from recordings of leg muscle activities) that also correlates with their increased instability from perturbations. Cumulatively, these observations likely explain why the ML instability elicited by walking balance perturbations generalizes in older but not younger.

In younger adults, we found unanticipated and complex associations that are at odds with our primary hypothesis. As the first example, we found weak but significant correlations between the ΔMoS_AP for lateral waist-pull versus mediolateral optical flow perturbations. We were surprised to find a correlation in ΔMoS_AP between contexts designed to elicit ML instability. This outcome alludes to the role of intersegmental dynamics in transferring ML to AP instability. Intuitively, those intersegmental dynamics may be interpreted as a more widespread consequence of ML perturbations on walking instability. Conversely, a distribution of the ML instability elicited by perturbations to the AP direction in younger adults could be interpreted more favorably as a strategy to better mitigate the risk of falls. We also found a weak but significant correlation between ΔMoS_ML and ΔMoS_AP for treadmill-induced slip perturbations – the only perturbation context in this study designed to elicit AP instability. Thus, unlike our other perturbation contexts, any ML instability elicited by treadmill-induced slips was an indirect consequence of the perturbation. Due to the intersegmental dynamics previously disclosed, associations between ΔMoS_AP and ΔMoS_ML during walking would not be unexpected. Indeed, other authors have found correlations between MoS_AP and MoS_ML during steady-state walking(Buurke & den Otter, 2021). Nevertheless, we found no such generalization between ΔMoS_AP and ΔMoS_ML to lateral waist-pulls or optical flow. Moreover, the apparent consequence of these intersegmental dynamics was absent for older adults. This latter finding may suggest an age-related loss in the ability to successfully distribute the instability elicited by treadmill-induced slip perturbations.

A key component for enhancing or limiting the generalizability of perturbation-induced effects on stability is fundamental differences in neuromuscular control and motor response patterns between each paradigm. Similarities in recovery motor responses have been shown to be distinct factors in the ability to transfer stability and performance between different paradigms(Werth et al., 2022). These notions are consistent with the findings of this research study, where the AP effect of lateral waist-pull perturbations correlated with those of mediolateral optical flow perturbations. As noted previously, a possible explanation for the generalized ML effect of the perturbations in the OA cohort could be a reduced motor repertoire. In a young to middle-aged cohort of participants, König et al. 2022 demonstrated that differences in the number of muscle synergies utilized in the stability response to two different tasks affected the transfer of stability(Konig et al., 2022). An important line of future research will be to examine the role of neuromuscular control, to include muscle synergy analyses, in governing generalization or lack thereof in the stability response between balance tasks.

The absence of a widespread “goldilocks” paradigm to elicit instability, despite some generalization between perturbation contexts, is clinically important. Fully characterizing an individual’s multi-directional instability to balance challenges in their communities will require a multifaceted approach and a variety of perturbation paradigms designed to emulate various contexts in which falls could occur. Such approaches may also require the patient-specific deployment of these perturbation paradigms for precision diagnostics. Ultimately, the knowledge gained would allow for the creation of personalized rehabilitation programs designed to target the unique instabilities of individual patients at risk of falls. However, we fully acknowledge the overwhelming challenge of implementing this recommendation in the context of the time, space, financial, and personnel constraints placed on our front-line healthcare workers. Overcoming this challenge will require interdisciplinary dialogue between various stakeholders (e.g., engineers, rehabilitation scientists, public health specialists, physicians, nurses, and clinic/hospital management, etc.) to create adequate individualized solutions. In the meantime, for a specific interest in instability to perturbations in older adults, a single context may allow for relevant and actionable information – at least for that based on ML instability. This could be extremely useful, due to the importance of lateral stability in the prevention of falls for older adults (Hilliard et al., 2008; Rogers & Mille, 2003). In conclusion, this calls for a constant need to improve the ways in which we apply perturbation paradigms in the lab for better translation and innovation in our efforts to build accurate patient instability profiles.

Thus far, we have interpreted our results to allude to reactive responses driven by feedback mechanisms prompted by the applied balance perturbations. However, we cannot preclude the possibility that participants altered their motor planning to reflect some level of generalized anticipatory control driven by feedforward mechanisms. The clearest example of this is the widespread increase in ML MoS elicited by each balance perturbation. This observation is not unique to our study; previous research have similarly shown that ML MoS increases in the presence of balance perturbations (Golyski et al., 2022; McAndrew Young et al., 2012; Selgrade et al., 2020). If participants anticipated being perturbed, this would be a feedforward response which could possibly precipitate increased step width and MoS results similar to those observed. Nevertheless, we suspect that observed increases in lateral foot placement, at least in response to the mechanical perturbations, would arise at least in part from a feedback response to arrest the added momentum of the body.

There are several possible limitations relevant to our findings. The first is that we used only continuous sensory or discrete mechanical perturbations. Though, the selection of those perturbation types is ecologically relevant; intuitively, sensory impairments are more persistent in their impact while mechanical perturbations in the community are likely to be discrete events. Second, participants responded to all perturbations at their preferred walking speed, which they could not adjust as a strategy to accommodate instability. Indeed, we have shown that slowing down is an instinctive response to walking balance perturbations (Shelton et al., 2022). Third, we included only healthy younger and older adults, and our findings may or may not generalize to other populations facing greater levels of balance impairment. Fourth, we did not objectively match perturbation intensities across contexts. However, such matching was not necessary to test our central hypothesis about generalization of perturbation effects. Our statistical results confirmed that all perturbation paradigms significantly changed balance outcomes compared to unperturbed walking. Anecdotally, participants found the slip perturbations most difficult. Though, if this difficulty alone explained the generalization in younger adults between ΔMoS_ML and ΔMoS_AP of treadmill-induced slips, we would have expected the same results in older adults. Differences in MoS between perturbation paradigms due to time of application or difficulty would be unlikely to affect resultant correlations. Nevertheless, we recommend caution in making empirical comparisons of the magnitude of effect to different perturbation contexts. Finally, humans are able to rapidly adapt to balance perturbations over repeated exposures. We used randomization to prevent a training effect between perturbation contexts. However, participants also experienced ten perturbations for each of the two discrete contexts, and adaptation during these deliveries is likely. We opted to average over ten discrete perturbations to replicate the nature and requisite analysis for the two minutes of continuous optical flow perturbations. We also cannot exclude the possibility that adaptation to these perturbations differed between younger and older participants and that these effects contribute to between-group differences in generalization. Further study is warranted to study such adaptation across various perturbation paradigms. Due to equipment limitations, lateral waist-pulls could only be block randomized as either pulls to the left or to the right.

5.). Conclusion

We were unable to identify a single perturbation context that could be deployed to make broad generalizations about walking instability in the face of balance challenges. However, ΔMoS_ML of perturbations, at least that quantified using MoS, did generalize between most contexts in older adults and extended to those primarily designed to elicit AP instability (i.e., slips) or between sensory and mechanical perturbations (i.e., optical flow vs waist-pulls). Given the diverse and complex demographics of falls, screening for instability is very likely to require context-specific balance perturbations designed to emulate the variety of environmental contexts where falls can occur.

Table 1:

Generalizations between the mediolateral and anteroposterior effect of each perturbation paradigm or in the direction acted upon by each perturbation paradigm.

	Young Adults		Older Adults
Perturbation Context	r	p	r	p
Mediolateral Optical Flow	0.024	0.900	−0.012	0.954
Lateral Waist-Pull	0.125	0.520	−0.006	0.976
Treadmill-Induced Slips	0.468	0.011	0.001	0.995
Perturbation Direction
Optical Flow (ML) vs. Lateral Waist-Pull (ML)	0.021	0.913	0.462	0.013
Optical Flow (ML) vs. Treadmill-Induced Slips (AP)	0.155	0.422	0.347	0.185
Lateral Waist-Pull (ML) vs. Treadmill-Induced Slips (AP)	−0.006	0.976	−0.127	0.521

Highlights:

• Tested for the generalization of balance responses across perturbation contexts

• Included optical flow, treadmill slip, and lateral waist-pull perturbations

• Balance perturbation effects did not generalize across all contexts

• Only mediolateral perturbation effects correlated significantly in older adults

• Perturbation effects in younger adults were more context specific