Study design

This experimental study was a planned secondary investigation of a cross-over randomised controlled trial [14]. The study conformed to the Declaration of Helsinki and was approved by the University of New South Wales Human Research Ethics Committee (HC16227).

Participants

Participants were recruited via the Neuroscience Research Australia Research Volunteer Registry in July 2019 to March 2020. Eligibility criteria included adults aged 65 years or above who were living independently, able to walk unaided for 20 min or more and without osteoporosis or a neurological condition. All participants provided written informed consent.

Experimental protocol

Initially, participants were assessed for body weight, height and leg dominance (‘which leg would you kick a ball with?’) and asked to walk at usual gait speed along an 8 m walkway containing a 5.7 m GAITRite pressure sensitive mat (CIR Systems, Franklin, NJ, USA) three times to quantify usual step length, cadence and velocity. Participants were fitted with a full body harness, protective guards on knees, shins, foot instep and toes (placed over the shoes). Thirty-nine 14 mm diameter retro-reflective markers were affixed to participants’ anatomical landmarks in accordance with the Plug-in-Gait full-body marker set [31]. Participants started with either the walkway (n = 19) or treadmill condition (n = 19) based on a coin flip and completed both conditions.

Walkway

Participants familiarised themselves (3 min) with walking on a custom-built 10 m wooden walkway that concealed a trip-board (height: 14 cm, width: 50 cm) (S1 Fig in S1 File) [23] and interfaced with an eight infrared camera motion capture system (100 Hz; Vantage, Vicon, Oxford, UK). To ensure participants consistently walked at ~90% of their usual walking speed, stepping tiles were placed along the walkway at 95% of participants’ usual step length and they were instructed to step on these tiles in time with a metronome set at 95% of their usual cadence. Participants were informed that they may experience a hazard anywhere and at any time whilst walking on the walkway but to try to continue walking normally. We administered five unperturbed walking trials (each for 10 m) followed by two semi-random trip trials separated by at least one unperturbed walking trial to ensure no predictive behaviours. In the first perturbation trial, the trip-board obstructed the non-dominant left leg during the mid- to late-swing phase (≥50%) on the middle of the walkway (Fig 1B). In the second perturbation trial, the trip was delivered to the left leg during the early- to mid-swing phase at a different location on the walkway (Fig 1A). Trips were induced using a 14 cm height tripping board that flipped up from the walkway when the foot passed a hidden sensor at mid-swing.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1.

Images of trips on the walkway that induced an (A) elevating, (B) lowering recovery strategies, and (C) treadmill accelerations.

https://doi.org/10.1371/journal.pone.0312261.g001

Treadmill

Participants practised walking (3 min) on a split-belt treadmill (M-Gait, Motekforce Link, Amsterdam, NL) (without handrails) and interfaced with an eight infrared camera motion capture system (100 Hz; Bonita, Vicon, Oxford, UK). The treadmill belt velocity was fixed to 90% of participants’ usual walking speed (as assessed on the GAITRite mat). Participants were informed that they may experience a hazard at any time whilst walking on the treadmill but to try to continue walking normally. We administered five unperturbed walking trials (each lasting 30 s) followed by two semi-randomly presented perturbation trials separated by at least one unperturbed walking trial. In both perturbation trials, the non-dominant left leg was perturbed via treadmill belt acceleration at 8 m s^-2 to a maximum of 200% of usual walking speed for 30% of stride time [14] (Fig 1C; S2 Fig in S1 File). Perturbations were triggered when the hallux motion capture marker of the ipsilateral swing foot as it passed the hallux marker of the contralateral limb in the sagittal plane [29]. Our perturbation magnitude was determined following a pilot session with a middle-aged participant who reported muscle strain with a treadmill belt acceleration at 12 m s^-2 to a maximum of 200% of usual walking speed for 30% of stride time.

Outcome measures

Statistical analysis

A generalised linear mixed-effects model (GLMM) with robust estimation was used to examine the differences in responses on the treadmill and walkway. The model fit was inspected using Q–Q plots of the residuals and data were log-transformed to improve model fit. Unperturbed gait characteristics (MoS, gait speed, cadence, step length) between conditions (treadmill vs. walkway) were examined with the GLMM, adjusting for trial number (fifth/sixth trials) and order of condition (treadmill or walkway first) as fixed factors and participant number as a random factor. Two GLMMs were used to compare continuous kinematic outcomes between conditions (1. treadmill vs. lowering strategy, 2. treadmill vs. elevating strategy) for each step (Pre, Post1-5) adjusting for trial number (first/second trials), order of condition as fixed factors and participant number as a random factor. To control for the family-wise error rate, unadjusted P values obtained from the GLMM were corrected with the Holm-Bonferroni method [34]. Statistical analyses were conducted using IBM SPSS Statistics 26 (IBM Corp., Armonk, NY). P<0.05 was considered statistically significant.

Magnitude of destabilisation

Supporting our first hypothesis, both treadmill and walkway perturbations delivered initial disturbances (i.e., MoS during the first recovery step) to a similar extent. However, the destabilisation on the treadmill immediately dissipated from the second recovery step, while it lasted for multiple recovery steps on the walkway (irrespective of the strategy used) (Fig 2A). Previous studies also showed perturbations with a physical impediment when walking on a treadmill (backward ankle-cable pulls) [28, 35] necessitate more steps to recover stability compared to treadmill belt accelerations alone in older people [36]. The sudden drop and return of MoS around the Post1 step may simply reflect the rapid acceleration and return of treadmill belt speed included in the MoS calculation. Contrary to our second hypothesis, we found greater trunk angles following trips on the walkway (both strategies) than on the treadmill (Fig 4A). Greater trunk flexion angles, forward CoM velocity and displacement and more falls following the walkway trips indicate a greater balance challenge compared the treadmill belt accelerations. The maximum trunk angle on the treadmill was ~20° whereas trunk angles during the walkway trip recovery steps exceeded 30° (Fig 4A). As the head and trunk constitute over 50% of the total body weight [37], greater trunk flexion over the BoS likely contributes to instability and necessitates greater neuromuscular responses to avoid falling following the walkway trips [6, 20, 38].

Stepping kinematics

Distinct stepping kinematics were observed between conditions, supporting our third hypothesis. The anteriorly displaced CoM with an increased AP velocity was countered by longer step lengths during both walkway trip recovery strategies (Fig 2). Tripping over an obstacle requires the perturbed foot to be either quickly raised over or brought down for the contralateral foot to sufficiently clear the obstacle with an uncertain dimension [8]. The greatest vertical displacement observed during a recovery step on the treadmill was 10.3±3.5 cm (Post2), which was less than the greatest displacements observed in both the elevating (26.2±6.4 cm [Post1]) and lowering strategies (38.1±12.7 cm [Post3]) (Fig 2E) achieved via greater hip and knee angles in the corresponding recovery steps (Fig 4). The maximum height of recovery steps was considerably higher than the 14 cm trip-board. This may indicate an over-estimation of a novel obstacle to include a “safety margin” to avoid another obstacle collision [15, 39, 40]. In contrast, treadmill belt accelerations without any physical obstructions showed no increase in length or height in recovery steps.

In line with similar studies, our sample of older people had shorter step lengths and higher cadence while walking on the treadmill [41, 42]. This may reflect a “cautious gait” [43] in response to walking on an elevated, moving treadmill surface. However, we confirmed that the margin of stability and gait speed during unperturbed walking on the walkway and treadmill did not differ. Therefore, these gait parameters are unlikely to have contributed significantly to the observed differences in kinematic responses induced by the treadmill belt accelerations and walkway trips.

Implications for assessment and training using simulated trips

Whilst there is increasing interest in the clinical feasibility of fall prevention training involving simulated trips on a treadmill [24], the current findings suggest that treadmill-induced belt accelerations may not adequately simulate walkway trips. This can explain the lack of transfer from treadmill belt acceleration training to obstacle-induced trips [14]. Treadmill perturbations with higher belt accelerations during gait (e.g., 15 m s^-2) [30] may increase the initial destabilisation, forward trunk flexion, and potentially recovery step lengths compared to our findings based on a moderate acceleration (8 m s^-2). However, considering the lack of swing foot obstruction, larger belt accelerations may not induce the key tripping recovery response of heightened recovery steps because there is no need to further elevate the foot. As our middle-aged pilot participant reported muscle strain with a treadmill belt acceleration at 12 m s^-2, the safety of applying a high acceleration should be carefully considered.

Considering the potential utility of treadmills for perturbation-based training, future research should explore the potential for treadmill belt accelerations to induce responses that better resemble obstacle trips. This may involve increasing the belt acceleration properties, adding a short deceleration prior to a rapid acceleration [30] and accelerating both belts at single stance phase (rather than at a foot-strike). Ankle-cable pulls on a treadmill can physically obstruct the swing foot but repeated exposure to these perturbations lowered toe clearance during steady state walking [44]. Such gait adaptation may minimize the destabilizing effect of ankle-cable pulls but may not facilitate foot elevation to recover from an obstacle trip. Treadmill training linked with obstacle negotiation appears to modify recovery steps, in that Bieryla et al. found training involving treadmill belt accelerations with a 7.6 cm high obstacle near the feet during stance yielded elevated recovery steps and improved trunk recovery responses following a walkway trip [12]. Furthermore, it has been reported that sudden treadmill start training with a 5 cm high foam obstacle over two weeks reduced the rate of real-life falls in 82 women in the following year [17]. Obstacles appear to be crucial for replicating trip responses, the most common cause of falls in community-dwelling older people [21].

Study limitations

There are limitations that warrant consideration. First, study participants were relatively healthy and active, thus the findings may not generalise to frailer older people. Second, missing data impacted the walkaway trials more so than treadmill trials, due to greater perturbation impact (markers becoming dislodged), trunk flexion (markers being obscured) and smaller optimum capture volume relative to the long walkway. Third, the additional cognitive demands on participants to standardise their gait on the walkway (e.g., metronome, stepping targets) may have influenced the observed outcomes. Fourth, although MoS and gait speed were similar during unperturbed gait on the walkway and treadmill, shorter step lengths and higher cadence on the treadmill may have affected balance recovery responses. Future research could use the metronome to standardise both cognitive demand and cadence on the treadmill and walkway. Fifth, the treadmill belt accelerations (8 m s^-2) were delivered at the foot strike (double support stance) form but future studies may investigate different timings (e.g., mid-stance), accelerations (e.g., up to 15 m s^-2), duration and maximum speed [25]. Future research should investigate if different treadmill acceleration properties or use of additional devices may better simulate kinematic responses induced by obstacle-trips.