Participants and experimental design
A total of 44 healthy adults of various ages (24–72 years) participated in this study. Participants were not eligible to perform the experiments if they were suffering from any movement limiting neurological or musculoskeletal impairments or diseases of the lower limbs. After an initial briefing, all participants provided their informed consent. In the first stability recovery task, all participants were unexpectedly released from a static forward-inclined position (lean-and-release task). Following this, they were exposed to an unexpected trip-like perturbation while walking on a treadmill at a given speed. To ensure safety, participants were secured by a full-trunk safety harness attached to an overhead track allowing antero-posterior and medio-lateral movements but preventing any contact of the body with the ground (except for the feet). The present study was approved by the ethics committee of the German Sport University Cologne (ethical approval no. 141/2017) and was conform to all requirements for human experimentation in accordance with the Declaration of Helsinki.
Lean-and-release task
Participants’ stability recovery performance was evaluated using a lean-and-release task, that has been described in previous studies [8, 17]. Briefly, the participants were standing on a force plate (1080 Hz, 60 × 90 cm: Kistler, Winterthur, Switzerland) with their feet in parallel and flat on the ground. They were gradually inclined in the forward direction and held by a custom-built pneumatic break-and-release system via a horizontally running inextensible Teflon cable connected to a belt around the pelvis [10]. The targeted inclination matched an angle corresponding to a value of 23 ± 2% body weight and was controlled with the means of a load cell implemented in series with the supporting cable. The exact forward lean was chosen according to previous results of the reduced ability of older adults to regain balance within a single recovery step from cable loads of more than 23% body weight [17]. Once any anticipatory movement was attenuated (i.e. antero-posterior and medio-lateral weight shift corrections, checked real-time via cable load and ground reaction forces) the supporting cable was released without any further notice after an arbitrary period between 10 and 30 s. Prior to the measurement, participants were previously instructed to try regaining a stable stance with a single recovery step after being released using the limb of their choice [19]. To guarantee novelty of the task, no prior practice trials were performed.
According to previous findings [10], stability recovery performance was categorised into two stepping behaviours, i.e. single stepping versus multiple stepping. Participants were defined as single-steppers if they needed only one step to recover stability or if a follow-up step of the contralateral limb did not exceed the anterior displacement of the recovery limb’s foot. Consequently, multiple stepping was identified if participants required any additional step of the recovery limb or needed a safety harness support, i.e. more than 20% of body weight observed via a second load cell integrated into the harness suspension cable [20] (Fig. 1).
Single exposure to a trip-like perturbation during treadmill walking
The tripping-task used in the current study has been conducted previously [8, 18]. The protocol started with the participants walking unperturbed on a treadmill (pulsar 4.0; h/p/cosmos, Nussdorf-Traunstein, Germany) at a standardised speed of 1.4 m/s for 4 min followed by a baseline measurement (25 stride cycles of walking). Subsequently, they were exposed to an unexpected trip-like perturbation induced using a custom-built pneumatic perturbation system and encouraged to continue walking afterwards. Throughout one entire swing phase, the perturbation (restraining pull) was applied using a strap attached to the right ankle connected via a Teflon cable to the perturbation device. Although participants received prior information about the task, they were not able to anticipate the onset and removal of the perturbation. All participants were invited to familiarise only with unperturbed treadmill walking 4–7 days prior to the measurement day. To guarantee novelty of the task, no exposures to treadmill perturbations were performed prior to the actual measurement.
Data collection and processing
To determine the CoM trajectories and dynamic stability control during the two tasks, a reduced kinematic model was used [16]. Five retroreflective markers were attached to anatomical landmarks (seventh cervical vertebra, both greater trochanters and forefeet of the left and right legs, respectively) and tracked via a 10-camera optical motion capture system (120 Hz; Nexus 2.6.1; Vicon Motion Systems, Oxford, UK). Three-dimensional coordinates of the markers were smoothed using a fourth-order digital Butterworth filter with a cut-off frequency of 20 Hz [18]. Foot touchdown (TD) of the recovery step in the lean-and-release task was determined as the moment at which the vertical ground reaction force measured by a second force plate (1080 Hz, 60x90cm; Kistler) exceeded a threshold value of 20 N. For the tripping task, TD was defined as the impact peak of an analogue signal acquired using 2-D accelerometers (±50 g, 1080 Hz; model ADXL250; Analog Devices, Norwood, MA) positioned on the tibia of each leg [16]. The antero-posterior margin of stability (MoS) was calculated as the difference between the anterior boundary of the base of support (BoS) and the extrapolated centre of mass (XCoM), which includes both the position and the velocity of the CoM. The MoS and BoS were assessed at each TD during unperturbed, perturbed, and the first six recovery steps following the perturbation [18], as well as at TD of the first recovery step during the lean-and-release task [17]. The BoS was calculated as the distance between the toe markers of the trailing and stance limb at TD for both tasks. Furthermore, the rate of increase in BoS during the lean-and-release task was calculated as the ratio between the BoS at TD and the swing time until TD of the first recovery step.
Statistics
Normal distribution of all variables was confirmed by Lillifors-corrected Kolmogorov-Smirnoff tests (p > 0.05). To examine the relationship between the lean-and-release task and the tripping task performance across participants, Pearson product-moment correlation coefficients were computed for the MoS, the BoS, and the rate of increase in BoS. Since younger adults are not representative of high fall risk, subgroup comparisons (single-steppers versus multiple-steppers) regarding dynamic stability during the lean-and-release task as well as during the tripping-task were performed including only middle-aged and older adults. Independent samples t-tests were used to examine differences between single-steppers and multiple-steppers in the MoS, the BoS, and the rate of increase in BoS for the lean-and-release task. Subgroup comparisons for the tripping-task were performed using separate two-way mixed-measures ANOVAs with factors subgroups (single- versus multiple-steppers) and events (perturbed and the following six recovery steps) for the MoS and the BoS. In case of significant main effects or interactions, Duncan’s post-hoc corrections were applied. The level of significance was set at α = 0.05 and effect sizes were calculated using Hedges’g and partial eta square \( \left({\eta}_p^2\right) \). Effect sizes were considered small (\( {\eta}_p^2 \) =0.01; r = 0.1; g = 0.2), medium (\( {\eta}_p^2 \) =0.06; r = 0.3; g = 0.5), or large (\( {\eta}_p^2 \) =0.14; r = 0.5; g = 0.8). To identify age-related differences in the MoS, the BoS, and the rate of increase in BoS amongst the three age-groups (young, middle-aged, old) during the lean-and-release task, separate one-way ANOVAs were used. Separate two-way mixed-measures ANOVAs were used to detect age-related differences in the MoS, and the BoS during the tripping-task, with age-group (young, middle-aged, old) and events (perturbed and the following six recovery steps) as factors. Differences in age, body height and mass as well as physical activity between the three age groups were analysed using separate one-way ANOVAs. In cases of significant main effects or interactions, Duncan’s post-hoc tests were applied. All statistical and non-statistical analyses were performed using Statistica software (Release 10.0; Statsoft Inc., Tulsa, OK, USA) and MATLAB (2020b, MathWorks®, Natick, MA, USA).