We recruited two groups of participants, including control young participants and community-dwelling older adults. Inclusion criteria for the young participants were age between 18 and 30 and the ability to understand study instructions, while exclusion criteria included cognitive impairments (Montreal Cognitive Assessment (MoCA) < 20), mobility disorders, and history of dizziness, vertigo, sedating medication, or alcohol consumption within 24 h of testing. For older adults, inclusion criteria were age 65 years or older and the ability to understand study instructions. Exclusion criteria for older adults included disorders causing severe motor and balance deficits (e.g., stroke, Parkinson’s disease, severe arthritis, lower-extremity amputation, spinal cord pathologies, and diabetes), history of severe vestibular disorders, central nervous diseases, cognitive impairment (MoCA score < 20), vision problems affecting balance, and history of dizziness, vertigo, sedating medication, or alcohol consumption within 24 h of testing. All participants provided written informed consent in accordance with the Declaration of Helsinki, with approval from the University of Arizona’s Review Boards [28].

In addition to demographic information, several questionnaires were collected from older adult participants to account for confounding variables that can potentially influence balance. Questionnaires included Stopping Elderly Accidents, Deaths & Injuries (STEADI) [29, 30], fear of falling using Fall Efficacy Scale-International (FES-I) [31], cognition using MoCA [32], Charlson comorbidity score [33], depression using Patient Health Questionnaire (PHQ-9) with higher scores indicating greater depressive symptoms [34], pain in lower-extremity based on Visual Analog Scale (VAS) [35], and vestibular deficits using Dizziness Handicap Inventory (DHI). DHI evaluates dizziness-related impairment and vestibular symptoms, which are relevant for balance.

The demographic information, along with shin and thigh lengths, were measured during data collection to extract balance recovery outcomes. Participants were asked to wear lightweight sport shoes, to provide consistency in footwear and minimize associated confounding effects. There were two types of balance recovery sessions involving vibratory stimulation on either the ankle or hip muscles. To minimize potential learning effects from repeated balance recovery attempts, each participant was randomly assigned to only one of the balance recovery sessions (ankle or hip stimulation). In each session, after two practice trials with slow treadmill perturbation exposures, each participant completed 15 trials of treadmill perturbation within three bouts (Table 1). Participants were exposed to either low or high-frequency vibration or no vibratory stimulation in each bout. There were ~ 5-minute rest periods between trials to minimize potential residual effects of vibration and confounding influence of fatigue. Each bout included four trials of sudden backward belt movement with two difficulty levels, including two trials at a maximum speed of 0.35 m/sec and two trials at a maximum speed of 0.7 m/sec. To minimize anticipation effects, one forward belt movement trial with a maximum speed of 0.2 m/sec, was randomly assigned to each bout (Table 1). The order of bouts (no-vibration, low, and high frequency) and balance recovery difficulty (treadmill speeds) were randomized to minimize potential fatigue, learning, and vibration residual effects.

Vibratory stimulation was applied bilaterally to ankle area muscles including tibialis anterior, peroneus longus, soleus, and gastrocnemius, or hip area muscles including quadriceps, gluteus medius, and paraspinals. Vibratory stimulation devices were placed on the belly of the muscles, based on SENIAM guidelines for both the ankle and hip joint muscles [36]. These muscles were chosen for ankle and hip joints because proprioceptive information from these muscles is crucial for ankle- and hip-strategy balance mechanisms during static and dynamic balance, and they have also been the targeted muscles for vibratory stimulation in previous research [15, 21,22,23, 25, 37].

We considered two vibration frequencies for exciting muscle spindles. Previous evidence suggests that in healthy participants, 80 Hz vibrations of ankle muscles produce the maximal effect on postural balance, while frequencies below 40 Hz may not produce consistent effects and can vary between participants [38,39,40]. Accordingly, we used Gaussian noise, band-limited to 80 Hz for higher and 40 Hz for lower frequency stimulations. Magnetic actuator systems (C-2HDLF Tactor, Engineering Acoustics, FL, USA) and a Universal Controller (TDK, Engineering Acoustics, FL, USA) were used to provide the appropriate frequency ranges. The vibration amplitude was set to 1 ± 0.002 mm, a level found to effectively influence muscle spindle afferents [41, 42].

A modified treadmill setup (PhysioGait & PhysioMill, HealthCare International, Langley, WA, USA) was used to impose trip-like perturbations [10, 43]. To prevent an actual fall, the PhysioGait provides a protection harness that prevents knee or hand contact with the treadmill (Fig. 1). The harness was equipped with force transducers to measure the amount of weight tolerated with the harness. Participants were asked to stand motionless on the treadmill, which would start running unexpectedly. The perturbation involved a sudden backward movement of the belt to move the feet posteriorly and induce a forward loss of balance, similar to tripping [10, 44]. In response, the sensorimotor system executes a reactive stepping to expand the base of support and establish stable walking afterwards. Participants wore the harness, adjusted to prevent falling but not interfere with movement, and were instructed not to use their hands for support, as this would invalidate the trial. In each trial, the treadmill reached the max speed in ~ 37 ± 5 msec. Treadmill max speeds and acceleration were selected based on previous studies [10, 27]. Following a successful recovery, participants walked until they regained their steady-state walking, which included a minimum of 20 steps.

Treadmill perturbation setup during recovery stepping with vibratory stimulation applied to the ankle joint muscles

Failing to recover from the perturbation was identified when the entire body weight was supported by the harness [10, 43]. Recoveries with integrated weight support greater than 5% of the body weight × second, from toe off to heel strike of the recovery stepping, were classified as harness-assisted recoveries [43]. All other recoveries were considered successful and were used for calculating balance recovery outcomes. Three-dimensional acceleration and angular velocity of shins, thighs, and the trunk were measured using five wearable motions sensors (LEGSys™, BioSensics LLC, Boston, MA, USA, sampling frequency = 100 Hz), to derive balance recovery and gait outcomes [45, 46]. The signals from the sensors were filtered using first-order high pass butter-worth filter with a cutoff of 2.5 Hz, to remove noise and drift. Balance recovery outcomes included response time, recovery step length, trunk angle during toe-off and heel-strike of recovery stepping, and the required time for full recovery (Table 2).

Within each group of young and older adults, demographic information and clinical measures were compared between participants with ankle and hip exposures using analysis of variance (ANOVA) models for continuous variable and using Chi-square analysis for gender distributions. The association between balance recovery outcomes (Table 2) and vibration conditions (no vibration, 40 Hz, and 80 Hz vibration) was determined using repeated measures mixed effects models. Within these models, independent variables of vibration frequency and treadmill speed, along with their interaction effect, were considered as within-subject factors; sex was considered as a between-subject variable. Effect sizes were calculated using Cohen f or d calculations, using G*Power (version 3.1.9.7, copyright 2020, University of Dusseldorf, Germany). This analytical approach was applied separately for ankle and hip stimulations. Our primary focus was on main and interactive effects involving vibration conditions, with the main effects of speed on outcome measures not explicitly reported. All analyses were done using JMP (Version 14, SAS Institute Inc., Cary, NC, USA), and statistical significance was concluded when p < 0.05 based on the sample size and exploratory nature of the current study.