Participants
A secondary analysis was completed on previously reported data acquired from 12 participants aged 65-90 years (age = 73.8 ± 8.1; body mass = 66.9 ± 13.4 kg; height = 1.6 ± 0.1 m). All participants provided written informed consent as approved by the Hebrew SeniorLife Institutional Review Board. Exclusion criteria included neurodegenerative disease, the inability to walk unassisted or stand for at least one minute without support, or an inability to feel the vibrations produced by the instrumented insole (see below) at maximum output.
Vibratory insole
A full description of the vibratory insole (Fig. 1) has been previously reported [18]. Briefly, a urethane foam insole, customized with two 2.5 cm low voltage piezo-electric actuators placed 2.0 cm apart in the medial arch region of each insole, delivers white-noise-like vibrations below 100Hz to the plantar surfaces of the feet. Multiple insole sizes were developed to ensure comfort and proper fit for each participant. Once the proper size was determined, the insoles were inserted into the subject’s footwear and each participant wore the same pair of insoles and shoes throughout all study procedures.
Experimental protocol
Participants completed three separate study visits to test the effects of three different amplitude levels of sub-sensory vibrations, that is, 0 % (i.e., a no vibration “control” condition), 70 and 85 % of the standing vibratory sensation threshold of each foot. In order to examine the potential for adaptation to the vibration stimuli, standing postural control, foot sole vibration perception threshold and mobility were assessed during each visit within three separate testing sessions (Session 1, 2 and 3). The first session was completed in the morning and a one-hour break was provided in between each session.
The vibratory perception threshold for each foot sole was determined separately at the beginning of each visit. All thresholds were determined with the participants standing with their feet in the same position, approximately shoulder-width apart. Thresholds were obtained by a software program that automatically ramped the amplitude of insole vibration up or down. The participant was instructed to say “now” when they could, or could no longer feel the vibration. This procedure was repeated in multiple trials to narrow the boundary of sensation to a reproducible threshold [18]. On each visit, foot sole vibrations were then delivered at 0, 70 or 85 % amplitudes of the obtained vibratory perception threshold for each foot. This vibration level was set in a randomized order by study staff uninvolved with any other study procedure. As such, both participants and study personnel performing assessments were blinded to the experimental sub-sensory noise level.
Assessment of standing postural control
Standing postural control was assessed by measuring postural sway (i.e., center-of-pressure) fluctuations at 240 Hz with a force plate (Type9286B, Kistler, Amherst, NY). The force plate was placed with its mediolateral axis parallel to the laboratory wall. Participants stood on the plate so that they were facing the wall. Tissue paper was placed on the force plate and foot placement was outlined prior to the first trial. This outline was then used throughout the study to ensure consistent foot placement across trials. During each session, four 60-s trials were completed in each of two different experimental conditions: eyes open and eyes closed. Trial order was randomized. Participants were instructed to “stand as still as possible” prior to each trial. Before each eyes-open trial, participants were also instructed to visually focus on a target “X” placed on the wall in front of them at eye-level. The position of their feet on the force plate was measured and remained the same across all the trials.
To quantify postural sway complexity, MSE analysis was completed on both the AP and ML center of pressure (COP) time-series. Prior to calculation of MSE, empirical mode decomposition (EMD) was used to remove high-frequency noise and low-frequency trends [10, 19]. Specifically, fluctuations at frequencies over 20 Hz were removed, as they are unlikely to reflect physiologically-meaningful control processes [19] and thus the noise from the foot-sole vibration was removed. Fluctuations with frequencies less than 0.2 Hz were also removed to ensure that a sufficient number of dynamic patterns for the MSE analysis. EMD-filtered time series were then “coarse-grained” to capture system dynamics on different scales of time. This procedure divided the COP time-series into non-overlapping windows of length equaling a scale factor, τ, ranging from 1 to 40 data points, so that the coarse-grained series at the largest scale had 360 data points (i.e., 14400 points/40), which is sufficient to obtain reliable estimates of entropy [6]. We then computed the sample entropy of each coarse-grained time-series by choosing m = 2 and r = 15 % based on previous recommendations [6]. Fig. 2 shows the MSE curves generated by plotting sample entropy as a function of time-scale in a representative participant. Finally, the postural sway complexity index was computed as the area under the MSE curve (See Fig. 2, “no vibration” condition), such that larger area reflects higher sample entropy values over multiple time scales and thus, greater complexity.
We previously reported that sub-sensory vibratory noise reduced the average speed and magnitude of postural sway when standing [18]. We therefore included several traditional metrics of postural sway in the current analysis to enable comparison between sway metrics, as well as their relationship to system functionality as measured by the Timed-Up-and-Go (TUG) test (see Statistical Analysis section below). Traditional metrics included postural sway speed (i.e., COP distance traveled divided by duration of one trial) and area (i.e., the area of a confidence ellipse enclosing 95 % of the COP fluctuation).
Timed Up-and-Go Test (TUG)
Participants completed five TUG trials at each testing session, following the postural control assessment. The TUG test measures the time taken to stand from a chair, walk forward three meters, turn around, walk back and return to a seated position. The average TUG time was computed and used for analysis.
Statistical analysis
Analyses were performed with JMP Pro 11 software (SAS Institute, Cary NC). First, linear regression analyses were used to examine relationships between baseline foot sole vibration thresholds, postural sway outcomes (i.e., complexity, area, speed) and system functionality (i.e., TUG performance). These models focused on outcomes collected during the control visit only, during which no vibrations were delivered by the insoles. As this visit contained three testing sessions, the within-day testing session (i.e., 1, 2, and 3) was included as a model factor. Second, the effects of sub-sensory vibration on postural sway complexity were assessed using three-way ANOVAs. Model effects included vibration level (0, 70, and 85 %), standing condition (eyes-open, eyes-closed), within-day testing session (1, 2, and 3) and their interactions. Anterioposterior (AP) and mediolateral (ML) complexity were examined in separate models. Tukey’s post-hoc testing was employed to examine mean differences within statistically significant models. Third, linear regression analyses were used to examine the relationships between vibration-induced changes in postural sway outcomes and mobility (TUG times) (e.g., [(Complexity70 % ‐ Complexity0 %)/Complexity0 %] ∗ 100). Changes induced by vibrations at 70 or 85 % of vibratory threshold were examined separately. The significance level for all analyses was set to p < 0.05.