Abstract
Background:
Diabetic peripheral neuropathy (DPN) affects nearly 50% of patients with diabetes mellitus, leading to impaired balance and falls. Vibrotactile stimulation during walking is shown to improve gait performance, although the effects of targeted stimulations in the swing phase of gait are unexplored. This study investigated the effects of swing-phase vibrotactile stimulation on gait parameters in DPN.
Methods:
We conducted a prospective controlled study with the DPN group (DG) aged 18-75 years diagnosed with DPN and the control group (CG) with no diagnosis of diabetes. Participants completed 3-minute walk and Timed Up and Go tests with and without vibrotactile stimulation provided by smart shoes. The shoes measured gait parameters, including stride length, velocity, strike angle, heel clearance, swing-stance ratio, stride duration, and distance. Analysis included the Mann-Whitney U test, with a significance threshold set at P <.05.
Results:
A total of 28 individuals were analyzed (CG n = 18; DG: n = 10). DG participants were of higher age and BMI compared with controls (P < .01). At baseline, the DG demonstrated shorter stride length, smaller normalized stride length, lower walking velocity, reduced strike angle, and shorter distance covered in the 3-minute walk (P < .01). Between-group comparisons revealed increased heel clearance in controls (P = .03), which was attenuated after adjusting for age and BMI (P = .17).
Conclusion:
Our findings did not reveal an immediate effect of swing-phase stimulation on gait parameters in DPN patients. However, temporal gait parameters may require more demanding tasks or longer exposure to yield measurable benefits. Larger, age-matched trials using variability data are warranted to determine efficacy and identify responders in DPN.
Level of Evidence:
Level II, prospective comparative study.
Introduction
Diabetic peripheral neuropathy (DPN) is one of the most frequent complications of diabetes mellitus, affecting up to 50% of patients.1,2 DPN results in serious physical impairment often leading to altered gait mechanics, postural instability, and falls. 1 Peripheral nerve damage and its associated sensory loss results in measurable alterations in gait. Previous literature has shown distinctive patterns of abnormality including slower gait velocity, decreased stride length, and increased stride time.3 -5 However, despite the advances in diabetes care, current standard of treatment focuses on glycemic control to slow down the progression of neuropathy and on pain management strategies. 6 These approaches have limited impact on sensorimotor function and mobility impairments. Additional strategies to improve functional mobility are needed.
Given the well-characterized gait abnormalities noted with DPN, there is strong interest in therapeutic interventions aimed at improving gait to potentially reduce fall risk or other complications. Current interventions to improve gait include physical therapy, 7 custom orthotics, and neuromuscular stimulation. 8 Providing targeted sensory cues to patients has shown to be effective as it directly addresses sensory deficits in real time. 9 Recent study of vibrating insoles has reported improved dynamic balance and gait speed, in individuals with DPN. 10 Most prior studies of vibrotactile stimulation have applied vibration continuously or throughout much of the gait cycle. Because falls commonly occurs during the swing phase of gait when the foot must achieve adequate foot clearance, phase-specific cueing might be advantageous, mitigating cutaneous adaptation from continuous stimulation.11 -13 Although prior DPN studies show acute gait and balance benefits from vibrating insoles, the effect of targeted stimulation over the swing phase has not been investigated.
Therefore, in this study, we aim to assess the immediate effects of vibrotactile stimulation using a smart footwear system on gait parameters in individuals with DPN. We hypothesized that swing-phase plantar vibrotactile stimulation would increase gait speed in individuals with diabetic peripheral neuropathy compared with their baseline speed. As a secondary hypothesis, we assessed whether the magnitude of change in gait speed with stimulation differed between participants with diabetic peripheral neuropathy and healthy controls who were not expected to increase with stimulation.
Methods and Materials
Study Design and Population
This is a double-arm, prospective, controlled study conducted at a tertiary academic hospital in Boston, Massachusetts. The protocol was approved by the Mass General Brigham Institutional Review Board (IRB No. 2024P001353) and all participants provided written informed consent.
Individuals aged 18-75 years, with a body weight of ≤120 kg, who were able to ambulate independently without assistive devices, and had no major musculoskeletal impairment of the lower limbs such as active plantar ulceration, lower-limb amputation, other neurologic disorders affecting gait, or major peripheral vascular disease, were included in the study. A body weight of ≤120 kg was selected because of manufacturer guidelines for adequate function of the footwear sensors. The intervention group (DG) comprised patients with a clinical diagnosis of DPN supported by a score of ≥2 on the Michigan Neuropathy Screening Instrument (MNSI). 14 The control group (CG) included patients with no diagnosis of diabetes and no history of neuropathic symptoms.
Enrollment
Participants were identified from patients attending the outpatient Foot & Ankle Orthopaedic Clinic at our tertiary hospital. Eligibility was assessed in the clinic by two experienced podiatrists (S.R.S. and J.S.). For the DG group, 13 individuals met the inclusion criteria and were enrolled. Three were subsequently excluded because of footwear hardware issues or data capture failures, yielding 10 participants in the final analysis. Controls were recruited concurrently using the same inclusion/exclusion criteria. A total of 23 healthy volunteers were screened for participation, of which 5 individuals did not provide consent because of inability to complete all required tasks, and 18 were enrolled in the study. No enrolled participants withdrew consent.
Study Procedures
After confirmation of eligibility, patients were fitted with sensor-integrated footwear (NUSHU Research, Magnes AG; Supplementary Figure S1). Vibrotactile stimulation was delivered throughout the swing phase of gait (toe-off to heel strike). Gait events were detected in real time using embedded inertial measurement units, with toe-off and heel strike identified based on foot-to-ground kinematic profiles derived from the manufacturer’s validated algorithms. Each participant completed gait testing under 2 conditions: (1) no vibration and (2) swing-phase vibration, in which the actuator delivered stimulation during the swing phase of gait. Both groups were instructed to perform a 3-minute walk test on a flat surface, followed by the Timed Up and Go (TUG) test. 15 For the TUG, participants were timed while rising out of an armchair, walking a measured distance of 3 m at a comfortable pace, turning, and walking back to sit down. 16 Trials were performed in a fixed order, with the no-vibration condition completed first followed by swing-phase vibration. For each activity, 2 trials were completed (1 per condition). Participants were instructed to walk naturally at their usual pace for all trials.
The shoes contain inertial measurement units and a vibration actuator embedded in the heel. The sensors are powered through an inductive wireless rechargable battery. The system detects gait phases in real time and provides vibrotactile stimulation. According to the manufacturer’s documentation, the system identifies heel-strike and toe-off with event-timing precision on the order of tens of milliseconds, enabling phase-locked feedback. 17 The vibration actuator was set to the manufacturer’s maximum intensity (corresponding to ~0.7 g peak acceleration). Stimulation was delivered at the manufacturer’s maximum intensity to standardize input across participants. The optimal intensity for gait modulation in DPN remains unknown. Calibration procedures were performed according to the manufacturer’s proprietary algorithms.18 -20
All data were recorded via the smart footware and the manufacturer’s software and exported for analysis. The platform provides real-time gait-phase detection used to trigger swing-phase vibrotactile cues, while summary analytics are refreshed every 5 minutes in the software. Where relevant, bilateral signals were processed per foot by the platform’s validated algorithms and left-right values were averaged to obtain participant-level metrics.
Outcomes and Variables
Baseline demographics, including age, sex, and body mass index (BMI), were collected. The spatiotemporal gait parameters were recorded for both feet, consisting of speed (m/s), stride duration (s), distance (m), stride length (m), normalized stride length (stride length divided by participant height), stride velocity (m/s), strike angle (degrees), heel clearance (cm), and swing-stance ratio. Strike angle was defined as the foot angle at initial contact (heel strike) relative to the ground. Heel clearance was defined as the maximum vertical heel-to-ground distance during the swing phase. Toe height was unable to be evaluated because of the placement of the sensor at the heel of the shoe. The primary outcome included gait speed during the 3-minute walk test. Gait speed was chosen as the primary outcome as it is a validated overall marker of functional decline and fall risk integrating multiple aspects of gait performance, and while swing parameters were measured, they are more variable based on factors such as task at hand and age.
Statistical Analysis
All statistical analyses were performed using SPSS (version 28, IBM SPSS Statistics) and Python programming language (version 3.11.5). Baseline demographic and gait characteristics were summarized for study groups using Fisher exact test for categorical variables and the Mann-Whitney U test for continuous variables. Within-group pre-post comparisons were assessed using the Wilcoxon signed-rank test, and between-group comparisons of change scores (Δ) were conducted using the Mann-Whitney U test. Given the baseline differences in age and BMI between groups, we performed a sensitivity analysis using linear mixed-effects models (LMMs) to assess between-group differences in response to vibration. The LMM included gait parameters as the outcome, time (pre-post), group, and their interaction as fixed effects, age, and BMI as covariates. The time × group interaction term was used to test for differential response to stimulation, with results reported as β coefficients and 95% confidence intervals.
A priori power analysis was conducted using expected gait speed differences between DG and CG based on published values. 21 Assuming an effect size of 0.84, α = 0.05, and desired power = 0.80, the required sample size was estimated to be 19 participants per group. The study was concluded earlier than planned due to funding limitations, resulting in a smaller sample size than projected. To enhance interpretability despite the reduced sample, we report effect sizes (rank-biserial correlations for between-group changes) alongside P-values. A P-value of less than 0.05 was deemed statistically significant.
Results
A total of 28 individuals were analyzed (CG: n = 18; DG: n = 10). Three patients were not able to perform the TUG test because of technical issues and were not included in the final analysis for that test. DG participants were of older age (55.5 [51.0-73.3] vs 23.0 [20.5-27.3] years; P < .001) and had higher BMI (32.65 [26.20-38.83] vs 23.70 [21.52-26.70]; P = .004). Sex distribution did not differ (P = .19; Table 1).
Description of Baseline Values and Their Comparison Between the Groups (N = 28 Patients). a
Abbreviations: CG, control group; BMI, body mass index; DG, Diabetic Peripheral Neuropathy Group.
Data are expressed as median (IQR) or n (%). Boldface indicates significance (P < .05).
Baseline comparison of the study groups for the 3-minute test demonstrated shorter stride length (1.15 [1.00-1.19] vs 1.4 [1.33-1.46] m; P < .001) and smaller normalized stride length (0.61 [0.56-0.69] vs 0.78 [0.75-0.84]; P < 0.001) for the DG. Furthermore, walking velocity was lower in DG, the strike angle was reduced, and DPN patients covered less distance than healthy controls (P < .001 for all; Table 1). Between-group comparisons of the change (Δ) from pre- to post-vibration demonstrated an increased heel clearance in the healthy participants (P = 0.03) over the 3-minute test (Supplementary Figure S2). However, after adjusting for age and BMI using LMM, this difference was no longer statistically significant (β = −0.00, 95% CI −0.01, 0.00, P = .17; Figure 1). No other between-group differences in change scores reached significance in either the unadjusted or adjusted analyses.
Comparison of gait parameters between study groups during the 3-minute walk (3-Min) and the Timed up and Go (TUG) tests. Box plots display the median and interquartile range of the delta values between pre- and postvibration. P values and β coefficients with 95% CIs are derived from linear mixed-effects models adjusted for age and BMI; the time × group interaction term is reported.
For both DG and CG, there were no significant changes from pre- to post-vibration in either the TUG or 3-minute walk for any gait variable (all P ≥ .07). Median changes were small and directionally inconsistent (Table 2), indicating the vibration condition did not measurably alter spatiotemporal gait parameters in this sample.
Intragroup Comparison of Gait Parameters for the Timed Up and Go (TUG) and 3-Minute Walking Tests (N = 28 Patients). a
Abbreviations: DG, Diabetic Peripheral Neuropathy Group; CG, Control Group; TUG, Timed Up and Go.
The Wilcoxon signed-rank test was used to compare the values between pre- and post-swing-vibration measurements.
Discussion
In this preliminary study, we evaluated the effects of swing-phase plantar vibrotactile stimulation on gait performance in individuals with DPN compared with healthy controls. As expected, DPN participants demonstrated substantial baseline impairments, including shorter stride length, reduced normalized stride length, slower walking velocity, and decreased strike angle during the 3-minute walk. There was a statistically significant increase in heel clearance among healthy controls after vibration stimulation, with the 3-minute walking test, although no significant differences were noted with TUG testing. Contrary to our hypothesis, acute exposure to swing-specific vibration did not elicit significant improvements in the gait parameters of patients with DPN or healthy controls.
In healthy older adults, continuous foot-sole vibration has been shown to reduce postural sway and gait variability and modestly improve TUG performance.22,23 However, phase-specific and patterned stimulations have gained attention, because of their ability to selectively improve phase-specific parameters of gait. For instance, providing swing-phase electrical stimulation to 25 patients with stroke-induced unilateral lower extremity paresis, Sehle et al 24 reported improved walking speed in the 10-meter walking test (P = 0.01). To date, no studies have directly investigated swing-phase stimulation and gait parameters in patients with DPN. However, Ravanbod et al 25 reported no change in center-of-pressure sway during sit-to-stand or turning tasks after brief exposure. In contrast, continuous vibrations in this population have shown immediate improvements in gait speed and postural balance during level walking and stair negotiation, with the largest gains appearing through the whole plantar surface stimulation. 10 In our study, swing-phase stimulation did not yield statistically significant changes in gait parameters in either group. One plausible explanation could be that the testing conditions may have lacked sufficient complexity to elicit measurable improvements. As shown by Hijmans et al, 26 random vibrations in individuals with DPN improved standing balance only under dual-task conditions, highlighting the interaction between cognitive demand and sensory feedback. Our unadjusted analysis revealed an increase in heel clearance in the healthy group (P = .03), which may reflect the larger sample size in this cohort (n = 18) or the sensitivity of this metric to swing stimulations. However, this finding did not remain significant after adjusting for age and BMI differences between groups (P = .17), suggesting that the observed effect may have been confounded by demographic differences rather than representing a true differential response to stimulation. This highlights the need for larger, age-matched trials to determine whether swing-phase stimulation can enhance clearance, particularly in populations with neuropathy.
At baseline, the DG showed a characteristically more conservative gait consistent with prior evidence that DPN is associated with reduced velocity and shortened steps, even at preferred pace. 3 The lower strike angle in DPN patients reflects distal sensorimotor deficits and decreased dorsiflexion, a finding that has been directly shown to be associated with decreased gait speed and stride length. 27 In contrast to reports that DPN patients often exhibit longer stride and stance times, our cohort showed no between-group differences in stride duration or swing-stance ratio. 28 This discrepancy might be due to the task- and speed-dependent nature of these parameters. The absence of a significant heel-clearance difference aligns with findings from Suda et al, 29 who reported no difference in mean clearance values. However, a higher toe clearance trajectory variability was found in their study with severe DPN patients when compared with healthy individuals. Finally, group imbalances in age and adiposity likely contributed to the observed contrasts. Rössler et al 30 reported reference values including 545 healthy patients, showing that aging is associated with slower gait and shorter stride length and with increased temporal variability. These changes were specifically pronounced in those aged >80 years. Elevated BMI is also linked to slower gait and shorter steps in overweight/obese populations, which could further depress spatial measures in DPN. 31 Taken together, our baseline findings align with established spatial gait alterations in DPN while underscoring the need to incorporate variability-based end points in future analyses.
This study has several limitations that should be considered when interpreting the findings. First, the study was underpowered; the target sample size of 19 participants per group was not reached because of recruitment challenges. Second, the DG and CG groups differed in age and BMI at baseline. To address this, we performed sensitivity analyses using LMM adjusting for these covariates, which attenuated the only significant between-group finding (heel clearance) to nonsignificance. This underscores the importance of age-matched designs in future studies, as residual confounding by age, adiposity, and other unmeasured factors (eg, glycemic control) likely influenced between-group contrasts. Third, the intervention exposure was acute and single-session, and we did not assess dose-response, retention, or longer-term adaptation, all of which may be necessary for timing-specific plantar stimulation to yield measurable gait changes.
In conclusion, our findings show that swing-phase plantar stimulation did not acutely alter gait parameters in DPN patients. An initial between-group difference in heel clearance response was not sustained after adjusting for baseline age and BMI differences. Larger, age- and BMI-matched trials using variability clearance end points are warranted to determine efficacy in DPN patients.
Supplemental Material
sj-pdf-1-fao-10.1177_24730114261427468 – Supplemental material for Swing-Phase Plantar Stimulation and Gait Parameters in Patients With Diabetic Neuropathy: A Preliminary Study
Supplemental material, sj-pdf-1-fao-10.1177_24730114261427468 for Swing-Phase Plantar Stimulation and Gait Parameters in Patients With Diabetic Neuropathy: A Preliminary Study by Iris Hoffmann, Justin Luk, Kendal Toy, Juliet Moncho, Jennifer A. Skolnik, Sara E. Rose-Sauld, Soheil Ashkani-Esfahani and Atta Taseh in Foot & Ankle Orthopaedics
Footnotes
Appendix
Acknowledgements
The authors gratefully thank all study participants for their time and commitment.
Ethical Considerations
The study protocol was approved by the Mass General Brigham Institutional Review Board (IRB No. 2024P001353).
Consent to Participate
Written consent was obtained from all participants
Author Contributions
Conceptualization: Iris Hoffmann, Atta Taseh, Soheil Ashkani-Esfahani
Methodology: Iris Hoffmann, Atta Taseh, Justin Luk, Kendal Toy, Juliet Moncho
Formal analysis: Atta Taseh, Iris Hoffmann
Investigation (data collection/experiments): Justin Luk, Kendal Toy, Juliet Moncho, Atta Taseh
Data curation: Atta Taseh, Iris Hoffmann
Visualization: Atta Taseh
Resources: Jennifer A. Skolnik, Sara E. Rose-Sauld, Soheil Ashkani-Esfahani
Project administration: Iris Hoffmann, Kendal Toy
Supervision: Jennifer A. Skolnik, Sara E. Rose-Sauld, Soheil Ashkani-Esfahani
Writing – original draft: Iris Hoffmann, Atta Taseh
Writing – review & editing: All authors
Funding acquisition: Soheil Ashkani-Esfahani, Iris Hoffmann, Kendal Toy
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was partially supported by Magnes AG, Zurich, Switzerland who provided shoes for this study.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Disclosure forms for all authors are available online.
Data Availability Statement
Data will be made available upon reasonable request from the journal and with appropriate Institutional Review Board (IRB) approval.