Effect of Haptic Cueing on Long-Range Autocorrelations in Parkinson's Disease Gait Variability

Effect of Rhythmic Haptic Stimulations on Long-Range Autocorrelations in Parkinson's Disease Gait Variability : A Comparison With Auditory Stimulations

Parkinson's Disease (PD) patients suffer from gait impairments responsible for falls and bad quality of life: reduced speed and stride length, randomness in the temporal organization of stride duration variability (reduced Long-Range Autocorrelations (LRA)). For years, auditory cueing has been used to modulate PD gait and its effect on LRA is known. Less is known regarding the effects of haptic cueing on PD gait and especially on LRA. This pilot study will compare the spatio-temporal gait parameters and LRA of PD patients tested under three conditions: walking without cueing, walking with auditory cueing and walking with haptic cueing by means of rhythmic vibrations on the patients' wrists.

BACKGROUND Parkinson's disease (PD) is the second most common degenerative neurological disease. PD induces gait disorders that lead to increased risk of falls. These falls seriously affect patients' quality of life and generate significant health care costs. Unfortunately, gait disorders do not respond well to drug treatments and their management is mainly based on rehabilitation treatment. The rehabilitation approach comprises two steps: a functional assessment of locomotor capacities followed by completion of a therapeutic physical exercise program. Like heart rate, stride duration varies in the short and long term according to a complex dynamic of temporal variations. These variations present long-range autocorrelations (LRA): the stride duration does not vary randomly but in a structured way. The study of LRA is based on complex mathematical methods requiring recording of at least 256 consecutive gait cycles. LRA are altered in PD patients whose gait rhythm is excessively random. Alteration of LRA is correlated with neurological impairments (Hoehn & Yahr scale and UPDRS) and patients' locomotor stability (ABC scale & BESTest). Measurement of LRA would be the first available objective and quantitative biomarker of stability and risk of falling in patients with PD. Guidelines concerning rehabilitation programs for PD patients are based on education (prevention of falls and inactivity,...), physical exercises, functional training (double task, complex tasks,...), learning, and adaptation strategies such as the use of rhythmic sensory cueing. Auditory Cueing (AC) has been used for years for clinical and research purposes and its effects on spatio-temporal gait parameters and LRA are known. Less is know regarding Haptic/Somatosensory Cueing (HC). A few research were conducted to study the influence of HC on PD spatio-temporal gait parameters but to the best of our knowledge, none has yet addressed its effects on LRA. The aim of this present study was to compare PD saptio-temporal gait parameters and LRA under three conditions : walking without cueing, walking with AC and walking with HC. METHODS 2.1 Participants : 10 patients suffering from idiopathic Parkinson's Disease were recruited from the local community and from the Neurology and the Physical and Rehabilitation Medicine outpatient clinics of the Cliniques universitaires Saint-Luc (Woluwe-Saint-Lambert, Belgium). 2.2 Functional assessment: Before the expermientations starts, all participants underwent a non harmful assessment including clinical tests and questionnaires PD patients: Age, height, weight, sex, most affected side, Movement Disorder Society-Unified Parkinson Disease Rating Scale (MDS-UPDRS), Mini Balance Evaluation Systems Test (Mini-BESTest), Simplified version of the Activities-specific Balance Confidence Scale (ABC-Scale), modified Hoehn & Yahr scale, Mini Mental State Examination (MMSE). 2.3 Procedure : Every participants walked in three conditions in a randomized order. Each condition lasted ±10 minutes in order to get 256 gait cycles mandatory to assess the presence of LRA using the evenly spaced averaged version of the Detrended Fluctuations Analysis (DFA). The first condition was the control condition (CC), patients walking without any cueing on a rectangular track with rounded corner of 63.2 meters in CUSL at their comfortable walking speed. The second condition was the Auditory Cueing Condition (ACC) and consisted in walking on the same rectangular track using auditory cueing by the mean of a smartphone app called Soundbrenner. This app allowed to precisely deliver rhythmic auditory stimulations through earphones paced 10% faster than each patient's preferred step frequence assessed before the experiment. The last condition was the Haptic Cueing Condition (ACC) and consisted in walking on the same rectangular track using haptic cueing by the mean of a vibratory device called Soundbrenner Pulse and attached to each patient's wrist located on their most affected side. The Soundbrenner app on the smartphone was connected to the Soundbrenner Pulse by Bluetooth to deliver rhythmic vibratory stimulations also paced 10% faster than each patient's preferred step frequence, the same frequence as during ACC. 2.4 Data acquisition: Two Inertial Measurement Units (IMU) (IMeasureU Research, VICON, USA) were taped on patients' both lateral malleoli. This system allowed to record ankle accelerations at 500 Hz. The data were then put on a computer and each peak of acceleration, corresponding to each heel strike, was detected by software internally developed to determine all stride durations. 2.5 Gait assessment: Data were extracted from 256 consecutive gait cycles which is required for LRA computation. 2.5.1 Spatiotemporal gait variables: Mean gait speed, gait cadence and stride length were measured as follow: Mean gait speed (m.s-1) = Total walking distance (m)/ Acquisition duration (s) Gait cadence (#steps.min-1) = Total number of steps (#)/Acquisition duration (min) Step length (m) = Gait speed (m/s)*60/Gait cadence (steps/min) 2.5.2 Stride duration variability : Stride duration variability can be assessed 2 ways: in terms of magnitude or in terms of organization (how stride duration evolves across consecutive gait cycles). 2.5.2.1 Magnitude of the stride duration variability : To determine the effect of the RAS on the magnitude of the stride duration variability during 256 gait cycles, the mean, the standard deviation (SD) and the coefficient of variation (CV = [SD/mean] * 100) were assessed. 2.5.2.2 Temporal organization of the stride duration variability : Temporal organization of stride duration variability were assessed by LRA computation using the evenly spaced averaged version of the Detrended Fluctuation Analysis (DFA) to obtain α exponent. The presence of LRA can be shown with α exponent values between 0.5 and 1. Data were treated by the mean of CVI Labwindows (C++). 2.6 Statistical analyses : Statistical analyses were conducted using Sigmaplot 13. If the normality test passed, a one-way repeated measures ANOVA was applied to determine the effect of the various walking condition on spatiotemporal gait parameters (gait speed, gait cadence, stride length) and on linear and nonlinear measures of stride duration variability (CV, SD, H and α exponents). If a significant difference between groups was detected with the ANOVA, a Holm-Sidak post hoc test was performed to compare each mean with the other means to isolate the groups from each other. Effect size between conditions regarding all parameters was assessed using Cohen's d.

Patients walked in three conditions in a randomized order. During control condition, patients walked without cueing on a rectangular track of 63.2 meters at their comfortable gait speed. During Auditory Cueing Condition (ACC) patients walked on the same track using auditory cueing by the mean of a smartphone app called Soundbrenner thath deliverered rhythmic auditory stimulations through earphones paced 10% faster than each patient's preferred step frequence. During Haptic Cueing Condition (HCC) patients walked on the same track using haptic cueing by the mean of a vibratory device called Soundbrenner Pulse and attached to each patient's wrist located on their most affected side. The Soundbrenner app on the smartphone was connected to the Soundbrenner Pulse by Bluetooth to deliver rhythmic vibratory stimulations also paced 10% faster than each patient's preferred step frequence, the same frequence as during ACC.

Long-Range Autocorrelations computation using the evenly spaced version of the Detrended Fluctuations Analysis

Change from baseline in long-range autocorrelations during each intervention condition (3 x 10 min walking)

Change from baseline in coefficient of variation during each intervention condition (3 x 10 min walking)

Inclusion Criteria: Idiopathic PD according to the UK Brain Bank Criteria modified Hoehn & Yahr scale <= 3 Mini-Mental State Examination > 24/30 Ability to walk 256 consecutive strides Exclusion Criteria: Other pathology interacting with gait Significant hearing problems not allowing to hear the auditory cueing Significant somatosensory problems not allowing to feel the haptic cueing on the skin