Effect of rhythmic auditory stimulation (RAS)^®^ with and without melody on Parkinson’s disease (PD) patients with deep brain stimulation (DBS): A study protocol

Authors: Kyurim Kang, Reina Arakawa, Isabella Sterner, Joseph Seemiller, Lauryn Currens, Rebecca Khamishon, Yousef Salimpour, Kelly Mills, Alexander Pantelyat

Abstract

Parkinson’s Disease (PD) is characterized by motor impairments, including gait abnormalities that contribute to instability and an increased risk of falls as the disease progresses. While deep brain stimulation (DBS) of the subthalamic nucleus (STN) or the globus pallidus internus (GPi) can be highly effective for managing tremor, rigidity and limb bradykinesia, its impact on gait remains limited, highlighting the need for dedicated gait therapies. Rhythmic auditory stimulation (RAS)^®^ has been used to improve PD gait parameters, such as cadence, velocity and stride length. This exploratory pilot study aims to investigate behavioral and neurophysiological response to two RAS^TM^ pure rhythmic RAS, which uses metronome-generated beats, and melodic RAS, which incorporates composed music along with rhythmic beats. We will record local field potentials (LFP) from the DBS devices already implanted in individuals with PD as part of their routine care, along with gait parameters, to assess neurophysiological and behavioral responses to RAS. By recording neural signals from the STN and GPi with assessment of gait parameters, this innovative approach will provide insights into how pure and melodic RAS paradigms modulate neural activity in the basal ganglia, informing the design of future multi-session or powered efficacy trials.

Introduction

Parkinson’s Disease (PD) is a complex progressive neurodegenerative disease associated with significant morbidity and adverse impacts on quality of life. While PD pathology involves degeneration of dopaminergic neurons in the substantia nigra pars compacta that underlies motor symptoms, the disease presents with a highly heterogeneous variety of symptoms [1] related to both dopaminergic and non-dopaminergic mechanisms. Common motor symptoms of PD include asymmetric resting tremor, rigidity, bradykinesia, and changes in posture and gait, including postural instability [2]. Non-motor symptoms in PD include hallucinations, depression, anxiety, neurogenic bladder dysfunction, sleep disorders, and anosmia [3].

Therapeutic management of PD typically involves pharmacological interventions aimed at restoring normal dopaminergic tone in the basal ganglia-thalamo-cortical networks that underly movement control. In cases where pharmacotherapy fails to adequately address motor symptoms, Deep Brain Stimulation (DBS) is considered for selected patients, typically targeting the subthalamic nucleus (STN) or the globus pallidus internus (GPi) [4,5].

While many studies have shown that STN-DBS or GPi-DBS in PD is advantageous for tremor, rigidity, and bradykinesia [5], fewer studies have noted clear benefits for posture and gait abnormalities [6], and these symptoms represent a significant unmet need. As PD progresses, addressing gait abnormalities becomes crucial as worsening motor impairments contribute to greater instability and a higher risk of falls. Therefore, additional approaches are necessary to enhance the effect of existing treatment options.

Rhythmic auditory stimulation (RAS)^®^

Rhythmic auditory stimulation (RAS)^®^ is a Neurologic Music Therapy^®^ technique that involves rhythmic and/or musical stimulation to improve different aspects of movement, including gait [7–10]. This anticipatory time cue is effective as both an immediate entrainment stimulus, providing rhythmic cues during movement, and as a facilitating stimulus to achieve and maintain a more functional gait pattern [10,11]. The standardized RAS^TM^ protocol comprises the following 1) determining the individual’s baseline cadence (steps/minute); 2) adjusting the metronome rhythm to match the patient’s cadence; 3) instructing the individual to synchronize their footsteps with the beat; 4) gradually increasing the pace by 5–20% above the individual’s baseline cadence; and 5) gradually fading out the RAS [10]. RAS has been shown to improve several gait parameters in PD including gait velocity and stride length and may potentially reduce falls [12–16].

PD patients are known to exhibit deficits in internal timing regulation, leading to challenges performing automated actions that require consistent pacing, such as walking [17]. A prevailing hypothesis across research groups suggests that RAS may facilitate network level interaction among the auditory system, cerebellum, basal ganglia, and frontal executive areas, leading to enhanced fronto-temporal-cerebellar cortical striatal network activation [18,19]. This interaction is posited to enhance intra-network connectivity, potentially compensating for basal ganglia dysfunction in PD [19,20]. By providing rhythmic sound cues with predictable intervals, RAS enables gait synchronization with an external rhythm, which may aid patients in improving movement speed and trajectory and muscle recruitment [13,20–23]. Furthermore, research has shown that beta modulation in the STN increases when patients receive metronome-based auditory cues while stepping in a seated position, supporting more consistent step timing [24].

While several studies have suggested that RAS can modulate excessive cortical activity [19,25,26], the extent to which RAS modulates subcortical activity specifically in the STN or GPi remain less well understood.

RAS versus RAS with melody

RAS intervention involves either a musical piece with accentuated and superimposed rhythmic beats [7,8,13,20] or a metronome delivering purely beat-based stimuli [14,27]. Researchers have studied various components within these approaches. For example, Capato and colleagues (2020) [27] investigated the effect of pure beat RAS in PD patients with and without freezing of gait (FOG). Similarly, Murgia and colleagues (2018) [14] explored differences of artificial metronome sounds and “ecological RAS,” which utilized recorded footstep sounds. While numerous studies have examined the effects of pure beat RAS or musical RAS on gait independently, research directly comparing pure beat RAS and melodic RAS is limited. This gap poses an important limitation for a complete understanding of RAS and its optimal effects, particularly in the context of individualized clinical applications, as highlighted by Rodger et al. (2016) [17].

To specifically evaluate the contribution of rhythm and melody in RAS, this pilot study will compare two distinct RAS a “pure RAS” approach utilizing metronome-generated drumbeats and a “melodic RAS” approach incorporating original composed music with superimposed metronome drumbeats. Both paradigms will be assessed for their impact on behavioral outcomes (e.g., gait parameters) and neurophysiological responses (e.g., Local Field Potentials, LFP).

Local field potentials (LFP)

LFP represent synaptic potentials aggregated across groups of neurons near the recording electrode in the DBS lead, are directly associated with transmembrane currents, and provide valuable insights into correlated neural activity [28]. While electroencephalography (EEG) provides an overview of cortical activity, LFP offer a more localized measure of neural activity, capturing signals from subcortical tissue, such as the STN and the GPi [29,30]. Previous studies using LFP found that certain frequency bands are often associated with specific activated neural networks both physiologically and pathologically [31]. These identified frequency bands can provide a clinically relevant biomarker of brain activity; in fact, beta frequency band activity has been successfully used to develop a recently approved closed loop DBS system, which triggers stimulation in response beta frequency band changes [32,33].

PD patients typically show a pathologically increased oscillatory synchrony in the STN in the 8–35 Hz range (alpha and beta range), whose suppression via dopaminergic medication or DBS leads to symptom improvement [34]. While the alpha range (8–12 Hz) was previously shown to correlate with the intensity of dyskinesia, it has been mostly evaluated in association with cognitive function and emotion (Yin et al., 2022). Thus, the present study will primarily focus on the motor band, specifically the beta frequency band ranging from 13 to 35 Hz.

By leveraging advanced techniques to record LFP directly from the DBS device already implanted in participants, we aim to explore how pure rhythmic RAS and melodic RAS modulate gait performance and basal ganglia activity, thereby generating preliminary behavioral and neural evidence to guide future multi-session or efficacy trials.

Materials and methods

This exploratory study design protocol was developed in accordance with the SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) 2013 guidelines (S1 Table).

Participants

We will recruit a total of 20 participants diagnosed with idiopathic PD according to 2015 MDS Criteria [2] and implanted (10 in the STN and 10 in the GPi) with the Medtronic® Percept™ PC Model B35200 neurostimulator with BrainSense™ technology (Percept™ PC) or Medtronic® Percept™ RC (rechargeable neurostimulator). Participants must have had at least 1 additional follow-up after their initial programming visit for inclusion. The study procedures, risks, and data collection process will be explained to all participants and written informed consent will be collected prior to study participation by the study team member. Table 1 indicates inclusion and exclusion criteria for this study.

We plan to recruit eligible patients from the Johns Hopkins Parkinson’s Disease and Movement Disorders Center who meet the inclusion criteria described above. To optimize our recruitment strategy, all patients with PD and Medtronic Percept DBS under the care of the clinician study team member will be referred to the study.

Recruitment was initiated on April 19, 2023. As of the current stage, we have recruited 7 participants with STN deep brain stimulation and 1 participant with GPi stimulation. We aim to complete recruitment of the targeted number of participants by July 30, 2026, with data collection anticipated to conclude by August 31, 2026. The results are expected to be available by January 31, 2027.

Study design

Fig 1 provides the study timeline and design (SPIRIT schedule). This pilot study is a repeated measures design, where each participant is exposed to the control condition (without RAS) as well as both intervention conditions (with Pure RAS and with Melodic RAS). This within-subject design allows for direct comparison of outcomes across conditions within the same individual, enhancing our ability to detect changes attributable to the intervention.

Each visit will be conducted by both a study clinician and a research team member. The study clinician will monitor and manage the Medtronic Percept^TM^ tablet, respond in case of adverse events, and assess the gait and balance items of the MDS-UPDRS III (3.9 through 3.13). The researcher will measure the cadence (steps per minute), velocity (meters per second), and stride length (distance in meters covered between successive placements of the same foot) of the participants’ gait and keep track of the different conditions. The assessment will begin with the participant’s deep brain stimulation (DBS) randomly assigned to either the ON or OFF condition. The randomization code will be generated in Excel by a member of the research team. To ensure allocation concealment, the code will be stored in a password-protected excel file and shared only with the study clinician responsible for operating the tablet. The initial stimulation status will be known only to the study clinician, while the researcher, data analyst, and participant will remain masked. Because participants may recognize their stimulation status based on perceptible differences in symptom response between the ON and OFF states, this is considered a single-masked study. An important and novel aspect of this study design involved composing a musical piece to serve as the melodic RAS intervention. In selecting the type of music for the melodic RAS intervention, the NIH Music-Based Intervention (MBI) Toolkit offers a framework for study design guidance [35]. According to the MBI Toolkit, the following elements must be considered in music used for clinical Frequency, Tempo, Melody, Playback volume, Pitch, Timbre, and Rhythm. Additional considerations include the familiarity of the music and its personalization. As the aim of this study is to compare the effects of RAS with and without melody, it is essential to standardize these elements across participants. To achieve this, a musical piece was created that incorporates all of the above considerations.

The tempo will be determined during the pre-intervention phase for each participant using a baseline 10-meter walk assessment and will be applied during the intervention phase. Using online composing software Noteflight, we will dynamically set the beats per minute (bpm) of pre-composed music to match the participant’s normal gait tempo and repeat with a 10% faster metronome beat. The remaining elements of the composition will include melody, pitch, timbre, and rhythm. This approach will allow for precise adaptation of musical elements while maintaining the core structure of the composition (S1 File).

For the melodic RAS, the composed music incorporated key musical elements that align with therapeutic goals, drawing on findings from previous research [36–38]. Table 2 outlines the primary components of the composed melodic RAS music.

Study protocol.

This study protocol has been registered on ClinicalTrials.gov (Identifier: NCT05763732). After participants provide consent, they will complete the Barcelona Music Reward Questionnaire [39] to assess their perceived level of reward experienced with music, helping us understand how individual music preferences and emotional connections to music might influence RAS intervention outcomes. The protocol consists of two conditions (DBS ON and DBS OFF), each of which is divided into a Pre-RAS phase, a RAS/Intervention phase and a Post-RAS phase. The order of stimulation states (DBS ON or DBS OFF) will be block randomized among participants.

Fig 2 illustrates the study protocol. LFP recordings will be collected across all stages. Each bracket in Fig 2 is considered an “event” and is subsequently saved as a.json file. Consequently, the protocol encompasses 26 events, including 1) pre initial walk, 2) initial walk, 3) post initial walk, 4) pure RAS walk, 5) rest, 6) melodic RAS walk, 7) rest, 8) pure-RAS 110% tempo walk, 9) rest, 10) melodic-RAS 110% tempo walk, 11) pre final walk, 12) final walk, 13) post final walk; 14) – 26) repeat 1) – 13) for either DBS ON or OFF). In the DBS ON condition, participants will receive optimized stimulation unless the active contact is at a lead end, preventing brain sensing. In such cases, parameters are modified to permit sensing electrode placement above and below stimulation, while closely mirroring the original optimized settings. After an initial rest period of 2 minutes during which the patient will be sitting in a chair, the study clinician performs the MDS-UPDRS-III rating scale part 3.9 through 3.13. Following this assessment, participants will begin a series of 2-minute walks with breaks in between.

The first walk, titled “Initial Walk” in Fig 2, will be used to set the baseline cadence, where participants will be asked to walk for 2 minutes on a set path without any intervention. The cadence measured during this walk will be used to determine the tempo at which the subsequent RAS interventions are provided. Measuring the baseline cadence for each participant in this fashion will allow for individual adjustments for each participant, rather than setting a common tempo for all participants [13]. When comparing the different RAS interventions, the LFP measured during this initial walk will serve as the baseline measurement. This reference point will allow for comparison of neural activity under different conditions, facilitating an understanding of how each type of RAS influences subthalamic or pallidal LFP activity and associated motor performance.

After another 2-minute resting period, participants will enter the intervention phase. The first half consists of walking to metronome beats set to the base cadence. This includes the pure RAS intervention, where participants will walk to metronome beats (“Pure RAS”) and the melodic RAS intervention, where participants will listen to an original composition at the same tempo with superimposed metronome beats while walking. The same task will be repeated, but with a 10% increase in tempo relative to the base cadence, following a standard RAS protocol [10,11]. To account for the sequence of having either the pure or melodic RAS walk first, the order in which the participants first walk with melody or without melody will be randomized.

Finally, during the post-RAS phase, participants will undergo the same procedures as in the pre-RAS phase. By comparing the outcomes between the pre-RAS and post-RAS phases, we aim to determine whether any immediate effects of the interventions persist beyond their immediate application.

In DBS OFF, there will be a 10-minute washout period to allow the participant’s brain circuits to adjust to the absence of stimulation. Aside from the stimulation state, DBS OFF phase will follow the same protocol as the DBS ON phase.

After completing the interventions, participants will be asked to fill out a brief post-session survey to assess the subjective impact and perceived differences between the pure RAS and melodic RAS interventions (S2 File).

Data collection

The study will take place in the conference room at the Johns Hopkins Outpatient Clinic, which provides ample space for participants to walk back and forth during the sessions.

Demographic data.

Participants’ demographic data will be obtained from their electronic medical record. Key variables include age, gender, years since PD diagnosis, medication status, asymmetry of symptoms if relevant, DBS target and years since DBS surgery. Additionally, participants will be asked to report the time of their last medication dose on the day of data collection.

Behavioral data.

During the 2-minute walk test, cadence, velocity, and stride length will be collected as participants walked repeatedly back and forth over a 5-meter distance. The cadence from the initial walk will be used to determine the tempos for the pure-RAS, pure 110% RAS, melodic-RAS, and melodic 110% RAS conditions.

The study clinician will assess MDS UPDRS part III items 3.9 through 3.13 directly before the initial walk in the pre-RAS phase and directly after the rest after final walk in the post-RAS phase (Fig 2). This will be repeated during DBS ON and DBS OFF conditions.

LFP data.

The Medtronic Percept™ PC allows for the measurement of LFP through DBS leads implanted in the brain, even when the DBS is turned OFF. There are four electrodes on each lead in each hemisphere (left: 0–3; 8–11). To use BrainSense^TM^ Streaming technology and record LFP, the electrode configuration must be compatible. During sensing, only the middle two contacts (left: 1 or 2 and 9 or 10) can be used for actual stimulation [40], and the recording electrodes must be set to above and below the active electrodes (left: 0 and 3 or right 8 and 11) (Fig 3).

If the patient’s clinical settings utilize electrodes 0,3,8, or 11 for stimulation electrodes, the BrainSense Group configuration must be adjusted so that the simulating contact is one of the two middle contacts. This configuration will be used during the entirety of the study session. Following the session, participants’ original clinical DBS settings will be restored. The setup process will begin with an impedance test, followed by a signal test during which the LFP is read. The clinician then will select the contact pair with the largest peak in the beta frequency range, which will determine a frequency band of interest (5 Hz range of the selected frequency). The power of this frequency band across time will be recorded at a sample rate of 2 Hz and provided in the output as BrainSense LFP. During the study visit itself, streaming technology will be used to collect time domain data and to observe changes in LFP in response to different activities.

Another advantage of the streaming function is that it does not impose a strict limit on recording duration. However, to reduce the risk of potential issues—such as longer processing times for JSON exports or possible data loss—we will record in 10-minute segments for each session. Thus, new sets will be made for each of the different Pre-RAS Phase, Intervention Phase, and Post-RAS Phase. In this setting, the LFP data will be streamed from both hemispheres, and the time domain data will be sampled at a rate of 250 Hz. After the end of the session, the “Export Json Session Data” will be selected. The.json file report can subsequently be opened and analyzed using MATLAB.

Data analysis

Both behavioral measures and LFP activity will be compared across the walking initial walk (no RAS), during pure RAS (at base cadence and at 110% base cadence), during melodic RAS (at base cadence and at 110% base cadence), and final walk (no RAS) in DBS ON and DBS OFF state.

Behavioral data.

Cadence, velocity, and stride length will be assessed during a 2-minute walking task involving repeated back-and-forth walks along a 5-meter walkway for all walking conditions. For each walking trial, the participant’s cadence will be calculated as the average of three recordings taken from the middle segment of the walking bout, starting at 40 seconds. This approach is intended to capture cadence data that reflects steady-state walking, rather than the initial phase. The cadence measured during the initial walk will be used as the baseline to establish the tempos for the pure-RAS, pure 110% RAS, melodic-RAS, and melodic 110% RAS conditions. Furthermore, rhythmic entrainment (how well a listener is able to match a given tempo) will be measured by the percentage change divergence between the participant’s measured cadence and the given tempo during their intervention phase (Fig 4A). The lower the absolute value of this percentage change, the better entrained the participant is to the given tempo.

LFP data.

There is currently no single industry standard for LFP analysis, nor an established measurement for studying the effect of RAS on LFP during walking. Additionally, while suppression of beta power has been shown in multiple studies to be an increasingly accepted biomarker for improvement in PD, this has more commonly been shown for symptoms such as bradykinesia [40] and rigidity [41], rather than gait. One paper by Storzer et al. reported on the potential relevance of a small band around 18 Hz that could be relevant for freezing of gait [42]. However, there is currently not enough literature to warrant determining the severity of gait impairment by purely using LFP. Thus, this study will employ various methods of LFP analysis to identify the outcome measure that most accurately reflects the effects of RAS. The goal is to establish an LFP metric that could be used as a biomarker for assessing gait improvement in future research. All LFP analysis will be performed in MATLAB.

Similar to other neurophysiological measures, LFPs are susceptible to potential artifacts. In the case of LFP, common sources of such artifacts are cardiac activity and movement artifacts [43]. There are different methods to address these artifacts. In this study, we will be using the open source Perceive toolbox (https://github.com/neuromodulation/perceive) [43] to both correct for artifacts and support other parts of the LFP analysis. Although no existing tool or subject matter expert can fully eliminate all artifacts, this toolbox has demonstrated efficacy in reducing artifact-induced power, thereby enhancing the analysis of LFP dynamics [44]. Furthermore, to reduce between-participant variability related to electrode placement and local signal amplitude, relative power will be analyzed instead of absolute power.

To minimize data-driven inference and enhance reproducibility, LFP analyses will follow a hierarchical, staged

Stage 1 (Primary analysis): Initially, our focus will be on beta power [45,46] (Fig 4B). Existing literature also indicates that subbands within the LFP spectrum may be more closely linked to symptom improvement, with some studies suggesting that specific subbands are associated with particular PD symptoms. Therefore, we will examine the effect of various RAS interventions on the overall LFP beta power (13–35 Hz), as well as on the low beta band (13–20 Hz) and high beta band (21–35 Hz). Power will be averaged across movement epochs for each condition and stimulation state [47].

Stage 2 (Secondary analysis): Another outcome measurement used in previous studies is the peak approach. To verify the stability and localization of the beta response, a peak-frequency approach will be applied by identifying the frequency of maximal power within the beta range for each participant and condition. Peaks are generally defined as the “maxima in normalized spectral power” [48]. However, the method to determine peaks has differed across studies. Typically, when focusing on peaks, the power spectrum density (PSD) is used to find peaks either visually [40] or through an original algorithm [49]. PSD plots will be visually inspected for the presence of clear beta peaks (Fig 4C). When a distinct peak is identified, peak-centered power will be computed as the mean power within ± 2.5 Hz around the peak frequency [50].

Stage 3 (Exploratory analysis): A broadband power analysis will then be performed to characterize the full spectral profile of RAS-evoked neural activity and to identify additional rhythmic components (e.g., low-frequency or gamma activity) that may inform hypotheses for future studies.

While the beta band has been widely used in LFP analysis, the gamma band has received less focus. Some studies report that the gamma range is associated with a prokinetic effect in PD (Florin et al., 2013; Swann et al., 2016). Therefore, we will also investigate if there is a relevance of RAS on the low gamma band, limited to the 31–95 Hz range. This limitation is due to the 250 Hz sampling frequency of the time domain data collection of the Percept™ PC or RC device.

Statistical analysis

All analyses will be conducted in R. Given the small sample size (10 STN and 10 GPi) and repeated-measures design, the described analyses are exploratory and intended to characterize condition- and stimulation-related effects, rather than perform confirmatory hypothesis testing. Separate linear mixed-effects models (LMMs) will be fitted for behavioral (Gait cadence, velocity, stride length, entrainment level) and neural (LFP: relative beta power and peak-centered power) outcomes, and separate models will be run for STN and GPi targets to reduce model complexity and account for physiological differences between groups.

For each target (STN and GPi) and outcome domain (Behavioral and LFP), LMMs will include the following

(1) Fixed Condition (six initial walk, pure RAS, melodic RAS, pure RAS-110, melodic RAS-110, and final walk) and DBS state (ON vs OFF); (2) Order effects as fixed DBS order (ON-first vs OFF-first) and RAS order (pure-first vs melodic-first); (3) Fixed Trial sequence (1–6) to control for potential fatigue or learning effects; and (4) Random Participant, modeled as a random intercept to account for repeated measures within individuals.

LMMs will be fitted using the lmerTest package with Kenward–Roger approximation for degrees of freedom and small-sample F-tests. Estimated marginal means and pairwise contrasts among RAS conditions and DBS states will be obtained using the emmeans package. False discovery rate (FDR) correction will be applied across these contrasts to control for multiple comparisons.

For each contrast, standardized mean differences (SMDs) and 95% confidence intervals (CIs) will be reported to describe effect magnitude and uncertainty. Bias-corrected bootstrap CIs (2,000 iterations) will be used for key comparisons to improve robustness under small-sample conditions. Sensitivity analyses will include (1) comparing random-intercept versus random-slope models where convergence allows; (2) leave-one-out participant analyses to examine influence; and (3) pooled models including both targets (STN and GPi) with Target as a fixed factor to confirm the consistency of effects across DBS sites. In addition to inferential analyses, descriptive change values (Δ and % change from the initial walk) will be computed to provide a clear representation of the direction and magnitude of effects across conditions. These values will be reported descriptively and will not be used for inferential analyses.

Exploratory neural–behavioral coupling analysis.

To explore relationships between neural activity and motor behavior, associations between LFP features (e.g., beta-band power, low-frequency synchronization) and gait parameters (e.g., cadence, velocity, stride length, entrainment level) will be examined using repeated-measures correlation (rmcorr) and complementary linear mixed-effects models (LMMs). Repeated-measures correlation will quantify the within-participant association between LFP and gait across RAS conditions while accounting for inter-individual variability. In parallel, LMMs will be fitted with gait parameters as dependent variables and LFP features as predictors, including Condition as a fixed factor and Participant as a random intercept. To differentiate within- and between-participant effects, person-mean centering will be applied to neural predictors. Analyses will be conducted separately for STN and GPi datasets, and key results will be summarized as correlation coefficients or standardized slopes with 95% confidence intervals. The neural–behavioral coupling analyses are exploratory in nature. Given the modest sample size, statistical findings will be interpreted with appropriate caution. However, the high-density neurophysiological recordings provide substantial within-subject data, allowing for careful interpretation of consistent trends.

Sample size justification

Because no directly comparable studies exist for estimating effect sizes in this population, we used a precision-based approach to determine the sample size for this pilot study. With a total of n = 20 participants (10 STN and 10 GPi, analyzed separately), the expected 95% confidence interval (CI) half-widths for standardized effects are approximately 0.44–0.47 SD for pooled estimates, and 0.62–0.85 SD for within-target pre–post changes, depending on baseline–post correlations (r = 0.6–0.3; worst-case ~1.01 SD if r = 0). These precision levels are consistent with recommendations for early-phase feasibility studies and are sufficient to stabilize variance estimates within each target group. This sample size therefore provides reasonable bounds on plausible effects and supports effect-size estimation needed to design and power a subsequent full-scale trial.

Handling of missing data

There are multiple factors that contribute to the complexity of LFP data collection in this study. The LFP signal itself is highly sensitive to artifacts and recording variability, PD is a multifaceted condition with heterogeneous motor states, and unpredictable issues can arise when working with patients during recording sessions. To ensure data integrity, after each data collection session, the JSON output file will be cross-checked against session notes to confirm that recorded events are correctly matched to their corresponding experimental conditions.

Casewise deletion will be applied for any events that do not correspond to the predefined 26 recorded events. For instance, if an LFP segment was recorded during non-experimental activities (e.g., MDS-UPDRS clinical rating), that segment will be excluded. Data will also be discarded if a clear confounding factor is present, such as recordings obtained shortly after medication intake, which could alter neural activity patterns.

Each LFP segment is considered an output uniquely associated with its specific condition and time point; thus, within-participant imputation will not be performed for missing data. Missing or unusable data will instead be handled through the LMM framework, which accommodates unbalanced data under a missing-at-random assumption. To assess how robust the results are to violations of the Missing-At-Random (MAR) assumption, we will perform sensitivity analyses using approaches such as pattern-mixture modeling. To maximize the usability of available recordings, analyses will incorporate all valid data. When only one hemisphere provides usable LFP signals, analyses will be stratified as appropriate into (1) the average of both sides, (2) the more affected side (as determined by MDS-UPDRS), and (3) the less affected side. This approach ensures that partial but valid data contribute meaningfully to the analysis while maintaining methodological rigor and transparency.

Data management plans

All study data will be collected and stored in compliance with institutional and federal data protection regulations, including HIPAA. Behavioral, neurophysiological (LFP), and clinical assessment data will be de-identified and coded using a secure participant ID system. Data will be entered into and managed through a secured Johns Hopkins OneDrive server.

LFP data captured during triggered events will be stored as the implantable pulse generator (IPG) and downloaded to the clinician programmer. Data will be then exported as.json files, which can be imported into MATLAB for preprocessing and analysis. The files will be subsequently stored in a secure location such as JH OneDrive, accessible only to IRB-approved study personnel.

All digital data files contain only numerical behavioral performance results and no identifying information. Any additional digital files, including raw or processed neurophysiological data, will be housed and analyzed within a dedicated JH OneDrive folder shared exclusively with the approved study team. Access to these storage locations is restricted and must be granted by the Principal Investigator. All data encryption and security protocols meet or exceed HIPAA compliance standards. Data will not be shared outside of Johns Hopkins University. Regular data audits will be conducted to ensure accuracy, completeness, and compliance with the approved protocol every 6 months.

Data monitoring

A formal data monitoring committee will not be established for this study, as it is an early-phase, non-pharmacologic feasibility study involving a small sample size and low anticipated risk. However, the study team will conduct ongoing monthly internal monitoring for adverse events, protocol adherence, and participant safety. Any serious adverse events will be reported immediately to the IRB.

Safety considerations

Given the involvement of individuals with implanted DBS systems, safety protocols have been established in close consultation with neurology, neurosurgery, and neurologic music therapy collaborators. RAS stimuli will be carefully calibrated to avoid excessive auditory stimulation or entrainment effects that could interfere with motor function or induce discomfort. No changes will be made to the participants’ clinical DBS settings, except in cases where the top or bottom contact is used clinically and one of the middle contact levels will need to be used to allow LFP recording or streaming; in these cases, a temporarily stimulation profile will be made uniquely for the study session, and they will be switched back to their original stimulation profile after the study session. All sessions will be conducted under clinical supervision, and participants will be monitored for any adverse effects, including fatigue, dizziness, or changes in mood or motor symptoms during the session.

Interim data analysis and quality assurance

No formal interim analyses are planned, given the exploratory nature and small sample size of this study. However, ongoing internal monitoring will be conducted to ensure data integrity and participant safety. After each session, LFP data stored on the tablet will be exported as a JSON file and imported into MATLAB for verification. This process will confirm successful data capture and transfer, enabling the research team to detect and resolve any inconsistencies or errors. Interim data will be accessible only to study personnel listed on the IRB for quality control purposes.

Protocol amendments and study termination

Any modifications to the study protocol, including changes in study design, eligibility criteria, outcome measures, or data collection procedures, will be documented and submitted for prior approval by the IRB. Substantial amendments will be clearly indicated in future publications and version-tracked using protocol registries, where applicable. In the event of study termination—whether due to safety concerns, logistical challenges, or lack of feasibility—all data collected to that point will be securely archived and analyzed in accordance with the original study aims, unless otherwise restricted. The reasons for termination will be transparently reported in the final manuscript and any public data repositories, as appropriate.

The status and timeline of the study

This study is currently in the preparatory phase, with preliminary data collected from seven participants with STN DBS implants with Percept PC approved by the IRB at the Johns Hopkins University School of Medicine (IRB #00374716) (S3 File). These initial recordings have informed feasibility and protocol refinement. Preliminary feasibility data were used solely for internal planning purposes and are not analyzed, reported, or interpreted in the present protocol manuscript. Data analysis pipelines for advanced LFP processing are currently under development, including custom scripts for signal extraction and event-related analyses. Full data collection is anticipated to continue over a 12-month period following recruitment, with subsequent phases dedicated to data cleaning, statistical analysis, and dissemination of findings. Preliminary results may be shared at relevant scientific conferences, and the full study outcomes will be submitted for peer-reviewed publication upon completion.

Dissemination plans

The results of this study will be disseminated through peer-reviewed publications in relevant neuroscience, rehabilitation, and music-based intervention journals. Findings will also be presented at national and international scientific conferences, including those focused on neuromodulation, auditory neuroscience, and music therapy. To promote transparency and reproducibility, de-identified datasets, analysis scripts, and relevant materials (e.g., auditory stimuli, software interface screenshots) will be shared via open-access platforms such as the Open Science Framework (OSF), contingent upon ethical approval and participant confidentiality considerations. In addition, lay summaries will be provided to participants and community stakeholders when appropriate.

Discussion

This exploratory study will aim to investigate the impact of pure- and melodic-RAS during DBS ON and OFF on patients with PD who have the Medtronic Percept^TM^ PC or RC DBS device with brain leads in the STN or GPi. The main innovation of this study lies in the use of LFP to deepen our understanding of the neurophysiological effects of RAS in PD patients with DBS. Additionally, the Medtronic Percept^TM^ PC or RC DBS device allows LFP to be recorded during the DBS OFF stage, which is an important and novel addition in comparison to other existing DBS devices. Finally, the use of an original composition for RAS with the flexibility to change the tempo in the context of RAS is also novel, allowing the researchers to focus on the effect of melody on RAS while keeping other factors in a musical piece (such as familiarity, timber and rhythmic structure) constant.

The goal of the study is to recruit 10 participants each with STN and GPi DBS. This number of participants is considered feasible for the following First, there are many DBS devices available on the US market, including Boston Scientific and Abbott. The Percept device was first developed in 2020, and many patients still use its precursor, the Activa PC. Second, the implantation of the DBS device is commonly split among the STN and the GPi to address specific predominant symptoms of PD. Third, for some patients with DBS, their degree of gait impairment has progressed to a point where they no longer fulfill the inclusion criteria for the study. These difficulties have been experienced during the recruitment process. Although the limited sample size precludes confirmatory statistical testing, above-described exploratory analyses will identify potential patterns of RAS-related changes in motor performance and neural activity. In addition to the exploratory analyses, reporting of variance, effect sizes, and confidence intervals will offer quantitative estimates of the magnitude and variability of these effects. These findings will inform methodological refinements and guide the design of larger, multi-session trials aimed at clarifying the mechanisms and full therapeutic potential of RAS in PD.

Importantly, by examining the relationship between LFP activity and gait parameters across different RAS conditions (e.g., pure tone vs. melodic cues), this study aims to identify the types of auditory stimulation that elicit the most adaptive neural responses—such as suppression of pathological beta oscillations or enhancement of movement-related gamma activity. These findings may contribute to the development of individualized cueing strategies based on patients’ neural profiles, potentially optimizing therapeutic outcomes. Furthermore, this line of research could inform the design of closed-loop DBS or neurofeedback systems that dynamically adjust auditory cues or stimulation parameters in response to real-time LFP signals during walking. Such approaches may improve gait performance, reduce freezing of gait, and facilitate the translation of RAS interventions from controlled clinical settings to everyday mobility contexts.

Supporting information