Efficacy of sensorimotor cortical paired associative magnetic stimulation prior to exercise on balance and gait in Parkinson’s disease: study protocol for a randomised controlled trial in China

Abstract

Introduction

Balance and gait dysfunction in Parkinson’s disease (PD) involves impairments in both motor execution and sensory processing. Conventional transcranial magnetic stimulation (TMS) protocols targeting only the motor cortex may not fully address these complex deficits. A dual-site TMS protocol targeting the sensorimotor network could be more effective.

Methods and analysis

This three-arm, double-blind, randomised controlled trial will recruit 72 participants with PD. Participants will be randomly assigned to receive sensorimotor cortical paired associative TMS (sm-PAS) followed by exercise, single-site motor cortex TMS followed by exercise or sham TMS followed by exercise. All participants will complete a 2-week intervention with five sessions per week. The sm-PAS protocol delivers paired stimuli to the supplementary motor area and the temporoparietal junction with an individualised interstimulus interval. The primary outcome is balance and gait measured by the Mini-Balance Evaluation Systems Test. Secondary outcomes include motor symptoms, quality of life and mechanistic measures. The latter include cortical excitability measured by paired-pulse TMS and cortical activation measured by functional near-infrared spectroscopy. Assessments will be conducted at baseline, after 1 and 2 weeks of intervention and at 1 month follow-up. The primary estimate is the between-group difference in the change of Mini-Balance Evaluation Systems Test scores from baseline to Week 2, comparing sm-PAS group with both sham TMS and single-site motor cortex TMS groups. Data will be analysed using linear mixed-effects models following intention-to-treat principle.

Ethics and dissemination

Ethics approval has been obtained from the Biomedical Ethics Committee of West China Hospital, Sichuan University (No. 2025–2041). Findings will be disseminated through peer-reviewed publications.

Trial registration number

ChiCTR2600116787.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• First randomised controlled trial to evaluate a dual-site, sensorimotor cortical paired associative transcranial magnetic stimulation for balance and gait in Parkinson’s disease.

• Rigorous double-blind, sham-controlled design with allocation concealment minimises bias.

• Multimodal assessment integrating clinical, neuroelectrophysiological and neuroimaging outcomes to elucidate mechanisms.

• The 1-month follow-up limits conclusions regarding long-term durability.

• Mechanistic outcomes are exploratory and may be underpowered.

Background

Parkinson’s disease (PD) is the second most common neurodegenerative disorder worldwide. In China, it affects 1.37% in individuals aged over 60 years, representing an estimated 3.62 million people.¹ As a chronic and progressive disease, PD causes both motor and non-motor symptoms. It leads to significant disability in up to 60% of individuals.² Notably, balance and gait impairments are primary contributors to this disability. They significantly increase the risk of fall,³ reduce mobility, limit daily activities, diminish quality of life and impose substantial socio-economic burdens.⁴

Exercise is a cornerstone of PD rehabilitation. It effectively improves balance and gait, enhances independence and improves quality of life.⁵ Concurrently, transcranial magnetic stimulation (TMS) has emerged as a widely adopted non-invasive neuromodulation technique in neurological rehabilitation.^{6 7} Studies have demonstrated its capacity to improve balance,^{8 9} gait^{9 10} and motor function^10–12 in PD. Emerging evidence further suggests that combining TMS and exercise may produce synergistic benefits. This combination could yield superior improvements in balance and gait performance compared with exercise alone.¹³

However, conventional TMS protocols for PD rehabilitation have primarily targeted the motor cortex. This singular focus may be insufficient as PD is not solely a motor disorder. A growing body of evidence highlights significant sensory abnormalities contribute to motor disturbances.¹⁴ The basal ganglia filters irrelevant or erroneous sensory signals and integrates complex environmental information to modulate motor outputs.¹⁵ As highlighted in a prior study, sensory deficits in PD may arise from increased neural noise and enlarged neuronal receptive fields within the basal ganglia.¹⁶ Therefore, motor dysfunction in PD involves a dual impairment in both motor execution and sensory processing.¹⁷ This dual-impairment model suggests that targeting sensory networks could enhance motor outcomes. Consequently, a TMS protocol that simultaneously modulates both sensory and motor cortical regions may offer a more powerful approach to ameliorating complex balance and gait deficits, especially during challenging daily activities that demand seamless sensorimotor integration (SMI), such as navigating obstacles or turning.

SMI is the critical neurophysiological process by which the central nervous system integrates sensory feedback with motor commands to produce coordinated and adaptive movement.^{18 19} The temporoparietal junction (TPJ) and supplementary motor area (SMA) are anatomically and functionally interconnected, establishing a core processing stream for SMI. The TPJ is a critical functional hub for multisensory integration and spatial cognition.²⁰ It integrates visual, auditory, somatosensory and vestibular inputs to form a unified representation of body position in space.^{21 22} The SMA in turn uses this bodily representation for motor preparation, planning, timing and sequence encoding and constitutes a key region for balance and gait control.^{23 24} The disruption of TPJ–SMA communication may lead to errors in anticipatory postural adjustment and freezing of gait.^{25 26} Consequently, TMS targeting both the SMA and TPJ cortex may enhance SMI capacity and could represent a potential therapeutic approach for improving balance and gait in PD.

Cortico-cortical paired associative TMS is an effective dual-site TMS protocol grounded in the Hebbian principle.²⁷ By delivering paired pulses to two interconnected cortical areas with a specific interstimulus interval (ISI), it can induce associative plasticity and modulate connectivity within targeted brain networks.^{28 29} Given the established role of SMI in motor and gait control, we hypothesise that a novel sensorimotor cortical paired associative TMS (sm-PAS) could enhance SMI capacity. This protocol targets key nodes of the sensorimotor network: the SMA, which is crucial for motor programming, and the TPJ, a hub for multisensory integration and spatial cognition. Applying sm-PAS prior to exercise may prime the sensorimotor system, thereby augmenting the benefits of subsequent exercise.

This paper describes the protocol for a randomised controlled trial designed to test this hypothesis. The primary objective is to investigate the additive benefits of pre-exercise sm-PAS on balance and gait impairments in PD, compared with single-site motor cortex TMS (m-TMS) and sham TMS (S-TMS). The secondary objectives are to: (1) assess the effects of sm-PAS on motor function and quality of life; and (2) elucidate the underlying cortical mechanisms using a combination of neuroelectrophysiological and neuroimaging techniques.

Methods

Study design

This parallel, three-arm, randomised controlled trial will employ a repeated-measures design with concealed allocation, blinding of participants and assessors, and intention-to-treat analysis. This study protocol is developed in compliance with the Standard Protocol Items of the Recommendations for Interventional Trials guidelines. The flow diagram to be followed in this study is shown in figure 1.

Figure 1. The CONSORT flowchart of the trial. AEs, adverse events; A_max%, maximal amplitude percentage; CONSORT, Consolidated Standards of Reporting Trials; fNIRS, functional near-infrared spectroscopy; MDS-UPDRS-III, Movement Disorder Society-Unified Parkinson’s Disease Rating Scale Part III; Mini-BESTest, Mini-Balance Evaluation Systems Test; m-TMS, single-site motor cortex transcranial magnetic stimulation; PD, Parkinson’s disease; PDQ-39, 39-item Parkinson’s Disease Questionnaire; SAEs, serious adverse events; sm-PAS, sensorimotor cortical paired associative transcranial magnetic stimulation; S-TMS, sham transcranial magnetic stimulation.

Participants

Participants diagnosed with PD will be recruited from the Department of Rehabilitation Medicine at West China Hospital, Sichuan University. Recruitment will occur from July 2026 to September 2028. Recruitment has not yet started. All eligible participants or their legally authorised caregivers will provide written informed consent (online supplemental file 1) before enrolment. During the consent process, we will thoroughly inform potential participants or caregivers about study details, including procedures, potential risks and benefits, confidentiality measures and the voluntary nature of participation. The inclusion and exclusion criteria are detailed in table 1.

Table 1. Inclusion and exclusion criteria.

Inclusion criteria

Exclusion criteria

Age <75 years. Diagnosis of PD according to the Diagnostic Criteria for Parkinson’s Disease.61 Hoehn-Yahr stage III–IV. Presence of balance and gait impairments, as indicated by the Mini Balance Evaluation Systems Test less than 20 points.49 Mini-Mental State Examination score more than 24 points, indicating no significant cognitive impairment62 and adequate ability to comprehend and follow therapist instructions. Receiving stable pharmacological treatment for PD. Absence of other neurological disorders. The individual or their legal guardian has the capacity to provide informed consent.

Contraindications to transcranial magnetic stimulation according to the guidelines.60 Severe cardiac, pulmonary, hepatic or renal insufficiency, or uncontrolled malignant hypertension precluding participation in exercise. Severe psychiatric conditions, severe malnutrition, severe osteoporosis, unhealed fractures or other significant musculoskeletal disorders that could interfere with completion of exercise. Lower limb dysfunction attributable to causes other than PD.

PD, Parkinson’s disease.

Sample size calculation

Sample size is calculated using G*Power (V.3.1) with the repeated measures, within-between interaction module, based on the change in Mini-Balance Evaluation Systems Test (Mini-BESTest) scores from baseline to Week 2. A minimal clinically important difference of 4.0 points for the Mini-BESTest is considered meaningful for identifying clinical effects of balance and gait rehabilitation in individuals with PD.³⁰ Assuming postintervention changes of 4 points in sm-PAS group, 2 points in m-TMS group and 0 points in S-TMS group, the SD of group means is calculated to be 1.633. The pooled baseline SD of 9.17 is derived from a previous randomised controlled trial (intervention group A: 22±9.47, n=30; intervention group B: 20±8.16, n=30; control group: 23±9.85, n=28).³¹ With the correlation between repeated measures set at 0.5, the SD of the change scores is 9.17, yielding a Cohen’s f effect size of 0.18 for the group-by-time interaction. With a significance level of 0.05, a statistical power of 80% and an anticipated dropout rate of 20%, the required total sample size is 72 participants, with 24 participants per group.

Randomisation and blinding

An independent researcher, who will not be involved in delivering the intervention, will randomly assign participants to one of the three groups: sm-PAS group, m-TMS group or S-TMS group. This researcher will generate the random allocation sequence using IBM SPSS Statistics V.31. The allocation sequence will be concealed in sequentially numbered, opaque, sealed envelopes, which will be stored in a locked cabinet. On enrolment of an eligible participant, the therapist will contact the independent researcher to reveal the allocation.

Both participants and outcome assessors will remain blinded to the group assignment throughout the trial. After each TMS session, all participants will rate perceived stimulation intensity on a 0–10 scale, which they complete immediately after stimulation. Following the final intervention session, blinding integrity will be formally assessed by asking participants to guess their group assignment and rate their confidence in that guess. The Bang Blinding Index will be calculated to quantify the success of blinding.

A dedicated therapist will deliver all TMS and will not be involved in outcome assessment or exercise delivery. This therapist will also be prohibited from communicating with the exercise therapist. The TMS suite and exercise space are located in separate areas of the hospital, with no shared workspace. Exercise sessions will be provided by a separate therapist who is blinded to group allocation and who will never be present during TMS delivery.

In case of serious adverse events (SAEs), unblinding of the affected participant will be performed after consultation with the principal investigator. Any such event will be reported to the Medical Ethical Committee.

Interventions

Participants in the sm-PAS group will receive sm-PAS within 1 hour before exercise. Those in the m-TMS group will receive m-TMS within the same time frame. The S-TMS group will receive S-TMS within 1 hour prior to exercise. A rapid magnetic stimulator (YINGCHI, China) equipped with two figure-eight coils (outer wing diameter 88 mm) will deliver the magnetic stimulation.

Before stimulation, we will individually determine the resting motor threshold (RMT) recorded from the first dorsal interosseous. RMT is defined as the minimum stimulation intensity required to elicit motor evoked potentials (MEPs) exceeding 50 μV in at least 5 out of 10 trials at rest. Following Rossini’s procedure,³² we will perform this measurement over the primary motor cortex (M1) of the less affected hemisphere under electroneuromyography guidance.

During stimulation, participants will sit comfortably in a reclined chair with neck and head support. Based on the findings of Mi et al,³³ stimulation will be delivered at 10 Hz with an intensity of 90% RMT over the target cortex, administered as five sessions per week for 2 weeks. Each session will consist of 20 trains of 5-second duration, separated by 15-second intertrain intervals, totalling 1000 pulses over 6 min.³³

For sm-PAS, cortico-cortical paired associative magnetic stimulation will be applied over the SMA³³ and TPJ³⁴ of the more affected hemisphere with an individualised ISI.

The ISI is defined as the interval that produces the maximal facilitatory effect. In this study, the ISI specifically denotes the interval at which TPJ-stimulation exerts its maximal facilitatory effect on SMA. Operationally, this corresponds to the interval at which the amplitude of MEPs evoked by SMA stimulation reaches its peak following sm-PAS.

Since direct electrophysiological recording of MEPs from the SMA is not feasible, and the ISI_TPJ–SMA has not yet been established, we will adopt the ISI range of 4–15 ms derived from the well-characterised ISI of the posterior parietal cortex (PPC)–M1 (ISI_PPC–M1) pathway reported by Koch et al³⁵ as a reference. This approach is grounded in the following evidence.

First, the TPJ, SMA and M1 are anatomically and functionally interconnected, forming a core TPJ-SMA-M1 processing stream for SMI. The SMA serves as a gateway between the low-level and high-level neural systems,³⁶ exhibits dense connectivity within and beyond the motor system and plays a central role in coordination.²³ Neuroimaging and neuroelectrophysiological studies have consistently demonstrated functional connectivity between the TPJ and SMA.^{37 38} It is only when the TPJ integrates real-time multisensory information to update the body schema that the SMA can generate accurate motor programme, particularly in unfamiliar environments.²⁵ Moreover, the SMA has dense, direct anatomical and functional connections with M1 that are essential for the voluntary movement execution.^{39 40} Together, these findings establish the TPJ-SMA-M1 pathway as the core processing stream for SMI.

Second, the late peak in the PPC-M1 facilitation profile is likely mediated by a relay in SMA, indicating that the indirect latency measured at M1 provides a reference proxy for the conduction delay to the SMA. Koch et al demonstrated that a conditioning pulse over the PPC potentiates ipsilateral M1-MEPs with a biphasic time course: an early peak at 4 ms ISI and a late peak at 15 ms ISI.³⁵ The extremely short latency (4ms) is consistent with a fast, direct, likely disynaptic pathway linking the PPC directly to M1.³⁵ The longer latency (15 ms), however, suggests an indirect, polysynaptic pathway. Given that the SMA receives comparably dense inputs from parietal region and constitutes a higher-order motor area within the same parietofrontal stream, this 15 ms indirect pathway serves as a physiologically plausible template for the TPJ-SMA connection.⁴¹ Therefore, the individualised ISI_TPJ–SMA will be estimated via stepwise adjustment from 4 ms to 15 ms in 1 ms increments.

To determine this interval, single-pulse stimulation will be delivered to M1 at 90% of individual RMT to measure the amplitude of unconditioned M1-MEPs (A_uc-M1-MEPs). Paired-pulse stimulation will then be applied, with a conditioning stimulus over the TPJ followed by a test stimulus over M1, both delivered at 90% of the individual RMT, using ISIs ranging from 4 to 15 ms³⁵ in 1 ms steps, to measure the amplitude of conditioned M1-MEPs (A_TPJ-M1-MEPs). 10 unconditioned and 10 conditioned MEPs are collected in random order with a 5-second intertrial interval. The maximal facilitatory effect will be quantified as the maximal percentage increase in MEPs amplitude (A_max%), defined as [(A_TPJ-M1-MEPs – A_uc-M1-MEPs) / A_uc-M1-MEPs] × 100%.⁴² That ISI that yields A_max% will be designated as the individual ISI and used in the intervention. If A_max%≤0 or exhibits a fluctuation >30%, a default ISI of 15 ms will be used.

Target localisation will be performed using a standardised 10–10 electroencephalography-based landmark approach, as a neuronavigation system is not available. The right SMA is located at FC2 and left SMA at FC1,³³ as shown in figure 2. The TPJ is defined as the midpoint between P6 and CP6 on the right hemisphere, or between P5 and CP5 on the left hemisphere,⁴³ as shown in figure 2.

Figure 2. Localisation of the SMA and TPJ. (A) Positions in the international 10–10 EEG system. (B) Positions on a two-dimensional head model. The SMA and TPJ are indicated by red and blue circles, respectively. EEG, electroencephalography; SMA, supplementary motor area; TPJ, temporoparietal junction.

To reduce interindividual variability attributable to head size and cranial geometry, lengths of nasion-inion and preauricular will be individually scaled in accordance with the 10–10 electroencephalography system’s proportional rules. The 10–10 electroencephalography coordinates will be transformed to Montreal neurological institute space via group level probabilistic atlases. Although this approach cannot resolve individual-level sulcal or gyral differences, it provides a reproducible and widely accepted localisation framework. Coil position reproducibility across the 10 sessions will be maintained by marking the coil centre on a transparent scalp grid applied the first session. At each subsequent session, the coil will be repositioned if the centre deviates ≥5 mm from the mark. Head movement and intersession drift will be further minimised using a custom head support and chin strap.

In the m-TMS protocol, active stimulation will be applied solely over the SMA, while sham stimulation will be delivered over the TPJ. For S-TMS, sham stimulation will be administered over both the SMA and TPJ. During active stimulation, the coil is placed tangentially to the scalp over the target cortex, ensuring that the induced magnetic field lines are oriented perpendicular to the scalp surface (figure 3A). The coil handle is oriented posteriorly and laterally at a 45° angle away from the midline. At the coil centre, the currents in the two loops flow in the opposite direction to the handle. This configuration induces a cortical current that flows in the opposite direction to the coil currents—that is along the handle direction.⁴⁴ This induced current preferentially stimulates the SMA and flows ipsilaterally along the handle, thereby minimising current spread to the contralateral hemisphere.⁴⁴ For sham stimulation, all stimulation parameters are identical to those used in active stimulation, except that the coils rotate 90° away from the scalp (figure 3B). This validated approach reproduces the auditory and somatic sensation of active stimulation without inducing effective current in the brain.⁴⁵

Figure 3. Positioning of the coil for active and sham stimulation. (A) Positioning of the coil for active stimulation. (B) Positioning of the coil for sham stimulation.

All participants receive exercise within one hour poststimulation, consistent with evidence indicating a 1-hour window of enhanced neural plasticity after TMS.⁴⁶ The exercise programme will be delivered by a dedicated therapist with 5 years of experience in PD rehabilitation. Each exercise session will last 1 hour and follow the American Physical Therapy Association guideline.⁴⁷ The exercise programme is detailed in online supplemental file 2. For each participant, all TMS and exercise sessions will be scheduled at the same time of day and approximately 1 hour after the usual levodopa intake, ensuring a stable medication-on state and control for dopaminergic fluctuations. Additionally, all participants will continue to receive standard pharmacological therapy, occupational therapy or speech therapy as needed. All such co-interventions will be documented. No additional balance and gait rehabilitation interventions will be allowed during trial participation.

Outcomes

All outcome measures will be assessed by two independent researchers who are blinded to the group assignment and have over 5 years of experience in the administration of these measures. Assessments will be conducted at a consistent time of day while participants are in the medication-on state, at baseline, after 1 week and 2 weeks of intervention, and at 1 month postintervention. Primary time endpoint is the change from baseline to week 2 postintervention, with changes at week 1 and at the 1 month follow-up serving as secondary exploratory time-point comparisons.

The primary outcome measure is the Mini-BESTest, which will be used to assess balance and gait. Secondary outcome measures include the Movement Disorder Society-Sponsored Unified Parkinson’s Disease Rating Scale Part III (MDS-UPDRS-III) for evaluating motor symptoms and the 39-item Parkinson’s Disease Questionnaire (PDQ-39) for assessing quality of life. Additionally, we will employ neuroelectrophysiological and neuroimaging techniques to explore the mechanisms underlying TMS-induced cortical plasticity related to balance and gait.

Mini-BESTest

The Mini-BESTest is a validated clinical tool for assessing balance and gait, demonstrating high utility in individuals with PD.⁴⁸ It comprises four domains: (1) anticipatory postural adjustments, (2) reactive postural control, (3) sensory orientation and (4) dynamic gait. The test includes 14 items, each rated on a three-point ordinal scale (0=severe impairment, 2=normal), yielding a maximum total score of 32, as two items are assessed separately for the right and left sides. When sensitivity and specificity are jointly maximised, the optimal cut-off score is 20 out of 32.⁴⁹ Using this threshold, the Mini-BESTest shows adequate ability to identify fallers based on 6-month retrospective and prospective fall history.⁵⁰ The test exhibits high test–retest and inter-rater reliability in PD populations,⁴⁹ correlates strongly with the Berg balance scale⁵¹ and outperforms all gait velocity measures in identifying fallers in PD cohort.⁵² A between-group difference ≥4 points will be considered as clinically meaningful improvement in balance and gait.³⁰

MDS-UPDRS-III

The MDS-UPDRS-III is extensively employed to evaluate motor function and intervention efficacy in PD.⁵³ It covers key motor domains, with total scores ranging from 0 to 132.

PDQ-39

The PDQ-39 is a reliable and valid self-administered instrument for assessing health-related quality of life in PD.⁵⁴ It covers 39 items across eight domains: mobility, activities of daily living, emotional well-being, stigma, social support, cognition, communication and bodily discomfort.⁵⁵

Neuroelectrophysiological measure

Neuroelectrophysiological assessments will be performed at four time points to determine the A_max%, which quantifies the maximal facilitatory effect induced by sm-PAS and thus serves as an index of sm-PAS-induced neuroplasticity. At each time point, paired-pulse stimulation will be delivered using the individualised ISI established at baseline, and the A_max% will be calculated accordingly.

Neuroimaging measure

We will perform neuroimaging examinations using functional near-infrared spectroscopy (fNIRS). fNIRS is an emerging neuroimaging technique well-suited for measuring cortical activity. It quantifies cortical haemodynamics by measuring changes in the concentrations of oxygenated haemoglobin (HbO₂) within cortical vasculature.⁵⁶ Based on the principle of neurovascular coupling, an increase in HbO₂ serves as a reliable biomarker of neural activation.⁵⁷

To measure TMS-induced cortical plasticity related to balance and gait, we will employ a 48-channel fNIRS system (NirSmart-TB, Huichuang, China). It operates at wavelengths of 730 nm and 850 nm. The system includes 21 light sources and 18 detectors interconnected via optical fibres to an elastomeric headgear designed for optimal head-fitting. Optodes are arranged in a grid pattern covering bilateral regions of interest according to a standard electrode placement system. We will determine the Montreal neurological institute spatial coordinates for each channel using the Patriot localisation system. Channels will then be categorised into specific functional regions based on a probabilistic Brodmann area atlas. The layouts of the sources, detectors, channels and regions of interest are illustrated in figure 4A–B.

Figure 4. fNIRS channel layout and measurement tasks. (A) The layouts of the sources, detectors and channels. (B) The regions of interest. (C) Measurement tasks. fNIRS, functional near-infrared spectroscopy; ROI, regions of interest; TUG, timed up and go.

We will acquire haemodynamic signals at 11 Hz across all channels using a block design. Following an initial 5-minute resting state, participants will complete two consecutive assessments: a 40-second baseline measurement followed by three trials of timed up and go test (TUG) (figure 4C). Each TUG trial will be followed by a 40-second sitting interval to allow task-induced haemodynamic response to return to baseline. During the resting and baseline periods, participants will remain seated comfortably with eyes closed, instructed to stay relaxed and minimise head movement. Participants will perform the TUG at a self-selected comfortable speed across three trials. We will quantify cortical plasticity related to balance and gait by calculating the mean difference in HbO₂ between the TUG tasks and the baseline (ΔHbO₂).

fNIRS measurements will be performed by an experienced operator in a quiet, dimly lit room. Prior to the measurement, participants will receive standardised instructions to ensure procedural familiarity. Throughout the assessment, kinematic data during TUG will be simultaneously recorded using a three-dimensional computer vision-based system (HoloMotion, China). Auditory cues generated by the computer will guide participants during task performance. Raw fNIRS data will be preprocessed using the Homer2 package in MATLAB R2018b (MathWorks, USA). Detailed fNIRS data preprocessing procedures⁵⁸ are provided in the online supplemental file 3, and kinematic data preprocessing procedures⁵⁹ are provided in the online supplemental file 4.

Adverse events assessment and management

We will record adverse events (AEs) and SAEs in detail, including start date, end date, description, severity and frequency, to determine safety. Prior to each TMS session, we will screen participants in accordance with established safety guidelines⁶⁰ to exclude potential contraindications.

Before each TMS session, a seizure-risk screening questionnaire is administered. During TUG and exercise, the trained therapist and caregiver will provide fall prevention. Supine and standing blood pressure and heart rate will be measured before and after each exercise session. Orthostatic hypotension is defined as a systolic drop of ≥20 mm Hg or a diastolic drop of ≥10 mm Hg accompanied by dizziness. Medication-on state will be confirmed by self-report and timing since levodopa intake. Stopping criteria include severe headache, seizure-like activity, syncope, fall resulting in injury or participant request. All AEs or SAEs will be adjudicated by an independent blinded clinician using a standardised causality scale (unrelated/possibly related/probably related/definitely related) to attribute events to either TMS or exercise. Emergency equipment and immediate neurology consultation are readily available, given the close proximity of the emergency department and the inclusion of PD neurologists on the research team.

Should any AEs or SAEs occur during the trial, the investigator will promptly implement appropriate clinical management. In the event of SAEs, the investigator will report them to the Ethics Committee in a timely manner. All AEs and SAEs will be documented in the case report form, along with an assessment of their relationship to the intervention. If the intervention is suspended, we will report the reason for the suspension in the case report form. Data on AEs and SAEs will be appropriately analysed and included in the final study report.

Data management and monitoring

We will record all baseline data and raw data on case report forms completely, accurately and legibly at the time of acquisition. We will manage data using a Microsoft Excel spreadsheet database. All research data will be entered into this electronic database by the last day of each month. Two independent investigators will cross-verify the entries to ensure data accuracy and integrity. The database will subsequently be imported directly into IBM SPSS Statistics V.31 for statistical analysis.

On enrolment, each participant will be assigned a unique numeric study identifier to enable anonymous tracking throughout the study. Only the principal investigator will have access to the key that links study data to individual participants. Signed consent forms and hard-copy data collection documents will be stored in a locked cabinet at the research centre, accessible solely to the principal investigator.

Statistical analysis

Continuous variables will be presented as means and SD, and categorical variables as frequencies and percentages. Baseline demographic and clinical characteristics will be compared between groups with one-way analysis of variance (for normally distributed continuous variables), the Kruskal-Wallis test (for non-normally distributed continuous variables) or the χ² test (for categorical variables). The normality of continuous data will be assessed using the Shapiro-Wilk test.

Primary and secondary outcomes will be analysed using a linear mixed-effects model to evaluate the effect of the three interventions on the change from baseline of outcome measures across the four time points. The primary estimate is the between-group difference in the change of Mini-BESTest scores from baseline to Week 2, comparing sm-PAS group with both S-TMS and m-TMS groups. The fixed effects will include group, time and the group-by-time interaction. If the interaction is statistically significant, post hoc pairwise comparisons between groups at each postbaseline time point will be conducted using the estimated marginal means. To account for baseline differences and improve precision, the baseline value of the respective outcome will be included as a fixed covariate in all models. In addition, age, disease duration, levodopa equivalent daily dose and Hoehn and Yahr stage will be entered as a-priori selected covariates. A random intercept per participant will be included to account for between-subject variability. The residual covariance structure will be assumed unstructured to allow flexible correlations among repeated measures within subjects. For non-normally distributed outcomes, the linear mixed-effects model will be fitted on the original scale with robust standard errors to guard against violations of normality. For missing-not-at-random dropout, a pattern-mixture model with delta adjustment will be used for sensitivity analyses, compared against primary likelihood-based model assuming missing-at-random.

For the primary outcome and clinical secondary outcomes, a Tukey-Kramer adjustment will be applied to control the type I error rate for multiple comparisons across groups and time points. For the neuroelectrophysiological and fNIRS outcomes, a false discovery rate correction will be performed instead. Multiple linear regression will be performed to analyse the correlation between changes in neuroelectrophysiological or neuroimaging outcomes and change in primary outcome over the same period. Statistical analyses will be performed using IBM SPSS Statistics V.31, with a significance level of 0.05.

Patient and public involvement

Patients and the public are not involved in the design, conduct, reporting or dissemination plans of this research.

Ethics and dissemination

Ethics and trial registration

This study will be performed in accordance with the Declaration of Helsinki. Ethics approval has been obtained from the Biomedical Ethics Committee of West China Hospital, Sichuan University (approval no. 2025–2041) on 4 November 2025, with an update on 4 June 2026. The study has been registered in the Chinese Clinical Trial Registry (https://www.chictr.org.cn, registration No. ChiCTR2600116787) on 15 January 2026 and updated on 8 September 2026.

Dissemination plan

The results are expected in September 2028. These findings will be submitted and published in international peer-reviewed journals in 2029.

Discussion

To our knowledge, this is the first randomised controlled trial protocol to evaluate the clinical efficacy of a novel dual-site paired associative TMS protocol termed sm-PAS. Moving beyond conventional single-site approaches, sm-PAS targets both sensory and motor cortices to synergistically modulate SMI networks and enhance the benefits of exercise on balance and gait in PD.

In this protocol, we employ an individualised paired-associative timing strategy, using neuroelectrophysiological measures to determine the optimal ISI between TPJ and SMA stimulation, thereby enabling more precise and personalised facilitation of SMI. To evaluate sm-PAS-induced neuroplasticity and explore the underlying cortical mechanisms, we will measure the maximal facilitatory effect on M1 excitability via neuroelectrophysiology and acquire neuroimaging data to examine cortical activation patterns and functional network connectivity. By integrating clinical outcome measures with these multimodal neural data through multiple linear regression analyses, we aim to investigate the relationships between clinical improvements and neural changes.

We hypothesise that this concurrent modulation of SMI may produce a superadditive effect superior to isolated stimulation. It could therefore offer a novel conceptual framework and a potential therapeutic paradigm for PD rehabilitation. Furthermore, given that sensorimotor dysfunction is a common feature across numerous neurological conditions, such as traumatic brain injury, stroke, spinal cord injury and multiple sclerosis. A positive outcome from this trial could support the translation of this sm-PAS methodology to the rehabilitation of other neurological disorders.

This trial has several strengths. First, it is the first randomised, double-blind, sham-controlled clinical trial that uses an individualised Hebbian-timing strategy to investigate the efficacy of sm-PAS in individuals with PD. It may offer a novel and more effective strategy for improving balance and gait. Second, the integrated multi-modal assessment combines clinical, neuroelectrophysiological and neuroimaging measures. This will help elucidate the underlying mechanisms of intervention effects. Third, the trial will provide detailed stimulation parameters, serving as a valuable reference for designing future TMS protocols in PD.

Several limitations should be acknowledged. First, the intervention duration is limited to 2 weeks, and the follow-up assessment is restricted to 1 month postintervention. This timeframe is chosen to minimise participant burden and optimise resource utilisation, particularly given the difficulty of obtaining informative long-term follow-up data. However, it precludes conclusions regarding the long-term durability of the observed effects. Second, the method for determining the individualised ISI_TPJ-SMA relies on the facilitatory effect on M1-MEPs and may not directly reflect the optimal timing for the TPJ-SMA pathway. Future studies could apply EEG to directly measure the conduction delay between TPJ-SMA. Third, the present TMS target localisation method cannot account for individual sulcal or gyral variability. Fourth, all interventions and assessments will be performed in the medication-on state. While this approach enhances safety and feasibility, it may partially mask cortical plasticity effects and attenuate the measurable impact on balance and gait, a limitation that should be addressed in future trials in the medication-off state. Fifth, neuroelectrophysiological and fNIRS outcomes are exploratory and may have limited statistical power. All mechanistic findings will therefore be interpreted with caution and without drawing definitive causal inferences.