Parkinson's disease (PD) is a progressive disorder marked by the degeneration of dopaminergic neurons in the substantia nigra projecting to the basal ganglia (BG). As these neurons degenerate, individuals with PD frequently experience the cardinal motor symptoms of slowness of movement, tremor, rigidity, and postural instability. The estimated prevalence and incidence are expected to grow as a result of aging populations (1).

Dopamine-based pharmacologic treatments such as levodopa remain the gold standard for symptomatic treatment of PD (2) and are robust and effective in improving motor function, particularly in the early stages of the disease. However, some symptoms such as gait and balance dysfunction may be poorly responsive to dopaminergic medication (3), and many people who have been treated with levodopa for prolonged periods may experience complications such as dyskinesias and motor fluctuations (2). Deep brain stimulation (DBS) is an effective treatment for advanced PD (4) but utilized in as few as 2% of the PD population (5) for reasons including the invasiveness of surgical intervention and associated potential complications (6), medical comorbidities that prevent surgery, lack of advanced medical care, relatively mild symptoms, and good response to medication. The exact mechanisms underlying DBS effects are not yet fully understood, but likely involve suppression of pathological neural oscillations [e.g., exaggerated beta oscillations (7, 8)] in the BG circuit (9).

Inspired by the success of DBS in alleviating PD symptoms, non-invasive brain stimulation (NIBS) is being increasingly explored. As with DBS, NIBS techniques can apply electric currents to the brain to modulate neural activity (10, 11) and affect downstream behaviors (12). NIBS can be safely and economically tested within a wide range of the PD population, from early to advanced disease stages. Although NIBS lacks the ability to directly target focal areas for maximum effectiveness of the stimulation compared with DBS, it does not rely on implantable hardware. Hardware that must be implanted has severe constraints on design as it must be small in size, strongly conserve battery power, and have strict temperature regulation. In contrast, NIBS is not affected by these limitations to the same degree and can utilize external (and potentially portable) stimulators. Thus, NIBS techniques can employ more complicated stimulus waveforms such as random noise (RN) and multisine signals that can be delivered to achieve different effects, as we show here, as compared to electrical pulses used in DBS.

Galvanic vestibular stimulation (GVS) is one type of NIBS technique that applies weak electric currents to the mastoid processes behind the ears to modulate the firing rates of the vestibular afferents. In human studies, GVS has been utilized primarily as a means to activate the vestibular system in order to study balance and head movement responses (13). A pioneering study to investigate GVS effects on PD patients was conducted in 2005 (14) by applying 24-h continuous noisy GVS to six idiopathic PD patients and one patient with akinesia. The stimulation improved short-range heart rate variability, speed of transitions between rest and activity in the trunk, and reaction time in a Go/NoGo task.

Since then, GVS is being increasingly investigated for the treatment of PD symptoms, motivated by anatomical and functional evidence supporting close connections between the vestibular nuclei, thalamus, and BG (15–19). Prior GVS studies in PD have reported improvement in autonomic system regulation, postural balance and gait, and motor task performance (Supplementary Table 1). Notably, six out of the nine (66.7%) GVS studies have used RN stimuli while the other studies (33.3%) used direct current (DC) stimuli. The predominance of DC stimuli is likely because it has been long-used in balance research to induce body sways using GVS (20). Similarly, a RN stimulus has been adopted as it was used in the original GVS study in PD (14) and has been supported by the stochastic resonance theory stating that the addition of an appropriate level of random noise can paradoxically enhance the response of the nervous system to a weak signal (21–23). Notably, the GVS frequencies used in these studies have been limited to <30 Hz as this reflects the frequency range of most physical movements, and therefore likely reflects the physiological range of endogenous vestibular activation (24). However, we do not know if RN is the most effective stimulus and if different stimulus frequencies significantly influence the motor effects.

Here, we assessed the motor performance of PD participants in a simple reaction time task using several band-limited multisine GVS stimuli. Specifically, we compared whether the multisine stimuli can result in better task performance compared to the more traditional RN stimulus. We next sought to answer the following questions: (1) is there a single band-limited stimulus that brings about the most robust and largest effects across individuals?; (2) does the most effective band-limited stimulus vary across individuals?; and (3) how much improvement in motor performance can be evoked by varying stimulation frequency within an individual? Increasing evidence demonstrates that the same transcranial electrical stimulation can induce substantial variability in individual responses (25–28) due to various factors including methodological differences in the study protocols and participants' physiological traits (e.g., age and sex) and brain states (e.g., emotional and mental states) (28, 29). Here, we posit that a data-driven approach—whereby individual responses to different stimuli are assessed—may be a strategy to partly ameliorate these innate differences. Finally, for the first time, we aimed to address the question of whether GVS effects interact with levodopa medication by recruiting the same PD participants both off-/on-dopaminergic medication.

None of the participants reported any adverse effects nor awareness of any differences between stimulation conditions.

To our knowledge, this is the first study investigating SRT task performance of PD and HC participants while applying GVS across a wide range of frequencies. Overall, our results suggest that RPT can be improved by GVS in PD patients. However, as bradykinesia is a key feature in PD, we cannot disentangle whether or not the RPT improvements were a result of faster decision-making and/or faster movement. This will require examining the simultaneously acquired EEG and will be the topic of another report. We demonstrated that the motor improvement is significantly dependent on the GVS parameters used. Surprisingly, we found that RN-GVS, despite its popularity, did not actually evoke the best task performance in the PDMOFF and HC groups, with ms-γ (30–50 Hz) proving superior in reducing RPT (Table 5). The performance of the off-medicated PD participants during ms-β and ms-γ were comparable to the baseline performance when they were in the on-medication state. We found that the GVS frequency that resulted in the shortest RPT varied considerably across participants, suggesting that a one-size-fits-all stimulus will not be as effective as a personalized stimulus. For most of the PDMOFF and HC participants, the best GVS frequency varied in the range of 8–100 Hz. The worst task performance was found during RN or ms-θ for more than half of the participants in these two groups. These results provide evidence that further work is required to tailor GVS parameters for maximum efficacy.

Whether or not RPT is actually delayed in PD has been controversial (39–41), partially due to methodological heterogeneity and different clinical characteristics of the participants (42). In this study, the difference in baseline RPT between the PDMOFF and HC groups did not reach statistical significance (P = 0.068; Supplementary Table 2). Instead, the most interesting finding was that responses to different GVS stimuli showed a similar trend between the two groups (Figure 3). This finding may suggest that there are some mechanisms underlying the GVS effects that are common between these groups.

In contrast, we found that the PDMON group showed relatively different responses to GVS. Normally, dopamine is active both phasically and tonically during motor performance. Levodopa has complex effects in PD, which may result in relative normalization of tonic dopamine firing, yet impairment of phasic firing (43). While phasic dopamine firing is normally associated with rewards in reinforcement learning paradigms, it may also be involved in internal representations of desired actions with actual sensory feedback during motor performance (44–46). Thus, many studies have suggested that movement-related phasic changes can be observed in nigrostriatal dopamine neuron firing (47–49), and that dorsal striatal phasic dopamine signaling is associated with specific kinematic features of movement (43). Complex effects of dopaminergic medication in PD have also been reported in fMRI studies (50–54) showing that levodopa medication does not simply restore brain connectivity aberrant in PD. Rather it induces functional connectivity changes distinctive from those identified to be different between PD patients and healthy controls (51, 54). Taken together, this is an important point to consider for future GVS studies, as prior studies (Supplementary Table 1) included only medicated PD patients, and the information on the dosage and timing of the medication was rarely reported.

Given the functional role of pathological beta oscillations in PD (55), the result of particular interest was that ms-β resulted in the largest decrease in RT among the tested stimuli in the PDMOFF group whereas it did not improve motor performance for the HC group. In this regard, there is some evidence to support the concept that beta-frequency stimulation may have clinical effectiveness in PD patients. In a transcranial alternating current stimulation (tACS) study conducted on 10 PD and 10 HC participants (56), 20-Hz stimulation at the primary motor cortex (M1) yielded a significant decrease in beta-band cortico-muscular coupling in PD patients but not in HC. A TMS study showed that 20–33 Hz stimulation at M1 elicited significant suppression of the motor evoked potential (MEP) in PD patients and the amount of suppression was correlated with their UPDRS III scores (57). It should be noted, however, that there have been only a handful of studies that utilized beta-frequency NIBS in PD patients, and it is difficult to determine from our results whether the ms-β effects observed in the PD participants were related to the pathological beta-band activity. Thus, further neuroimaging studies are strongly suggested to be carried out to validate our results and elucidate the mechanisms of action.

The frequency-dependent GVS effects we observed may be related to the overlap between the neural processes affecting RPT and neural pathways affected by external vestibular inputs. One of the main vestibular pathways is the direct ascending projection from the vestibular complex to the thalamus, primarily targeting the ventral anterior, ventral lateral, ventral posterior lateral, ventral posterior medial, intralaminar nuclei, and the geniculate bodies (58–60). Strong activations in these regions by vestibular stimulation (18, 59, 61–63) suggest a critical thalamic contribution to processing vestibular information (60, 63). The ventral parts of the thalamus are also closely connected with M1, premotor cortex, and BG (15, 58, 63, 64), modulating a range of aspects in motor control (15, 63, 65). Thus, we conjecture that GVS effects on RPT can be in part explained by vestibular inputs affecting the motor thalamus. This may also explain the mild GVS effects on the PDMON participants as the BG inputs to the motor thalamus would vary at different dopamine levels.

It is also possible that GVS affected the striatum, a region described as an integrative center for sensory information and involved in motor planning and execution. Although the largest inputs to the striatum are from the cortex, recent studies have elucidated the subcortical pathways critical for interpreting and responding to environmental stimuli (66, 67). Electrophysiological studies in animal models and neuroimaging studies in humans have shown that vestibular stimulation activates the head of the caudate nucleus and putamen (16–19, 62, 68, 69), likely through the parafascicular thalamic nucleus (PFN) (64, 70). In addition, it has been recently proposed that the striatal tail may play a role as a multisensory integration center (71), and thus it is possible that there are vestibular inputs to this region as well.

Our observations of different motor effects evoked by varying GVS frequency are consistent with many animal studies (72, 73). Surprisingly, canal and otolith afferents in macaque monkeys responded to GVS as a function of frequency such that the response gain (i.e., spikes/s/mA) increased more than twice when the stimulation frequency varied from 0.1 to 25 Hz (72). This seminal finding opposes the common idea that high-frequency GVS would result in smaller gains because the tissues between the electrode and vestibular afferents may act as a low-pass filter. Similarly, the firing rate of the PFN increases when the frequency of stimulation applied to the semicircular canal nerve is >100 Hz (73). Taken together, these findings could explain in part the efficacy of the frequency range we observed in most of the PDMOFF and HC participants.

There are several limitations in our study. Considering the study design and our primary objective to examine different types of GVS stimuli, we did not try to replicate previous findings demonstrating that GVS results in better motor performance compared to baseline performance seen during off-stimulation. Although we think the practice effect on the task performance is unlikely for a simple, over-learned motor task like ours, the possibility was not completely ruled out when the baseline measurement always preceded active-GVS. Similarly, our study was not designed to measure the after-effects of GVS. Post-stimulation behavior effects of GVS are largely unknown (13). Studies that examined GVS aftereffects stimulated participants for more than 30 min, and the results are conflicting (74–77). The issue of whether stimulation effects last after the cease of stimulation is not only limited to GVS but is one of the main controversial topics for transcranial electrical stimulation (78). As online stimulation effects differ depending on stimulation parameters (e.g., frequency, intensity, duration, target sites) and experimental tasks, the presence and duration of after-effects appear to be influenced by the stimulation parameters and tasks (79, 80). Although after-effects are infrequently reported, evidence from tES studies shows the presence of after-effects when stimulation was applied at >0.5 mA for longer than 10 min (78, 81–83). Given that we applied GVS for a short duration at a low current intensity with a 2-min inter-block off-stimulation break, we suggest that any effects carried over from previous stimulation were relatively mild compared to the online-stimulation effects. Validation of GVS after-effects and their relationships with stimulation parameters will be areas/topics of interest for future work. Note that, since GVS can utilize portable stimulators, reliance on much more subtle after-effects is not as important as other technologies that are not as easily portable (e.g., TMS). Finally, several studies support the notion that GVS effects are mostly spatially restricted to the vestibular organs. For instance, the auditory effects of GVS are rare (20) despite the proximity between the auditory and vestibular systems. GVS evokes circumscribed cortical activation of vestibular areas, and effects on the somatosensory cortex are only seen at specific frequencies (62). At higher intensities, the stimulation of the vestibular system can be self-reported by feelings of vertigo. A recent computational modeling study of the electric field generated by GVS (84) suggests that the bilateral and bipolar configuration, as used in our study, results in the most spatially-restricted current flow to the vestibular organs. However, some current may diffuse to the medulla, pons, and cerebellum. Although we note that both the electrodes (11 mm) and current intensity used here (0.43 ± 0.19 mA) were less than those used in the computational model (30 mm and 1 mA), we cannot completely discount that some of our results may be via modulation of extra-vestibular structures.

In conclusion, our findings provide key information necessary for the future development of GVS techniques to induce robust and effective therapeutic effects in PD. Future research is warranted to confirm similar behavioral effects of GVS applied at frequencies beyond the assumed physiological ranges and to establish potential mechanisms.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Clinical Research Ethics Board at the University of British Columbia. The patients/participants provided their written informed consent to participate in this study.

SL designed and conducted the study, including patient recruitment, data collection, and data analysis. MJM participated in the study design, data analysis, supervision, and funding acquisition. All authors have participated in manuscript writing and editing and approved the final manuscript.

SL was supported by Rina M. Bidin Foundation Fellowship in Research of Brain Treatment. WL was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. 2021R1C1C1009436). MJM was supported by John Nichol Chair in Parkinson's Research and the Canadian Institutes of Health Research (CIHR) grant (453374).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.