Recordability of the vestibular cerebellar evoked potential

Abstract

OBJECTIVE:

Vestibular cerebellar evoked potentials (VsCEPs) are short-latency, vestibular-dependent responses recorded from electrodes placed superficially near the cerebellum. This study aimed to replicate previous findings and further explore optimal recording parameters for air-conducted VsCEPs.

DESIGN:

Responses were collected similarly to previous investigations for comparability. Two reference electrode locations (nose and contralateral earlobe) were compared.

STUDY SAMPLE:

VsCEPs were measured in 15 healthy individuals without prior history of otologic or neurological impairment (mean age = 24.7; sd = 3.8, 2 male).

RESULTS:

We report a response rate that never exceeded 60%. Responses were most often present over the posterior scalp contralateral to the stimulated ear. A small but significant latency difference was found between reference locations. Larger amplitudes were observed over midline and contralateral areas for the left ear only. There was no difference in amplitude between reference locations.

CONCLUSIONS:

The low recordability and high inter-subject variability suggests VsCEPs are not yet suitable for clinical application. However, results should be treated with caution as the stimulus level used in the current study was lower than previous investigations. Thus, we may be below the response threshold of subjects. Future investigations should explore other parameters to improve recordability and stability of VsCEPs.

1. Introduction

Vestibular evoked myogenic potentials (VEMPs) are clinical tests that measure function of the saccule (cervical VEMP or cVEMP) and utricle (ocular VEMP or oVEMP) and can be elicited using brief intense sounds or bursts of vibration (Colebatch et al., 1994; Todd et al., 2007). These stimuli provoke a muscle reflex that can be detected with surface electrodes placed superficial to the belly of the ipsilateral sternocleidomastoid muscle (i.e., cVEMP) or superficial to the contralateral inferior oblique muscle (i.e., oVEMP). The identification of stable and clinically practicable recording techniques has led clinicians to incorporate both the cVEMP and oVEMP into routine vestibular test batteries used to assess patients presenting with dizziness and/or unsteadiness.

In addition to cervical and ocular areas, authors have speculated that evoked potentials of vestibular origin can also generate and be measured from subcortical and cortical sources as well (Todd et al., 2008). For example, it has been reported that stimulation of the otolith end organ results in activation of the cerebellum (Buttner-Ennevera, J.A., 1999; Buttner, U. et al., 1999). In order to maintain balance, the cerebellum must perform multisensory integration of information from the visual, somatosensory, and vestibular sensory systems (Morton et al., 2004). The cerebellum also initiates and monitors a number of central processes that include compensation when the vestibular system becomes unilaterally impaired (Igarashi et al., 1983; Dieringer & Precht, 1977).

When appropriate stimuli are utilized, surface electrodes placed proximal to the occipital protuberance (i.e., the “inion”), can enable investigators to record a novel short latency response referred to as the vestibular cerebellar evoked potential or VsCEP (Todd et al., 2017). A number of investigators have been instrumental in the definition of optimal stimulus and recording characteristics (e.g., Todd et al., 2018a, 2018b; Govender et al. 2020). The researchers who first recorded the VsCEP contended that it was a neurogenic evoked potential (Todd et al., 2017). However, early experiments revealed that surface electrodes placed at the inion could be used to record sound evoked myogenic potentials (Bickford et al., 1964; Todd et al., 2017). Bickford described this potential as the inion response, consisting of four main positive and negative deflections between 10 and 50 msec post-stimulus onset. This response has shown to have a relatively high response rate in healthy controls and hypothesized to be primarily driven by the otolith organs (Townsend and Cody, 1971). In addition to the inion response, other myogenic responses have been documented in this region including the postauricular, splenius capitis, and median neck extensor muscles (Bickford et al., 1964; Gulec et al., 2013; Rosengren et al., 2019; Sakakura et al., 2005; Townsend and Cody, 1971). Thus, sound evoked myogenic responses recorded from posterior neck muscles could demonstrate latencies that overlap those of the VsCEP. As a result, a stimulus that initiates a cVEMP or oVEMP could also concurrently trigger multiple myogenic responses with overlapping epochs, a concept that has been proposed by other investigators (Deriu et al., 2010; Todd et al., 2017).

There have been only a handful of contemporary reports of this evoked potential. These seminal investigations have focused on attempts to determine whether the response is neurogenic, myogenic, or a combination of both (Todd et al., 2018a, 2018b, 2019; Govender et al., 2020). In support of a neurogenic origin, the VsCEP is measurable when posterior neck muscles are relaxed which is a recording condition not compatible with the recording of VEMPs. In addition, there is evidence that VsCEP amplitude modulates when recorded concurrently with an optokinetic stimulus, leading authors to speculate that the processes underpinning this evoked potential originate from the cerebellum due to its fundamental role in visual-vestibular integration (e.g., Todd et al., 2018a, 2018b). While noteworthy, these reports have shown that the VsCEP has a peak-to-peak amplitude much greater than one might expect for a neurogenic (sensory) evoked response (i.e., amplitudes up to 60 microvolts have been reported in some individuals). Following sensory stimulation, amplitudes of that magnitude are uncommon. Examples would include component N18 of the cortical somatosensory evoked potential used in patients with myoclonic epilepsy or pattern reversal associated with visual cortical evoked potentials (Rothwell et al., 1984).

The contemporary VsCEP literature has also explored a number of stimulus and recording parameters. For example, previous investigations have elicited this response with using air- and bone-conduction, stimulus levels and profiles. (Todd et al., 2017; Todd et al., 2018b; Govender et al., 2020). However, the majority of reports demonstrated a mixed response rate averaging around 50 – 70%.

The anatomic origins of the VsCEP remain elusive and the fundamental characteristics for recording the VsCEP still require clarification. As such, the objectives of this study were two-fold: 1) to examine the reproducibility of the VsCEP which is an area of limited prior research, and 2) to compare two reference electrode locations in an effort to optimize VsCEP recording parameters.

2. Methods

2.1. Participants

The current study protocol was approved by the university’s institutional review board (IRB #212012). All participants consented prior to enrollment in the current study. Fifteen adult participants (mean age = 24.7; sd = 3.8; 2 males) younger than 40 years of age were enrolled into the study protocol. Participants were only included if they reported no significant history of hearing, balance, or known neurological conditions.

2.2. Stimulus

A 500 Hz Blackman-gated tone burst (2-0-2 msec) was presented monaurally at 125 dB peak-sound pressure level (dB pSPL) with an ER3A insert earphones to the left and right ears of every participant. The stimulus rate was presented at 5.45 per second. The stimuli and presentation level were similar to those used during clinical VEMP testing.

2.3. Procedure

All data were collected using Neuroscan SCAN software (Version 4.5). For all VsCEP recordings, bandpass filters were set to 5 – 500 Hz (comparable to Todd et al., 2018) with an averaging window of −20 to 80 msec and a sampling rate of 20 kHz. A notch filter at 60 Hz was applied to reduce electrical interference. Each evoked potential consisted of a minimum of 130 averaged sweeps. All conditions were randomized to reduce any order effect. Each signal averaged waveform was replicated once (two total waveforms per condition) so that an assessment of waveform replicability could be obtained.

2.4. VsCEP

A mild abrasive gel was used to reduce skin impedance at the electrode recording sites using conventional electrode preparation techniques. Ten Ag/AgCl electrodes were used for recording. Figure 1 shows the non-inverting electrode sites which were identical to those reported by Govender et al. (2020). Methods for identifying the inion and placing an electrode at this position (lz) are described below using the 10-20 landmarks. A second electrode (CBz) was placed at midline (5% of the nasion-inion distance) inferior Iz (approximately 2 cm). Using 10% of each participant’s nasion-inion semi-hemispheric distance, additional electrodes were placed laterally left and right of midline in a 2x2 electrode grid, each spaced approximately 3 cm from the other. Figure 1 shows the locations of all non-inverting electrodes including Iz, CBz, Iz-3, Iz-6, CBz-3, CBz-6, Iz+3, Iz+6, CBz+3, and CBz+6 with negative numbers indicating electrode sites to the left of midline and positive numbers indicating locations to the right of midline. The non-inverting (“reference electrode”) inputs were placed on the earlobe opposite to the stimulated ear (CA) and on the tip of the nose (N). A ground electrode was placed at the forehead midline (i.e., Fpz).

Figure 1.

Montage for active electrodes. In each row, electrodes were separated by 10% skull hemi-circumference measured from Iz to Nz, approximately 3 cm. Top and bottom rows were separated by 5% Iz to Nz distance, approximately 2 cm.

Each participant was semi-recumbent in a comfortable recliner chair, with the head and neck supported with a pillow. This position was chosen to record the neural response and to minimize any activation of myogenic contributions to the response. Either the left or right insert earphone was placed into the participant’s ears. The participant was asked to focus on a one-inch neutral visual target (1 meter in front of their direct line of sight) while acoustical stimuli were presented through the earphones. A total of four runs were collected for each reference electrode site (2 runs for the left ear and 2 runs for the right ear, each 1 – 1.5 minutes, 8 – 12 minutes total).

2.5. Analysis

An expert examiner (DR) identified all evoked responses using a custom MATLAB program. The presence or absence of a response was determined by visually identifying peaks 10-12 msec after stimulus onset and 15-17 msec after stimulus onset, p1 and n1 respectively (Govender et al., 2020). A response was considered present if both peaks (i.e., p1 and n1) were visualized in the first run and its replication. We tabulated for analysis p1 latencies and p1-n1 peak-to-peak amplitudes (peak to trough).

All analyses were conducted in R statistical language for computing (v 4.3.0; R Core Team 2023). Statistical significance was defined as p < .01 for all analyses. To explore response rates, we used chi-squared tests to evaluate the effects of fixed factors (ear, reference site, recording electrode site, and participant) on frequency of missing responses. Each factor was analyzed separately using the chisq.test in base R.

To explore the effects of stimulus ear and recording parameters, latencies and amplitudes were analyzed formally, each as the dependent variable in a linear-mixed effects model. Fixed factors included ear (left/right), reference site (contralateral earlobe, tip of the noise), and recording electrode site. Recording electrode sites are described relative to the ear being stimulated, and were defined as being contralateral, ipsilateral, or midline. There were 4 contralateral recording electrode sites, 4 ipsilateral recording electrode sites, and 2 midline recording electrode sites (Figure 1). Participant was entered as a random intercept. Models were constructed using the lmer function of the lme4 package (Bases et al., 2015) and were analyzed using the anova function in base R. For follow-up testing, significant main effects and interactions were explored using the emmeans function of the emmeans package (Lenth 2019), adjusting for family-wise error rate (Benjamini & Hochberg, 1995). Assumptions underlying linear-mixed effects modelling were evaluated with the check_model function of the performance package (Lüdecke et al., 2021).

3. Results

3.1. VsCEP waveforms

All visually identified repeatable waveforms are described below for each condition which includes 10 non-inverting (recording) electrode sites, two inverting electrode locations (i.e., earlobe and nose reference), and a single ground electrode. Grand average waveforms responses using the contralateral earlobe reference across each ear are shown (Figure 2).

Figure 2.

Grand average waveforms obtained using the earlobe contralateral to the stimulated ear as a reference (n = 9). The ear stimulated is indicated by color (red = right, blue = left). The shaded region indicates the standard deviation for responses obtained. Electrode locations are based on their individual position on the scalp.

3.2. Response rate

For the response to be considered “present”, two peaks (p1 and n1) had to be visually detected (see section 2.5). Across all conditions (electrode sites, reference locations, and ears), the response rate never exceeded 60% (9/15) in any condition. Figure 3 reveals the response rate and missing data rate for each fixed factor (recording electrode site, reference site, ear, and participant). Analysis revealed significant effects of electrode site (χ²(9) = 53.21, < .001) and participant (χ²[14] = 352.94, < .001). Specifically, some participants had higher response rates than other participants; response rates ranged from 0 to 100% across participants. In addition, contralateral recording electrode sites had higher response rates (~64%) than did the ipsilateral (<30%) or midline electrode sites (~58%). There was no effect on response rates of reference site (χ²[1] = 1.54, p = .21) or ear (χ²[1] = 2.48, p = .12), suggesting reference site and test ear did not contribute to missingness, but electrode site and participant did contribute to missingness in response rates.

Figure 3.

Number of responses that were deemed to be missing (grey) or present (black) based on electrode location (top left panel), reference site (top right panel), stimulated ear (bottom left panel), and participant (bottom right panel). Analysis revealed response rates varied significantly with electrode site and participant, but did not vary significantly with reference site or stimulated ear.

3.3. Latency measures

VsCEP p1 latency measures consisted of an average p1 latency of 11.92 msec [sd = 0.77] with a range of 10.4 to 13.3 across all conditions (see supplemental digital content 1). The model met assumptions necessary for analysis. Results revealed a significant main effect of reference site (F[1, 206] = 89.03, p < .0001), but no other significant main effects or interactions (all effects p > .01). Follow-up analysis revealed that response latency was prolonged with the nose reference site (estimated marginal mean = 12.4 msec, standard error = 0.11) compared to the contralateral earlobe reference site (estimated marginal mean = 11.5 msec, standard error = 0.11; estimated marginal mean difference = 0.85 ms; t(206) = 9.41, p < .0001). These findings reveal that, as expected, p1 latency did not differ across electrode site, or ear. Furthermore, the effect of reference site, although statistically significant, was small (<1 msec).

3.4. Amplitude measures

VsCEP amplitude relative to the stimulated ear is shown in Table 1 and Figure 4 for all recording sites and reference electrode locations. The model of amplitude measures violated the assumption of residual normality (Schielzeth et al., 2020). However, analysis proceeded because linear mixed-effects models are robust to this type of violation. Analysis revealed a significant effect of electrode (F[9, 201] = 6.34, p < .0001) and of ear (F[1, 202] = 24.29, p < .00001), in addition to a significant electrode by ear interaction (F[9, 201] = 3.13, p < .005). There was no significant main effect of reference site and no other significant interactions (all effects p > .01).

Table 1.

Comparing differences in amplitude between ear stimulated, electrode site for each ear, and reference site (CA = contralateral ear, N = nose).

Condition	Mean (μV)	SD	Variance	Median (μV)
Right Ipsi	7.5	±4.3	18.7	5.9
Left Ipsi	9.9	±5.2	26.9	9.4
Right Contra	13.7	±8.7	74.8	13.1
Left Contra	26.1	±22.5	504.7	17.3
Right Midline	10.2	±6.6	43.5	8.6
Left Midline	17.1	±14.0	195.3	14.4
CA Reference	18.9	±19.4	376.9	11.6
N Reference	21.9	±17.4	302.2	16.8

Figure 4.

VsCEP P1-N1 amplitudes for each relative electrode location and stimulation ear. Amplitude organized with electrode location based on either ipsilateral (Ipsi) or contralateral (contra) evoked pathways. This includes both nose and contralateral ear reference locations.

Results of the follow-up analyses of the electrode by ear interaction are displayed in Supplemental Digital Content 2. In brief, analysis revealed no differences in amplitude between electrode sites for stimuli presented in the right ear (all comparisons p > .80). However, in the left ear, response amplitudes were larger for contralateral electrode sites compared to midline or ipsilateral electrodes (all comparisons p < .01). Another set of follow-up analyses, displayed in Supplemental Digital Content 3, also revealed that amplitude responses were larger for the left ear than the right ear, but only for contralateral electrode sites (p < .01) and none of the ipsilateral (p > .60) or midline locations (p > .01). Combined, these data demonstrate that amplitudes were larger at midline and at electrode sites contralateral to the stimulated ear, but only for the left ear (Figure 4). In addition, with stimulation to the left ear, amplitudes were larger when compared to the right ear, but only at the contralateral electrodes. Amplitude measurements did not vary between, or interact with, reference electrode locations (contralateral earlobe and nose).

4. Discussion

As part of our study, we sought to: 1) examine the reproducibility of previously reported findings for the air-conducted VsCEP, and 2) compare two reference electrode locations in an effort to optimize VsCEP recording parameters. Similar to previous findings, we found that our subjects never demonstrated more than a 60% response rate and VsCEPs were more likely to be present over the posterior scalp contralateral to the stimulated ear. Additionally, a small, yet significant, difference was observed for p1 latency between reference electrode locations. Further, VsCEP amplitudes were larger at electrode sites contralateral to left ear stimulation suggesting a lateralization of responses in some subjects. Authors found no differences in peak-to-peak amplitude across the two reference locations.

4.1. VsCEP best recorded over electrode sites midline and contralateral to the stimulated ear

We observed the highest response rates over the midline and contralateral electrode sites. These data are consistent with previously reported findings (Govender et al., 2020). Taken together, we found that regardless of which reference location was used, present responses were more likely to be recorded over midline and contralateral electrode sites. The results of this study suggest that regardless of the origin of the VsCEP, the pathways underpinning this response primarily favor a midline and contralateral recording electrode montage.

Furthermore, it must be acknowledged that this study’s stimulus level was lower than previously studies (Govender et al., 2020; Todd et al., 2017). Thus, we cannot rule out that the stimulus level chosen for this study (125 dB pSPL) was below threshold for some subjects, which could have contributed to the lack of present responses. However, previous investigations have shown similar response rates (50 – 70%) despite using a more intense stimulus (Govender et al., 2020; Todd et al., 2017).

4.2. The VsCEP response amplitude demonstrated a left ear preference

The average amplitude measurements with left ear stimulation appeared to be larger than right ear stimulation for contralateral electrode sites. While this finding supports lateralization of the response found in previous studies, it should be noted that the laterality is reversed. For example, the data from Govender et al., 2020 support a right ear preference. It is not clear what is contributing to this difference in lateralities across studies. It is possible that participant handedness could be contributing to the variability across studies since it is known that vestibular-driven activation of cortical pathways favors the side that aligns with handedness (Dieterich et al., 2003). Despite the reversed laterality, one commonality between the current data and published data is the strength of the contralateral response. Interestingly, previous animal models found that amplitudes of cerebellar evoked potentials in rats were also larger in the contralateral condition suggesting that cerebellar sensory pathways cross from the pontine reticular nuclei into the cerebellum (Lorenzo et al., 1977).

4.3. Differences between reference location (contralateral earlobe and tip of the nose)

Between the two reference sites used in this study (earlobe and nose), there were no clear differences found in VsCEP response rate or peak-to-peak amplitude (Figure 3). While nose reference latencies were significantly longer than when using contralateral earlobe, it was a small effect (<1 msec). Nonetheless, it is possible that the signals recorded from both reference locations were different. For example, a reference electrode that is too far away from the source (in this case, the nose) could have amplified unwanted signals which may have affected p1 latency by introducing phase shifts and reduced common-mode rejection (Dumitru et al., 2024). That said, the reference electrode site did not affect the ability to measure the VsCEP.

4.4. Study limitations, considerations, and future directions

The most obvious concern with clinical use of VsCEP is that our investigation is consistent with previous studies yielding present responses 50% – 70% of the time in participants with normal vestibular function when using air-conduction stimuli (Todd et al., 2017; Todd et al., 2018b; Govender et al., 2020). Furthermore, there were several instances where two peaks were observed and other times only one peak. This study and previous investigations raise questions regarding clinical utility of the VsCEP. It is unclear what the significance of these observations is in the broader context of VsCEP literature.

Additionally, it must be acknowledged that previous investigations have used different stimuli including bone conduction stimulus, which is generally accepted to result in more efficient stimulation of the otolith end organs (Rosengren et al., 2019). Moreover, previous VsCEP studies utilizing air-conduction used higher intensities ranging from 130-138 dB peak SPL (Govender et al., 2020; Todd et al., 2017). Although the response rate found in this study was similar to previously reported findings, we cannot rule out that stimulus level did not affect our ability to record the VsCEP.

The search for the origin of the VsCEP is ongoing. As mentioned previously, myogenic responses (e.g., inion potentials) are recorded near and/or overlap with the recording sites used in this study. Similar to previous investigations, the peak-to-peak amplitudes observed in this study were larger than one might expect from a neurogenic response, and while there is emerging evidence (through source analysis) to suggest the VsCEP localizes to the cerebellum (Todd et al., 2021; Govender et. al., 2024), this cannot be confirmed until findings are replicated across multiple laboratories and/or using animal models.

Lastly, should alterations in VsCEP be the direct result of cerebellar changes, these findings would support a novel evoked potential that reflects higher level processing of vestibular sensory information. In fact, there is growing support to move beyond traditional electrophysiologic measures (e.g., myogenic potentials) to further expand our understanding subcortical and cortical vestibular processing.

5. Conclusions

We analyzed the recording of vestibular cerebellar evoked potentials in a group of young adults. According to our findings, VsCEPs were best recorded at midline and contralateral recording sites with equal recordability when the contralateral earlobe or nose tip were used. Left ear stimulation resulted in larger amplitudes compared to the right ear for the contralateral recording electrode sites. Finally, we have shown that while the VsCEP is recordable, a number of issues regarding its recording capability and origin must still be investigated before VsCEP can be clinically used in patients with cerebellar lesions.

Data Sharing Policy:

We will honor any requests for data access for nonprofit research purposes.