Evidence-based diagnostic use of VEMPs

Each cycle of a sound wave (ACS or BCV) is able to create a pressure wave—and thus a change in acceleration—in the endolymph surrounding the otolith maculae (for details see S2 [online supplementary material]). Accordingly, a stimulus frequency of 500 Hz is equivalent to 500 changes in acceleration per second, making it an ideal “jerk” stimulus for the transient vestibular system [21]. The conventional “accelerometer” mode of the otolith organs (Fig. 3a) would, however, be too sluggish to encode this high-frequency acceleration with adequate temporal precision. Instead, the utricle and the saccule work as seismometers in this situation—similar to the mechanism by which technical seismometers detect earthquakes (see Infobox 1 for further information about the mode of action of a seismometer): the neuro-epithelium oscillates with the stimulus frequency, while the otoconia remain stationary because of their inertia (Fig. 3b; [34]) resulting in a relative motion between the otoconial membrane and the neuro-epithelium, albeit with reversed roles of the two layers as compared to the accelerometer mode. Finally, this relative motion causes deflection of the hair cell stereocilia and, thus, hair cell depolarization (see above and [21]).

Experimental evidence for the “seismometer” mode of otolith organs derives from a sophisticated guinea pig model showing that application of sound or vibration to the guinea pig’s skull resulted in periodic oscillations of the utricular macula and de-/repolarizations of the utricular hair cells in sync with the stimulus frequency up to several kHz (for details see S3 [online supplementary material] and [65, 66]).

It is the dual mode of action as accelerometers and seismometers that allows the otolith organs to encode both constant/low-frequency linear acceleration and high-frequency sound and vibration. The 500 Hz ACS or BCV stimulus applied in clinical VEMP testing is ideally suited to activate the transient (= dynamic) otolithic system [21].

How does sound- and vibration-induced depolarization of vestibular hair cells activate vestibular afferents? And which types of afferents: otolith or semicircular canal? Those with irregular or regular resting discharge? The answers to these questions were mainly found by performing extracellular recordings of primary vestibular neurons in Scarpa’s ganglion (SG) during application of ACS or BCV to a guinea pig’s skull (Fig. 2d). The measurement set-up, in particular the stimulus parameters, were chosen to match the conditions during VEMP testing in humans [10, 13, 15, 18, 21, 22].

These experiments revealed that 500 Hz sound and vibration selectively activate type I vestibular hair cells in the utricular and saccular striola and their postsynaptic vestibular afferents with irregular resting discharge (Fig. 4). The combination of type I vestibular hair cells, ultrafast calyx synapses, and irregular otolith afferents is ideally suited for a stimulus- and phase-locked coding of these dynamic vestibular stimuli (500 changes in acceleration per second) with high temporal precision (see above and S1, S2 [online supplementary material] for more details).

In contrast, the resting activity of semicircular canal (both regular and irregular) and regular otolith afferents is virtually not altered during application of 500 Hz ACS or BCV at stimulus levels usually employed for clinical VEMP testing. Most probably, the sound- and vibration-induced endolymph flow in the semicircular canals is too weak to deflect hair cell stereocilia in the canal crista as long as the bony labyrinth is intact [75]. This may however change when a third mobile window is introduced into the bony wall of the inner ear in addition to the round and oval windows (see section “Superior canal dehiscence” and S4 [online supplemental material] for details).

Short ACS and BCV stimuli can be used for recording both o‑ and cVEMPs (for detailed stimulus parameters see below and [63, 68, 74]). A stimulus frequency of 500 Hz is generally recommended for most clinical applications because most of the neurophysiological data in the guinea pig model were obtained for 500 Hz sound or vibration [21, 74]. Please note that stimulus frequencies <500 Hz are able to activate irregular semicircular canal in addition to irregular otolith afferents—even in the normally encased bony labyrinth. As this might compromise the otolithic specificity of VEMPs, these frequencies are not recommended for VEMP testing in clinical practice [19].

Beside sound and vibration, galvanic currents applied to the mastoid via large surface electrodes can be employed as VEMP stimuli (galvanic or gVEMPs). So far, this technique has mainly been applied in experimental settings, and not in clinical routine [28].

Air-conducted sound: ACS (clicks or tone bursts) is presented to both ears sequentially with calibrated headphones or insert phones, while the ipsilateral cVEMP or the contralateral oVEMP response is recorded (Fig. 5). The test subject has to be protected from excessive sound exposure during presentation of ACS stimuli. When using ACS, the test subject has to be protected from excessive sound exposure.

To achieve this aim, a careful calibration of the sound source is of paramount importance. As a rule of thumb, a sound pressure level (SPL) of 130 dB(A) applied for 1 s is regarded to be safe. The peak (p) SPL should not exceed 140 dB(A). Detailed information concerning the calculation of sound exposure and prevention of noise trauma has been provided by [7, 8, 68, 74]. If the sound pressure level is indicated in “dB nHL” (normal hearing level) by the manufacturer of the VEMP platform, it has to be converted into “dB pSPL” or “pe (peak equivalent) SPL” in order to calculate safe sound exposure levels. The individual conversion factor depends on multiple variables, such as headphone type, stimulus parameters (see below), and ear canal volume [39, 68]. Conversion factors for the most common test conditions are often provided in the manufacturer’s instructions (e.g., [38]), otherwise they have to be determined during calibration of the sound source.

The use of ACS stimuli requires functional integrity of the middle ear.

As a rule of thumb, a 500 Hz ACS tone burst of 95–100 dB nHL is required to obtain stable cVEMPs in a healthy subject, while 5–10 dB more are needed for oVEMPs, reflecting the higher sensitivity of irregular saccular as compared to utricular afferents to ACS in the guinea pig model [18, 72, 74].

Finally, the neurotologist should always be aware that the use of ACS stimuli requires intact middle ear function. An air–bone gap as small as 10 dB is able to absorb so much sound energy in the middle ear that the remaining energy finally reaching the labyrinth might be too small to evoke reliable VEMP responses [86].

Bone-conducted vibration: BCV is usually applied to the midline of the forehead at the hairline (Fz; Fig. 5a), from where the vibration propagates through the skull and causes simultaneous linear acceleration in both labyrinths with approximately equal intensity. Therefore, a time-saving simultaneous recording of right- and left-sided VEMP responses is possible in this set-up [42]. Furthermore, BCV does not put the test subject at risk of excessive sound exposure, as irregular otolith afferents are much more sensitive to BCV (thresholds of ~0.02 g) than ACS (thresholds >80 dB SPL) [21]. Vibration can also be applied in patients with conductive hearing loss, as the sound energy is transported to the labyrinth via bone and connective tissue, and not through the middle ear.

A conventional bone-conduction device (e.g., RadioEar B71) is usually too weak for application at Fz; therefore, more powerful electromechanical vibrators are recommended (e.g., minishaker type 4810, Bruel and Kjaer) along with an appropriate amplifier (e.g., power amplifier type 2718, Bruel and Kjaer) [41, 42]. It should however be noted that the latter are not certified medical devices; therefore, their use is currently restricted to scientific applications. Tapping Fz with a standard reflex hammer (which is a medical device) is a cost-effective alternative for delivering BCV to Fz in clinical practice (for technical details see [37, 41, 49]).

The amplitude of a VEMP response is not determined by the total duration of the stimulus, but by its abrupt start (“jerk”; [3, 48]). The latter is the adequate stimulus for activating the transient channel of otolithic function (see above). Therefore, short stimuli (up to 6 ms) with short rise times (0–2 ms) are generally recommended for evoking o‑ and cVEMPs in clinical practice. In general, 50–200 stimuli at a repetition rate of 5 pulses per second (pps) are required to obtain robust VEMP responses [74]. When using the reflex hammer, 10 taps to Fz are usually sufficient to evoke a stable signal, as this is an ideal “jerk” with a rise time of 0 ms [17, 37].

For recording of oVEMPs, the active electrode is placed on the skin covering the inferior oblique muscle, i.e., on the infraorbital rim in one line with the pupil, the reference electrode is located 1–2 cm further below (Fig. 5; [42]). The patient is asked to look upwards (“look at your forehead”, Fig. 5a) during the recording as this increases the relatively small oVEMP n10 amplitude (5–10 µV). This effect is caused by a combination of bringing the belly of the inferior oblique muscle closer to the recording electrode on the skin and increasing the tonic activity of the muscle during upward gaze [73].

The active electrode for recording cVEMPs is placed on the skin above the mid-third of the anterior arm of the SCM, the reference electrode is located either on the medial end of the clavicula (Fig. 5b) or on the sternum (in case only one reference electrode is used for the two sides). The ground electrode for both o‑ and cVEMPs can be fixed to the chin or sternum. A certain baseline activity of the SCM is required for recording cVEMPs, as these are inhibitory responses. To this end, the test subject is asked to lift their head straight ahead while lying in a semi-recumbent position on the examination table. If the baseline activity of the SCM is too low with this maneuver, the test subject is asked to lift the head and turn it to the contralateral side [74].

Amplitudes and latencies of the oVEMP n10 and the cVEMP p13n23 responses are the most important outcome parameters in clinical practice. For a better comparison between right- and left-sided amplitudes, the asymmetry ratio (AR) is calculated:

An AR value >40% indicates an asymmetry between right- and left-sided dynamic otolithic function [42, 43, 86].

The “raw” cVEMP p13n23 amplitude (in µV) correlates linearly with the baseline activity of the SCM (in µV) [5]. Therefore, it is recommended to calculate a dimensionless “corrected” p13n23 amplitude: the raw amplitude (obtained from the unrectified averaged trace of the electromyograhic [EMG] recording) is divided by the baseline activity of the SCM. The latter is determined as the average amplitude of the rectified or root mean square (RMS) EMG recorded in parallel to the unrectified EMG in the interval between −20 ms and stimulus onset (0 ms). For details, see Figs. 5 and 6 in [74]. The corrected p13n23 amplitude allows for a better comparison of cVEMP responses recorded from different sides, in different individuals and during different recording sessions.

Normal values for VEMPs depend on many variables, such as the test subject’s age and various stimulus parameters (e.g., frequency, intensity, duration, rise times) [1, 3, 60, 67, 72, 85]. Therefore, it is strongly recommended that each neurotological center defines its own normal values for each set of stimulus parameters and different age groups [74].

In line with the afferent innervation of the labyrinth (Fig. 2d), patients with a superior vestibular neuritis display reduced oVEMP n10 amplitudes on the contralateral side (Fig. 1e), while right- and left-sided cVEMPs are symmetrical. On the other hand, reduced ipsilateral cVEMP p13n23 amplitudes and symmetrical oVEMP n10 responses are a hallmark of inferior vestibular neuritis (Fig. 1f; [12, 33, 44, 51, 52, 56, 61, 64, 82]). This “double dissociation” of o‑ and c‑VEMPs in superior versus inferior vestibular neuritis has two important implications. First, it indicates that the two otolith-driven reflexes do not originate in the same subset of vestibular receptors. If they did, one would expect that either both responses or none at all would be affected in superior/inferior neuritis. Second, the double dissociation shows that the predominant specificity of oVEMPs for contralateral utricular function and cVEMPs for ipsilateral saccular function is sufficient to distinguish between utricular and saccular dysfunction in clinical practice—although the neuronal projections of utricular and saccular afferents are not completely separated [11].

VEMPs are an efficient and reliable tool for detecting isolated disorders of the otolith organs [16, 30]. If vestibular testing assesses only semicircular canal function (e.g., by calorics or vHIT), these disorders may easily been missed and classified as “non-vestibular” or “functional”. Therefore, the results of semicircular canal testing have to be complemented by selective tests of otolithic function such as VEMPs in clinical practice.

This is especially important in patients with mild traumatic brain injury or blast trauma who often suffer from isolated dysfunction of the otolith organs despite intact semicircular canal function [2, 47].

A dehiscence in the bony roof of the superior semicircular canal creates a third mobile window in the bony labyrinth (beside the oval and round windows) resulting, e.g., in sound-induced vertigo and nystagmus (Tullio phenomenon; for details see Infobox 1​ and S4 [online supplementary material]; [35, 83]).

A highly increased oVEMP n10 response recorded from the inferior oblique muscle is a powerful indicator for a superior canal dehiscence (SCD) in the contralateral labyrinth with a sensitivity and specificity >90% depending on stimulus parameters and control populations (Fig. 6a; [45, 54, 80, 88]). Furthermore, patients with an SCD display a contralateral oVEMP n10 response to 4000 Hz ACS or BCV, which is absent in normal subjects with an intact bony labyrinth (Fig. 6b; diagnostic accuracy = 100%). As the latter test does not depend on comparison between right- and left-sided responses, it is ideally suited to diagnose bilateral SCD [55].

The tremendous effect of SCD on the oVEMP n10 amplitude can be explained by activation of the otherwise “silent” vestibulo-ocular projections of superior canal afferents (see section “Neuronal projections”). While semicircular canal hair cells and afferents do not respond to sound or vibration in the normally encased labyrinth, the third mobile window in SCD induces an endolymph flow in the superior semicircular canal [75] finally resulting in activation of type I vestibular hair cells and their postsynaptic irregular afferents. As these project to vestibulo-ocular neurons—just like utricular afferents—they increase the oVEMP n10 amplitude recorded from the inferior oblique eye muscle [4, 27, 79].

Following closure of the superior canal dehiscence in a guinea pig model of SCD, superior semicircular canal afferents do no longer respond to ACS and BCV, which is in line with the normalization of VEMP responses after successful surgical closure of SCD in humans [27, 87].

Most neurotological tests, such as caloric irrigation and vHIT, infer vestibular (dys‑)function from the video-oculographic (VOG) assessment of the vestibulo-ocular reflex, or put simply: you have to look into your patients’ eyes to determine their balance function. This may, however, be difficult or even impossible in certain groups of patients, e.g., those with congenital/infantile nystagmus (superimposition of the VOG recording by the infantile nystagmus), extremely poor vision (blind nystagmus; inability to fixate a target during vHIT), or reduced ability to cooperate (e.g., blink and lid artifacts). In these cases, o‑ and cVEMPs offer a simple, quick, and reliable alternative to obtain some basic information about superior (oVEMPs) and inferior vestibular nerve (cVEMPs) function (Fig. 2d; [14]). Both responses can be recorded in patients with infantile nystagmus [53]. Due to the short duration of the examination (40 s for a trial with 200 stimuli at 5 pps), VEMPs are also a suitable diagnostic tool for patients with limited ability to cooperate, e.g., children and patients with cognitive impairment.