|Figure 1: Schematic of the utricle and saccule. These sensory organs in the inner ear primarily respond to linear acceleration such as due to orientation to gravity, but the saccule and utricle are also somewhat sensitive to sound. This is the basis of the VEMP test.|
We will use the terminology "cVEMP" to denote vestibular evoked myogenic potentials elicited from the sternocleidomastoid muscle.
Other types of VEMPs are oVEMP, tVEMP, etc. When we use the terms "oVEMP" or tVEMP or whatever, the small letter indicates that a muscle other than the SCM is being monitored - - such as ocular or triceps. When we use the unqualified "VEMP", we mean any vestibular evoked myogenic potential (i.e. cVEMP, oVEMP, tVEMP, etc). "sVEMP" might have been a better name for the SCM VEMP, as there are many muscles in the cervical spine area, but it is too late now.
The cVEMP test is thought to determine if the saccule, one portion of the otoliths, as well the inferior vestibular nerve and central connections, are intact and working normally. Both of the otolith organs have a slight sound sensitivity and this can be measured. This sensitivity is thought to be a remnant from the otolith organs use as an organ of hearing in lower animals.
A recent review article regarding cervical and ocular VEMP testing was published by Fife et al (2017).
|Figure 2. cVEMP circuitry. Sound stimulates the saccule, which activates the inferior vestibular nerve, lateral vestibular nucleus, 11th nerve nucleus, and then the sternocleidomastoid muscle (mostly ipsilaterally).|
In general, one needs to consider the input, central processing, and output for physiological responses. Things are a bit messy and not nearly as simple as suggested in some of the review articles on this topic.
The pathway supposed to account for the cVEMP response is shown in figure 2 above. Sound stimulates the saccule, traverses the vestibular nerve (mainly inferior, but a bit also in the superior) and vestibular ganglion to reach the vestibular nucleus in the brainstem. From there, impulses are sent to the neck muscles via the medial vestibulospinal tract (MVST), then the spinal accessory nucleus, and the accessory nerve. For most muscles, the net effect of saccule stimulation is inhibition, but excitation from electrical stimulation of the saccule has also been reported in some muscles in animals (Uchino et al., 1997; Kushiro et al., 1999).
While the clinical literature concerning cVEMPs suggest that they are almost entirely saccular in origin, the animal literature does not support this claim. Many studies of monkeys show that all 5 vestibular end organs phase lock to loud clicks (Xu et al, 2009). The utricle responds to sound as well (see ovemp page). Thus when you put in a loud noise, you are stimulating everything in the ear. Fife et al (2017) stated that "it is not known whether oVEMP/cVEMP responses accurately identify vestibular function specifically related to the saccule/utricle". We think this is fair, as of 2017. We have unproven claims.
Concerning the output circuitry, similarly there is no one:one connection between a particular vestibular endorgan and a particular muscle. According to Forbes et al (2018), "divergent pathways originating from one vestibular end organ branch to multiple neck motoneurons, while convergent pathways combine aﬀerent signals from multiple end organs prior to termination in the spinal cord ". In other words, everything is connected to everything.
It would also seem very possible that in certain disorders of the ear, such as SCD, or cholesteatoma, the semicircular canals might also be a source of cVEMPs. In these disorders, the canals become sensitive to sound, and sound can make the eyes move through activation of the semicircular canal. It seems very likely that it might also activate the neck as well as the vertical eye muscles (i.e. inferior rectus and inferior oblique, such as are measured with the oVEMP test).
According to Naranjo, VEMP amplitudes are increased by threat and fear (Naranjo et al, 2016) showing that there are also other inputs to consider. The implication is that cVEMPs are not simple saccule reflexes, but have multiple inputs.
So overall, things are very very messy. It is fairly reasonable to assume that the main source of the cVEMPs in normal people is the saccule. It is probably not correct to assume this is the case for patients with ear disease, and due to this, it is also not safe to assume that VEMP wiring is as simple as shown in figure 2 above in all cases.
According to Oh et al (2016), the cVEMPs are mediated by the vestibular nuclei and uncrossed medial vestibulospinal tract descending in the lower brainstem and spinal cord. Therefore, lesions involving the vestibular nuclei can present abnormalities of cVEMPs and medullary lesions involving the descending MLF or the spinal accessory nucleus impair cVEMPs.
Sound evoked cVEMPs recorded from the neck are claimed to be almost completely unilateral. (Colebach et al, 1994; Uchino et al, 1997; Kushiro et al, 2000; Murofushi et al, 1996; Wilson et al, 1995), but in clinical practice this can't be counted upon (see example below). It also is likely wrong when the semicircular canals are activated by sound as discussed above.
The output for cVEMP responses is by definition the sternocleidomastoid (SCM) muscle, which is innervated by the accessory (11th) cranial nerve. If either one of these structures were disturbed, one would expect alterations in the cVEMP as well. One would expect there might be many other muscles that activate the neck and are relevant to posture that are also activated by sound.
Additionally, when the semicircular canals are sensitive to sound, such as in SCD or some canal fistulae, it would also seem very likely that vertical eye muscles on both sides of the head might be activated by the VEMP protocol. Thus the ipsilateral wiring assumptions may be incorrect in certain ear diseases.
|Figure 3 : Equipment used to record a VEMP, a Bio-Logic Navigator Pro.|
VEMPs are recorded using an evoked response computer, a sound generator, and surface electrodes to pick up neck muscle activation or other muscles if this is of interest. Figure 3 above illustrates the basic rather minimal equipment needed. In the author's laboratory, a Bio-Logic Navigator Pro does nearly all of the work, and sends the results to a desktop computer.
cVEMP testing is not hard, but there are a lot of technical pitfalls. The basics can be learned by a technician in about 30 minutes. It is a very big response, and as long as the person doing the test is attentive to details (getting the sound in both ears with proper placement of inserts or headphones, having the person lift their head through the entire trial, electrodes in the right place with proper impedance), it is very straightforward. The details (see below) take much longer to learn.
The figure below illustrates a cVEMP test printout. The cVEMP response consists of an initial positivity (p1 or p13) followed by a negativity (n1 or n23), see figure 4 below. It is an evoked potential. Although P1 is positive, it is shown negative on many cVEMPs, because of electrode placement (basically putting them on backwards). The most reliable measure of the cVEMP response is the amplitude (Isaradisaikul et al, 2008).
Later cVEMP components have a lower stimulus threshold and are thought to be non vestibular. This is not well understood and we are frankly dubious about this idea - -they might simply be part of the same spike train. As VEMP's are easily elicited without the need for EMG rectification, and EMG spikes occur at roughly the same latency as the waves in a cVEMP, coherent spike trains are a reasonable alternative explanation. In other words, later waves may all be part of the same response. In our opinion, a VEMP system that does rectification is not necessary. To be fair though, there are some that differ from this opinion (Lee et al, 2008). Rectification can correct in part for technical errors involving head positioning, at the price of adding more "noise" to the entire system.
Higher than normal thresholds or low amplitudes may be found in persons with saccule disorders as well as conductive hearing loss. Reduced amplitudes are commonly found in vestibular nerve disturbances. Lower than normal thresholds as well as asymmetrical amplitudes are found in persons with Tullio's phenomenon, which is dizziness induced by sound. Prolonged latencies of P13 may be found in central disturbances (Murofushi et al, 2001), but practically these are very rarely encountered, and technical error is the source of most prolonged latencies.
|Normal cVEMP using a Bio-Logic Navigator Pro.. The main potential, P1, is located at about 13 msec. Each side is about 250 microvolt in size (which is far above the lower limit of normal, about 70). There is an electrical artifact seen at 0 msec for the left side. This can be ignored.|
The general rule of thumb with hearing and cVEMP's is that conductive hearing loss obliterates air conducted VEMP's, and that sensorineural hearing loss does little or nothing to VEMP's. The two plots below illustrate this.
|Figure: cVEMP obtained in an individual with a modest left sided conductive hearing loss, using a Bio-Logic Navigator Pro. The VEMP on the right was normal, and the VEMP on the left, entirely absent. P1 designates the potential that occurs at 13 msec (often called P13)|
|Figure: cVEMP obtained in an individual with a profound right sensorineural hearing loss, using a Bio-Logic Navigator Pro. This shows that (sensorineural) hearing is not necessary to obtain a VEMP. This recording was obtained using binaural stimulation (see comments regarding this method).||Audiogram in person with the cVEMP shown to the right. There is a profound hearing loss, presumably sensorineural.|
A potential pitfall in a person with a complete sensorineural hearing loss, is that one has no way of determining whether they also have also have a conductive component to their hearing loss. For example, someone might have far advanced otosclerosis. Thus, one could falsely conclude that there was no vestibular function on one side based on an absent VEMP. Bone VEMP's are one way to get around this problem.
Basic head positioning and electrode layout
Best practice, as shown below, is to apply EMG electrodes to the middle third of anterior neck muscles (sternocleidomastoids) and the supine patient holds their head up unsupported, using the anterior neck muscles. Subjects are instructed to tense the muscle during runs of acoustic stimulation, and relax between runs. If the neck muscles are not activated, no cVEMP is produced. The reflex scales to tonic EMG -- once again -- if you don't activate the neck muscle, you don't get a response. A corollary, is that if you get a response without neck muscle activation, it isn't a VEMP (maybe a PAM ? see below).
The EMG electrodes should not be the kind that use alligator clips to attach to foil, because the head moves around during the cVEMP. These types of electrodes generate massive amounts of noise with movement. Instead, electrodes should be used that are "hard wired" to the cable.
People with fat necks have lower responses, as the signal from the muscle is underneath the fat and has to go a longer distance. Similarly, people with long necks have later responses, as the signal again has to go a longer distance (Chang et al, 2007)
Some patients are unable to hold their heads up at the angle shown. In this case, some experts recommend simply tilting the entire body up by about 30 degrees, so that there is less torque needed by the patient to hold their head up (Colebatch, personal communication). We think this is an excellent idea, but at present there are no amplitude norms for this procedure.
|Figure: Recommended patient positioning for VEMP testing. During the test, the patient lies flat on their back, lifting the head off the table.|
Another (but bad) method of obtaining activation is to have patients sit upright with their chin turned over the contralateral shoulder to tense the SCM muscle. We think this is a bad idea because you have no way of knowing how much tension the patient is producing, it is tiring, and patients can do something different on the right than on the left, without you knowing. It is also well known that head position on body can change cVEMP responses. This adds another variable.
Because the response is generally ipsilateral (carefully note the qualification), one theoretically can use bilateral stimuli and bilateral recording to reduce the number of trials (Huang et al, 2006; Young, 2006). We tried this method in our clinical practice, but discarded it for reasons explained below. It has the advantage of using the same stimulus when recording each side, which reduces some of the considerable variability. The limits of normal for the amplitude with the head-raising technique are roughly 70 to 700. Regarding the upper limit, there is no disease that we know of that can be diagnosed by larger than normal cVEMPs on both sides, but in our clinical experience, we have seen this primarily in persons with hyperacusis.
|Monaural VEMP that shows the basic pitfall of binaural/bilateral recording. Because there is a substantial response from the contralateral side on the left, one could assume wrongly that there was a present VEMP even in a person with a vestibular nerve section. We do not recommend binaural stimulation.|
We think that binaural stimulation (i.e. tone bursts in both ears) is generally a bad idea. The figure above illustrates why we have discarded this technique. While cVEMP's are "generally ipsilateral", this doesn't mean that they are always, 100% ipsilateral. Artifacts can also go across the midline, which reduces much of the value of using a binaural stimulus. Because the sternum is rather close to the sternocleidomastoid muscles, there can also be artifact due to "volume conduction" -- meaning electrical activity from one side getting confused with the other (Li et al, 1999). There is also the problem that in SCD, the canals may be activated by sound, and then all bets are off regarding laterality.
If you really care that what you are measuring reflects the side you think you are measuring, don't use a bilateral cVEMP. In our opinion, this pretty much eliminates the technique. The bottom line is don't use a bilateral VEMP.
Sound stimulus for cVEMPs
Loud clicks or tone bursts (typically 95-100 DB nHL or louder) are repetitively presented to each ear in turn at 200 msec intervals (5/second). Accoridng to Singh et al (2018), 125 dB SPL is a safe intensity for testing normal subjects. The implication is that abnormal subjects -- e.g. persons with a damaged ear -- might be more vulnerable. This has not been addressed as yet. The optimum frequency lies between 500 and 1000 Hz. The sternum is generally used as a reference and the forehead as a ground. There is some evidence that linked wrists might be a better choice for a reference though (Li et al, 1999).
We generally use a Bio-logic Nav-Pro set up with the parameters documented here.
Note that binaural presentations of sound is not recommended (by us anyway). It is faster but it reduces ones ability to localize the side of lesion because of crossover. Our recommendation, based on some unfortunate clinical experiences where binaural stimulation lead us astray, differs from others in the literature (e.g. Young, 2006).
Myogenic potentials are amplified, bandpass filtered (30-3K Hz), and averaged for 200 presentations. The response evoked in the neck EMG is averaged and presented as a cVEMP (see figure 2). The latency, amplitude, and threshold for the p13-n23 wave is measured. The amplitude is the most reliable measure (Isaradisaikul et al, 2008). Latencies are less reliable but are useful in deciding whether a particular waveform is a "cVEMP" or just noise.
Because of the high intensity of the sound used to evoke these responses, carefully checked inserts should be used. Headphones are less reliable than inserts, because small errors in headphone placement can result in substantial changes in sound intensity, and loss of a cVEMP. When the head is being held upright, headphones can shift easily.
The cVEMP is generally quick and easy to obtain because it is a strong potential and only requires about a minute of stimulation to get 100 presentations. We usually use 200 presentations ourselves. This means that you can easily repeat the cVEMP test. In our opinion, a minimum of two repetitions should be obtained on each side, to be sure that the VEMP is reproducible or absent, as the case may be. Three is generally aimed for. An exception to this can be made if the first two repetitions are of large amplitude and nearly identical (e.g. see figure 2). We generally use monaural recordings.
We recommend tone bursts rather than clicks -- here is the reasoning. A similar response is produced using tone bursts instead of clicks (Murofushi, Matsuzaki et al. 1999; Welgampola and Colebach, 2001; Cheng, Huang et al. 2003). Either 500 or 1000 hz tones are presented at a 5/second rate. They suggested using an intensity of 120 db SPL. (Note Singh et al reccomend 125 SPL). A stimulus duration of 7 msec was found optimal. The advantage of the tone burst stimulus compared to a click is that it requires lower absolute stimulus intensities. This is important if you are using equipment that doesn't produce optimally loud stimuli (see below). We suspect that it produces longer and more variable latencies. At the present writing however, the clinical value of measuring latency (other than being sure you have a VEMP), remains somewhat elusive. Amplitudes are much more reliable.
Rauch et al (2004) also advocate using tone bursts, and suggest a 500 hz frequency is optimal [See subsequent paragraph where they suggest 2000 Hz is even better] . They suggest monitoring ongoing EMG activity to ensure that the SCM muscles are activated as without muscle activation, a cVEMP does not occur. In their study, a cVEMP was judged absent when no replicable response was observed and enough responses were averaged for residual noise to be less than 3 uv. They suggest that thresholds are more useful than amplitudes. We do not agree -- we feel that amplitudes are more useful than thresholds, given a well standardized protocol. The literature suggests that amplitudes are more reliable than thresholds too (Isaradisaikul et al, 2008).
Practically, one cannot do both -- 3 repetitions as well as thresholds, because of fatigue. We have also found many patients with low thresholds, but no dehiscence. See more discussion of the problems inherent in doing thresholds below under technical pitfalls.
Noij, Herrmann, Gunan and Rauch (2018) reported that 2000-Hz tone bursts improve the detection of SCD. They stated "ROC analysis indicated that for both 2,000-Hz cVEMP threshold and for 2,000-Hz TWI, 100% specificity could be achieved with a sensitivity of 92.0%. With 2,000-Hz VEMPn and VEMPid at the highest sound level, 100% specificity could be achieved with a sensitivity of 96.0%. CONCLUSION: The best diagnostic accuracy of cVEMP in SCD patients can be achieved with 2,000-Hz tone burst stimuli, regardless of which metric is used." These conclusions were reached using research grade instrumentation, and whether or not they are equally useful in the clinical world remains to be seen.
Nearly all VEMP problems are caused by operator error. The VEMP is a relatively new test, and so far, manufacturers have not built into the protocols methods of quality assurance. In fact, after an FDA review in the US, manufacturers had to PULL some protocols they had put in to be helpful. Our government restricting innovation.
When cVEMP's make no sense in the overall clinical context, we think it is a good idea to just repeat them on a later visit, and inservice the technician if there is a big discrepancy
Assuring neck muscle activation is the biggest problem. While one can run a cVEMP very successfully with the patient's head being held up vs. gravity, this is tiring. A common quality control problem in the cVEMP is an overly nice technician who allows the patient to put their head down during the test. cVEMPs can be run with the head being actively turned to one side, thus fatiguing only one side rather than both, but this procedure also has it's pitfalls, as it is less reliable and produces smaller potentials (Wang and Young, 2006). A suggestion for any manufacturer who might be reading this is to add a method of determining if the head is off the table (a simple pressure switch would work). A few more comments about VEMP biomechanics are here.
In persons who can't cooperate, assuring neck muscle activation is a gigantic problem. Consider, for example, attempting to do a cVEMP in an infant. How do you ensure that the neck muscle is activated ? This is an intrinsic problem with doing cVEMP's in very young children, and perhaps it should just not be attempted at all. There are some reports of doing oVEMPS in children however. At a minimum, cVEMP's done in persons who are (perhaps) not cooperative should be done with equipment that can monitor the EMG.
Another technical "gotcha" in the cVEMP is a sound not getting to the ears. This is very very important ! Common things that go wrong here are asymmetrical placement of inserts, wax occluding one side, or a defective sound generator (the Bio-logic ones seem to go bad very often, when used for VEMPs -- it has happened to us 3 times in just 2 years !). Because cVEMP's are not far above the threshold provided by most evoked potential equipment, little things like not putting the insert in as far in one ear as the other can make a big difference. Just 10 dB can be important. Regarding checking of the inserts, we suggest a sound-check with every single VEMP. The easiest way to do this is to run an initial cVEMP without lifting the head -- this both assess for PAM artifact (see below), as well as can provide a good time to do a sound check. It would make sense for the cVEMP protocol to include a threshold check at 500 hz, using the inserts and transducer of the evoked potential machine, but so far, this is not available.
Thresholds can also be problematic. There are several major problems. The first is that there is a subjective element to picking a threshold. One person's response might be another person's noise. The second is that if you do thresholds, you can't do a lot of repetition (because people get tired). The third is that they don't always work -- low thresholds are NOT always accompanied by radiographic evidence of superior canal dehiscence. We have encountered several patients with thresholds of 60, but normal temporal bone CT's.
Please follow this link for a discussion of norms.
|cVEMP artifact (squiggles at start) due to technical error in placement of sound generator too close to electrodes.|
Electrical artifact. cVEMP's are huge (compared to ABR or ECochG) and there should not be much random electrical artifact. If you see stimulus artifact, or sinusoidal undulations to the trace, it is very likely that the electrodes are bad, someone put the sound generator box too close to the electrodes, the impedance was too high, or the evoked potential machine needs service. In our experience with the latter, the thing that routinely goes bad with the evoked potential machine are the connectors and the insert driver.
In the cVEMP trace above, there is stimulus artifact. After it went away when the sound generator was moved, we realized that the sound generator cannot be very close to the EMG leads. The sound generator produces a magnetic field which induces a strong artifact, if one places it close to one of the electrical leads.
Other artifact. Occasional patients have "VEMP like potentials" (? VLP), that are not VEMP's.
In some instances, this is caused by the "posterior auricular muscle (PAM)", a sound evoked twitch of the ear. The PAM is a small muscle behind the ear that can "wiggle" the ear. Actually, there are three of these muscles -- posterior, superior and anterior auricular muscles - -all vestigial. PAM is innervated by the facial nerve. The latency of the response is about 11 msec, making it overlap with the VEMP. (Funahashi et al, 1992). The literature about PAM is confused -- some authors may have mistaken the much larger cVEMP's for PAM responses, and vice versa. See Gibson (1978) for a review of the older literature.
To rule out PAM potentials, you can run your cVEMP initially or perhaps in between other runs, without contracting the SCM.
In other instances, this is due to "volume conduction" -- electrical activity from one side showing up on the other side. Volume conduction problems can be best solved by monaural stimulation. For reasons developed above, we think that a run or two of monaural stimulation is a very good idea.
Logical inference. While cVEMP's are "generally ipsilateral", this doesn't mean that they are always, 100% unilateral. If you are trying to "rule out" any vestige of remaining vestibular function -- don't use a binaural stimulation cVEMP. As nearly always one is trying to localize the side of lesion, we think that binaural cVEMP's are best avoided. An exception is when one is trying to diagnose bilateral vestibular loss, when it is OK.
One can also broadly criticize all studies of VEMPs in that the core dogma driving the enthusiasm in doing VEMP tests is implausible. The core dogma is that things are very simple, allowing one to make diagnostic inferences. While not emphasized much in the literature, VEMP tests likely involve input from many senses, and are not confined to saccule or utricle inputs as has been suggested. It is also implausible that the wiring is "unilateral". Thus we have a large clinical database that has not produced very impressive results, perhaps based on an oversimplified idea about how the ear is wired up to the nervous system.
It is good to have variant methods because it is an indication of experimentation, through which the field may advance.
The term "rectification" is used in a a confusing way by some to indicate normalization -- the unrectified VEMP signal is divided by an averaged pre-stimulus rectified surface EMG (Lee et al, 2008). The reasoning is that the size of the surface EMG signal may vary according to electrode placement or differences in muscle activation or body build. Chang et al (2007) showed that people with larger amounts of neck subcutaneous tissue (i.e. fat), have smaller cVEMPs. This is reasonable and just common sense, and a good reason for normalization.
Variability due to these technical problems might be reduced by adjusting for these factors. This technique would seem quite reasonable for adjusting for differences in electrode placement as well as neck anatomy. It would seem to us to be not quite as reasonable for adjusting for differences in muscle activation, because this might easily change during the 2 minute cVEMP itself.
We have experimented with this technique ourselves and decided that it was worse than using a very careful electrode placement and having patient's lift their head up straight (ensuring equal activation of both SCM). In other words, we discarded it.
On the other hand, Lee et al reported that rectification was better (2008). They did not use our favored technique -- instead of lifting the head up straight, they had subjects lift and turn their heads. Because this method introduces the variability due to head turning, in our opinion, their experimental design was simply designed to fail. Overall, it seems to us that their method may be better than the (poor) method of head-turning as it controls for intermittent effort which is a problem with the head-turning methodology. We presently continue to hold to the opinion that a careful, unrectified head-straight lifting VEMP is superior. However, as technology improves, we may change our mind.
In this regard, we would like to see a cVEMP output that provided the clinician more data -- the "raw" cVEMP amplitude, the "raw" integrated and rectified EMG, and the "normalized" values. We would also like to see a cVEMP machine that printed out the entire raster of EMG activation during the trial -- to assure that the patient was activating their neck and that the electrodes remained firmly attached.
There is room for other improvement -- for example, a VEMP stimulus that had "gaps" in the sound stimulus might allow a clever algorithm that used the gaps to compute a more useful normalization signal. Also, the term "rectified VEMP", should be replaced with the more reasonable term "normalized VEMP". We don't think the term "corrected" VEMP is a reasonable one, as "correction" covers too much ground.
The number of studies using this test has grown remarkably, and we have devoted a separate page to oVEMPs.
We have studied acoustically generated VEMP's in the gastrocnemius and triceps (Ruddissil et al, 2008; Cherchi et al, 2009). These responses occur at much longer latency than SCM VEMP's. The size of the triceps evoked potential scales with force (Cherchi et al, 2015). Triceps potentials are reduced in cervical cord lesions (Shirley et al, 2015). Their clinical utility remains to be established, but they might be useful to diagnose spinal cord lesions or cervical vertigo.
Skull taps and bone conduction tones can also be used to elicit VEMPs. Taps can be delivered to the forehead or lateral skull, with some differences in polarity and sidedness of the resulting potential. Bone conduction tone bursts also can evoke VEMPs, using frequencies of about 200 Hz. Clinical bone vibrators generally require additional amplification to produce strong enough stimuli for VEMP testing. Bone conducted VEMPs are not as well lateralized as click evoked VEMPs. (Sheykholeslami, Murofushi et al. 2000)
McNerney and Burkhard (2012) compared air to bone conducted VEMPs, using 500 hz for both. A TECA unit was used for the testing. BC stimuli were set to 120 db pFL, using a B071 bone vibrator. We are not entirely sure what a "pFL" is -- presumably something related to bone rather than air conduction. We are not at all sure how pFL levels of 120 db relates to the ordinary problem with clinical bone vibrators that only produce 60 db.
Well, anyway they found that amplitudes were less variable for bone conduction. This is unsurprising as bone conduction stimulates both ears and one would think that averaging across two ears would be more reliable than one. They also found that bone masking reduces amplitudes of air conducted VEMP's. To us this suggests that a component of air-conducted VEMP response is either due to startle or multisensory convergence.
Deninis et al (2015) reported on 12 normal persons, using a Bruel and Kjaer Minishaker, using 100 or 500Hz stimuli, applied through a perspex rod to the mastoid. The plastic rod is needed to provide a distance between the powerful magnetic field of the minishaker and the electrodes. This specialized hardware is necessary because clinical bone vibrators will not produce these powerufl stimuli (138 dB peak force level). They concluded "BC 500 Hz shows similar properties to AC 500 Hz, consistent with an overlapping spectrum of afferents being excited by both, with bilateral effects for the BC stimulus. "
To summarize, although BC VEMP's are undoubtedly more reliable, their obvious problem lies in their inability to confine the stimulus to one ear.
Galvanic stimuli can also produce a VEMP (Watson and Colebatch 1998). This technique bypasses the mechanical part of the ear and thus might be used to separate end-organ from nerve and more proximal lesions. Technically, averaging of electrical consequences of galvanic input is difficult because there is a large electrical artifact associated with the stimulus itself. This has been handled by subtraction methodology. A typical protocol uses 4 ma cathodal current pulses, of 20 ms duration, administered at 2 Hz, with 128 stimuli total (Bacsi et al, 2002).
More recently, Wu et al (2019) used a protocol with 1 msec, 5mA impulses. The tested both cVEMP and oVEMP. The reduced duration presumably reduces current flow into the head, and also is probably less painful. Logically, to prevent artifact, it would seem better to record from muscles that are far away from the current input (ear), such as in the legs or arms. Wu et al used substraction to reduce electrical artifact.
Galvanic and acoustic VEMP's activate different populations of vestibular afferents. Galvanic stimulation excites distal vestibular nerve afferents, particularly those with irregular discharges, but any selectivity on this action on afferents arriving from the different vestibular end organs is uncertain. (Bacsi, Watson and Colebatch, 2002). Because galvanic stimuli activate the entire vestibular nerve, and are not confined to the otoliths, one cannot assume that galvanic VEMPs are actually a indicator of otolith function.
As the entire vestibular nerve is stimulated by galvanic input, one would expect that galvanic VEMPs would be insensitive to partial nerve lesions (i.e. failed vestibular nerve section), but also quite sensitive to complete vestibular nerve loss. Thus an absent galvanic VEMP might be used as a rationale to avoid doing more surgery.
For the same reason, galvanic VEMPs should also not be able to differentiate between endorgan (saccule) damage and inferior vestibular nerve damage because one would expect that the galvanic VEMP would be present even if the inferior vestibular nerve were damaged.
Galvanic VEMPs may nevertheless prove useful using threshold or latency information. Galvanic VEMPs are not suppressed by anodal current, which suggests that VEMPs do not require the irregular afferents (Bacsi and Colebach, 2003). Galvanic VEMP's may be more reliable than acoustic VEMP's and thus be a better method of monitoring vestibulospinal connections through the spinal cord than acoustic VEMP's. More study is needed.
A closely related test to the VEMP was described by Halmagyi and others (2003). Event triggered averaging is used to detect electro-oculographic responses to loud clicks -- intensities ranging from 80 to 110 Db. 128 clicks were delivered at a rate of 5/s from 60 to 110 db, in 10 db steps. Normal subjects have no or a very low amplitude response of < 0.25 deg at 110 db. The latency was 8 msec. This test is not generally available, but appears promising. The technology is very similar to VEMP, and perhaps might even be obtained with similar instrumentation. It may be a good candidate to replace the Tullio test.
What does it test ?
Figure 2 illustrates the pathway for the acoustic cVEMP response, which includes the saccule, the inferior vestibular nerve, the vestibular nucleus, the medial vestibulospinal tract, the accessory nucleus, the 11th nerve, and finally the sternocleidomastoid muscle. Abnormal VEMPs might be caused by abnormalities in any of these structures. The sound induced VEMP also requires conduction of sound to the inner ear, which means that an intact middle ear is needed.
While VEMPs are presently attributed to the saccule, the data presented so far suggests that hearing is not necessary for VEMPs. This does not exclude the possibility that hearing is sufficient for a low-level VEMP, as it is unusual to encounter a human subject with a well documented vestibular lesion that is confined to the saccule. There are some data however suggesting that vestibular nerve section abolishes VEMPs, which would be against this idea (Watson and Colebatch, 1998). It is also possible that hearing is synergistic with vestibular input -- i.e. you get more of a response with multisensory convergence. We are presently of this opinion, but these are issues that need to be worked out.
The VEMP literature is rapidly increasing and it seems that there are considerable valuable diagnostic information to be obtained.
See the following sections for brief discussions of VEMPs, often illustrated with traces from our practice, in the following (follow links for some)
- Superior canal dehiscence
- Vestibular nerve disorders
- Ramsay Hunt
- Vestibular Neuritis
- Acoustic Neuroma
- failed VNS
- Bilateral vestibular loss (such as due to gentamicin)
- Benign paroxysmal positional vertigo (BPPV)
- Central vestibular disorders
- Meniere's disease
- Hearing disorders
For some of the more popular disorders where VEMPs are used, we have separate pages, and have combined data from other VEMP protocols as well.
There is an disturbing pattern in the literature to attribute isolated abnormal cVEMPs to "inferior vestibular neuritis" . In our opinion, this is circular reasoning at worst, and putting the cart before the horse at best. We would like to see autopsy confirmation that this entity exists and correlates with VEMP testing -- this may take a long time.
VEMP tests are commonly performed by an audiologist or an electrophysiology technician. Oddly enough, cVEMP's are sometimes even done by physical therapy practices. Audiologists are often associated with otolaryngology practices (ENT doctors), while electrophysiology technicians are often associated with neurology practices. cVEMP's are easy to obtain, but there are many pitfalls involving hearing.
We strongly suggest having an audiologist or audiology technician do this test.
We do not think that this test should be done by people who are not familiar with hearing -- i.e. physical therapy practices or general neurology practices, because a good working understanding of how hearing interacts with VEMPs is essential. We shudder to think of what might go on in a practice where testing is being done without any knowledge of whether the patient has a conductive hearing deficit, or sound-checks being done on the equipment.
Regarding interpretation, it is simply not reasonable for general audiologists, who have no neurological background, to make inferences about brain function. Similarly, in our opinion, most physical therapists do not have appropriate audiology training to interpret cVEMP's. Because the interpretation process involves both peripheral and CNS pathways, we think that a team combining an experienced audiologist and otoneurologist is optimal.
See this page for a discussion of what we have found to work.