This material is intended for clinicians and vestibular scientists. See also: bedside version -- HIT test
Timothy C. Hain, MD • Page last modified: March 9, 2021
One would EXPECT that the VHIT gain in normal people should have a gain 1.0, and there should be few if any saccades during or just after the head thrust. The VHIT gain assumption of 1.0 is wrong because VHIT gains are "all over the place".
The above plot, which we call the "roadkill graph", shows all VHIT tests through 4/2020, from more than 3300 patients tested at Chicago Dizziness and Hearing (using the Interacoustic device, not the GN otometrics device). There are many patients with gains > 1. It is almost certain that these values are not "real" but rather are due to measurement error. Thus the VHIT test is not a "clean" test with a normal gain of 1.0, but rather it is "all over the place". Notice that there are also some "real" abnormal VHIT tests shown here -- with low gain on one or both sides. There is useful diagnostic information in VHIT testing, but it is obscured by a cloud of noise, similar to almost all other vestibular tests. This graph was produced using an R-script, shown here.
Why does this graph look like "roadkill" ? Well, there seem to be some "forbidden areas" on this plot -- we just don't have many individuals with gains of 1.0 on one side, and gains < 0.25 on the other. Presumably there are some physiological constrants with superimposed noise. It is likely that this graph shows four groups -- persons who are normal (the big splatter in the middle), people with bilateral vestibular weakness (the tail), and people with unilateral vestibular loss (the wings). The big cloud in the middle is the variablity.
The "roadkill" graph also illustrates the diagnostic usefulness of the VHIT -- the tail and the arms of the roadkill, are diagnoses. The tail is bilateral loss, and the arms are unilateral weakness.
Why is the VHIT gain so variable ? Different devices, goggle slippage, looking at the wrong places.
There are presently 3 different VHIT devices and they don't provide the same results.
Figure from Jankey et al (2017) showing scatter in VHIT gains in normal subjects. PG is the micromedical device, IG is the "eyeseecam" device, and AG is from GN otometrics.
Pursuing this idea, Jankey et al (2017) wrote about "effects of device on .. vHIT gain". They observed that three different VHIT devices all do things a bit differently. They plotted this out in their figure 2, shown above. PG is position gain, IG is instantaneous gain, and AG is area under the curve gain (similar to position gain). Or another way to say this is that position gain is the Visual Eyes device from Micromedical, Instantaneous gain is the EyeSeeCam device from Interacoustic, and Area under curve is the "Impulse" device from GN Otometrics.
This plot shows very clearly that VHIT gains average about 1, but their variability is astounding. About half of normals have VHIT gains greater than 1 ! If this were really true the eyes would have to go backwards after getting to the target. This doesn't happen often so it isn't true. This means that this variability is technical in origin, and the gains cannot be greatly trusted. .
So the take home message here is that the 95% limits on VHIT gain looks to be roughly 15% in either direction. We have also observed a similar variability in our clinic population (we use the Interacoustics -- instantaneous gain -- device). Although this variability may seem daunting, the VHIT device is still very useful -- if one diagnoses unilateral weakness from patterns of the velocity trace and saccades. So don't throw the baby out with the bath water, just don't take the VHIT gain numbers too seriously.
Goggle slippage is a contributor to VHIT gain variability
Slippage artifacts Figure 2 from Suh et al, 2016
We think ourselves that goggle slippage (i.e. inertia of the camera/google) is the big one. Suh et al (2016) wrote about the effect of goggle slippage on the VHIT.. They commented " Slippage-induced artifacts are presumed to be because of the slingshot-like movement of the goggles during head movement in three different phases (lagging, overshooting, and bouncing of the goggles). " They noted that "too loose" straps resulted in high gains as well as numerous other artifacts in the trace. We think they are absolutely correct. We have also noticed that when subjects wear masks (during the Covid-19 pandemic), artifacts resembling those described by Suh et al became much more.
Looking in the wrong place also contributes to gain variability.
There is also an effect of the point of regard -- i.e. it has been shown previously using other devices that people who are looking at near targets have higher gains. This is due to the physics of the situation. Generally speaking though, this is not a big factor because the test is nearly always done with the point of regard at roughly a meter distance.
Figure from van Dooren et al (2018), showing trend for higher VHIT gain with refractive error. Note the scale here -- there are no points below 0.5. A VOR gain of, lets say, 1.3, is enormous.
Van Dooren et al (2018) examined the question as to whether or not use of contacts or spectacles might affect VOR gain. The figure above, from their paper, illustrates their results. In essence, the effect of refraction was drowned out by the very large variability of VHIT gain. They concluded that it was not necessary to correct for spectacle wearing, arguing that lack of a significant effect was the same as no effect. Of course, the problem with this argument is that had they studied more subjects, it seems highly likely that significance would have been achieved. The difference in gain, from their trend line, looks to be roughly 10% at 5 diopters. Note that these are normal subjects, but their gains are "all over the place", as seen in other studies.
Finally, perhaps the calibration procedures are not so accurate on some or all of these devices. We also wonder whether the video tracking on these devices can be relied upon.
We don't think that this wild variability is "baked in".
Rey-Martinez, Burgess and Curthoys (2018) suggested in a case report in an open access journal that "we should consider the hypothesis of a physio-pathologic cause for the enhanced eye responses to vHIT testing of some patients with vestibular function. It is easy to suggest considering hypotheses when one is dealing with one or two cases, but the simplest and most likely assumption, given the wild variability shown above, should just be that the VHIT is a messy test, and that little changes in gain are meaningless.
Effects of Age on VHIT -- not so much: See this page for details.
A tentative conclusion:
Absolute gains for VHIT tests are "all over the place. Relative gains (i.e. comparing the left and right head thrust) are probably much more relevant, and looking for corrective saccades is also likely much more relevant.
We think the VHIT device is very useful, but we depend on relative gains, patterns of saccades, and patterns of the eye velocity profile. We do not depend greatly on the gains. When we see gains > 1, we assume that there was goggle slip and we look for other clues.
Norms for vertical canal plane VHITs are even noiser than horizontal canal VHITs. We choose not to discuss these. We think this could improve with better technology (i.e. an ability to track torsion), but that will not happen any time soon. So we just don't think that vertical canal plane VHIT tests can be normed.