Vestibulo-ocular reflex gain improvements at peak head acceleration and velocity following onset of unilateral vestibular neuritis: Insights into neural compensation mechanisms

1 Introduction

When an acute unilateral peripheral vestibular deficit (aUPVD) occurs, the altered vestibulo-ocular reflex (VOR) function can be easily revealed using rapid head rotation impulses, either driven mechanically by a motor or manually applied by hand [9, 27]. With this approach, response-gains for head impulses towards the ipsi- and contra-deficit sides can be separated into phasic (acceleration-dependent) and tonic (velocity-dependent) components, in order to examine how these components of VOR responses change during the compensation processes for an aUVPD, and whether these changes could be associated with known changes in neural processing in, for example, the vestibular nuclei. It is unknown whether changes over time in these phasic and tonic components are different during recovery from vestibular loss due to vestibular neuritis. We reasoned that studying changes in these components over time might provide useful insights into the underlying neural compensatory mechanisms for vestibular loss.

The question also arises if the time courses of the recovery processes of these phasic and tonic components of VOR gain are different for ipsi- and contra-deficit head rotations following onset of aUPVD. This question was raised by Palla and Straumann [27] who noted that if bilateral linear processes were driving VOR responses then both ipsi- and contra-lateral gains should be equally affected by the deficit. They found that contra-deficit response gains were larger than ipsi-deficit gains [27]. They argued, based on the model of Lasker et al. [17], that the observed gain differences were due to rapid changes in non-linear pathways enhancing VOR responses more during contra-deficit head impulses. In this case it might be expected that the recovery processes of phasic and tonic velocity-gains of the VOR should be different for ipsi- and contra-deficit directed head impulses, specifically faster for the contra-deficit phasic responses.

There are also reasons for assuming that a patient’s age would also influence VOR responses following an aUPVD. First, VOR gains of normal subjects decrease slightly with age [22, 28]. Second, there is a trend for ipsi-deficit video head impulse test (vHIT) gains in the elderly to be lower than in young adults at onset and during recovery from a vestibular deficit due to vestibular neuritis [31]. However, other authors [29] have argued that these age effects are not complementary as the characteristics of catch-up saccades during vHITs are different in the elderly with and without peripheral vestibular loss. In an attempt to resolve this difference, we examined the interaction between changes in phasic and tonic vHIT velocity-gains and age following onset of acute vestibular neuritis. Thus based on the considerations noted above, in the current research, we asked the following questions:

Are phasic and tonic velocity-gains of VOR responses on the ipsi-deficit side elicited by yaw rotations of the lateral canals equally affected by aUPVD due to vestibular neuritis and do both gains similarly fail to improve to a normal gain by the same amount?

By using a ramp-step increase in stimulus head velocity, Minor and Lasker [17, 24] computed VOR velocity gains at peak acceleration and steady state velocity following unilateral vestibular neurectomy or canal plugging. These authors provided evidence that the dynamics of the tonic-velocity gain (vG) of the VOR resembled the gain and phase dynamics of regularly firing vestibular afferents, whereas dynamics of the phasic-velocity or acceleration gain (aG) of the VOR resembled the gain and phase of irregularly firing vestibular afferents. When these gains were examined in normal humans and primates, aG was slightly greater than vG (1.0 versus 0.85, respectively) [8, 24]. However, this balance between aG and vG changed drastically following a peripheral vestibular lesion due to labyrinthectomy with the largest changes occurring during compensation of aG on both the lesion and intact sides, with greater changes on the intact side. For this reason, Minor et al. [23] argued that signals underlying aG, are critically important for recovery of VOR gain after a peripheral vestibular deficit [23]. Furthermore, changes to aG and vG were similar to those when the lesion was induced by canal plugging rather than by a neurectomy [16, 17], suggesting that the aG and vG changes were not dependent on the presence of background afferent nerve resting activity.

Improvement in VOR gains following vestibular neuritis can occur both through peripheral recovery and central compensation. On average, peripheral recovery accounts for a 35% improvement and central compensation for a 40% improvement in ipsi-deficit vHIT responses [4]. Given that the response properties of irregularly firing vestibular afferent fibres closely follow those of the phasic-dependent aG and those of regular afferent fibres follow response properties of tonic-dependent vG as suggested by Hullar and Minor [15] then an increase in the proportion of irregular to regular vestibular nerve afferents could be one mechanism underlying a possible increase in aG during peripheral recovery. Sadeghi et al. [30] showed, however, that following labyrinthectomy such a change only occurred for contra-lesion vestibular afferents tending to rule out this mechanism for ipsi-lesion peripheral recovery. At the level of the vestibular nuclei, Newlands and Wei [26] stated that type I vestibular nucleus neurons (those excited by ipsilateral rotation) have sensitivities to rotation in the compensated state which are not different ipsi- and contra-lesionally. In contrast, type II vestibular nucleus neurons (inhibited by ipsilateral rotation) have higher sensitivities ipsi-lesionally. As this later action is mediated by a commissural pathway [32], it represents a central compensatory action. Furthermore, because type I neurons project to the oculomotor nuclei [14] and type II neurons have inhibitory connections to ipsi-lateral type I neurons [32], an ipsi-deficit increase in aG could result via this pathway.

In view of these features described above: 1) a loss of excitatory responding vestibular afferents ipsi-lesionally; 2) an increase in phasic, acceleration-dependent, inputs contra-lesion: and 3) the changes in type II ipsi-lesion vestibular nuclei neurons; all of which could possibly underlie changes in aG and vG, following onset of vestibular neuritis, our working hypothesis was similar to that of Minor and Lasker [23]. Namely, that compensatory increases in acceleration-dependent (phasic) VOR gain (aG) would be more rapid than velocity-dependent (tonic) VOR gain (vG), and that vG would be more affected on the deficit-side. This would be in keeping with the concept that the effect of vestibular neuritis (VN) on VOR gains is more similar to the effect seen with vestibular neurectomy rather than the effect seen with semicircular canal plugging as it is the vestibular nerve that is affected by neuritis [6]. As far as we are aware, this is the first report on acute and compensatory changes in phasic and tonic velocity-gains for head impulse tests of VN patients. We also compared these gains with those reported for healthy normal subjects [28].

2 Methods

2.2 Data analysis

Data were analyzed offline using a custom made algorithm developed using MATLAB (Mathworks Inc. Natick, MA) and described in detail elsewhere [7]. Head and eye velocity recordings were first filtered using a 3^rd order, dual pass 40 Hz low-pass Butterworth filter. Individual trials were accepted if peak head velocity was greater than 80°/s. Once peak head velocity was identified, the first sample, prior to peak head velocity, with a velocity less than 20°/s was specified as the impulse onset. A mean eye and head velocity were calculated during a 40 ms interval between 20–60 ms prior to this impulse onset time (see interval 4 in Fig. 1). The mean eye and head velocities over this 40 ms period were subtracted from the rest of the trial’s eye and head velocities samples, respectively. Covert saccades, overt saccades, anti-compensatory quick eye movements (ACQEM), and artifacts (as defined below) were identified using a custom algorithm [7]. A covert saccade was identified when eye acceleration was larger than 5000°/s², had an opposite sign to peak head velocity and occurred after peak head velocity but before head velocity zero crossing. In this case, the eye velocity response was cut off when eye acceleration first crossed zero acceleration prior to the 5000°/s² threshold. An overt saccade was identified with the same criteria but occurred after the zero crossing of head velocity. An ACQEM or an artifact was excluded if the criteria for a covert saccade were fulfilled, and peak eye acceleration and peak head velocity were in the same direction. The cut-off used for an ACQEM or artifact was the same as described for covert saccades. If peak eye acceleration exceeded 5000°/s² prior to peak head velocity, we assumed an artifact was present and rejected the impulse response from analysis.

OPEN IN VIEWER

Fig. 1

Head velocity (thin grey line), eye velocity (thin black line) and head acceleration (thick grey line) traces from a representative subject for a contra-deficit (normal) side response to a head impulse test. Eye velocity is inversely plotted, that is, in the same direction as head velocity. 0 ms indicates the onset of the trial (dashed vertical line, 20°/s head velocity). Peak head acceleration (A) had a latency of 35 ms, on average. Peak head velocity (B) had a latency of 63 ms, on average. Peak eye velocity (C) had a latency of 74 ms, on average. The end of the trial, as defined by the first zero-crossing of head velocity after peak head velocity (D), provided no covert saccades or artifacts were present and had an average latency of 137 ms, on average (see vertical dashed line in Figs. 2). A doubled headed horizontal arrow (marked 3) indicates the whole response interval used for wvG analysis. Similar arrows marked 1 and 2 indicate the eye velocity gain analysis intervals (peak±12 ms) used for the aGs and vGs, respectively (see methods). Another doubled headed horizontal arrow (marked 4) indicates the 40ms interval used, ending 20 ms before trial onset, to determine baseline head and eye velocity for each trial. Partially reproduced from [7] with permission.

Head and eye velocity traces were clipped for analysis from response onset (see Fig. 1) to the response termination defined as the first zero crossing of head velocity after peak head velocity (see interval 3 in Fig. 1). For trials in which a covert saccade, ACQEM or artefact occurred, the response termination was identified prior to head velocity zero crossing as described above. If a head velocity zero crossing was not reached within 120 ms after peak head velocity, the response termination was set at when head acceleration first crossed zero after peak head velocity.

VOR gain was calculated over 3 different time intervals as the ratio of the area of eye velocity to head velocity, after accounting for the delay of eye velocity with respect to head velocity. The 3 different intervals used are labelled 1–3 in Fig. 1:

Acceleration gain (aG) was calculated over a window of±12 ms around the time of peak head acceleration (35 ms on average) when goggle slippage effects correlated with the first derivative of acceleration (jerk) are at a minimum (see Fig. 3, MacDougall et al. [18]). No shift of eye velocity with respect to head velocity was used for ESC and ICS version 1 data because the average shift between the peak head and eye acceleration signals (1 ms) for the contra-deficit side at aUPVD onset was less than the vHIT software sample interval (4 ms). This shift was set to 8 ms (2 sample intervals) for the version 3 software of ICS as the average shift between the head and eye acceleration signals was on average 7.9 ms for this software version due to different filtering of eye and head velocity being applied in comparison to version 1 software of ICS. The interval we used to calculate aG is similar to that used by Minor and coworkers [17, 23]. As described above, we used an interval 23–47 ms, on average, from stimulus onset when eye velocity was increasing from 60 to 120°/s (see interval 1 in Fig. 1). Minor and coworkers used an interval 20–40 ms from stimulus onset, also when eye velocity was also increasing from 60 to 120°/s. They also termed the gain so measured the acceleration gain [17, 23].

Velocity gain (vG) was calculated over a window of±12 ms around peak head velocity (Gv) when head acceleration is zero and acceleration effects on eye velocity gain would theoretically be zero. A shift of 8 ms (2 sample intervals) was used to calculate eye velocity area after determining that the average shift of peak eye velocity with respect to peak head velocity was 6.5 ms for the contra-deficit side at aUPVD onset. This shift was increased to 16 ms (4 sample intervals) for the version 3 software of ICS as the same shift was on average 15.5 ms for this software version. Minor and coworkers [17, 23] also measured vG as the ratio of the mean eye and head velocity over an interval when head velocity had plateaued, that is when head acceleration was zero. Their interval was longer (100–300 ms) due to differences in head velocity stimulus profiles.

Whole-window velocity-gain (wvG) was calculated over the “whole” interval from stimulus onset to response termination which was clipped as described above (see arrow labelled 3 in Fig. 1). Similar shifts in eye velocity were applied for the gain calculation as described above for vG gain. The wvG method, without the shift correction, is the standard ICS gain calculation method.

OPEN IN VIEWER

Fig. 3

Mean acceleration and velocity gains at onset of an aUPVD due to vestibular neuritis, 3 weeks and 7 weeks later. The height of the column represents the mean value, and the vertical bar on the column the sem. ^* marks an ipsi-deficit response with a gain value at 7 weeks which is significant less than the normal reference data of Pogson et al. [28], Ω marks an acceleration gain, aG, which is greater than the velocity gain vG, at the same examination time, and ▴ marks gain values at 3 and/or 7 weeks which are greater than gain values at aUPVD onset. The significance levels of the differences are listed in the legend to Table 1.

For each impulse trial, we calculated the area gain as follows:

AREA Gain = ∑ i = 1 n EVi ∑ i = 1 n HVi(1)

where n (generally equal to 7 for aG and vG, except when the eye velocity responses were clipped, see above) is equal to the number of eye velocity (EV) and head velocity (HV) samples in the time interval used, and i is indexing from 1 to the total number of data points. We used the area gain calculation as it leads to the lowest mean variance in comparison to regression and point-to-point gain calculations [7].

Apart from the shifts of eye velocity we have applied to compare amplitudes of head and eye velocity at their peaks, the aG, vG and wvG methods described above are very similar to methods 2, 4 and 1, respectively, of Pogson et al. [28]. We used the data of Pogson et al. [28] to compare aG, vG, and wvG gains reported here with those of healthy normal subjects. The methodical differences are as follows: 1) Not taking the shifts into account will lead to slightly lower gain values for vG (our method 2) and wvG (our method 3) by 2.6% [7]. 2) The time windows for Pogson et al. [28] methods 2 and 4 are±20 ms which cause a window overlap of 15 ms as peak head acceleration occurs on average at 35 ms and peak head velocity at 60 ms (see Fig. 1). To avoid this overlap, we used time windows of±12 ms. There is no difference in onset times between method 1 of Pogson et al. [28] (60 ms prior to peak head velocity) and the onset time we used (head velocity exceeds 20°/sec) as the difference between our onset time and the time of peak head velocity is 60 ms on average [7].

4 Discussion

The major finding of this study was to show that, ipsi-deficit, VOR gains measured at peak acceleration (aG) were always greater than VOR gains measured at peak velocity (vG) during the weeks following onset of acute unilateral vestibular neuritis. Several authors [17, 23] have also termed the acceleration gain (aG) the phasic velocity VOR gain. Likewise the velocity gain (vG) has been termed as the tonic velocity gain. Our major finding lends support to the findings of Lasker et al. [17] who argued that phasic-velocity (or acceleration dependent) response components of the VOR had a major role in the compensation for a unilateral labyrinthectomy because aGs had improved more rapidly than vGs after lesion onset. In our study there was no difference in the aG and vG improvement after deficit onset, however aG values were always larger than Gv values, presumably permitting greater participation in the recovery processes. However, further modelling studies are required to support this conclusion.

Differences in aG and vG values can be expected based on the obvious clipping of eye velocity which occurs prior to peak head velocity in vHIT responses (see Fig. 2). The most parsimonious explanation for this effect is that it is due to the limits of disinhibition of the contra-deficit input to the ipsi-deficit vestibular nuclei and the reduction in amplitude of the ipsi-deficit afferent response. As the response is clipped around peak head velocity, where there is no contribution from phasic, acceleration dependent responses, this suggests that the clipping effect is velocity dependent. In contrast, at peak acceleration both phasic and tonic inputs could contribute to the computed aG gain as head velocity is not zero at peak head acceleration (see Figs. 1 2).

The major changes to ipsi-deficit aG and vG gains occurred between onset and 3 weeks after onset of vestibular neuritis causing an aUPVD. This time course is similar to that observed using ‘whole window’ velocity gains (wvG) to measure VOR improvement, as documented in Table 1 and elsewhere [12]. VOR deficit side vHIT gain improvement has an exponential time-course with a time constant of 5.9 weeks, leading to a gain value within 10% of steady state at 13 weeks [1]. The VOR improvement reported here is not directly comparable to improvement following vestibular neurectomy. Improvements in cases of neurectomies can only be due to central compensation as the peripheral nerve is dissected. VOG improvement, in cases of vestibular neuritis, has two components: central compensation and peripheral recovery. The peripheral loss in this study was on average 91% at onset of the aUPVD as measured by a caloric examination. The improvement was 24% to a 67% remaining peripheral loss, on average at 7 weeks, for our current data. By comparing the improvement in patients who recover some peripheral function with those who do not and have an initial loss of 90% at onset, we have estimated that central compensation could provide approximately 40% of the improvement in patients with partial peripheral recovery [4]. Given that this could provide more total gain improvement than needed, this might explain the relative overshoot of contra-deficit gain values we observed at 3 weeks in comparison to those at 7 weeks (Table 1 and Fig. 3). Based on the contra-deficit recovery processes described by Palla and Straumann [27], this change could occur because the CNS first attempts to boost the contra-deficit VOR gain close to 1, and subsequently reduces the response asymmetry with respect to ipsi-deficit gain. Our hypothesis concerning the role of central compensation on the ipsi-deficit side, however, differs from that of Palla and Straumann [27] who argued that ipsi-lesional VOR gain recovery is a result of spared peripheral vestibular function on the deficit side. As described above, we observed considerable ipsi-deficit gain recovery due to central compensation when only 10% or less of peripheral function (as determined by CP values of caloric tests) was available [4]. To explain which of these different compensatory mechanisms underlies ipsi-deficit vHIT gain recovery of patients with vestibular neuritis, responses of patients with 90% or greater CP due to vestibular neuritis need to be compared with those of patients with a vestibular neurectomy.

To our knowledge, there have been no previous reports comparing phasic (acceleration dependent) and tonic velocity gains (aG and vG) for cases of vestibular neuritis. Reports do exist for changes in the whole window velocity gain (wvG) over time following onset of vestibular neuritis. Therefore, a comparison with previous research can only be made for wvGs (see Table 1). Our results are comparable with previous reports of wvG gains bearing in mind that the canal paresis values at onset, indicating the degree of peripheral loss, and time interval between tests after deficit onset may differ. For example, Martin-Sanz et al. [21] observed ipsi-deficit wvG values of 0.63 at onset and 0.74 at 1 month, which are only comparable to our value at 7 weeks (Table 1). However, the gain values at onset are not comparable to ours despite being recorded within 24 hours of hospital admission. This difference may be due to the difference in mean CP values (47% for Martin-Sanz et al., [21] versus 91% at onset in the current study). There is a negative correlation between CP values at onset and wvG vHIT gains [4]. Thus higher CP values on average will be associated with lower wvG gains as in the current study. Similar wvG results to ours have been previously reported by Fu et al. [12]. Gain values of 0.47 at 10 days and 0.69 at 6 months after onset of vestibular neuritis are very similar to the values reported here (see Table 1; 0.42 within 5 days of onset, and 0.66 at 3 and 7 weeks later). Thus, based on these similarities and exceptions, we assume that similar results for aG and vG values would have been obtained by the authors of the previously mentioned studies.

We noted no differences between the recovery times for ipsi- and contra-deficit aG and vG values. aG and vG on both sides showed most improvement over the first 3 weeks with less improvement in values at 3 compared to 7 weeks (Table 1). There are minor differences in contra-deficit gain recovery compared to the major differences in ipsi-deficit gain recovery. Previously we observed a similar result for contra-deficit gains using less subjects (37) for whole window gains at 3–4 weeks [1]. We expected to observe more significant differences for contra-deficit gains with a larger group of subjects (61). This was not the case.

There are a number of limitations to this study. First, we have measured the phasic velocity, or as termed here, acceleration gain, aG, using vHIT techniques. For tests in the horizontal canal plane, these techniques are particularly susceptible to goggle slippage during the ascending head velocity phase [18]. The data of MacDougall et al. [18] shows, however, that the biphasic pulse associated with such slippage has peaks at the onset and end of the monophasic acceleration phase. That is, these slippage peaks have a profile more like the time derivative of head angular acceleration (jerk), and not like the ascending monophasic head acceleration profile itself. The peak head acceleration occurred, on average, in our data at 35 ms with respect to zero latency at 20 deg/s head velocity (see Fig. 1). This time point is approximately at the zero crossing of the biphasic “bump” or jerk artifact as would be expected if it is dependent on the time derivative of head acceleration (see Fig. 3, MacDougall et al. [18]). Thus, this comparison indicates that when computing gain across our acceleration gain time-window of peak head acceleration±12 ms, the effect of the goggle slippage or “bump” artifact would have averaged out. Furthermore, in our average traces of Fig. 2, it can be noted that a marked presence of such an artifact is not apparent in our data. Secondly, based on our previous results [1], we have assumed that peripheral functional loss measured with the caloric test is correlated with that measured with head impulses. It might be argued that irregular phasic responding vestibular afferents contribute less than regular afferents to the eye movement response during caloric stimulation, and that the contribution of irregular afferents becomes more observable under head impulses. If so, after vestibular neuritis, a caloric test might underestimate the contribution of spared irregular afferents to VOR gains. Countering this reasoning, our previous investigations [1] determined that low frequency rotating chair deficit response amplitudes and high frequency head impulse response amplitudes were equally well correlated with CP values. However, the time course of head impulse response recovery was more similar to that of CP values [1]. These results suggest spared irregular fibres should be equally apparent as spared regular fibres in caloric responses.

We also found no differences in aG and vG values with respect to subject age. In a previous study with 30 subjects, we determined that wvG gains at vestibular neuritis onset tended to be less in the elderly (60–74 years of age) compared to the young (23–35 years). Our current study with 61 patients could not confirm this finding. Coupled with the finding that gain improvements are similar across canal planes [3], these findings on the effect of age and gain side suggest that vestibular rehabilitation should concentrate on improving the effects of head movement velocity and acceleration on eye movements and balance control similarly across all age groups.

We have assumed that changes in acceleration and velocity VOR gains over time in the yaw planes are attributable to 2 processes: partial recovery of lateral canal peripheral function, and central compensation in yaw directed VOR brainstem pathways. This assumption may require modification for two reasons. Firstly, the semi-circular canals are not orthogonal to one another [5, 9]. When the head is upright as in vHIT tests, this anatomical non-orthogonality leads to lateral canals being stimulated by 0.91 of the total head yaw acceleration, and the posterior canals by 0.32 of yaw head acceleration and anterior canals a negligible amount [9]. Crucially, posterior canal function is affected significantly less by vestibular neuritis [10, 13] with the wvG for the posterior canal being slightly reduced to 0.69 on average for ipsi-deficit impulses compared to 0.4 for the lateral canal [2]. The normal lower limit of wvG for the posterior canal is 0.6, mean 0.78 [28]. Thus, theoretically, a third method of central compensation would be to use the largely unaffected posterior canal sensitivity to yaw rotations to drive horizontal VOR neural circuits. However, in 40% of the cases the posterior canal response is affected by vestibular neuritis with an average gain of 0.43 not different from that of the deficit lateral canal [3], leading to variable amounts of central compensation if the posterior canal response would be involved in the central compensation of the lateral canal deficit.

5 Conclusion

The key findings of this study were that following vestibular neuritis, acceleration dependent (often called phasic-velocity) VOR gains (aG) are less affected than tonic-velocity dependent VOR gains (vG) and aG remained larger than vG during the first 7 weeks of peripheral recovery and central compensation. These findings might be considered predictable on the basis of changes in the proportion of contra-lesion irregular (phasic) afferents [30], and aG and vG changes following labyrinthectomy in animals [16, 17].