We used data from 66 participants: 28 healthy volunteers and 38 patients (22 patients with vestibular neuronitis and 16 patients after vestibular neurectomy). Subsequently, we divided participants into four groups according to their different vestibular behaviors, results, and complaints.

The inclusion criteria for the vestibular neuronitis group included a history of the first vertigo attack and a confirmed acute UVL without any neurological or cochlear deficit. According to the HINTS plus protocol (13–15) and normal magnetic resonance imaging (MRI) results, peripheral vestibular deficit was proven. For the surgery group, subjects with MRI-confirmed vestibular schwannomas [Koos classifications 1–2 (13)] with normal or near-normal vestibular function before the surgery were included. The inclusion criteria for the control group included no evidence of any balance problems currently or in the past, no history of vestibular disease, and normal hearing thresholds. We examined participants from January 2018 to February 2022.

To assess the features of HSN during vestibular compensation, we defined the criteria for poorly compensated patients as those who had a unilateral vestibular deficit [>26% unilateral weakness (UW) by a bithermal caloric test] and who had at least three of the following criteria: complaints of blurred vision (subjectively often described as slow or lazy eyes) during daily life head turns (e.g., turning the head to the left and right before crossing the street) or dizziness, defined as a total score higher than 16 on the Dizziness Handicap Inventory (DHI) (14), having a gait disturbance defined as a timed up and go (TUG) test score more than 10 s (or in need of assistance), or having spontaneous nystagmus (SPN). According to our criteria, we planned to establish a poorly compensated group after V4 was completed because there is no test to confirm sufficient and finished dynamic compensation. Six patients fulfilled the criteria during all follow-ups, and post-hoc formed the poorly-compensated group.

Post-hoc, we established four groups for comparison: the neuronitis and surgery groups (compensated patients), the poorly compensated group, and the control group. We analyzed 20 subjects with vestibular neuronitis (neuronitis group, five women, 15 men, mean age 45 years), 12 patients with vestibular schwannoma after unilateral vestibular neurectomy (surgery group, two women, 10 men, mean age 51), and six subjects who were poorly compensated from both groups (poorly compensated group, two from neuronitis and four from the surgery group, five women and one man, mean age 49 years). The control group consisted of 28 volunteers (15 women, 13 men, mean age 48 years).

We examined all subjects at a tertiary referral center, University Hospital.

We measured average slow phase velocity (aSPV) [°/s] of spontaneous nystagmus (SPN) and head-shaking-induced nystagmus (HSN), unilateral weakness (UW) during the caloric test, and video head impulse test (vHIT) gains.

VOG VisualEyes™ 525 (Interacoustics, Denmark) and VHIT EyeSeeCam (Interacoustics, Denmark) were used to perform a complex vestibular examination on all participants.

The patient was asked to sit upright, and visual fixation was denied. The tracing was recorded for 40 s (sitting position, head still, and goggle closure). If any nystagmus occurred, the VOG software measured its slow phase component.

The patient was asked to lie in a caloric position, and each ear was irrigated with warm (50°C) and cold (24°C) air for 60 s and nystagmic response was recorded for 120 s.

The patient was asked to sit upright, and visual fixation was denied. Eye movements were observed for 10 s to obtain a baseline. The examiner moved the patient's head (pitched forward 30°) briskly to the left and the right, aiming for a frequency of ~2 Hz and a head displacement of roughly 40–60°, 20 cycles (duration of 10 s), and then stopped abruptly. VOG was recorded for 120 s; induced nystagmus was evaluated. If there were more than two repetitive nystagmus beats after a headshake, they were analyzed.

The patient was asked to sit upright, with visual fixation of a spot at a distance of approximately 1 m, unpredictable and passive head turns, a peak head velocity of between 150° and 250°/s, and the amplitude of a head turn being 10–20°.

The subjective functional status of participants was assessed using the DHI questionnaire, which represents the functional, emotional, and physical aspects of subjectively reported disability (14).

The Timed Up and Go test measures functional mobility to estimate the risk of falling and the ability to maintain balance while walking. The patient was asked to sit in a chair; after the examiner said “go,” the timer started, and the patient got up from the chair, walked a distance of 3 m, turned and walked back to the chair, and sat down again and the timer stopped. We also evaluated the need for assistance. One limitation of the TUG test was the subjective connotation of the “normal walking speed.” Some could interpret this as a brisk walk, while others interpreted it as a leisurely pace.

We scheduled four examinations: the first (V1) within the 1st week after unilateral vestibular loss (UVL), the second (V2) after 4–6 months, the third (V3) after 12 months, and the final (V4) after 24 months.

Our study used continuous quantitative variables (aSPV of SPN and HST in degrees/s, caloric weakness in %, vHIT gains). To evaluate the potential of HSN to reflect vestibular compensation after UVL, we had to exclude potential sources of bias. First, compensation for previous vestibular loss, which could be present in the schwannoma group before the surgery, was possibly and might have already started or even completed. A longer compensation period could give false results during the scheduled 2-year period. Second, normalization of vestibular function (functional recovery) at follow-up in the neuronitis group would not assess a compensation process. Finally, differences in behavior exist in the poorly compensated group.

To minimize bias, we first established two study groups in which normal or near-normal vestibular function was expected before unilateral vestibular loss (measured in the surgery group and expected in neuronitis without a history of imbalance). We excluded patients who had abnormal vHIT gains and significant caloric weakness present before surgery (the surgery group consisted of small Koos 1 or 2 tumors).

Second, we post-hoc excluded neuronitis patients with normalized vestibular function (normalized caloric test) and performed a post-hoc analysis of patients with significant vestibular loss (caloric weakness >26%) during all scheduled follow-ups.

Finally, we post-hoc established a poorly compensated group according to the defined criteria to reflect differences in behavior between poorly and well-compensated patients after UVL.

As part of the descriptive statistics, we evaluated the normality of the data distribution using the Anderson–Darling test for the average velocities of the slow phase of SPN, HSN, caloric weakness, and gains from the vHIT regression analysis in all groups. Because some continuous variables did not show a normal distribution, we reported medians and lower and upper quartiles for all continuous variables. To compare patient data to control group values, we used an unpaired two-tailed t-test in case of confirming the normality of the data distribution. Otherwise, a non-parametric two-tailed Wilcoxon rank-sum test was used.

We used linear regression analysis to statistically evaluate the evolution of examination results over time. Because the exploratory analysis showed exponential dependencies, we log-transformed time. We then calculated the slope of the regression line, hereafter referred to as the trend, for each patient. We tested the set of individual trends using the one-sample t-test or the Wilcoxon test for significant differences from 0.

The relationship between the examination methods was assessed using the correlation analysis of individual trends. The individual trend was calculated using linear regression, and, as the data showed an exponential dependency, time was logarithmized. The relationship between the methods was calculated using Pearson's test in the case of a normal distribution and Spearman's test in the absence of a normal distribution. A significant positive correlation between the methods indicates that faster improvement in one method leads to a faster improvement in the other.

To compare the sensitivity between the examinations, we contrasted the measured values with the cutoffs as follows. The abnormal cutoffs for HSN (2.03°/s) and SPN (1.70°/s) were determined as 97.5% of our control group, 26% UW for the caloric test, and a gain of 0.78 for vHIT (15). We assessed sensitivity separately for each visit (V1–V4). Individual tests were not corrected for multiple comparisons because this would increase the likelihood of false negative results. R-project software was used for statistical processing (R Development Core Team 2022) (10).

Before including a subject in this study, we received written informed consent signed by a volunteer/patient.

The result trends from all vestibular tests are presented in Table 1 and visualized in Figures 1, 2 for the neuronitis group and the neuronitis and surgery groups, respectively.

We contrasted the measured values to the following cutoffs to compare the sensitivity and specificity (to identify vestibular loss) between the different vestibular tests. Abnormal cutoffs for HSN (2.03°/s) and SPN (1.70°/s) were determined to be 97.5% of our control group (as continuous variables did not show a normal distribution); the abnormal cutoff for the caloric test was used as 26% of UW and for vHIT was a gain of 0.78 (15).

The specificity for identifying controls was 94% for HST, 93% for SPN, 94% for caloric weakness, and 75% for vHIT.

To assess a trend to identify a vestibular loss (vestibular asymmetry), we calculated the sensitivity for each visit (V1–V4) separately. In the neuronitis group, the sensitivity of HST was 100% at V1, 79% at V2, 24% at V3, and 6% at V4. The sensitivity of SPN was 100% at V1, 47% at V2, 12% at V3, and 0% at V4. The sensitivity of vHIT was 100% at V1, 84% at V2, 94% at V3, and 83% at V4. The caloric test sensitivity was 100% on each session.

In the surgery group, the sensitivity of HST was 100% at V1, 93% at V2, 42% at V3, and 20% at V4. The sensitivity of SPN was 100% at V1, 57% at V2, 17% at V3, and 0% at V4. The sensitivity of vHIT and caloric weakness was 100% at all sessions.

In the poorly compensated group, the sensitivity of HST was 100% at V1 and V2, 66% at V3, and 50% at V4. The SPN was 100% at V1 and V2, 83% at V3, and 17% at V4 (a weak SPN of at least one aSPV was present in all poorly compensated groups).

To summarize, the caloric test and vHIT showed almost stable sensitivity to detect vestibular loss, whereas SPN and HST demonstrated a strongly decreasing ability to detect asymmetry, particularly in compensated patients.

Several mechanisms have been proposed to explain HSN (5, 9). The etiology of HSN is thought to be related to both Ewald's second law and an asymmetry in the central velocity storage mechanism (17). In healthy subjects, each head turn is immediately opposed by a contralateral head turn during a head shake, resulting in a balanced ratio of excitation/inhibition from both sides.

In vestibular loss, there is a need to reestablish tonic and phasic vestibular function. Evidence suggests that second-order vestibular neurons tend to modulate their neuronal resting discharge and reach prelesion levels (16, 18–20), placing the site of neural rearrangement in commissural pathways (21), leading to clinical improvement and static compensation. Due to inhibitory saturation, the centrally restored pacemaker discharge is insufficient to restore the whole ipsilesional dynamic range (22). Therefore, during HST, head turn toward the healthy side is opposed during ipsilesional head turn only by inhibitory cutoff from the healthy side. Excitation is more effective than inhibition as a vestibular stimulus. During 20 cycles of HST, non-linearity arises and may charge the velocity storage mechanism in an asymmetric manner. When the head abruptly stops, HSN will result from the discharge of the asymmetrically charged velocity storage mechanism.

Studies suggest that, when lower frequency sinusoidal stimuli are used, the performance of the VOR often recovers over time, and asymmetries are only noted at higher rotational velocities with increasing head velocity (23) as a result of linear and nonlinear VOR pathways (24, 25).

Head-shaking test was extensively tested for its sensitivity and specificity in identifying different vestibular diagnoses and in patients with peripheral and central vestibular lesions (3, 4). The presence of HSN in peripheral vestibular lesions varies between 34% (26), 40% (27), 90% (28), and 100% (17) and was also reported in benign positional paroxysmal vertigo (29, 30), in central disorders of 23% and was found to be present in 10–14% of healthy controls (3, 26), in 74% of dizzy patients (3), and 15% dizzy but normal patients with electronystagmography (31). The well-established literature review by Burgio (32) reported that HST is neither sensitive nor specific enough to be used as a screening test for vestibular loss. Another study found a closer connection between HSN and poorly compensated UVL than functional asymmetry (33).

To conclude, research results appear to be inconsistent. Some studies used active head movements, while others used passive head movements; some used scleral search coil, ENG, or VNG, while others used only Frenzel goggles. The study size, patient inclusion criteria, and the HST method also differed. We found different HST amplitudes (half a distance) in the literature, varying from 15 to 45° (6, 8, 10, 17, 34), resulting in different velocities (120–360°/s).

The authors attempted to explain discrepancies between the studies and suggest that some may arise from partial UVL, which may not have sufficient asymmetry, to elicit HSN, or that the central velocity storage may be reduced or lost after UVL. Others point out that HST evaluates a wider frequency range than the standard caloric or rotary chair stimulation (31).

We support the suggestion that HSN should always be interpreted in relation to other tests, such as SPN, VHIT, and a side of caloric weakness (35).

As the evidence appears inconclusive, the role of HST in clinical practice is still unclear.

To summarize our results, HSN was not related to vestibular asymmetry (loss) at all follow-ups, but it did show a strong time-related intensity decrease, reflecting an improvement in dynamic ability and a decrease in self-perceived handicap. The decreasing trend in HSN sensitivity to identify vestibular loss also supports the ability of HSN to reflect dynamic vestibular compensation after UVL. Similar results were revealed by Angeli et al. (33), suggesting that HSN had a stronger correlation with poorly compensated unilateral peripheral loss than functional asymmetry.

According to evidence, ~20% of patients with chronic stable UVL continued to experience chronic postural imbalance (the syndrome of chronic vestibular insufficiency) (36). We had a similar report of 15% poorly compensated patients, showing a well-detectable HSN and SPN as well as a higher TUG and DHI score at all follow-ups.

We provide the first long-term evidence of the features of HSN, showing the ability of HSN to reflect dynamic vestibular compensation and decreasing sensitivity to identify chronic and stable vestibular asymmetry. We found no theoretical explanation for the decreasing trend in the sensitivity of an HST to identify vestibular loss (high sensitivity in the 1st year after UVL and low sensitivity in 2 years). We found no literature based on animal electrophysiological experiments on chronic adaptive changes of the VOR explaining long-term (years) characteristics in velocity storage and on long-term dynamic changes of the VOR more than a few months after UVL. We hypothesize that the sensitivity of HST to identify vestibular asymmetry could be decreased by individual adaptive changes in velocity storage, including the use of non-linear/linear pathways to compensate for dynamic asymmetry during a long-term compensation process.

Despite the small sample size (due to strict inclusion/exclusion criteria such as “no-allowed-recovery” and “no-allowed-prior-deficit”), the power size of the main message (time-related decrease in HST intensity) was strong (for the neuronitis group: Spearman's ρ = −0.8436362; p = 3.228042e-19, two-tailed test: effect size and confidence limits, d = −3.14 [−4.04 −2.24]), (for schwannoma group ρ = −0.8912293; p = 3.819181e−19, two-tailed test: effect size and confidence limits, d = −3.93 [−5.13 −2.73]) and the power of the study with aforementioned effect was = 1.