Evaluation of the Video Ocular Counter-Roll (vOCR) as a New Clinical Test of Otolith Function in Peripheral Vestibulopathy

Key Points

Question  Can a video oculography–based ocular counter-roll measurement be used as a clinical test of otolith function?

Findings  In this case-control study of 56 individuals, video ocular counter-roll measurement with a simple bedside head tilt maneuver could detect patients with loss of vestibular function and healthy controls. The technique had a diagnostic accuracy of 83%.

Meaning  The video ocular counter-roll (vOCR) test can be performed with a simple bedside maneuver to detect or track loss of otolith function.

Importance  Video-oculography (VOG) goggles have been integrated into the assessment of semicircular canal function in patients with vestibular disorders. However, a similar bedside VOG method for testing otolith function is lacking.

Objective  To evaluate the use of VOG-based measurement of ocular counter-roll (vOCR) as a clinical test of otolith function.

Design, Setting, and Participants  A case-control study was conducted to compare vOCR measurement among patients at various stages of unilateral loss of vestibular function with healthy controls. The receiver operating characteristic curve method was used to determine the diagnostic accuracy of the vOCR test in detecting loss of otolith function. Participants were recruited at a tertiary center including the Johns Hopkins outpatient clinic and Johns Hopkins Hospital, Baltimore, Maryland. Participants included 56 individuals with acute (≤4 weeks after surgery), subacute (4 weeks-6 months after surgery), and chronic (>6 months after surgery) unilateral vestibular loss as well as healthy controls. A simple bedside maneuver with en bloc, 30° lateral tilt of the head and trunk was used for vOCR measurement. The study was conducted from February 2, 2017, to March 10, 2019.

Intervention  In each participant vOCR was measured during static tilts of the head and trunk en bloc.

Main Outcomes and Measures  The vOCR measurements and diagnostic accuracy of vOCR in detecting patients with loss of vestibular function from healthy controls.

Results  Of the 56 participants, 28 (50.0%) were men; mean (SD) age was 53.5 (11.4) years. The mean (SD) time of acute unilateral vestibular loss was 9 (7) days (range, 2-17 days) in the acute group, 61 (39) days (range, 28-172 days) in the subacute group, and 985 (1066) days (range 185-4200 days) in the chronic group. The vOCR test showed reduction on the side of vestibular loss, and the deficit was greater in patients with acute and subacute vestibular loss than in patients with chronic loss and healthy controls (acute vs chronic: −1.81°; 95% CI, −3.45° to −0.17°; acute vs control: −3.18°; 95% CI, −4.83° to −1.54°; subacute vs chronic: −0.63°; 95% CI, −2.28° to 1.01°; subacute vs control: −2.01°; 95% CI, −3.65° to −0.36°; acute vs subacute: −1.17°; 95% CI, −2.88° to 0.52°; and chronic vs control: −1.37°; 95% CI, −2.96° to 0.21°). The asymmetry in vOCR between the side of vestibular loss and healthy side was significantly higher in patients with acute vs chronic loss (0.28; 95% CI, 0.06-0.51). Overall, the performance of the vOCR test in discriminating between patients with vestibular loss and healthy controls was 0.83 (area under the receiver operating characteristic curve). The best vOCR threshold to detect vestibular loss at the 30° tilt was 4.5°, with a sensitivity of 80% (95% CI, 0.62%-0.88%) and specificity of 82% (95% CI, 0.57%-1.00%).

Conclusions and Relevance  The findings of this case-control study suggest that the vOCR test can be performed with a simple bedside maneuver and may be used to detect or track loss of otolith function.

Video-oculography (VOG) goggles have been integrated into the bedside assessment of patients with vestibular disorders. The emphasis has been mostly on the function of semicircular canals using the video head impulse test (vHIT), but a similar VOG method for the clinical assessment of otolith function is lacking.^1,2 The otolith organs detect inertial forces on the head and provide vestibular inputs essential for a wide range of neurophysiologic functions, such as the vestibulo-ocular reflex (VOR) during head tilt or head translation, sensing head motion and position, postural control, and spatial orientation.^3-9 Because otolith pathways are commonly affected in vestibular disorders, an easy-to-use, bedside VOG test of otolith function would be valuable in the evaluation of these patients.^10-12

When the head is tilted laterally in the roll plane (ear to shoulder), the resulting response is a torsional rotation of the eyes in the opposite direction of the head tilt, known as the ocular counter-roll (OCR).^13-15 During the movement of the head, the dynamic component of the OCR consists of a torsional nystagmus with a slow phase in the opposite direction of the head tilt, which is driven by the activity from the semicircular canals and otolith organs.^16,17 During a sustained head tilt, however, the static OCR maintains the ocular torsion away from the side of the head tilt, which is primarily generated by the otolith inputs from the utricles.^3,18-20

In this study, we sought to examine whether VOG-based OCR (vOCR) measurement can be used to detect loss of otolith function at the bedside. For vOCR recording, we used a video method that can track torsional eye position accurately in real-time using an iris-recognition algorithm.¹⁵ In a previous study, the vOCR measurement using this video method was able to identify patients with chronic vestibular loss, and the accuracy was comparable to the vestibular-evoked myogenic potentials (VEMP), a widely used laboratory test of otolith function.^15,21 The vOCR was symmetrically reduced between the side of vestibular loss and the healthy side in patients with chronic unilateral vestibular loss. In this study, to evaluate vOCR as a clinical test, we applied a simple bedside maneuver during which the head and trunk were tilted laterally en bloc by the examiner. We evaluated the accuracy of this bedside method in detecting abnormal vestibular function at various stages following vestibular loss.

We recruited patients with unilateral loss of vestibular function from a resection of vestibular schwannoma (Table). Patients were enrolled consecutively from the Johns Hopkins outpatient clinic or during their hospital stay at Johns Hopkins Hospital, Baltimore, Maryland, after their vestibular schwannoma was resected. All participants gave written informed consent and the study was approved by the Johns Hopkins Medicine Institutional Review Board. Participants did not receive financial compensation. The study was conducted from February 2, 2017, to March 10, 2019. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for case-control studies. Patient recruitment was based only on unilateral loss of vestibular function regardless of whether they had any vestibular symptom or sign of vestibular recovery. Individuals with any other neurologic abnormality (eg, ischemic lesions noted on brain magnetic resonance imaging) or history of vestibular or hearing loss on the nonlesion side were not included. Forty-five patients were initially recruited but 4 were later excluded due to difficulty completing the maneuvers after the surgery or poor VOG data quality. Based on the time from resection of vestibular schwannoma, patients were divided into 3 subgroups: within 4 weeks after surgery (acute, n = 13), between 4 weeks and 6 months after surgery (subacute, n = 13) and more than 6 months after surgery (chronic, n = 15). All patients had loss of hearing and signs of vestibular loss on the side of the surgery: mean (SD) horizontal vHIT gain for acute on the lesion side, 0.28 (0.19) vs 0.80 (0.13) on the nonlesion side; subacute on the lesion side, 0.33 (0.11) vs 0.69 (0.18) on the nonlesion side; and chronic on the lesion side, 0.44 (0.16) vs 0.83 (0.30) on the nonlesion side. A healthy control group (n = 15) was also recruited with no known neurologic, auditory, or vestibular disorders and with normal vestibular evaluation (mean [SD] horizontal vHIT gain on both sides, 0.91 [0.10]). All 4 groups of participants were age-matched (Table). A total of 56 individuals were included in the analysis.

A bedside tilt maneuver was applied for vOCR measurement, during which the head and trunk were tilted en bloc 30° while the participant was sitting upright (Figure 1). We used this maneuver because, when the head and trunk are tilted together, the static OCR is primarily driven by the otolith inputs, whereas if the head is tilted on the trunk, a change of inputs from the neck proprioceptors in addition to the vestibular inputs may contribute to the static OCR response.^22-25 The degree of tilt was measured by a linear accelerometer mounted on the VOG goggles and the operator received numeric feedback on the head roll and pitch angles in real time so that a roll angle of 30° and a pitch angle of 0° could be maintained during the tilt maneuver. We chose 30° lateral tilt because it is large enough to produce a measurable OCR and yet within the comfortable range of positions to maintain during the tilt maneuver.

Participants sat upright on a chair, fixing on a visual target (2.5 cm in diameter), 135 cm away at eye level. The torsional position of both eyes and the position of the head were recorded simultaneously using the VOG goggles. The recording of ocular torsion and head positions began with the head and trunk in the upright position for 30 seconds, followed by 3 right and 3 left lateral tilts in random order, each lasting 30 seconds, separated by 30-second periods with the body back in the upright position (Figure 1). There was also a final 30-second recording during which the participant was returned to the upright position from the last tilt position. A soft collar was used to restrict the movement of the neck on the trunk. The soft collar was equipped with a protractor to ensure that the head and trunk were aligned, and both were tilted at 30° during the maneuver. In addition to the fixation target straight ahead, 2 lateral targets were used so that the straight-ahead fixation could be maintained during the tilt by instructing the participants to fix on the lateral target. All 3 targets were in line with the lateral targets placed at equal distances from the central target. The distance between the lateral and central targets was calculated as the difference between the sitting surface (participant’s hip level) to the lateral canthus of the eyes divided by 2 (ie, multiplied by the sine of 30°), hence taking into account the height of the trunk for maintaining straight-ahead fixation during 30° tilts. Thus, the taller the trunk, the farther the lateral targets had to be displaced to the sides.

We used a videonystagmoscopy system (the RealEyes xDVR goggles; Micromedical Technologies Inc) for vOCR measurement. To measure the tilt angle simultaneously, an accelerometer (MPU-92/65 sensor; InvenSense) was mounted onto the goggles and connected through an Arduino board (Teensy 3.2) to the input port of the cameras. We used custom software previously developed to measure ocular torsion.¹⁵ To determine the angle of ocular torsion with this method, a template-matching method is implemented to compare the iris pattern at any point in time during head tilt with a reference image obtained at the beginning of the recording with the head in the upright position (Figure 2). Further technical details of this method have been published previously.¹⁵

Ocular torsion and tilt position data were analyzed in Matlab, version 9.7, 9.8, and 9.9 (Mathworks Inc). The vOCR value was measured as the difference between the average ocular torsion of both eyes during the static tilt from the preceding reference value in the upright position. The vOCR was measured with lateral en bloc tilts to each side (ie, right and left tilts). The final vOCR value was the mean of 3 measurements in each tilt direction. We only used the ocular torsion values during the static tilt position to avoid canal stimulation during vOCR measurement. Because the static tilt was maintained manually and there could be a few degrees of difference in the actual tilt position vs the desired position of 30° during the test, the vOCR value was corrected based on the actual amount of static tilt so that the values could be compared across participants:

Corrected vOCR = (measured vOCR × 30)/(measured tilt).

To account for the difference between the vOCR values on the lesion side (ie, the side of vestibular loss or LS) and the nonlesion side (NLS), the vOCR asymmetry ratio was calculated as:

vOCR asymmetry ratio = (NLS – LS)/(abs [LS] + abs [NLS]).

Statistical analyses were done in Matlab, versions 9.7, 9.8, and 9.9 and R, version 4.0.4 (R Foundation for Statistical Computing) statistical software using the effect size and 95% CI. To compare the mean vOCR during head tilt and account for multiple comparisons, we conducted 2-way repeated measures analysis of variance on the influence of 2 independent variables: participant groups, including healthy controls and patients with acute, subacute, and chronic vestibular loss; and the side tested, including lesion and nonlesion sides in patients and 2 randomly assigned sides among controls (Table). The asymmetry ratios among the participant groups were compared using 1-way analysis of variance. Variables in the analysis of variance were compared using the η² test, which quantifies the variation in the results by a given variable. Generally, η² values used to categorize the variations are 0.01 (small), 0.06 (medium), and 0.16 or greater (large).

The accuracy of the vOCR test in discriminating between vestibular loss and normal function was evaluated using receiver operating characteristic (ROC) analysis by including all participants (41 patients and 15 healthy controls). In this method, the sensitivity (true-positive rate) is plotted against 1-specificity (false-positive rate) for different thresholds of a test parameter. Thus, each point on the ROC curve corresponds to the true- and false-positive rates at a particular decision threshold. In the ROC analysis, the area under the curve is a measure of the test’s ability to distinguish between patients with vestibular loss and healthy controls, and the best threshold can be determined as the point on the ROC curve representing the best sensitivity and specificity for the test. For ROC analysis, vOCR was the index test and the target condition was loss of vestibular function. The clinical reference was the known state of vestibular function in patients and healthy controls.

The population included 28 men (50.0%) and 28 women (50.0%); mean (SD) age 53.5 (11.4) years. Other demographic information and vOCR values for all patient and control groups are provided in the Table. The groups that represented the time from loss of vestibular function were significantly associated with the vOCR values (2-way repeated measures analysis of variance; η² = 0.18; 95% CI, 0.006-0.35) (Figure 3A). Post hoc comparisons showed that the mean (SD) vOCR on the lesion side in the acute loss group was lower than in the chronic loss and control groups (acute, 2.19° [1.79°]; subacute, 3.36° [1.41°]; chronic, 4.00° [1.91°]; and control, 5.37° [1.35°]). In between-group comparisons, decreases were noted with acute vs chronic: −1.81°; 95% CI, −3.45° to −0.17°; and acute vs control: −3.18°; 95% CI, −4.83° to −1.54°; however, the difference from the subacute patient group was not significant (acute vs subacute: −1.17°; 95% CI, −2.88° to 0.52°) (Figure 3A). Also, the vOCR on the lesion side in the subacute group was lower than the control group (subacute vs control: −2.01°; 95% CI, −3.65° to −0.36°), but it was not significantly different from the chronic group (subacute vs chronic: −0.63°; 95% CI, −2.28° to 1.01°) (Figure 3A). The vOCR on the lesion side in the chronic group was not lower than the control group (chronic vs control: −1.37°; 95% CI, −2.96° to 0.21°). The vOCR on the nonlesion side was not significantly different by group (mean [SD], acute, 3.80° [1.93°]; subacute, 3.89° [1.36°]; chronic, 4.00° [1.64°]; and control, 5.43° [1.55°]) and between groups (acute vs chronic: −0.20°; 95% CI, −1.84° to 1.44°; acute vs control: −1.63°; 95% CI, −3.27° to 0.01°; subacute vs control: −1.54°; 95% CI, −3.18° to 0.10°; subacute vs chronic: −0.11°; 95% CI, −1.75° to 1.53°; and chronic vs control: −1.43°; 95% CI, −3.01° to 0.15°).

There was a significant association of the measurement side and the vOCR values (η² = 0.34; 95% CI, 0.11-0.52). There was also a significant interaction between the side of vOCR measurement and participant groups (η² = 0.24; 95% CI, 0.04-0.41). Similarly, the participant group had a significant association with the asymmetry ratio as the measure of difference in vOCR between the lesion and nonlesion sides (mean [SD] acute, 0.30 [0.38]; subacute, 0.08 [0.17]; and chronic, 0.02 [0.11]; η² = 0.22; 95% CI, 0.004-0.46) (Figure 3B). Post hoc comparisons showed that the lesion-side vOCR was lower than the nonlesion-side vOCR in the acute patient group (−1.61°; 95% CI, −2.32° to −0.89°) (Figure 3A). No significant difference was found between the VOCR value on either side in the subacute, chronic, or control groups (subacute: −0.52°; 95% CI, −1.24° to 0.19°; chronic −0.001°; 95% CI, −0.67° to 0.66°; and control: −0.05°; 95% CI, −0.72° to 0.61°). The vOCR asymmetry in the acute group was higher than in the chronic group (0.28; 95% CI, 0.06-0.51), but there was no significant difference between the acute and subacute groups (0.22; 95% CI, −0.01 to 0.45) or between the subacute and chronic groups (0.07; 95% CI, −0.16 to 0.29).

We used ROC analysis to evaluate the performance of the vOCR test (index test) using the mean vOCR measurements from both sides in patients and healthy controls (Figure 3C). The performance of the test in discriminating between patients with vestibular loss and healthy controls, determined by the area under the curve, was 0.83. The best vOCR threshold to detect vestibular loss was 4.5°, with a sensitivity of 80% (95% CI, 0.62%-0.88%) and specificity of 82% (95% CI, 0.57%-1.00%) (Figure 3C). At this threshold, the true-positive value was 33; true-negative, 12; false-positive, 3; and false-negative, 8.

Figure 4 shows examples of vOCR measurement in patients with acute, subacute, and chronic vestibular loss as well as a healthy control individual. In the patient with acute loss, there was a vOCR asymmetry with 1.5° on the side of vestibular loss and 5° on the healthy side. In the patient with subacute loss, there was also an asymmetry, with a vOCR of 1.7° on the side of vestibular loss and 4.4° on the healthy side. In the patient with chronic loss, there was no significant asymmetry and vOCR was reduced at approximately 2.5° on both sides. In the healthy control individual, vOCR was approximately 5° on both sides. Thus, with 30° tilt, a vOCR value less than 4.5° indicated vestibular loss with more reduced values in patients with acute vestibular loss.

In this study we investigated whether the vOCR can be used as a clinical test to detect loss of otolith function. Our findings show that the vOCR test can be easily applied with a bedside tilt of the head and torso en bloc. Using a 30° tilt, vOCR values less than 4.5° could detect patients with loss of vestibular function compared with healthy controls with an accuracy of 83%. In patients with acute loss, vOCR was lower on the side of vestibular loss, while the asymmetry was reduced in subacute and chronic groups. These findings suggest that the vOCR deficit is pronounced early following vestibular loss but can recover over time.

Under normal circumstances, when the head is stationary, the utricular inputs are in a state of equilibrium, similar to the balance between neural inputs from the semicircular canals at rest.^2,26 Removal of the utricular inputs from one side can cause imbalance, leading to a pathologic OCR response, skew deviation of the eyes, and a roll head tilt toward the side of vestibular loss.^12,27-31 Such otolith-related findings are compensated as presumably the balance in neural activity between the 2 vestibular nuclei is restored to a new set point over time, just as the spontaneous nystagmus from canal imbalance also declines following unilateral vestibular loss.^11,32,33 Consistent with such recovery, in this study we found symmetric reduction in vOCR on both sides in patients with chronic vestibular loss, but the vOCR deficits in those with acute vestibular loss were asymmetric with more reduced values on the side of vestibular loss. Future studies should investigate whether such recovery in vOCR corresponds with improvement of daily function and vestibular symptoms in a same group of patients. If a significant association is found, vOCR can be used to track recovery in addition to detecting loss of otolith function. The noninvasive and quick application of the vOCR test can facilitate follow-up examinations for tracking recovery in patients.

The differences in vOCR among patients and healthy controls show that the vOCR can be valuable as a clinical test of otolith function to detect the extent and pattern of otolith hypofunction. Otolith pathways are commonly affected in vestibular disorders and, although vHIT has been widely used for clinical assessment of canal function, a VOG method for evaluation of otolith-ocular function has been lacking. Similar to vHIT, vOCR is measured using high-speed VOG goggles and automated analysis software providing immediate results. Thus, if combined, these tools can be used as one VOG battery for evaluation of both otolith and canal functions. Such a VOG battery can also help to detect the side of otolith hypofunction as the vOCR on the nonlesion side can be reduced especially in patients with chronic loss of vestibular function (Figure 3 and Figure 4). In such cases with symmetric vOCR deficit, vHIT findings can be helpful to detect the side of vestibular hypofunction, whereas asymmetric vOCR deficits would be consistent with unilateral loss of otolith-ocular function with a lower vOCR value on the side of vestibular loss (Figure 4). Based on our results, the vOCR threshold to detect loss of otolith function was 4.5°. Future studies should examine patients with other causes and varying degrees of peripheral vestibular loss to establish more accurate thresholds for detecting loss of otolith-ocular function.

In this study we did not include dynamic vOCR during lateral movement of the head because it would result in stimulation of both the canals and otoliths, and thus the outcome would not be specific to the otoliths. A previous study compared vOCR during a static head position with the laboratory-based, vestibular-evoked myogenic potentials measurements of both utricular and saccular functions.²¹ Among various measurements, the click and tap ocular VEMP correlated with the vOCR values, consistent with all being measures of otolith-ocular or utricular function.²¹ Various methods have been reported for OCR measurement, using either VOG or scleral search coils, but the real-time vOCR measurement with a bedside maneuver was not used as a clinical test of otolith function.^{3,15,21,33-36} Another method for assessing otolith function is through the VOR generated by translation of the head.^37-39 The normal VOR response during translation of the head is usually less than compensatory and it is often supplemented by saccade, which depending on the viewing distance, can facilitate bringing the eyes to the target.⁴ Also, the manual head translation for VOR measurement can be contaminated by unintended head rotation that stimulates the semicircular canals.³⁹ Thus, using the translational VOR as a bedside test of otolith function is not as feasible as applying static head tilt for vOCR measurement.

The present study has limitations. The sample size was small and the case-control design can overestimate the measures of test performance. In addition, vOCR was recorded at different time points in separate groups of patients and there was no direct measure of compensation or the outcome of interventions such as vestibular physical therapy to link with plausible vOCR recovery among patients. The heterogeneity of the vOCR deficit among the patient groups can also affect the accuracy of vOCR thresholds in the ROC analysis as the deficit becomes more symmetric in later stages of recovery. In addition, our sample was limited to patients with vestibular nerve resection and only 1 tilt angle for vOCR measurement (ie, 30°). Future studies should examine patients with larger tilt angles and include patients with other causes and varying degrees of peripheral vestibular loss to establish more accurate thresholds for detecting loss of otolith-ocular function.

In this study, our findings suggest that vOCR measurement can be easily applied as a bedside test with a simple tilt maneuver to detect or track loss of otolith-ocular function at different stages of compensation. Although patients with acute vestibular loss had lower vOCR values on the side of loss, patients with more chronic loss had symmetrically reduced vOCR values. Such a symmetrical vOCR deficit can be a result of long-term central adaptive processes and may also correspond with the state of recovery in patients.

Accepted for Publication: February 3, 2021.

Published Online: March 25, 2021. doi:10.1001/jamaoto.2021.0176

Correction: This article was corrected on August 12, 2021, to fix an error in the key to Figure 3.

Corresponding Author: Amir Kheradmand, MD, Vestibular and Ocular motor (VOR) Laboratory, Department of Neurology, The Johns Hopkins Hospital, 600 N Wolfe St, Path 2-210, Baltimore, MD 21287 (akherad@jhu.edu).

Author Contributions: Drs Sadeghpour and Otero-Millan had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Otero-Millan, Carey, Kheradmand.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Sadeghpour, Otero-Millan, Kheradmand.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Sadeghpour, Otero-Millan, Kheradmand.

Obtained funding: Kheradmand.

Administrative, technical, or material support: Sadeghpour, Fornasari, Carey, Kheradmand.

Supervision: Kheradmand.

Conflict of Interest Disclosures: Dr Otero-Millan reported receiving grants from the National Institutes of Health (NIH) during the conduct of the study and software used in this study (not patented) was licensed to Labyrinth Devices, LLC in 2016 and Interacoustics A/S in 2018. Dr Zee reported receiving royalties from Oxford University Press outside the submitted work. No other disclosures were reported.

Funding/Support: This work was supported by grants from the National Institute of Deafness and Other Communication Disorders (NIDCD; K23DC013552), National Eye Institute (NEI; K99EY027846), and the Leon Levy foundation.

Role of the Funder/Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank the pictured individual for granting permission to publish this information.