Patient subjects and healthy controls
We prospectively consented 66 consecutive and symptomatic patients diagnosed with CVD (n=30) or VM (n=36) from the Taichung Tzu Chi Hospital Neurology Dizziness Clinic in Taiwan between January 1, 2013 and December 31, 2013. Healthy volunteers with age and gender matched with CVD were recruited to be the healthy controls (n=30). All the patients who had prolonged vestibular symptoms (≧ 24 hours) without clear diagnoses underwent brain MRI. Patients with CVD were diagnosed by the results of brain MRI and had verified brainstem and/or cerebellar lesions confirmed by neuroradiologists. Patients with VM were diagnosed using the accepted diagnostic criteria of VM as published in the International Classification of Vestibular Disorders (ICVD) (Additional Table 1) . Each of the 66 patients were experiencing vertigo, dizziness or unsteadiness at the time of their clinical exam.
Any patient with history of a peripheral vestibular disorder (i.e. BPPV or vestibular hypofunction) was excluded. In the VM group, those with a history of structural brain lesions were excluded. In the CVD group, any patient with a preceding history of migraine was included. Healthy controls were excluded if they reported vestibular symptoms. MRI with MR angiography was performed in VM patients 3-6 days after clinical exam to rule out brainstem, cerebellar lesions, or vertebrobasilar TIA when their clinical exam was suggestive of a central lesion (i.e. video-oculographic confirmed gaze-evoked nystagmus).
A board-certificated neurologist performed a structured clinical examination on each patient to include: 1) Subjective Visual Vertical (SVV) using the bucket test, 2) ABCD2, 3) headache and vertigo history, 4) focal neurological (i.e. proprioception, vision) exam, 5) video-oculography examining for spontaneous, gaze evoked, positional, and head shaking-induced nystagmus, and 6) clinical head impulse testing. Healthy control subjects completed only the SVV test.
Subjective visual vertical using bucket test
Deviation of the SVV is a clinical sign of a deficit involving the graviceptive pathways . Tests of SVV are widely used in neuro-otological examinations to detect otolithic dysfunction, and to help diagnose central vestibular disorders [24, 25]. The test is performed with subjects seated upright looking into an opaque plastic bucket, with the head placed inside the rim of the bucket to prevent visual orientation cues. A straight, yellow diametric line is placed on the interior and bottom of the bucket. On the exterior, the bottom of the bucket held a protractor (180°), with a zero line at 90° corresponding to the true vertical. A weighted string was suspended from the center of the bucket bottom and served as the plumb line for which the reading was made . For each measurement of SVV, the examiner rotated the bucket to an initial displacement, and from there the subject rotated the bucket clockwise or counterclockwise to an endpoint, stopping when the inside line appeared to be vertical (Additional Fig. 1). The examiner noted the position of the plumb line on the protractor. Three trials were performed, with an inter-trial interval of one minute. Mean values of the SVV deviations were calculated for all subjects. We defined the normal range of SVV deviation as determined by the bucket test to be 0 ± 2.3° based on the literature .
The ABCD2 exam (Additional Table 2) combines points for Age, Blood pressure, Clinical features, Duration of symptoms, and presence of Diabetes as means to help predict the risk of stroke after having a transient ischemic attack. ABCD2 scores ≧4 are defined as having a higher risk of a future cerebrovascular event .
Spontaneous and gaze-evoked nystagmus (i.e. the nystagmus induced by eccentric gaze 30 degrees from central position) in upright and seated position was first examined in room light. Next, video-oculography (Synapsys, France) was employed and these two exams were repeated, along with Dix-Hallpike test, supine roll test, and the head-shaking nystagmus test.
Focal neurological signs
Formal neurological examinations were performed on all patients, including clinical assessments for cognition and cranial nerves, manual muscle tests, deep tendon reflex and Babinski’s sign, sensory tests (pinprick, light touch, vibration, and joint position sense), tests for limb ataxia (finger-to-nose test, heel-to-shin test, finger tapping, and foot tapping), and gait. The definition of having abnormal focal neurological signs was the presence of one or more of the following, (i) dysfunction of cranial nerves, (ii) weakness or upper motor neuron signs, (iii) sensory defects, or (iv) limb ataxia.
Postural imbalance or unsteadiness, which may appear in the acute stage of various vestibular disorders, were not included as focal neurological signs. Alteration of consciousness or other cognitive impairments were also excluded as a positive focal neurological sign, as they were more likely to be affected by diffuse cortical or non-neurological processes such as metabolic disturbances or drug effects.
Head impulse test
Horizontal head impulse testing was performed at the bedside. Patients were instructed to look at the examiner’s nose, and the examiner quickly turned the patient’s head with small amplitude, moderate velocity, and high acceleration head rotations while observing the patients’ eyes. The existence of a re-fixation saccade was defined as “positive.” Repetitive head impulses of unpredictable timing and direction were applied in attempt to reduce the presence of covert saccades and increase the test sensitivity .
All data was assessed for normality. Student’s t-test was used to examine continuous data between groups. For non-parametric data, we used the Mann-Whitney U-test (i.e., absolute deviation of SVV). Chi-squared analysis and the Fisher’s exact test compared the categorical variables between the groups. In order to determine how well the SVV bucket test distinguished CVD from VM, we used the receiver-operating characteristic (ROC) curve and calculated the area under the ROC curve (AUC). For investigation of diagnostic accuracy, the sensitivity, specificity, positive likelihood ratio and negative likelihood ratio of our different clinical measures were determined and compared using the McNemar test. The comparisons were set at a minimum significance level (α) to 0.05, but for multiple comparisons between CVD and VM we performed a Bonferroni correction with an α-level adjusted to a minimum of 0.0041. Statistical significance was assessed with SPSS (version 23) (IBM SPSS Inc., Chicago, IL, USA).