The vestibular system, in conjunction with other systems, maintains equilibrium and balance. Anatomically, it can be separated into peripheral components, which reside within the inner ear and petrous temporal bone, and central components, which are located primarily within the caudal portion of the brainstem but interact with structures in the cerebellum, midbrain, and cranial cervical portion of the spinal cord. Exclusive of seizures, vestibular dysfunction is one of the most common indications for advanced imaging (MRI or CT) of the head in small animal neurology patients.
Diagnosis of vestibular disease and, in large part, determination of whether to pursue advanced imaging and the type of advanced imaging to pursue are highly dependent on findings of the neurologic examination. Other than neurologic examination, no clinical tests exist that can establish the presence of vestibular dysfunction. In addition, a definitive diagnosis of vestibular disease associated with dysfunction of central components often requires more expensive and aggressive diagnostic testing and treatment.1 Therefore, it is important that neurologic examination findings indicative of vestibular disease, as well as localization of the dysfunction to central or peripheral structures, be highly reliable and strongly predictive of eventual definitive diagnosis.
Key neurologic examination findings generally agreed upon as indicative of vestibular dysfunction include head tilt; abnormal nystagmus; vestibular ataxia characterized by leaning, falling, or rolling to 1 side, with or without neck and trunk deviation with concavity to the same side; and vestibular strabismus, defined as a unilateral ventral or ventrolateral strabismus induced on extension of the head and neck.1–6 Other lesion locations, specifically the thalamus, are also reportedly associated with many of these signs,7 although the relative frequency of these signs in the context of isolated thalamic lesions of different etiologies is not well understood. Also, little is known about the frequency of occurrence of individual vestibular signs as related to the identification of lesions within vestibular structures.
Two studies3,8 have been conducted to investigate, at least tangentially, the relationship between interpretation of neurologic examination findings and eventual localization and etiologic diagnosis of vestibular disease in dogs. Neither of these studies included a comparison group of dogs without vestibular disease, and without such a group, one cannot determine which test findings (in this situation, interpretation of examination and MRI results) have value in the diagnosis of vestibular disease.
The purpose of the study reported here was to explore the reliability of interpretation of neurologic examination results for the identification and localization of vestibular disease in a large number of dogs. On the basis of previous research,3,8 we expected that there would be excellent agreement between clinical localization of lesions to central vestibular structures and identification of lesions of central vestibular structures on MRI and that there would be fair to good agreement between clinical localization of lesions to peripheral vestibular structures and identification of lesions of peripheral vestibular structures on MRI. Our goal was to determine whether specific features of the clinical examination were reliably associated with the presence and location of MRI-identified lesions affecting the vestibular system in dogs.
Materials and Methods
Animals
Electronic medical records of the Texas A&M University Veterinary Teaching Hospital were reviewed to identify dogs that underwent MRI of the head for diagnosis of a neurologic problem between September 5, 2011, and September 30, 2015. This period was selected because it corresponded with the availability of 3-T MRI. Dogs were required to have a completely recorded neurologic examination to be included in the study.
Medical records review
Two investigators (CED and CEB) reviewed the medical records and extracted information regarding patient signalment, neurologic examination findings at initial evaluation, MRI findings, and CSF analysis results, when available. Neurologic examination findings were tabulated, with abnormalities scored as present or absent, by use of a simplified version of the neurologic examination form in clinical use at the institution (Supplementary Appendix S1, available at avmajournals.avma.org/doi/suppl/10.2460/javma.252.7.830). Lesions identified via MRI were categorized by anatomic region (cerebral cortex, basal nuclei, thalamus, caudal brainstem [comprising the midbrain, pons, and medulla], cerebellum, spinal cord segments C1 through C3, middle or inner ear structures, CN VIII, nonneurologic lesion, or no lesion). Anatomic regions that were visibly deformed because of compression from a mass or herniation of other anatomic regions were considered to have a lesion. For example, midbrain compression due to caudal transtentorial herniation secondary to a forebrain mass effect would be scored as both a forebrain lesion and a caudal brainstem lesion.
The MRI-based diagnosis for each dog was initially considered the primary differential or presumed imaging diagnosis and was categorized as intra-axial or extra-axial mass, ischemic infarct, large intracerebral hemorrhage, microbleed, tympanic bulla disease with no evidence of extension to central structures, otitis media or interna with extension to meningitis or encephalitis, or meningoencephalitis of any other origin. Microbleeds were defined as ≥ 1 well-defined round or ovoid intraparenchymal signal voids < 5 mm in diameter on T2*-weighted images that were isointense or mildly hypointense on T2 spin-echo images, isointense on T1-weighted images, and nonenhancing on T1-weighted post–gadolinium administration images. For tympanic bulla disease with no evidence of extension and ischemic infarcts, a coincident CSF analysis with no evidence of pleocytosis (≤ 5 WBCs/μL) was required to substantiate diagnosis. For meningoencephalitis, a coincident CSF analysis with findings of pleocytosis (> 5 WBCs/μL) was required. For dogs for which multiple differential diagnoses were given equal probability or no CSF sample was available, or when diagnoses did not fall into any of the aforementioned categories, cause was classified as other. Dogs could have had ≥ 1 lesion category recognized on a single MRI examination.
To identify an appropriate subset of dogs for independent observer review, dogs were initially categorized as having signs of peripheral, central, or no vestibular dysfunction in accordance with the method adapted from de Lahunta and Glass9 as described by Troxel et al3 (Supplementary Appendix S1). Vestibular ataxia was defined as ataxia characterized by leaning, falling, rolling, or circling to 1 side, with or without flexing of the neck and trunk with concavity to the same side (adapted from a report by Rossmeisl1). Vestibular strabismus was defined as an induced ventral to ventrolateral strabismus (dropped globe) present unilaterally and apparent only when extending the head and neck (adapted from a report by Rossmeisl1). Induced unilateral ventrolateral strabismus, although commonly included in lists of clinical signs of vestibular disease,1–5,9–12 is excluded in other references13 and remains somewhat controversial in clinical practice. For this reason, dogs in which the only potential sign of vestibular dysfunction was vestibular strabismus were initially evaluated separately. In brief, patients with any of the findings associated with vestibular dysfunction were classified as having vestibular dysfunction. Dogs were then classified as having central vestibular dysfunction if they had deficits that would not be possible with a purely peripheral disorder, such as mental status change; CN deficits other than those of CN VII or VIII or Horner syndrome; postural reaction deficits; or cerebellar signs (eg, dysmetria or intention tremors). The remaining dogs with signs of vestibular dysfunction were classified as having peripheral vestibular dysfunction.
MRI protocol
All MRI examinations had been performed with a 3.0-T scannera with dogs anesthetized and positioned in sternal recumbency. A standard series of images was obtained for all dogs (Supplementary Appendix S2, available at avmajournals.avma.org/doi/suppl/10.2460/javma.252.7.830). Contrast-enhanced T1-weighted images (magnetization-prepared, rapid gradient-echo, fat-saturated sequence) had been obtained following a single IV injection of gadopentetate dimeglumine (0.2 mL/kg [0.09 mL/lb]).
CSF analysis
Clinicopathologic reports of findings for CSF samples collected from the cisterna magna at the time of MRI were retrospectively evaluated. A normal CSF result was defined as a protein concentration ≤ 25 mg/dL and total nucleated cell count ≤ 5 cells/μL. Samples with results that exceeded these values but were judged by pathologist review to be most consistent with hemodilutional effects with no other abnormalities were also classified as normal. All other reported findings were classified as abnormal.
Agreement in diagnostic test results
On the basis of the initial neuroanatomic classification assigned by 1 investigator (CEB) by use of the described rubric, a subset of 53 dogs was chosen by a different investigator (GTF), who had no knowledge of other aspects of their health status. This subset included 24 dogs classified as having peripheral vestibular dysfunction, 24 classified as having central vestibular dysfunction, and 5 classified as having nonvestibular dysfunction. All but 1 dog with peripheral vestibular dysfunction were selected, and the other groups were arbitrarily selected from eligible dogs. A small number of dogs with no vestibular signs were selected to ensure that observers were blinded to the initial diagnosis, and an equal number of dogs with central and peripheral vestibular disease were selected to establish an approximate true prevalence of 50% in the evaluated population (1:1 ratio of central vs peripheral disease).
The 50% prevalence was desired because the value of Cohen κ, the statistic used to measure agreement among results of CSF analysis, neurologic examination, and MRI examination, can vary with prevalence and we wished to equally weight the sensitivity and specificity of diagnosis. Purely random selection would have resulted in approximately 43 dogs with clinical central vestibular disease and only 7 with clinical peripheral vestibular disease. Furthermore, a 1:1 ratio also eliminated the potential bias associated with observers guessing that the lesion type was the one most common in the population.
Neurologic examination findings for the subset of dogs were distributed in anonymized, randomized, tabulated form to 3 independent observers (JM, JML, and KMA), who were asked to score each dog as having signs of peripheral, central, or nonvestibular dysfunction on the basis of that recorded information. A consensus clinical diagnosis for a given dog was defined as agreement in scores between ≥ 2 observers.
The MRI scans for the subset of dogs were distributed independently and anonymously to 3 independent observers (JM, JML, and AKV), who were asked to note the location of any lesions on the entire MRI series and to report whether the lesions were expected to cause peripheral vestibular disease, central vestibular disease, or neither. No instruction was given to observers regarding which anatomic regions should be considered as causing vestibular disease, but each was independently debriefed after evaluation to determine the classification criteria used, and all used the same regions.
Lesions associated with peripheral vestibular disease included any abnormalities affecting the inner ear, petrous temporal bone, CN VIII, or middle ear, which has been reported to be a common abnormal finding in patients with peripheral vestibular signs.3,8 Lesions that could be expected to produce central vestibular disease included the caudal brainstem (midbrain, pons, and medulla), cerebellum, cerebellopontine angle, and first 3 segments of the cervical portion of the spinal cord. These could be primary or secondary to disease at another location (eg, compression and herniation). Nonvestibular lesions were considered those affecting exclusively the forebrain (all cerebral gray and white matter, thalamus, or hypothalamus) or only extracranial structures (eg, atrophy of the muscles of mastication). Because of a previous report7 of vestibular signs in patients with exclusively MRI-identified thalamic lesions, the same observers were asked separately to report presence or absence of a thalamic lesion for each dog.
Statistical analysis
Categorical data are reported as frequencies, proportions, and 95% mid-P exact CIs. Proportions were compared between groups with the χ2 or Fisher exact test by use of statistical software.b Quantitative data were assessed for normality of distribution by use of descriptive statistics, histograms, and the Anderson-Darling test.c Because the normality assumption was violated, these data are reported as median and interquartile range. Comparisons of quantitative data among groups were performed by use of the Kruskal-Wallis test followed by pairwise comparisons with the Mann-Whitney U test with Bonferroni correction.
Logistic regression was used to evaluate associations between dog-level factors and a diagnosis of vestibular disease (vs no vestibular disease) within all available data (n = 496) and also with a diagnosis of central vestibular disease (vs peripheral vestibular disease) within the subset of dogs with vestibular disease (185). Univariate analysis was performed to identify all predictor variables with a value of P < 0.20, and those variables were then evaluated in multivariable models.d A backward stepwise approach to model building was used, and variables were removed one by one on the basis of the largest Wald P value until all remaining variables had a significant (P < 0.05) value. Interaction terms were not evaluated in the multivariable models, and model fit was assessed with the Hosmer-Lemeshow goodness-of-fit test.
Agreement among diagnostic test results (CSF analysis, neurologic examination, and MRI examination) and among the 3 observers was assessed by calculation of the Cohen κ statistic14 with the aid of a spreadsheet application.e Cohen κ values of ≤ 0.20, 0.21 to 0.40, 0.41 to 0.60, 0.61 to 0.80, and 0.81 to 1.00 were classified as poor, fair, moderate, good, and very good agreement, respectively.15 Values of P < 0.05 were considered significant for all analysis, except where specified otherwise.
Results
Four hundred ninety-six dogs met the study inclusion criteria. Of these, 185 (37%) had ≥ 1 sign of vestibular dysfunction as defined for the study; the remaining 311 (63%) had no neurologic abnormalities or had abnormalities attributed exclusively to nonvestibular sites. Of the 185 dogs with ≥ 1 sign of vestibular dysfunction, 160 (86%) had ≥ 1 sign of central vestibular dysfunction on the basis of the initial grading rubric used. Vestibular signs included head tilt (n = 103 [56%]), vestibular nystagmus (102 [55%]), abnormal nystagmus of any type other than listed (78 [42%]), vestibular ataxia (67 [36%]), circling (25 [14%]), neck or trunk flexion (21 [11%]), vertical nystagmus (16 [9%]), cerebellar ataxia (15 [8%]), and cerebellar tremors or titubation (15 [8%]). Most dogs with vestibular dysfunction (102 [55%]) had ≥ 1 sign. Considering the 83 dogs with only 1 sign of vestibular disease, vestibular strabismus (39/83 [47%]) and head tilt (24/83 [29%]) were most likely to occur as solitary signs of vestibular dysfunction. Dogs with vestibular strabismus as a sole sign of vestibular dysfunction had a distribution of MRI-identified lesions similar to those with other solitary or multiple signs of vestibular disease on neurologic examination (Table 1).
Number (%) of dogs (n = 496) with neurologic disorders with or without various clinical signs of vestibular disease in which lesions were identified in various anatomic regions via MRI examination of the head.
| Anatomic region | No vestibular disease (n = 311) | Vestibular strabismus only (n = 39) | Any single vestibular sign excluding strabismus (n = 44) | Any 2 vestibular signs (n = 59) | Any 3 or more vestibular signs (n = 43) |
|---|---|---|---|---|---|
| Caudal portion of the brainstem | 67 (22) | 16 (41) | 19 (43) | 28 (47) | 15 (35) |
| Cerebellum | 68 (22) | 15 (38) | 18 (41) | 27 (46) | 16 (37) |
| Cerebral cortex | 109 (35) | 15 (38) | 15 (34) | 22 (37) | 16 (37) |
| Thalamus | 35 (11) | 9 (23) | 14 (32) | 14 (24) | 9 (21) |
| Spinal cord segments 1–3 | 31 (10) | 13 (33) | 11 (25) | 12 (20) | 5 (12) |
| Middle or inner ear or CN VIII | 34 (11) | 6 (15) | 9 (20) | 7 (12) | 7 (16) |
| No lesion | 119 (38) | 4 (10) | 6 (14) | 8 (14) | 10 (23) |
| Basal nuclei | 7 (2) | 2 (5) | 5 (11) | 6 (10) | 4 (9) |
Clinical interpretation of vestibular dysfunction
Several variables differed significantly between dogs with and without signs of vestibular dysfunction (Supplementary Table S1, available at avmajournals.avma.org/doi/suppl/10.2460/javma.252.7.830). Dogs with versus without these signs were more likely to be female (P = 0.04), be ≥ 10 years of age (P < 0.001), have an abnormal result of CSF analysis (P < 0.001), and have received an MRI examination in 2014 or 2015 (vs previous years; P = 0.02). The German Shepherd Dog breed was overrepresented (P = 0.01) among dogs with vestibular dysfunction (Supplementary Table S2, available at avmajournals.avma.org/doi/suppl/10.2460/javma.252.7.830).
MRI identification of central vestibular lesions
Of 185 dogs with clinical signs of vestibular dysfunction, 151 (82%) had a lesion identified on MRI; 110 of these 151 (73%) dogs had lesions affecting central vestibular structures, and 29 (19%) had lesions affecting peripheral structures (Table 2). Most of these dogs (110/151 [73%]) had lesions in > 1 anatomic region. Of the 192 dogs with no clinical signs of vestibular dysfunction, 95 (49%) had lesions in > 1 anatomic region.
Comparison of the distributions of various MRI-identified lesion types in dogs with neurologic disorders with (n = 151) or without (192) signs of vestibular disease, categorized by whether lesions were in a central vestibular or other location.
| Central vestibular location | Other location | ||||
|---|---|---|---|---|---|
| Lesion type, by dog group | No. of dogs | Proportion (95% CI) | No. of dogs | Proportion (95% CI) | P value |
| Dogs with signs of vestibular disease | |||||
| Microbleed | 5 | 0.05 (0.02–0.10) | 16 | 0.39 (0.25–0.54) | < 0.001 |
| Caudal occipital malformation syndrome or syringohydromyelia | 19 | 0.17 (0.11–0.25) | 0 | 0 (0–0.07) | 0.004 |
| Mass | 29 | 0.26 (0.19–0.35) | 14 | 0.34 (0.21–0.50) | 0.35 |
| Infarction | 12 | 0.11 (0.06–0.18) | 2 | 0.05 (0.01–0.15) | 0.35 |
| Hemorrhage | 7 | 0.06 (0.03–0.12) | 1 | 0.02 (0.00–0.11) | 0.68 |
| Secondary hernia | 17 | 0.15 (0.10–0.23) | 1 | 0.02 (0.00–0.11) | 0.03 |
| Middle or inner ear lesion | 17 | 0.15 (0.10–0.23) | 6 | 0.15 (0.06–0.28) | 0.90 |
| Otitis media or interna with extension | 3 | 0.03 (0.01–0.07) | 0 | 0 (0–0.07) | 0.56 |
| Meningoencephalitis of unknown cause | 37 | 0.34 (0.25–0.43) | 6 | 0.15 (0.06–0.28) | 0.02 |
| Multiple | 91 | 0.83 (0.75–0.89) | 19 | 0.46 (0.32–0.62) | < 0.001 |
| Dogs without signs of vestibular disease | |||||
| Microbleed | 3 | 0.03 (0.01–0.08) | 6 | 0.07 (0.03–0.15) | 0.17 |
| Caudal occipital malformation syndrome or syringohydromyelia | 21 | 0.19 (0.13–0.27) | 0 | 0 (0–0.05) | < 0.001 |
| Mass | 31 | 0.28 (0.21–0.37) | 31 | 0.38 (0.28–0.49) | 0.16 |
| Infarction | 2 | 0.02 (0.01–0.06) | 4 | 0.05 (0.02–0.12) | 0.41 |
| Hemorrhage | 2 | 0.02 (0.01–0.06) | 2 | 0.02 (0.01–0.08) | 1.00 |
| Secondary hernia | 22 | 0.20 (0.14–0.28) | 1 | 0.01 (0.00–0.07) | < 0.001 |
| Middle or inner ear lesion | 16 | 0.15 (0.09–0.22) | 18 | 0.22 (0.14–0.32) | 0.25 |
| Otitis media or interna with extension | 1 | 0.01 (0–0.05) | 0 | 0 (0–0.05) | 1.00 |
| Meningoencephalitis of unknown cause | 29 | 0.26 (0.19–0.32) | 13 | 0.16 (0.10–0.25) | 0.12 |
| Multiple | 80 | 0.73 (0.64–0.80) | 15 | 0.18 (0.11–0.28) | < 0.001 |
Of dogs with signs of vestibular disease, 110 had a central vestibular lesion as identified via MRI, and 41 had a lesion elsewhere. Of dogs with no signs of vestibular disease, 110 had a central vestibular lesion, and 82 had a lesion elsewhere. Dogs could have had > 1 lesion, so these totals do not reflect total numbers of lesions.
Categorically, certain lesion types were more likely to be identified in specific locations. For example, caudal occipital malformation syndrome or syringohydromyelia was identified (by definition) only in the region of the cerebellum and medulla. Microbleeds, identified in 21 dogs with vestibular disease and an MRI-identified lesion, were confined to the forebrain in 16 (76%) dogs. Meningoencephalitis (of any cause) commonly affected central vestibular structures as defined in the study protocol. Middle and inner ear lesions were identified with equal frequency with and without concurrent lesions affecting central vestibular structures. Of the 160 dogs initially classified as having central vestibular disease, 98 (61%) had an MRI-identified lesion in a central vestibular location.
Circling was more common in dogs with MRI-identified lesions outside central vestibular structures (10/41 [24%]) than those with MRI-identified lesions in central vestibular structures (13/110 [12%]), but this difference was not significant (P = 0.15). Vertical nystagmus was not more common in dogs with a lesion affecting central vestibular structures than in dogs with lesions identified in other locations and was not identified in any dog with a lesion of only peripheral vestibular structures. Dysfunction of CNs IX and X was more common when MRI-identified lesions were in central vestibular structures, rather than in other locations (Table 3). However, none of these variables remained significant in the multivariable logistic regression model, the final version of which included only the variables Yorkshire Terrier breed (vs other breeds), Chihuahua breed (vs other breeds), and CSF analysis (yes vs no).
Comparison of neurologic examination findings by anatomic location of MRI-identified lesions in 185 dogs with vestibular disease.
| Central vestibular lesion (n = 110) | Other lesion location (n = 41) | No MRI lesion identified (n = 34) | |||||
|---|---|---|---|---|---|---|---|
| Finding | No. of dogs | Proportion (95% CI) | No. of dogs | Proportion (95% CI) | No. of dogs | Proportion (95% CI) | P value* |
| Central abnormality | |||||||
| Abnormal mentation | 39 | 0.35a (0.27–0.45) | 17 | 0.41a (0.27–0.57) | 4 | 0.12b (0.04–0.26) | 0.01 |
| Absent menace response | 42 | 0.38 (0.29–0.48) | 17 | 0.41 (0.27–0.57) | 7 | 0.21 (0.09–0.37) | 0.12 |
| Circling | 13 | 0.12a (0.07–0.19) | 10 | 0.24a (0.13–0.39) | 2 | 0.06b (0.01–0.18) | 0.047 |
| Neck or trunk flexion | 16 | 0.15 (0.09–0.22) | 3 | 0.07 (0.02–0.19) | 2 | 0.06 (0.01–0.18) | 0.25 |
| Cerebellar ataxia | 11 | 0.10 (0.05–0.17) | 3 | 0.07 (0.02–0.19) | 1 | 0.03 (0.00–0.14) | 0.41 |
| Cerebellar tremor or titubation | 8 | 0.07 (0.03–0.13) | 1 | 0.02 (0.00–0.11) | 6 | 0.18 (0.07–0.33) | 0.61 |
| CN deficit† | |||||||
| III | 1 | 0.01 (0.00–0.04) | 0 | 0 (0–0.07) | 0 | 0 (0–0.08) | 1.00 |
| IV | 1 | 0.01 (0.00–0.04) | 0 | 0 (0–0.07) | 0 | 0 (0–0.08) | 1.00 |
| V | 15 | 0.14 (0.08–0.21) | 4 | 0.10 (0.03–0.22) | 2 | 0.06 (0.01–0.18) | 0.43 |
| VI | 1 | 0.01 (0.00–0.04) | 0 | 0 (0–0.07) | 0 | 0 (0–0.08) | 1.00 |
| IX | 8 | 0.07a (0.03–0.13) | 0 | 0b (0–0.07) | 0 | 0b (0–0.08) | 0.02 |
| X | 8 | 0.07a (0.03–0.13) | 0 | 0b (0–0.07) | 0 | 0b (0–0.08) | 0.02 |
| Any of the above | 20 | 0.18 (0.12–0.26) | 4 | 0.10 (0.03–0.22) | 2 | 0.06 (0.01–0.18) | 0.13 |
| Any of the above, unilateral† | 9 | 0.45 (0.25–0.67) | 3 | 0.75 (0.24–0.99) | 1 | 0.50 (0.03–0.97) | 0.64 |
| VII | 12 | 0.11 (0.06–0.18) | 4 | 0.10 (0.03–0.22) | 7 | 0.21 (0.09–0.37) | 0.28 |
| VII, unilateral‡ | 11 | 0.92 (0.65–1.0) | 2 | 0.50 (0.09–0.91) | 5 | 0.71 (0.33–0.95) | 0.16 |
| Limb abnormality | |||||||
| Any postural deficit | 80 | 0.75 (0.67–0.83) | 30 | 0.73 (0.58–0.85) | 22 | 0.65 (0.48–0.79) | 0.47 |
| Any paresis | 44 | 0.44 (0.34–0.53) | 17 | 0.43 (0.28–0.58) | 15 | 0.45 (0.29–0.62) | 0.97 |
| 1 limb affected | 12 | 0.11 (0.06–0.18) | 3 | 0.07 (0.02–0.19) | 1 | 0.03 (0.00–0.14) | 0.17 |
| 2 limbs affected | 19 | 0.18 (0.11–0.26) | 8 | 0.20 (0.09–0.34) | 9 | 0.26 (0.14–0.43) | 0.52 |
| 3 limbs affected | 7 | 0.06 (0.03–0.12) | 3 | 0.07 (0.02–0.19) | 2 | 0.06 (0.01–0.18) | 0.97 |
| 4 limbs affected | 47 | 0.44 (0.34–0.53) | 18 | 0.44 (0.29–0.59) | 10 | 0.29 (0.16–0.46) | 0.31 |
| Ocular abnormality | |||||||
| Slow or absent PLR | 5 | 0.05 (0.02–0.10) | 5 | 0.12 (0.05–0.25) | 2 | 0.06 (0.01–0.18) | 0.23 |
| Anisocoria | 8 | 0.07 (0.03–0.13) | 2 | 0.05 (0.01–0.15) | 0 | 0 (0–0.08) | 0.21 |
| Vertical nystagmus | 12 | 0.11 (0.06–0.18) | 3 | 0.07 (0.02–0.19) | 3 | 0.09 (0.02–0.22) | 0.79 |
| Any abnormality | 19 | 0.17 (0.11–0.25) | 10 | 0.24 (0.13–0.39) | 6 | 0.18 (0.07–0.33) | 0.60 |
P value is the result of the χ2 or Fisher exact test. No dog had CN II dysfunction.
Only 20 dogs had complete information for central vestibular lesions, 4 dogs had complete information for other lesion locations, and 2 dogs had no MRI-identified lesion.
Only 12 dogs had complete information for central vestibular lesions, 4 dogs had complete information for other lesion locations, and 7 dogs had no MRI-identified lesion.
PLR = Pupillary light reflex.
Values with different superscript letters are significantly (P < 0.05) different, as indicated by pairwise χ2 or Fisher exact tests with Bonferroni correction.
In the multivariable model, Yorkshire Terriers had approximately 10 times the odds (OR, 10.50; 95% CI, 1.29 to 85.60; P = 0.03) and Chihuahuas had approximately 5 times the odds (OR, 5.28; 95% CI, 1.10 to 25.50; P = 0.04) of central vestibular lesions, compared with other breeds (Supplementary Table S3, available at avmajournals.avma.org/doi/suppl/10.2460/javma.252.7.830). Dogs with central vestibular lesions identified on MRI were less likely to have a cisternal CSF analysis performed than other dogs (OR, 0.28; 95% CI, 0.14 to 0.56; P < 0.001).
Diagnostic test agreement
In assessment of agreement between neurologic examination–based and MRI-based diagnoses, 2 dogs were excluded owing to lack of consensus among the 3 independent observers, leaving 50 dogs (24 dogs classified as having peripheral vestibular dysfunction, 21 classified as having central vestibular dysfunction, and 5 classified as having nonvestibular dysfunction). Of the dogs for which the consensus clinical diagnosis was peripheral vestibular dysfunction, 7 (29%) had MRI-identified lesions of peripheral vestibular structures, 5 (21%) had intracranial masses, and 3 (12%) had meningoencephalitis (Table 4). Of the dogs for which the consensus clinical diagnosis was central vestibular dysfunction, 3 (14%) had MRI-identified lesions of peripheral vestibular structures. Two of these dogs also had lesions identified in central vestibular regions, and the other had only a left-sided otitis media (and presumed otitis interna) identified via MRI. Neurologic examination findings in this other dog included tetraparesis and proprioceptive placing deficits in all limbs (decreased in both thoracic limbs and absent in both pelvic limbs).
Comparison of the distributions of MRI-identified lesions between dogs with central vestibular dysfunction (n = 21) and dogs with peripheral vestibular dysfunction (24) as diagnosed through the consensus of 3 independent observers on the basis of neurologic examination findings.
| Central vestibular dysfunction | Peripheral vestibular dysfunction | ||||
|---|---|---|---|---|---|
| Lesion type | No. of dogs | Proportion (95% CI) | No. of dogs | Proportion (95% CI) | P value |
| Microbleed | 2 | 0.10 (0.02–0.28) | 4 | 0.17 (0.06–0.35) | 0.67 |
| Caudal occipital malformation syndrome or syringohydromyelia | 4 | 0.19 (0.06–0.40) | 1 | 0.04 (0.00–0.19) | 0.17 |
| Mass | 6 | 0.29 (0.12–0.50) | 5 | 0.21 (0.08–0.40) | 0.55 |
| Infarction | 4 | 0.19 (0.06–0.40) | 0 | 0 (0–0.12) | 0.04 |
| Hemorrhage | 0 | 0 (0–0.13) | 0 | 0 (0–0.12) | 1.00 |
| Secondary hernia | 4 | 0.19 (0.06–0.40) | 0 | 0 (0–0.12) | 0.04 |
| Middle or inner ear lesion | 3 | 0.14 (0.04–0.34) | 7 | 0.29 (0.14–0.49) | 0.30 |
| Otitis media or interna with extension | 0 | 0 (0–0.13) | 0 | 0 (0–0.12) | 1.00 |
| Meningoencephalitis of unknown cause | 5 | 0.24 (0.09–0.45) | 3 | 0.13 (0.03–0.30) | 0.44 |
| Multiple | 16 | 0.76 (0.55–0.91) | 13 | 0.54 (0.34–0.73) | 0.12 |
Dogs could have had > 1 lesion.
Six (29%) dogs with central vestibular dysfunction had intracranial masses, and 5 (24%) had meningoencephalitis. Five (21%) dogs with peripheral vestibular dysfunction had intracranial masses, and 3 (12%) had meningoencephalitis. A high proportion of dogs (16/21 [76%]) with central vestibular dysfunction and 13 of 24 (54%) dogs with peripheral vestibular dysfunction had lesions affecting > 1 anatomic region. Dogs with intracranial infarctions and secondary hernias identified via MRI were more likely to be classified as having central versus peripheral vestibular disease by consensus.
Agreement was poor to fair among CSF analysis, MRI examination, and neurologic examination findings for the diagnosis of central vestibular disease (Table 5). Considering only the 340 dogs for which results of CSF analysis were available, the percentage for which CSF findings matched MRI findings (ie, both abnormal or both unremarkable) was only 59% (κ = 0.38), which was not much in excess of chance agreement. Agreement between CSF analysis results and neurologic examination findings consistent with a central lesion (κ = 0.17) was poor.
Agreement among results of CSF analysis, MRI, and neurologic examination and among 3 observers for recognition of lesions in dogs that underwent MRI examination of the head for neurologic problems.
| Variable | No. of dogs evaluated | κ (95% CI) | P value |
|---|---|---|---|
| Presence of CNS lesion (yes vs no) | |||
| CSF and MRI abnormalities | 340 | 0.38 (0.28–0.48) | < 0.001 |
| CSF and neurologic examination abnormalities | 340 | 0.17 (0.08–0.27) | < 0.001 |
| MRI and neurologic examination abnormalities | 496 | 0.21 (0.13–0.28) | < 0.001 |
| Interobserver agreement | |||
| Central or peripheral dysfunction (record evaluation) | 47 | 0.72 (0.55–0.89) | < 0.001 |
| No lesion present on MRI (yes vs no) | 53 | 0.68 (0.52–0.84) | < 0.001 |
| Nonvestibular lesion present on MRI (yes vs no) | 53 | 0.50 (0.34–0.65) | < 0.001 |
| Central vestibular lesion present on MRI (yes vs no) | 53 | 0.67 (0.52–0.83) | < 0.001 |
| Peripheral vestibular lesion present on MRI (yes vs no) | 53 | 0.85 (0.70–1.00) | < 0.001 |
| Thalamic lesion present on MRI (yes vs no) | 53 | 0.67 (0.52–0.83) | < 0.001 |
Considering all 496 dogs in the study, the percentage for which MRI findings matched neurologic examination findings was 70%; however, the Cohen κ value for agreement was only fair (κ = 0.21). Agreement among the 3 independent observers ranged from moderate to very good (Table 5). Very good agreement was achieved for the recognition of a peripheral vestibular lesion on MRI (κ = 0.85), and good agreement was achieved for the classification of central versus peripheral dysfunction on the basis of neurologic examination findings (κ = 0.72). Agreement was lowest, albeit moderate (κ = 0.50), for diagnosis of nonvestibular lesions via MRI.
Discussion
In the present study involving a large group of dogs undergoing MRI of the head because of a neurologic problem, 37% of dogs had signs of vestibular dysfunction. These dogs were older than the general population of dogs undergoing MRI for the same purpose and were more likely to have abnormal results of CSF analysis. German Shepherd Dogs may be predisposed to the clinical manifestation of vestibular dysfunction. Lesions were identified via MRI in most dogs with clinical signs of vestibular dysfunction (82%), and these dogs were more likely to have MRI-identified lesions in multiple anatomic regions (73%) than those undergoing head MRI without signs of vestibular disease (49%).
The retrospective nature of the study did not allow evaluation of the reliability of neurologic examination for localization of vestibular dysfunction, only its interpretation, and errors could have been made in the examination, resulting in the misclassification of vestibular dysfunction. Clinicians may also have missed isolated signs of vestibular dysfunction, such as nystagmus induced only when dogs were in dorsal recumbency, given that some clinicians may not routinely perform this test. Dogs with just 1 sign of vestibular disease were no more or less likely overall to have lesions of vestibular structures identified via MRI than those with multiple signs of vestibular disease, suggesting that misclassification in the presence or absence of vestibular dysfunction due to inadvertent exclusion of dogs with just 1 sign that was not recognized by the clinician was unlikely to alter the proportion of dogs with lesions of vestibular structures identified via MRI.
Consistent with a previous report3 of clinical signs in dogs with vestibular disease, we found a small association between CN IX and X dysfunction and MRI-identified central vestibular lesions (vs lesions in other locations or no MRI-identified lesion); however, this association was no longer significant on multivariable analysis, so the clinical usefulness of this finding remains unclear. Also consistent with the previous report,3 no difference was identified in the prevalence of vertical nystagmus as a function of MRI lesion location, although this sign was uncommonly reported overall, and never in isolation of other neurologic abnormalities indicative of central vestibular dysfunction. This supported the assertion of de Lahunta and Glass9 that vertical nystagmus is often associated with central vestibular dysfunction. No specific abnormalities noted on the standard neurologic examination per-formed at our institution were predictive of an MRI-identified central vestibular lesion.
When classifying MRI-identified lesions as central or peripheral on initial evaluation of the entire cohort of dogs in the present study, lesions exclusively involving the thalamus were excluded from those considered likely to cause vestibular signs. This choice was made after reviewing previous reports7,16 regarding the occurrence of vestibular signs in dogs with thalamic lesions because those reports pertained to a single cause (infarction) and it was unknown whether the findings would pertain to other causes and whether all dogs had lesions confined to the thalamus. In 1 report,16 the 3 dogs with thalamic lesions also had lesions in the medulla oblongata. In the other report,7 the authors indicated that many of the 8 dogs with vestibular signs had lesion extension into the midbrain and cited this as a likely explanation for the vestibular signs observed.7
In the present study, no instruction was given to observers reviewing the MRI scans on how to classify thalamic lesions with respect to the vestibular system, but none of them chose to classify lesions exclusive to the thalamus as vestibular lesions. However, because the frequency with which exclusively thalamic lesions may be related to clinical vestibular signs is unknown, a post hoc analysis was performed to determine interobserver agreement on MRI identification of thalamic lesions, revealing good agreement (κ = 0.67).
Absolute agreement in the diagnosis of central vestibular lesions in the present study was 70% between neurologic examination interpretation and MRI findings. Nevertheless, the calculated Cohen κ value was only 0.21, indicating fair agreement between the 2 tests. Several possibilities may explain this finding. The data set could have contained dogs with clinically unimportant abnormalities (incidental findings) in vestibular regions or dogs with clinically important lesions of the vestibular system that failed detection on the standard MRI sequences. In addition, lesions in anatomic regions that contain parts of the vestibular apparatus may nevertheless have spared those structures, either by chance or by differential sensitivity of substructures (eg, gray vs white matter) in the same anatomic region. Without a gold standard test for detection of vestibular disease, we could not determine to which degree each of these possibilities existed. Instead, focus was placed on the reliability of neurologic examination interpretation and MRI lesion identification. Because interobserver agreement for these 2 tests was good, we believe that these observers were unlikely to have contributed meaningfully to the poor agreement observed between neurologic examination and MRI identification of central vestibular disease.
In the blinded review of neurologic examination findings, the 3 observers in the present study were able to agree in their distinction between central versus peripheral vestibular dysfunction. Although this review included only a small subset of the entire sample population, the use of multiple blinded observers suggested that this result was not attributable to interpreter bias and would likely be generalizable across dogs and observers. Our study also showed that retrospective localization of central and peripheral dysfunction through review of medical records was feasible and reliable. However, interobserver agreement will vary from population to population if the prevalence of the condition changes, except in the rare circumstances when the sensitivity and specificity are equal within each test. An attempt was therefore made to create a subset of dogs with a peripheral and vestibular disease prevalence of 50% each for this purpose so that sensitivity and specificity would equally impact the κ statistic. As a result, this statistic did not represent the population from which the dogs were selected because the prevalence of central vestibular disease was higher in that population. For this reason, researchers wishing to replicate the findings reported here should also attempt to create a study population with a vestibular disease prevalence of 50%, rather than focus on the actual prevalence, which could change over time.
Good agreement among 3 blinded observers was also observed in the present study for identification and classification of lesions likely to affect vestibular function, according to classical definitions, on multiplanar 3.0-T MRI brain scans. Specifically, agreement on involvement of middle and inner ear structures (ie, peripheral vestibular structures) was very good. Although agreement was less good for detection of central vestibular lesions and for identification of visibly normal MRI scans, it was still good as defined for study purposes. Consequently, it was unlikely that the discordance between results of lesion localization based on neurologic examination findings and results determined from MRI findings was attributable to poor, biased, or unreliable interpretation of sets of data.
In the dogs of the present study, a consensus clinical diagnosis of peripheral vestibular dysfunction made by use of classical definitions9 was associated with MRI identification of either an intracranial mass or presumptive meningoencephalitis one-third (8/24) of the time. Other dogs with clinical signs consistent with peripheral vestibular dysfunction alone had additional abnormalities identified, although the clinical importance of these findings in the context of vestibular disease is debatable. Therefore, it is reasonable to expect that approximately one-third of dogs with only peripheral signs of vestibular dysfunction may have a central vestibular lesion subsequently identified via MRI.
Additionally, our data suggested that the frequency with which lesions of peripheral vestibular and associated structures (middle and inner ear) were identified via MRI did not differ between dogs in which the lesion was localized via consensus clinical interpretation to central and peripheral vestibular structures. This may indicate that middle ear disease can be an incidental finding of head and brain MRI examination and may not always be the cause of clinical neurologic signs.
Compared with previous studies involving MRI findings in dogs3,8 and cats17 with vestibular disease, agreement between neurologic examination interpretation and MRI lesion identification in the present study was lower. Although the reasons for this discrepancy remain unclear, the strong agreement among blinded observers in our study regarding both neurologic examination interpretation and MRI identification of lesions suggested that unreliable clinical judgment or poor sensitivity of MRI for lesion detection was an unlikely cause. One difference between the present and previous studies was our initial inclusion of all dogs undergoing MRI of the head, instead of inclusion of only dogs with a final diagnosis of vestibular disease, as was done in other studies, resulting in a greater proportion of dogs with unremarkable MRI findings or multifocal intracranial disease in the study reported here. However, we believe that this situation more accurately represented the typical clinical situation involving dogs evaluated for vestibular disease. Overall, results of the present study suggested that the neurologic examination, in its entirety, allowed reliable localization of peripheral versus central neurologic dysfunction across interpreters, but no individual feature of the neurologic examination bore a strong relationship to the location of MRI-identified lesions.
ABBREVIATIONS
| CI | Confidence interval |
| CN | Cranial nerve |
Footnotes
MAGNETOM Verio 3T, Siemens, Malvern, Pa.
Epi Info, version 6.04, CDC, Atlanta, Ga.
MINITAB statistical software, version 13.32, Minitab Inc, State College, Pa.
IBM SPSS Statistics, version 23, International Business Machines Corp, Armonk, NY.
Microsoft Office Excel 2010, Microsoft Corp, Redmond, Wash.
References
1. Rossmeisl JH Jr. Vestibular disease in dogs and cats. Vet Clin North Am Small Anim Pract 2010;40:81–100.
2. Lowrie M. Vestibular disease: anatomy, physiology, and clinical signs. Compend Contin Educ Vet 2012;34:E1.
3. Troxel MT, Drobatz KJ, Vite CH. Signs of neurologic dysfunction in dogs with central versus peripheral vestibular disease. J Am Vet Med Assoc 2005;227:570–574.
4. Kent M, Platt SR, Schatzberg SJ. The neurology of balance: function and dysfunction of the vestibular system in dogs and cats. Vet J 2010;185:247–258.
5. Cook LB. Neurologic evaluation of the ear. Vet Clin North Am Small Anim Pract 2004;34:425–435.
6. Thomas WB. Vestibular dysfunction. Vet Clin North Am Small Anim Pract 2000;30:227–249.
7. Gonçalves R, Carrera I, Garosi L, et al. Clinical and topographic magnetic resonance imaging characteristics of suspected thalamic infarcts in 16 dogs. Vet J 2011;188:39–43.
8. Garosi LS, Dennis R, Penderis J, et al. Results of magnetic resonance imaging in dogs with vestibular disorders: 85 cases (1996–1999). J Am Vet Med Assoc 2001;218:385–391.
9. de Lahunta A, Glass E. ScienceDirect (online service). In: Veterinary neuroanatomy and clinical neurology. 3rd ed. St Louis: Saunders Elsevier, 2009;324–328.
10. Garosi LS, Lowrie ML, Swinbourne NF. Neurological manifestations of ear disease in dogs and cats. Vet Clin North Am Small Anim Pract 2012;42:1143–1160.
11. Dewey CW, Da Costa RC. Practical guide to canine and feline neurology. 3rd ed. Hoboken, NJ: Wiley-Blackwell, 2016;268.
12. Platt SR, Olby NJ, British Small Animal Veterinary Association. BSAVA manual of canine and feline neurology. 4th ed. Quedgeley, Gloucester, England: British Small Animal Veterinary Association, 2013;155.
13. Lorenz MD, Kornegay JN, Oliver JE. Handbook of veterinary neurology. 4th ed. Philadelphia: Saunders, 2004.
14. Fleiss JL, Levin BA, Paik MC. Statistical methods for rates and proportions. 3rd ed. Hoboken, NJ: Wiley-Interscience, 2003.
15. Altman DG. Practical statistics for medical research. London: Chapman and Hall, 1991.
16. Garosi L, McConnell JF, Platt SR, et al. Clinical and topographic magnetic resonance characteristics of suspected brain infarction in 40 dogs. J Vet Intern Med 2006;20:311–321.
17. Negrin A, Cherubini GB, Lamb C, et al. Clinical signs, magnetic resonance imaging findings and outcome in 77 cats with vestibular disease: a retrospective study. J Feline Med Surg 2010;12:291–299.
