The “hype” of hydrops in classifying vestibular disorders: a narrative review

Multiple literature search strategies in PubMed and Google Scholar were performed in February and March 2019 to identify papers reporting the prevalence of EH and/or neuro-otologic symptoms in three types of study populations:

The search strategy consisted of electronic database searching, reference checking, and citation searching. A variety of MeSH (Medical Subject Headings) terms and free-text keywords were combined to conduct multiple search strategies in PubMed (see “Supplementary materials”). Based on title and abstract, one reviewer (ML) selected and summarized English published studies deemed relevant within the scope of this review. To complement the initial search, the reference lists of the primary selected studies, as well as the “cited by” feature of Google Scholar [10], were used to screen for additional studies to include. This process was iterated until no new eligible studies were identified. A comprehensive literature overview was composed without any selection restrictions.

The following data items were tabulated from the selected publications: Author, Year of publication, study design, recruitment strategy and definition of the study population, sample size, age, EH assessment techniques, symptom description, and study outcomes. To explore the diagnostic value of EH in patients with MD, additional data were collected from the studies that investigated the diagnostic role of EH in patients with MD. The following items were collected: sensitivity, specificity, positive predictive values, and negative predictive values.

Differences in study characteristics were identified and processed after the consultation of a second reviewer (MH). The minimal criteria for the combined assessment of the collected data were: similarities in recruitment strategy and definition of the study populations, study outcomes, and EH assessment technique. To homogenize study outcomes, point prevalences, frequency rates, and period prevalences up to 1-year were combined. These were regarded to indicate best the proportion of a population that has active symptoms.

Although several EH assessment techniques exist [11, 12], this review combined the results of the most commonly reported grading method for hydrops MR evaluation in literature [13, 14]. This method categorizes hydrops in either “mild” or “significant.” Hydrops prevalences reported with the lowest cut-off point (mild) for cochlear and/or vestibular hydrops [13] were combined for prevalence estimation.

To determine the reliability of the estimated prevalences and to allow better transparency of the results in this review, a quality assessment of data synthesis was performed based on five predefined criteria (for the reliability criteria see “Supplementary materials”). For each criterion that was met, a score of one point was allocated. The reliability of the mean prevalences was estimated to be high (5) or low (1).

To graphically cluster symptom and EH occurrence based on their relative prevalence in vestibular patients and health, the estimated prevalences were plotted in a four-dimensional (x, y, r, α) bubble chart using the open-source, interactive graphing library for Python, Plotly 3.10.0.[15].

The estimated prevalence of symptoms and EH in patients diagnosed with a vestibular disorder and healthy reference populations were plotted on the x-axis and y-axis, respectively. The estimated prevalence in the general population was displayed by the radius (r) of the bubbles: large bubbles reflected common symptoms. The reliability of the results was presented by the transparency of the bubbles in which transparent bubbles represented a low level of reliability, and opaque bubbles, a high level of reliability.

In total, 69 publications were selected after database searching and citation and reference retrieval for estimating the prevalence of EH and neuro-otologic symptoms in different types of study populations [3, 9, 12, 18‐34, 36‐84]. Based on the available data, three vestibular disorders were selected for prevalence estimation: (1) The clinical diagnosis of Menière’s disease based on diagnostic criteria (Table 1) [16], (2) The diagnosis of vestibular schwannoma confirmed by MRI, (Table 2), and (3) The clinical diagnosis of vestibular migraine based on diagnostic criteria (Table 3) [17]. Other vestibular diagnoses, such as benign paroxysmal positional vertigo and acute unilateral vestibulopathy, were not reviewed since no EH imaging studies in relation to these vestibular disorders were found. No post-mortem studies were found published on the prevalence of EH in patients with vestibular migraine. Due to the lack of contrast-enhanced imaging studies in patients with a vestibular schwannoma, two studies were selected that used non-contrast-enhanced MRI to diagnose EH [18, 19].

The characteristics of the individual studies are presented in the supplementary materials.

Within the selected publications, symptom descriptions varied, and were sometimes absent or not specified. The definition of EH also varied between different post-mortem studies. Four post-mortem studies defined EH as a displacement of Reissner’s membrane or the membranous walls in the vestibule [3, 9, 20, 21]. Other studies only examined EH in the cochlear turns and did not explicitly define EH in the vestibule [22‐24]. All contrast-enhanced imaging studies used the same semi-qualitative assessment technique [13] and the lowest cut-off point (mild) for grading EH.

The results of the estimation of symptom and EH prevalences, together with the quality assessment of data synthesis for study populations of the three investigated vestibular disorders, are shown in Tables 1, 2, 3. Overall, neuro-otologic symptoms were more prevalent in vestibular disorders compared to healthy references with no vestibular history, as demonstrated in Tables 1, 2, 3, 4, 5.

The mean prevalence of histologically diagnosed EH ranged from 12.5% in healthy references to 99.4% in patients with MD. Radiologically diagnosed EH prevalences ranged from 25.9% in patients with vestibular schwannoma to 82.3% in patients with MD. The EH prevalence ranges varied from narrow (98.8–100%) to extreme broad (0–100%) demonstrated in Tables 1 and 5 by the prevalence ranges of EH histologically diagnosed in MD patients and radiologically diagnosed in healthy references, respectively. The reliability of the prevalence estimations varied from low (1) to high (5) (median = 2, IQR = 2). The prevalence estimations in the general population were most reliable (median = 4, IQR = 1.5) compared to the prevalence estimations in patients with vestibular migraine (median = 1.5, IQR = 1).

Figures 1, 2, 3 show the mean prevalence of symptoms and EH, in three vestibular diagnoses plotted against the prevalence in healthy references. Consequently, different clusters of features arose at different locations in the figures. Clusters with similar prevalences on the x-axis and y-axis were encircled. Clusters of features plotted above the oblique reference line are more prevalent in healthy references, and clusters plotted below the oblique line are more prevalent in the vestibular disorder. The greater the distance between the plotted features and the reference line, the more the feature distinguishes between a physiological condition and pathological condition.

An overview of the results of three publications that investigated the diagnostic role of EH observed on contrast-enhanced MRI in patients with MD is shown in Table 6. All studies used the same method for grading EH [13] and had similar study protocols. (For further study details see “Supplementary materials”). Next to this, the diagnostic performances of three other image features related to MD, “vestibular endolymphatic space contacting the oval window,” the degree of “perilymphatic enhancement,” and “Saccule to utricle ratio,” were presented.

The sensitivity of the semi-qualitative grading method [13] for “mild” vestibular or cochlear hydrops was reported from 60 to 94%. The specificity was reported from 10 to 96%. The positive predictive value was reported from 74 to 95% and the negative predictive value from 74 to 93%. When shifting the cut-off value to the grading “significant” hydrops, the sensitivity and the negative predictive value decreased, while the specificity and positive predictive value increased [27, 39]. The sensitivities of the other image features ranged from 50 to 85% and their specificities from 92 to 100%.

In all three vestibular disorders studied, depending on prevalence, different and distinctive clusters of symptoms emerged. As expected in patients with vestibular schwannoma hearing loss, and tinnitus were the most common symptoms; while in patients with vestibular migraine headaches, and vertigo were more common. In patients with MD, as expected, vertigo, hearing loss, tinnitus, and aural fullness were most common. Nonetheless, headaches were also quite frequent.

These symptom clusters can be used to differentiate between vestibular disorders. However, the results of this review demonstrated four difficulties regarding symptom-based classifications. First, neuro-otologic symptoms occur, in small prevalences, in healthy reference populations. Although symptoms are often considered to be a derivative of pathological states, they can occur outside the direct context of a (vestibular) disease. Second, within a defined pathological state, symptom prevalences can range vastly. These ranges may be caused by the various symptom definitions used in the reviewed publications. Next to this, patients often find it hard to describe their (dizzy) symptoms [86], which might also explain the extensive ranges in symptom prevalences. Third, only 75% of the patients with MD experience subjective hearing loss despite the obligatory audiometrically documented hearing loss for the diagnoses of MD. Perhaps vertigo complaints dominate the perception of hearing loss. Fourth, the expression of auditory symptoms (hearing loss, tinnitus, and aural fullness) in patients with vestibular migraine and the presence of headaches in a subset of patients with MD suggest there is a substantial symptom overlap between the two entities. It is uncertain whether this overlap is due to the inclusion of incorrectly diagnosed subjects in the study populations, or that the clinical phenotypes resemble each other [40]. This raises the question whether other classifications are possible or even more appropriate. Symptoms are subjective, prone to terminology problems [1], and they potentially overlap within vestibular entities [40] and are, therefore, not always optimal for classifying vestibular disorders. The use of additional (bio) markers would be helpful.

The presence of EH as a histological marker for inner ear disorders has been proposed [9, 87]. Especially in MD, EH is considered to have important diagnostic value since it has been histologically demonstrated in almost every case of MD [3, 9].

The results of this review reaffirmed the potential diagnostic value of EH as a histologic marker in discriminating patients with MD from healthy persons. However, it is important to highlight the fact that EH is also found in normal temporal bones. Apical hydrops occurs in 15% of the human cochleae but is assumed to be of no pathologic significance [88]. The prevalence of EH in other parts of the healthy labyrinth is poorly documented but seems to occur sporadically [20].

Furthermore, EH is found in almost 50% of the patients with vestibular schwannoma and is also well documented in other neuro-otologic conditions such as otosclerosis, chronic otitis media, labyrinthitis, and Mondini dysplasia [87, 89]. Therefore, EH is not a pathognomic feature and the probability of finding hydrops in an otologic population seems considerable. EH may be a valuable marker for MD, but cannot provide a definitive diagnosis without the support of clinical data [14]. Confirming EH should, therefore, not directly indicate the diagnosis of MD.

Recent developments in contrast-enhanced MR imaging enabled the radiological diagnosis of EH in vivo. Its diagnostic value has been broadly investigated and debated in the literature [14]. A striking finding of this review was the discrepancy in prevalence existing between the histological and radiological diagnoses of EH, documented in MD and vestibular schwannoma. It might be hypothesized that EH is better visible in a more advanced disease stage when examining post-mortem, given that the severity of EH seems to increase with the duration of MD [90]. However, this does not explain the relatively higher prevalence of radiologically confirmed EH compared to histologically confirmed EH in healthy references (31% vs. 12.5%). It might be possible that the radiological diagnosis of EH contributes to more false-positive results in healthy subjects and more false-negative results in patients with vestibular disorders. This implies that the discrepancy in prevalence could be attributable to a radiological inaccuracy in demonstrating EH, which seems plausible given the numerous influencing factors such as the MR sequences, image resolution, and interobserver variability [6, 14, 91, 92]. Also, the extensive ranges of the radiological diagnosis of EH found in this review justify the assumption of a radiological inaccuracy of approximately 25% with respect to the post-mortem examination. However, it is important to acknowledge that, although histology is seen as gold standard for confirming hydrops, this technique might also be influenced by the presence of artifacts.

Although hydrops imaging successfully identifies EH in subjects with definite MD resulting in relatively high sensitivity (60%–94%) [25, 27, 39], “mild” hydrops is consistently reported in other vestibular disorders and even in healthy subjects. Using hydrops imaging as a diagnostic test might result in false-positive outcomes leading to the overdiagnosis of MD. As a diagnosis of MD might imply medical treatment or even surgical intervention, having a diagnostic test with high specificity and positive predictive value is critical. However, studies that investigated the diagnostic role of EH in patients with MD reported a wide range in test specificity (10–96%), despite comparable study protocols. The reproducibility of the semi-qualitative grading technique can, thus, be questioned.

As shown in Table 6, increasing the cut-off value towards significant hydrops will decrease the false-positive results and improve specificity, but it will decrease the sensitivity. Hydrops imaging seems to not yield optimal diagnostic performance on its own, though in combination with symptomology, it could play an important role in the diagnosis of MD.

MR imaging is nowadays the most commonly used method to detect EH in clinic [7]. The high prevalence of EH (82.3%) detected in patients with MD underlines the potential diagnostic value of hydrops imaging. Nevertheless, multiple challenges are related to correctly interpreting the value of the presence of EH.

The first challenge is to determine whether EH is clinically relevant. MR imaging detected EH in approximately 28% of the contralateral asymptomatic MD ears. Comparable rates in autopsy series were found [39]. This suggests that EH occurs in symptomatic and asymptomatic forms. If hydrops imaging would be able to identify clinically relevant but ‘silent’ ears, it would be extremely valuable for the early detection, treatment, or even prevention of vestibular disorders in the future. However, it is still uncertain whether asymptomatic hydrops is a preclinical disease state or not. The percentages found in MR imaging could correlate with 35% incidence of patients developing bilateral MD within ten years [93]. However, it might as well correlate with the 31% of EH found in healthy subjects [14, 25, 27, 29, 39, 84].

The second challenge is to determine whether the found EH is equal to a pathological state:

The endolymphatic space seems to vary highly between healthy subjects [39, 84]. EH in healthy individuals might be a deviation of a personal baseline but might also be a physiological variation of normal anatomy. One could hypothesize that EH in principal is an asymptomatic process, and/or that a personal threshold needs to be exceeded to become symptomatic. It is also possible that EH is merely a byproduct of inner ear damage [9] and, although more prevalent, is not always indicative for pathology.

The third challenge is regarding implications for clinicians. It is uncertain how to handle the group of patients who present with only one symptom like hearing loss or isolated vertigo spells [31]. After all, the time delay between symptom onset of hearing loss and vertigo can reach up to more than five years [31]. The prevalence of EH in this group did not substantially differ (77.7%) from the prevalence in patients diagnosed with definite or probable MD (82.3%). Yet, it is unclear whether these patients should be diagnosed and treated the same as patients with the full-blown symptom spectrum of MD based on the presence or absence of EH.

The challenges mentioned above arise since the phenomenon of EH is not fully understood. EH may be an entity on its own, causing the clinical syndrome of MD, but also other processes (e.g., EH as a byproduct, EH as an asymptomatic process, EH as an own entity “Hydropic ear disease”) might be applicable [8].

Next to the challenges mentioned above, it should be noted that several disadvantages are related to contrast-enhanced MR imaging. Although contrast agents have not shown evidence of ototoxicity [6, 94], leakage into the cerebral spinal fluid and deposits in brain tissue have been reported [95]. Moreover, contrast-enhanced MR imaging is not feasible for every patient due to contrast allergy, renal failure, and impaired labyrinthine enhancement after IT injection [6]. A non-invasive imaging technique would be preferable.

The methodology of this review was constrained by the approach of the search strategy and the subjective aspect of study selection, quality assessment, generalization, and synthesis of the collected data. This hampers the reproducibility of this review.

The high level of heterogenicity in the collected studies did not allow a traditional meta-analysis. Furthermore, this review has not excluded any publications based on their quality or study limitations. The collected data from the selected studies were liable to suffer from selection bias. These limitations may have influenced the results of the prevalence estimation. Therefore, the mean prevalences in this review should not be interpreted as definitive epidemiologic data. The purpose of this narrative review was not to provide a methodologically and statistically robust model but to generate a first concept on clustering symptoms and EH according to their relative prevalences in health and vestibular disorders.

This review focused on the most commonly used grading method for hydrops MR evaluation [13, 14]. With this grading method, hydrops imaging seems not to yield optimal diagnostic performance [25, 27, 39]. No consensus has been reached on how to assess the radiological finding of EH. Therefore, other parameters like the “Saccule to utricle ratio,” “vestibular endolymphatic space contacting the oval window,” and the degree of “perilymphatic enhancement” were also investigated and showed promising results regarding test specificity, as shown in Table 6. [25, 39, 85]. This implies that it might be worth to explore features “beyond” hydrops. After all, other processes might also be related to inner ear disorders. Recently, an exploratory study detected differences in image features extracted from conventional MR images between patients with MD compared to controls [96]. The approach of extracting and analyzing quantitative image features, which are sometimes not perceptual for the human eye, is referred to as Radiomics [97, 98]. Physicians often rely on the presence of objective abnormalities to diagnose disease, for example, EH. However, in case of the absence of objective pathology, it is not always correct to rule out disease. Unknown, not (yet) perceptible factors may play a crucial role in disease development and have yet to be discovered. Techniques like Radiomics and morphological-based assessment [25, 39, 85] of the labyrinth may play an essential role in the future in (re)classifying and diagnosing vestibular disorders.