Introduction

Tinnitus is a common, and sometimes very distressing, long-term health condition that involves the sensation of a sound without any explanation of an external auditory stimulus. Idiopathic subjective tinnitus is where the experience of the phantom sound is heard by the individual alone and there is no known etiology. This is by far the most common form of tinnitus representing 95–98% of all tinnitus presentations. The most common risk factor for developing a subjective tinnitus is an otological problem leading to hearing loss, but ototoxic medication, head injury, orofacial, rheumatological, endocrine and metabolic problems may also play a causal role [1]. Objective tinnitus is the less common variant whereby an observer can hear the tinnitus. For objective tinnitus, often a medical diagnosis can be made which determines an internal sound source, such as pulsatile tinnitus caused by vascular origin, or myoclonus of the middle ear [2] or palatal muscles [3].

Given that otological conditions, especially hearing loss, present one of the major risk factors for tinnitus, neuroplastic responses to the reduction in sound input are deemed to play a pivotal role in the condition [4]. Sensory deafferentation affects the balance of excitatory and inhibitory activity leading to changes in the spontaneous stochastic neural firing rate, changes in the temporal firing pattern of otherwise stochastic spontaneous activity and spatiotopic reorganisation of frequency tuning (e.g. [5,6]). Changes are observed throughout the auditory nuclei within the ascending pathway, although it is still unresolved precisely what neural phenomenon is the correlate of the tinnitus sensation.

Synaptic plasticity related to tinnitus may not be limited to the central auditory system. Across the past several decades, clinical reports have described patients who are able to alter their tinnitus through somatic manoeuvres [7–9]. For instance, eye movements, jaw protrusion, jaw clenching, head and neck muscle contraction manoeuvres, or touch to the face or hands may modulate the loudness (or less commonly the pitch) of the tinnitus [10–13]. Several authors have put forward a physiological explanation for the somatosensory modulation of tinnitus—the dorsal cochlear nucleus (DCN) disinhibition model [8]. The fusiform cells of the DCN project directly to the central nucleus of the inferior colliculus as part of the extralemniscal auditory pathway [14]. These cells are unique in that they serve as a convergence point for integration of auditory input via cochlear nerve fibers with somatosensory input via the axons of cochlear nucleus granule cells [15–17]. These multisensory neurons are predisposed to cross-modal compensation following hearing loss. Findings within the animal literature indicate that projections from the trigeminal system to the DCN are increased and/or redistributed after hearing loss and can modulate spontaneous activity. Moreover, following noise-induced hearing loss, animals showed significantly lower thresholds, shorter latencies and durations, and increased amplitudes of response to trigeminal stimulation than their normal hearing counterparts. The finding that only neurons activated by trigeminal stimulation showed increased spontaneous rates after cochlear damage is indicative that somatosensory neurons may play a role in the pathogenesis of some forms of human tinnitus. The concept of disinhibition between sensory pathways is not restricted to tinnitus. It has also been influential in explaining other illusory percepts such as synaesthesia (e.g. experiencing vision from sound) and visual modulation of pain/touch phantom limb. In the case of phantom limb [18] and acquired synaesthesia involving blindness [19], the illusory experiences are invariably linked to sensory deafferentation, as seems to be the case for tinnitus.

In studying tinnitus in humans, the heterogeneity in terms of tinnitus aetiology, pathophysiology, and clinical features is well known. It is probable that different subtypes exist beyond the current subjective/objective distinction [20]. Indeed Baguley and colleagues [21] reiterated the challenge for an effective classification system informed by the underlying pathophysiological mechanisms. This would be a ground-breaking step towards personalised effective rehabilitation for individual tinnitus symptoms. Based on the finding that many patients can manipulate their tinnitus through somatic manoeuvres and the identification of neuronal pathways mediating somatosensory input to the DCN, the concept of ‘somatic tinnitus’ (or ‘somatosensory tinnitus’) has been proposed [7]. Whether somatic tinnitus is a clinically useful construct is not established. Nevertheless, Levine and colleagues [22] have proposed that treatment modalities directed toward the somatosensory system should be carefully assessed by targeting this tinnitus subgroup.

Perhaps with the exception of tinnitus in its objective form, one of the challenges to the international community is the lack of tools to effectively determine idiopathic subjective tinnitus subtyping. Individual studies invariably recruit only small numbers of patients. The lack of standardisation of core measures in the clinical history taking and evaluation of the condition makes it difficult to pool together data acquired by different studies to support meta-analysis of large datasets. The value of retrospective analysis of large datasets is illustrated in a recent exploratory study in which 1204 anonymised records were examined to determine whether comorbid orofacial complaints (specifically temporomandibular joint, TMJ, disorder) could provide a criterion for subtyping a somatic tinnitus [23]. Data presented in the form of descriptive statistics indicated that about 22% reported TMJ disorder. More people reporting TMJ problems could modulate their tinnitus by somatic manoeuvres than those without any TMJ problems (48% versus 30%, respectively). While these findings are interesting, they are somewhat preliminary as they report only the results of univariate descriptions of association and these are sensitive to confounding effects. Multivariate analyses would be needed to determine uniquely related variables because they take into account the interdependencies between different variables [24]. Another limitation relevant to the above review of the literature was that the data could not confirm a difference in hearing ability between tinnitus patients with and without TMJ complaints, perhaps because audiometric data were unavailable for about one-third of the cohort in the previous study [25]. Nevertheless, the authors made a case for conducting further studies in order to confirm these preliminary findings and to explore the relevance of the underlying TMJ pathology as a key defining variable for the putative somatic tinnitus subtype.

The objective of the current study is to investigate the characteristics of somatic tinnitus within a UK cohort. Unlike previous studies which have recruited from the clinical population (e.g. [23,24,25]), our cohort represents tinnitus in the general population. We first aim to describe the prevalence of somatic tinnitus in people without an objective tinnitus and second to determine whether these symptoms are linked to a particular profile of hearing loss, etiology, or other symptom characteristics using methods of model prediction. Our dataset has full audiometric profiles. These findings have implications towards the profiling of somatic tinnitus as a distinct subtype of tinnitus.

Materials and Methods

Standardised data collection and data coding

In order to ensure comparability of data collected across different studies, the BRU uses a minimum of standardised assessments during assessment and evaluation of treatment efficacy (if relevant to the study). These core assessments consist of i) a detailed tinnitus and medical history reporting, ii) a global tinnitus severity estimate, and iii) audiometric thresholds at 0.5, 1.0, 2.0, and 4 kHz in the left and right ears.

History taking always followed the Tinnitus Sample Case History Questionnaire (TSCHQ) developed by international consensus [30]. The TSCHQ includes one question on somatic manoeuvres: specifically “Does any head and neck movement (e.g. moving the jaw forward or clenching the teeth), or having your arms/hands or head touched, affect your tinnitus?”. The TSCHQ allows a yes/no response, but we also added a ‘don’t know’ response option since these phenomena tend not to be known simply through self-introspection without a qualified clinician to perform the somatic manoeuvres during the assessment (see [24]). For the purpose of investigating somatic tinnitus, we focused on a selection of the most relevant TSCHQ items for analysis. In addition to a question about somatic manoeuvres known to alter the tinnitus, these included demographic characteristics (age, gender, handedness), tinnitus characteristics (duration, etiology, pulsatility, constancy, loudness) and comorbidities (hearing difficulty, vertigo, TMJ complaints). The exact wording of the questions used can be found in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive statistics for the sample of participants that was included in the univariate and multivariate analyses.

https://doi.org/10.1371/journal.pone.0126254.t001

Global tinnitus severity was measured either using the Tinnitus Handicap Inventory (THI) [31] or the Tinnitus Handicap Questionnaire (THQ) [32], depending on the study. Both instruments yield a scaled score from 0 to 100. A decision was made to pool the scores to create a single continuous variable, given that these two outcome instruments show high convergent validity (r = 0.76) [33], have a similar (normal) distribution of scores and an almost identical clinically meaningful change score (20 and ≥21 respectively) [34,35].

Audiometric thresholds were converted into a pure tone average (PTA) at the four octave frequencies and this was then averaged across both ears to create a single continuous variable for the degree of hearing loss.

Results

Overall, 108 out of the 671 participants stated that they could manipulate their tinnitus through somatic manoeuvres. Thus the prevalence of the putative somatic tinnitus subtype in our research population of people with tinnitus was 16.1%.

Participant characteristics

In order to characterise the putative subtype more precisely, two steps were taken to reduce unwanted heterogeneity across the cohort prior to any statistical analysis. First, records were removed if the participant did not know whether somatic movements affected the tinnitus (n = 14) or again did not complete this item (n = 2). Second, to eliminate participants who might have a diagnosis of objective (pulsatile) tinnitus of a neurovascular origin, we removed records for all those who confirmed that their tinnitus pulsated in rhythm with the heartbeat (n = 39) or did not complete this item (n = 8). In summary, 608 records remained. The demographics and descriptive statistics are summarised in Table 1 and the raw data can be found in S1 File.

Table 2 shows the prevalence of somatic tinnitus according to demographics (age, gender, handedness), tinnitus characteristics (duration, etiology, pulsatility, constancy, loudness), comorbidities (hearing difficulty, vertigo, TMJ complaints), global tinnitus severity and hearing loss classification from the PTA. For global tinnitus severity, the typical classifications for the THI were used (0–16 No handicap; 18–36 Mild; 38–56 Moderate; 58–100 Severe/catastrophic; [36]), combining severe and catastrophic into a single category and extending these categories to the THQ. For etiology, the separate options of ‘whiplash’ and ‘head injury’ were combined due to small numbers of responses in each category.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Univariate analyses showing degree of association between reports of somatic modulation of tinnitus and other variables.

https://doi.org/10.1371/journal.pone.0126254.t002

Model specification included univariate and multivariate analyses. Statistical analyses were conducted with IBM SPSS Statistics for Windows, Version 22.0.

Multivariate relationships with somatic modulation

Logistic regression analysis is advantageous because it takes into account the interdependencies between different variables and avoids the problem of multiple comparisons that is inherent with a series of independent chi-square tests of association. In addition, the variables of age, time since onset, tinnitus severity and PTA were treated as continuous data thus avoiding the data reduction that had been necessary in the univariate analysis.

For this method to be valid it is important to first check for multicollinearity between predictor variables. Collinearity statistics were run using presence of somatic modulation as dependent variable and the same 13 independent variables listed in Table 2. Collinearity problems are indicated by tolerance values below 0.1 or VIF (Variance Inflation Factor) values above 10 [37]. In the present dataset, the tolerance values ranged from 0.454 to 0.968 and the VIF values ranged from 1.033 to 2.204. As such, multicollinearity is not a concern and a logistic regression is appropriate.

A logistic regression analysis was performed using the same 13 independent variables considered previously. The stepwise (backwards) model using likelihood ratios is considered appropriate when the analysis is exploratory in nature [37]. This method initially includes all variables and then removes variables one by one (removing first the variable with the least impact on the model) until no further variables can be removed without having a substantial effect on how well the model fits the data. The final model removed eight variables and retained five. This represented a significant fit to the data (χ² (5) = 26.02, p <. 001) and explained 7.5% (Nagelkerke’s R²) of the variance. The results are summarised in Table 3. The model retained the variables of age, pulsatility, loudness variability, TMJ disorder and gender. The first four were identified in the univariate analyses. Gender did not come out as significant in the univariate analyses and was of borderline significance in the multivariate model (p =. 055).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Logistic Regression Analysis of Somatically Modulated Tinnitus.

https://doi.org/10.1371/journal.pone.0126254.t003

Finally, a confirmatory approach (based on the forced entry method) was taken specifying only the four independent variables that were significantly associated with somatic modulation in the chi-square tests of association (i.e. age, pulsatility, loudness variability, and TMJ complaints). The model was a significant fit to the data (χ ² (4) = 24.40, p <. 001) and all four variables were significant predictors (age: Wald = 8.64, p =. 003; pulsatility: Wald = 4.39, p =. 036; loudness variability: Wald = 5.42, p =. 020; and TMJ disorder: Wald = 4.59, p =. 032).

Discussion

The main findings of this study are twofold:

1. The prevalence of self-reported ability to modulate tinnitus through somatic manoeuvres was 16% in a general cohort of people with tinnitus. We define this as somatic tinnitus.
2. Somatic tinnitus was significantly predicted by age, pulsatility, loudness and TMJ complaints.

There is little agreement about the prevalence of somatic modulation of tinnitus and except for a handful of dedicated specialists it has been largely neglected by clinicians and researchers over the years. Estimates of self-reported somatic tinnitus in the present study are comparable to at least one previous report that we could find in the published literature. Notably, within a specialist USA tinnitus clinic, about 20% of patients reported at interview that they could somatically modulate their tinnitus by head and neck movements or by muscle contractions, such as clenching the teeth together [38]. Prevalence is higher in cohorts where the assessment has involved the individual performing a guided series of different forceful contractions and judging its effect on loudness, pitch or lateralisation. For example, Levine [8] found that 68% of 70 consecutive patients seen in the clinic could somatically modulate their tinnitus. A similar picture is seen in clinical samples from Brazil [12] and Korea and Vietnam [39]. An even higher prevalence of 80% was reported in a nonclinical sample of people with tinnitus [40]. Again however, the manner in which somatic modulation was ascertained was by forceful head and neck contractions manipulated by a specialist neurologist and his team. Hence, the prevalence of somatic tinnitus is highly dependent upon whether self-report or expert guided manoeuvres are conducted. Self-report is likely to under-diagnose the condition. Nevertheless, the current study provides a conservative estimate based on a widely used screening questionnaire [30].

The results suggest that somatic tinnitus may represent a distinct sub-type. Whereas tinnitus in general is linked to older age and hearing loss [1], our results suggest that somatic tinnitus is more common in the younger group, but is not related to the severity of hearing loss or the severity of tinnitus itself. Moreover, the results suggest comorbidity with TMJ disorder. Whereas our study contrasted people with tinnitus who reported somatic modulation and found an association with TMJ disorder, Vielsmeier et al. [23] took the complementary approach of contrasting tinnitus with and without TMJ disorder and noted an association with somatic modulation. Their tinnitus group with TMJ disorder were also likely to be younger and female. A female bias for somatic modulation was also reported by Won et al. [39] but was not found in the present study (a trend was found in the reverse direction). However, an even gender distribution for somatic modulation still stands in contrast to the strongly male bias for tinnitus in general as found in our sample (67.6%) and elsewhere [1]. Symptomatic temporomandibular joint disk displacement peaks in teenage years and rarely onsets in those aged >45 [41]. Somatic causes may be less apparent in the older sample where tinnitus is more readily associated with hearing loss which onsets in the natural course of aging progress from aged 50 onwards. The onset of TMJ disorder is also linked to stressful life events [42]. Although we didn’t find a statistically significant association with stress it was the largest numerical etiological predictor of somatic modulation.

The putative mechanism for somatic tinnitus is a changed profile of inhibitory or excitatory connections between the auditory and somatosensory pathways. In the model of Shore and colleagues [43], increased somatosensory input reflects disinhibition in order to compensate for hearing loss. This model predicts a positive relationship between hearing loss and degree of somatic modulation that we failed to observe. Instead, somatic modulation may be related to changes in sensitivity within the somatosensory system (either in addition to or instead of such changes within the auditory system) such as those which characterise TMJ disorder (e.g.[44,45]). Future research in this area should explore the somatosensory profile of people with tinnitus in addition to those relating to hearing. For instance, temporomandibular disorder has two broad sub-types: one characterised by symptoms of pain and one characterised by intra-articular symptoms such as clicking of the jaw [46]. Knowing which of these symptom constellations is linked to tinnitus is important both clinically and theoretically. For instance, the trigeminal pathway to the DCN carries information about touch and proprioception but not pain [43] so this theory predicts that pain is unlikely to be the driving symptom. However, there are other cortico-cortical pathways which are sensitive to pain (in addition to touch and proprioception) that could potentially give rise to somatic tinnitus. Patients with TMJ disorder activate, in fMRI, the primary auditory cortex in response to tactile stimuli and this has been linked to connections between this region and the adjacent secondary somatosensory cortex [47]. In healthy controls, there are anatomical connections between these brain regions and, in a single case study of somatosensory impairment following thalamic lesion, these pathways became both structurally and functionally altered giving rise to auditory-induced somatic sensations [48]. The latter is conceptually opposite to somatically modulated/ induced tinnitus.

In general our findings have significant implications for the subtyping of tinnitus. Subtyping of tinnitus is important diagnostically, and in terms of improving treatment, and enhancing our theoretical understanding. Our findings provide evidence for a profile of somatic tinnitus. Theoretical models of tinnitus need to consider not only the integrity of the auditory system but also the integrity of, and its inter-relation with, the other senses.

Supporting Information