Methodological Aspects of Randomized Controlled Trials for Tinnitus: A Systematic Review and How a Decision Support System Could Overcome Barriers

Although a wide range of tinnitus management interventions is currently under research and a variety of therapeutic interventions have already been applied in clinical practice, no optimal and universal tinnitus treatment has been reached yet. This fact is to some extent a consequence of the high heterogeneity of the methodologies used in tinnitus related clinical studies. In this manuscript, we have identified, summarized, and critically appraised tinnitus-related randomized clinical trials since 2010, aiming at systematically mapping the research conducted in this area. The results of our analysis of the 73 included randomized clinical trials provide important insight on the identification of limitations of previous works, methodological pitfalls or gaps in current knowledge, a prerequisite for the adequate interpretation of current literature and execution of future studies.


Introduction
Tinnitus is traditionally described as the perception of a sound in the absence of corresponding external stimuli [1]. In a very recent consensus article, a more precise definition of tinnitus has been proposed: Tinnitus was defined as "the conscious awareness of a tonal and/or noise sound for which there is no identifiable corresponding external acoustic source" and tinnitus disorder was defined as "tinnitus plus tinnitus-associated emotional distress and functional disability" [2]. Tinnitus is considered an enigmatic situation and universally accepted answers to fundamental questions about its pathophysiology, course, and optimal treatment are still pending [1,3,4]. Its prevalence is estimated more than 10% in the general population; however, it is considered bothersome only in approximately 1% [3,4]. These numbers are of paramount importance, since, according to the recently released European Tinnitus Guidelines, clinical approach and decision ma-king should take into account not only tinnitus existence but also patient's reaction to tinnitus [1]. Tinnitus is considered as a symptom well tolerated by the majority of individuals; however, it might cause levels of annoyance which can be adequate to make tinnitus the determining factor for significant impairment of the perceived health status and the overall quality of life.
One of the few things considered common ground among tinnitus community is that no optimal and universal tinnitus treatment has been reached yet [1,5]. Despite the fact that a wide range of interventions including, but not limited to, drugs and medicinal products, sound amplification, sound therapy, psychological interventions, and transcranial magnetic stimulation have been applied, none of them is universally accepted as an adequate and globally effective solution for the whole spectrum of tinnitus sufferers [1].
Hence, there is a pattern across the tinnitus literature according to which, a varying subgroup of responders is found in most of the studies [5]. This could be attributed to statistical variance, but it could also be claimed that some therapeutic interventions could potentially be beneficial in a specific subgroup of patients with identifiable characteristics. Very few studies, however, attempt to create and identify a certain profile correlating with treatment response and the main research question is limited to whether an intervention is effective or not in a group of patients, rather than which factors influence treatment response [6,7].
Moreover, tinnitus related literature has some specific barriers, on top of the issues identified as problematic in medical literature in general, such as sample size calculation, study settings, statistical analysis, and selection bias [8]. These drawbacks include the heterogeneity of tinnitus patients, the fluctuation in tinnitus perception, the subjective nature of tinnitus and therefore the lack of objective outcome measures, the common existence of comorbidities, as well as their interaction with tinnitus perception and the different perception of tinnitus in different cultures, as well as in different times by the same individual [9][10][11].
Consequently, it could be stated that the reasons for the lack of an established and effective treatment are both native and intrinsic, as well as subjective. Aim of this paper is to summarize factors, objective restraints, methodological flaws, and research insufficiencies, in order to provide some explanation for the fact that no universal tinnitus treatment has been established yet.
A systematic review of the tinnitus literature including Randomized Clinical Trial (RCTs) has been conducted aiming towards the identification of common methodological flaws and insufficiencies. Finally, a brief overview of a Decision Support System (DSS) which is the core outcome of EU-funded project Unification of Treatments and Interventions for Tinnitus Patients (UNITI) is presented [12]. The DSS takes into account epidemiological data, audiological measurements, genetic background and socioeconomic data in order to choose the optimal treatment out of the most widely used treatments (sound therapy, sound amplification, Cognitive Behavioral Therapy and Structured Counselling) at an individualized level [12].

Materials and Methods
This paper is a systematic review of literature aiming at the extraction of the characteristics of the RCTs published from 2010 to 2020 in the field of tinnitus, drafted according to PRISMA guidelines [13].
The main goal of this paper is to map and evaluate the methodology of RCTs with regards to inclusion criteria, outcome measures, consideration of confounding factors like hearing loss and whether special care was taken in order to minimize the impact of tinnitus specific confounders such as fluctuations over time or unstable treatment response. This information may lead to useful conclusions about systematic bias, methodological flaws, drawbacks, and gaps in the body of RCTs targeting tinnitus treatment.
Review question was set as following: Which were the inclusion criteria: age, characteristics of tinnitus (primary complaint, laterality, chronic/acute, onset, intermittent cha-racter, etc.), procedure, outcome measures used, study methods (sample size, power analysis, randomization, etc.), and study timeline-duration of the RCTs published between 2010 and 2020. Additionally, the research aims were formulated according to the PICO template as following: People: Adults with history of tinnitus. Intervention: Any kind of intervention aiming towards tinnitus treatment, care and compensation.
Comparison: Non applicable. Outcome: Methodological aspects of tinnitus RCTs (inclusion criteria, study procedures, study outcome measures, study timeline).

Inclusion Criteria
Target population consists of adults with tinnitus. Trials targeting pulsatile tinnitus were excluded. Only RCTs published between 30 June 2010 and 1 July 2020, with at least 30 participants were included. The purpose of setting these inclusion criteria was to critically examine the methodological adequacy and patterns of barriers and limitations occurring in tinnitus literature in its highest level according to their type and level of evidence provided.

Exclusion Criteria
Protocols of RCTs, as well as RCTs mainly targeting entities other than tinnitus, even if tinnitus could occur as symptom in their context (e.g., sudden hearing loss, Meniere disease) were excluded. Any type of study other than RCT or studies written in a language other than English were excluded.

Information Sources
Information sources included 3 major databases (Medline, Central, and Scopus). All searches were conducted by two authors independently. The results were then handsearched [14].

Search
The search syntax used in Medline was broad: "tinnitus AND random*". Similar approach was used in the other two databases. Filters applied were "English language", "10 years" and "Randomized Controlled Trial".

Study Selection
Studies obtained from the aforementioned search have been reviewed independently by two authors. In that stage of analysis, the authors have identified duplicates or multiple reports of the same study, by first examining the titles and abstracts of the yielded studies and then their full text. Results of grey literature, trial registration platforms, and conferences were then added. The senior author resolved any disagreements.

Data Collection Process
Two reviewers have screened full-text articles and produced a matrix of relevant data independently [15]. Any ambiguities on data charting would be discussed and resolved by the senior author.

Extracted Data Items
• Study identification Author, year of publication.

• Methods and inclusion criteria
Population (sample size, whether specific age range was stated and if yes, the actual age range). Existence of an inclusion criterion determining a minimum tinnitus duration reported by the patients or not.
Existence of an inclusion criterion determining whether patients reported tinnitus as a primary complain or not.
Consideration of hearing loss both in inclusion criteria (range and thresholds of hearing loss provided) as well as in data analysis and interpretation.
Whether a limitation in regards to tinnitus laterality was present or not (whether only unilateral or bilateral cases were included, whether both unilateral and bilateral were included or whether no relevant criterion was set).

Methods and study procedures
The primary objective of the study. The treatments applied in the intervention and the control group.
Whether randomization procedure was properly and adequately described. Metric for this was the ability to reproduce the procedure based on the details provided.
Whether power analysis was presented (either ad hoc or post hoc) and the estimated power if available.
Follow up period duration and schedule of follow up visits.

Methods and outcome measures
Outcome measures used, and whether a primary outcome measure was clearly defined.

• Results
Results in terms both of clinically and statistically significant improvement in each one of the groups as well as of difference between groups. An estimation of whether intervention was considered effective is provided, based on whether interventional group showed statistical and clinical improvement in the primary outcome measure compared to baseline and also a statistically significant difference was found, compared to the control arm, in the case no intervention is conducted (no placebo, sham, or waiting list).

Risk of Bias in Individual Studies
In view of the nature of our research question and the objectives of this systematic review, potential bias and methodological pitfalls of individual studies make part of the core analysis and results interpretation.

Synthesis of Results
We present the included studies and summarize the extracted data items in form of tables, with emphasis on our predefined scientific queries. Further analysis in plain text and interpretation of the results shall be found in the corresponding discussion sections.

Study Selection
Seventy-three articles were identified. The screening procedure is reflected in Figure 1.

Study Characteristics
A detailed summary of the characteristics of the eligible studies can be found in the Tables 1 and 2.

Methods and Inclusion Criteria
Across studies, participants' screening and recruitment are focusing on participants' age and presence of tinnitus, while only a small proportion of studies use specific characteristics of tinnitus as inclusion criteria (Table 1). With regards to age, 20 studies set as participants' age upper limit the seventh decade of life, while 27 of them do not provide a clear report of this information. Seven studies have excluded participants for whom tinnitus was not their primary complaint, while there is a large heterogeneity among studies with regards to the tinnitus onset and minimum duration criterion; 2, 3, 6, and 12 months have all been used in the studies included in this review. In sixteen of them, no specification on time and/or duration is provided. Whether tinnitus is continuous or intermittent was not specified as an inclusion criterion in the investigated studies. Many studies defined a minimum annoyance level as inclusion criterion, measured either by means of simple visual analog scales (VAS) or through specific questionnaires (Tinnitus Functional Index (TFI), Tinnitus Handicap Index (THI), Tinnitus Handicap Questionnaire (THQ), and the Tinnitus Acceptance Questionnaire (TAQ)). Four studies excluded participants with hearing loss, while the grand majority of studies do not mention the hearing status of their sample. No study involving other than hearing aid fitting as study intervention had screened candidates for the use of hearing aids prior to their entry to the study.

Methods and Study Procedures
Median number of participants per RCT was as low as 54, whereas 11 of the included studies had a sample size of more than 100 subjects. The vast majority of the studies (46 out of 73) did not provide power analysis nor details on the sample size calculation prior to the study execution.
Across studies, randomization procedure was not described in a clear manner in 46 of the studies (62.16%) (absence of specific methodology, inadequacy of the reported method to the study design).
The duration of participants' follow-up ranged from immediate after treatment up to 26 months [18,98].

Methods and Outcome Measures
Outcome measures used across studies vary and range from simplistic VAS to specialized and cross-culturally validated questionnaires; 14 studies have used the TFI; 9 the THI; 9 the THQ; 2 the TAQ. The majority of studies use more than one questionnaire, covering tinnitus and its comorbidities (e.g., the Hospital Anxiety and Depression Scale or HADS and the Patient Health Questionnaire-Depression or PHQ-D) (for more details, see Table 2).

Tinnitus Duration and Intermittent Character
Tinnitus is a subjective symptom and in many cases it fluctuates over time [1]. Typically, patients report either fluctuations that might or might not be influenced by external factors or by their emotional status, for example levels of environmental noise or stress [1]. In the vast majority of studies, the outcome measures consist of questionnaires that are handed out at specific time points and supposed to evaluate a certain period of time [6]. This method has a fundamental flaw by default: even if patients are asked to provide information about the tinnitus severity over a defined time period (e.g., one week), the tinnitus severity at the moment when the questionnaire is filled out dominates. Tinnitus fluctuations over time or even periods without tinnitus are typically not sufficiently reflected due to memory and reporting bias. When patients are asked to fill in the questionnaire, the results will depend on their emotional status in general and particularly at the time information was provided, their overall attitude and tinnitus perception, and also on their tendency to focus on negative aspects. These confounding factors influence both the presence as well as the level of annoyance and consequently the tinnitus reporting. Therefore, they can be considered as an intrinsic difficulty that is present as a systematic bias across tinnitus related studies.
A potential solution to this could be the use of ecological momentary assessment, which is commonly integrated through mobile applications and allows ongoing recording of fluctuations in tinnitus severity as well as the correlation with certain incidents and behaviors which are captured at the same time (e.g., environmental noise, road traffic, etc.) [12]. This approach, if not well designed or capable of adjustments, may contradict efforts of habituation, since it requires that patients be frequently occupied with their tinnitus and its characteristics.
Only one of the studies took into account the duration of tinnitus within specific time intervals [80]. This finding is remarkable, as an expert consensus initiative from 2007 for tinnitus assessment and outcome measurement proposed that tinnitus patients should be asked about which percentage of their time they perceive their tinnitus?" [99].
Of course, even if asked, this information would be difficult to collect, due to recall bias and inability of patients to provide reliable information in regards to tinnitus duration due to different factors commonly mentioned, including fluctuation of tinnitus occurrence and perceived loudness if present, masking in noisy environments and lack of focus. Intrinsic issues already mentioned have not allowed a universally accepted answer to fundamental question in regards to tinnitus fluctuations, like whether patients with intermittent tinnitus tend to experience less annoyance or whether their treatment response is expected to be better [100]. Consequently, tinnitus duration and fluctuation and whether it was intermittent or not and under which conditions, was not taken into account as a factor in data analysis and interpretation neither, which is a possibly interesting point that should be considered by future studies.

Level of Perceived Annoyance
One of the few things that are considered common ground in tinnitus literature is the fact that the majority of people with tinnitus do not consider their tinnitus bothersome [4]. Overall prevalence, often replicated in the introductory parts of tinnitus studies considered as exceeding 10% in the general population is based on surveys in large samples [4]. On one hand, in the generation of these epidemiological data, some methodological considerations might arise about the criterion used to define tinnitus. On the other hand, the phrasing of the probably largest survey ("In the past 12 months, have you been bothered by ringing, roaring, or buzzing in your ears or head that lasts for 5 min or more?) seems clear enough in terms of adequate duration (excluding brief spikes), type of sound (noise rather than hallucinations), and time frame (one year and not whole life time). In any case, reproducibility of similar numbers in different countries confirms that these estimations should be close to reality [101].
As expected, the number of tinnitus sufferers seeking help by health professionals are much lower than the estimated prevalence [1]. This is of course easily explained by the fact that tinnitus is considered either as not a problem, or a small problem, often reported as non-bothersome. This also reflects to common clinical experience, according to which, a considerable group of patients with other chief complaints might mention tinnitus only when specifically asked. At the same time there is a subgroup of people with catastrophic tinnitus, who describe their tinnitus and the consequences as dramatical.
The discrepancy between the number of people with tinnitus and the number of those who seek medical help might also partly be due to the limited therapeutic options. A person suffering from tinnitus, who is told by the physician that there are no established possibilities to reduce the loudness of the tinnitus, might try to accept the situation without seeking further medical help.
As a result, it would be expected that tinnitus treatment studies, aiming to offer a solution to tinnitus sufferers, should select their participants accordingly and only include patients with a satisfactory level of annoyance, in order to fulfill a fundamental principle of medical research: ability to replicate their results to the target population.
According to the findings of our review, only 22 out of 73 RCTs clearly mention a minimum level of tinnitus annoyance in their inclusion criteria ( Table 1). All popular questionnaires (THI, TFI, and TQ) are used, as well as Visual Analog Scales. THI is used in 12 of the studies, thresholds for inclusion however vary from 18 to 38 with four additional intermittent values: 20, 25, and 30. Only one study has set both lower and upper limits, using TFI, in order to define a certain range [80].
Of course, it is reasonable to assume that individuals with non-bothersome tinnitus will not easily reach a tinnitus clinic and on top of this, be motivated for a usually demanding participation in an RCT. In addition, baseline values give an estimation of the overall annoyance.
Including a reasonable level of annoyance, using the outcome measures chosen for the specific RCT ideally not only setting lower but also upper limits, should be considered good practice in future RCTs.

Tinnitus Audiological Characteristics
Tinnitus frequency can vary from constant to less than weekly. There is also a considerable proportion of patients who state that their tinnitus is only detectable in the absence of any acoustic stimulation, typically before they fall asleep. On top of this, there is a wide range of sounds considered similar to the type of tinnitus sound. Typically, tinnitus pitch is better matched with high frequencies, although there is a considerable proportion of patients who either cannot easily identify a matching sound or better attribute to low frequencies [101]. Determination of tinnitus pitch, loudness, and minimum masking level can be useful in clinical practice, in spite of their questionable role and their fluctuating nature.
However, a robust relationship between tinnitus pitch, loudness and masking level and tinnitus prognosis and severity in terms of annoyance, functionality and handicap has not been established [102]. This means that in the studies investigating the effect of various treatments these characteristics are not useful as outcome parameter. This has been confirmed in our review, in which none of the studies used this type of data neither for outcome measurement, nor as a predictor for treatment outcome.

Tinnitus and Hearing Loss
It is widely reproduced in the literature that hearing loss is present in approximately 90% of tinnitus patients [4]. However, it is also common ground that existence as well as degree of hearing loss are not able to predict tinnitus occurrence and severity [1]. Since pathophysiology of tinnitus is complex and involves both auditory and brain function, it is impressive that hearing loss, although present in the vast majority of tinnitus patients, has not been thoroughly studied as a prognostic factor of tinnitus course, prognosis, and treatment response [1].
This gap is clearly reflected in the extracted literature. Only seven out of 73 studies clearly state in their inclusion criteria that tinnitus was considered as primary complaint by the participants (Table 1). It could be assumed that patients with hearing loss as their primary complaint would not be motivated to participate in tinnitus oriented RCTs. This means that an unknown proportion of study participants could have tinnitus, but not as primary complaint. The primary complaint could be hearing loss and tinnitus only the secondary complaint. Moreover, it is commonly seen in clinical practice that patients present with their primary complaint of tinnitus, but when they are clinically evaluated it is discovered that their main complaint and everyday handicap is their hearing loss. Consequently, and in accordance with common clinical experience, it could be hypothesized that these groups of patients are not homogenous in principle and combine patients in a wide spectrum between hearing loss and tinnitus as primary complaints-and all the shades in between. It is assumed that this could influence results and treatment response, especially in treatments like hearing aids. On top of this, even if identifying tinnitus as primary (but not only) complaint could potentially improve sample homogeneity, it could still exclude a significant group of patients who would consider hearing loss as their cardinal problem and also have adequately bothersome or even catastrophic tinnitus at the same time.
If just dealing with the existence of hearing loss is complex, taking the degree of hearing loss into account is even more challenging. More than one third of RCTs (25 out of 73) include a range of hearing loss in their inclusion criteria, whereas none of them analyzed the audiogram as a predictor for treatment response. Even those studies in which hearing levels were mentioned as inclusion criterion, they have typically vague descriptions of hearing functions, such as "normal hearing levels", hearing levels allowing conversation or mild, moderate, or severe hearing loss without an explicit definition or the respective thresholds.
In accordance, among the studies that evaluated other interventions than hearing aids, there was no study that took into consideration the use of hearing aids neither as an inclusion/exclusion criterion nor as a predictor nor as a confounder. The latter is potentially a hidden but significant risk of bias, since tinnitus improvement is considered as high as 55% in several case series analyzing hearing aids [5]. This could influence results in two ways; first, a selection bias, since only patients were included in the trial, in which hearing aid use was not effective to sufficiently decrease tinnitus annoyance; second, an unclear effect of a prolonged or recent use of hearing aids, which might influence the performance of an unknown proportion of hearing aids users both in the interventional and control arms. A clear exclusion of patients with a relatively recent hearing aid fitting should be considered as good practice in future RCTs.
The currently starting UNITI trial is strategically planned as an attempt to overcome the mentioned issues. Only patients with tinnitus as primary complaint will participate, and degree of hearing loss will be analyzed with sophisticated techniques as a potential confounder for treatment response. In addition, the efficacy of hearing aids as a sole measure to improve tinnitus will be tested for the first time in the context of a RCT against interventions based on other disciplines, like CBT [12].

Remarks on Study Methodology
Tinnitus interventions, in accordance with tinnitus pathophysiology, are heterogenous. In the RCTs collected, a wide spectrum of therapeutic strategies is performed ranging from transcranial magnetic and vagus nerve stimulation to internet-based CBT and altered/notched music (Table 2). A pattern that causes deviation from an optimal study design and is valid for various interventions is the inability to blind patients with respect to control interventions. For example, a blinded RCT comparing true and sham hearing aids is not feasible, since participants in the sham group will immediately recognize the sham devices, given that they will be unable to provide acoustic amplification.
Apart from intrinsic limitations and barriers, tinnitus literature also suffers from methodological insufficiencies which are common in other fields, as well. Median number of participants per RCT is as low as 54, whereas only 10 exceed 100 participants. Most of the studies (46 out of 73) did not provide power analysis; hence, the rest reported power over 80%. The same proportion of papers did not provide a clear, detailed, and reproducible description of their randomization procedure, a fact that clearly questions their qua-lity. Moreover, randomization procedure was found to be unclear in 46 of the studies (63.01%) under the sense, that relevant information provided was generic and not adequate for the procedure to be replicated.
Although out of the scope of this review, what needs to be underlined is that many of the RCTs concerning tinnitus are supported by pharmaceutical companies or hearing and tinnitus-related devices manufacturers. So, results must always be read with caution and extra consideration for potential biases or conflicts of interests.

Outcome Measures
Tinnitus is a condition affecting everyday life in many ways, causing a list of issues including, but not limited to annoyance, functional disturbances, tinnitus intrusiveness and acceptance, disability for certain actions and tasks like concentration, ability to ignore negative emotions, sense of control, malaise, and loathsomeness. Hall et al. (2019) have recently suggested proper outcome measures for each type of intervention. These recommendations could not have been applied to the body of the RCTs examined; however, lack of justification for the choice of a certain outcome measure is the rule [103].
Majority of studies (63 out of 73, 86.30%) use more than one outcome measures, a procedure that has been proposed to enable comparison across trials [99,104]. However, the use of multiple outcome measurements requirerequires the a priori definition of the primary outcome, which was the case only in 14 studies out of the 64. It is also interesting, that 34 of the RCTs use at least three outcome measures, which shows a relatively wide range of domains targeted and also increases the possibility of results ought to randomness.

Time Course
Relatively little is known about the course of tinnitus over time [7,[105][106][107]. There are a lot of factors contributing to this. Tinnitus installation is often prolonged and there is a considerable proportion of patients that cannot clearly identify an exact date of tinnitus onset. Especially in the cases where tinnitus habituation has occurred, patients may not clearly recall or may underestimate both time of onset or severity of their tinnitus, when they are asked about it or when they are filling retrospectively a relevant questionnaire. A considerable proportion with total relapse might not even contact health services, and therefore never be recorded, which means that estimation of the course of tinnitus over time is in current studies at based assessed retrospectively by questionnaires, which are subject to recall bias as well as to suboptimal phrasing of the relevant questions in the self-filled questionnaires. Consequently, there is lack of reliable information about the actual incidence, the course and the profile of the patients who experience tinnitus for a short period of time and then stop experiencing it. On top of this, patient trajectories differ strongly across countries depending in the health system.
Tinnitus is usually dichotomized into acute and chronic; however, recent European guidelines have also included the sub-acute type (from 3 to 6 months), in order to reflect the transition from acute to chronic tinnitus [1]. However, all these definitions are arbitrary, and little is known about differences in the pathophysiology of acute and chronic tinnitus and the time when this transition occurs. It is remarkable that only two of the RCTs identified focused on acute tinnitus [50,74].
The majority of the studies (58 of 73) clearly defines a minimum time interval from tinnitus onset, however variability in time intervals is large. Eleven trials set as minimum duration 3 months and 22 the 6 months interval, whereas a wide range of smaller or larger intervals occur. This variance may be relevant for the tinnitus course, since a recent systematic review has indicated a statistically significant decrease in the impact of tinnitus over time, although clinical significance could not be interpreted due to heterogeneity [108]. This practically means that in the comparison between RCTs differences in the tinnitus duration might matter.

Trial Design and Results
As expected, there is a large heterogeneity among the RCTs included. About one third of the included studies examine the efficacy of pharmaceutical agents either as sy-stemic or as topical administration. Second most common topic is various types of TMS, whereas 10 focus on CBT either face to face or online, 5 on non-CBT psychological interventions, 7 on HAs (alone or in combination with motivational interview), and 8 on sound therapy. Finally, acupuncture and laser beam have also been evaluated as monotherapy for chronic tinnitus [25,27,34,43,51,53,66]. The variance of the interventions with respect to their intended mechanisms, targets and duration should have led to different trial designs in terms of outcome measures as well as follow up schedule. For instance, TMS is usually implemented in strict and well-defined time periods (typically one to two weeks), whereas CBT is an intervention lasting several weeks and should be finished before the effect can be evaluated. At the same time, HAs have a continuous and possibly long-lasting effect. This is not reflected in the design of the studies, since criteria in regards to tinnitus onset, follow up duration and outcome measures are more or less equally distributed in these sub-groups of RCTs.
It is noteworthy that two third of control groups use different types of methods in order to be non-interventional: placebo, sham devices or interventions, participants from the waiting list or usual care. Ideally, the recommendation is to use best available treatment instead of placebo, at least for pharmacological studies. This is probably not applicable in the tinnitus field, since universally acceptable treatment is pending [109]. Hence, majority of the remaining studies use as control arms active interventions of the same discipline (different TMS protocols, HA fitting parameters, stimuli used for sound therapies). Very few RCTs compare two totally independent interventions. This design should be considered in future studies, because on top of efficacy superiority, it could potentially identify profiles of patients who could be more prompt to one intervention compared to other.
Although evaluation of results was not within the core scopes of this review, it should be mentioned, that roughly one third of the RCTs concluded that the intervention tested was considered effective. RCTs targeting CBT and different types of sound therapy re-presented more than one third of the RCTs with a positive outcome, whereas their proportion in the whole body of RCTs was significantly lower (13.6% and 10.96%, respectively). All other types of interventions had at least one clinical trial with a positive result (superiority against the control intervention).
One important aspect is that with regards to several RCTs in which different active interventions were compared, it remains unclear whether the results differ from placebo. Even if there were significant within arm comparisons for all investigated interventions, one cannot unambiguously differentiate between spontaneous improvement and an effect of the investigated intervention.
Moreover, with very few exceptions, only statistical and not clinical significance was examined, and any minimum benefit considered as significant was set ad hoc.

Conclusions
Tinnitus is a clinical enigma for many reasons, most of which are intrinsic. RCTs in tinnitus field suffer from many methodological flaws, including lack of strict and welldefined inclusion criteria, failure to eliminate confounders like hearing loss, stress, and tinnitus duration, variance in outcome measures and interventions, and barriers in design. Future RCTs should clearly set a minimum level of tinnitus related annoyance, range of age and tinnitus duration, hearing ability, as inclusion criteria, choose appropriate outcome measures for the intervention tested and interpret clinical on top of statistical signi-ficance.
Design of future trials should take into account issues, barriers and problems identified in this review. Patients with intermittent tinnitus should probably be included in different studies than those with tinnitus lasting all day. Although tinnitus onset might not be always clear, the majority of studies seem targeting chronic tinnitus (>6 months) and a gap of studies targeting acute tinnitus has been identified. A strict range of scores in the outcome measures should be set as inclusion criteria and the results of the trials should be clearly generalized in the relevant population. Outcome measures should be selected based on intervention and targeting related tinnitus dimensions, in line with current literature [92]. Comorbidity of hearing loss and its level should be also taken into account, not only as inclusion criterion, but also as a potential profiling factor for patients and treatment response. Tinnitus pitch, minimum masking level, and residual inhibition should also be considered as prognostic factors. Patients should be monitored for at least six months, in order to ensure stabilized results. Overall, trial designs and analysis should overcome the classic schema of comparing two interventions at two points: tinnitus is complex and heterogenous and requires identification of certain subgroups who are prone to certain treatments and of prognostic factors for treatment outcome.

Conflicts of Interest:
The authors declare no conflict of interest.