Study Design
The study, a three-armed, single-blind randomized clinical trial, will be conducted at the Philipps University in Marburg, Germany. The study has been approved by the respective ethics committees of the Department of Psychology and the Department of Medicine, Philipps-University Marburg, with registration numbers 2018-4k (Additional file 2) and 162/18 (Additional file 3) respectively. Furthermore, the trial is registered with clinicaltrials.gov (ClinicalTrials.gov Identifier: NCT03550430).
Sample size calculations were based on a repeated-measures multivariate analysis of variance (MANOVA) and conducted with G*Power 3.1.9.2. (24). The total sample size needed to detect a medium assessment-by-treatment interaction effect (f=0.25, Beta=0.80, Alpha=.05, Groups = 3, measurements = 4) is 171. However, to compensate for an estimated drop-out in treatment of 20%, the aim is to recruit 204 participants, i.e. 68 participants per group. Post-hoc analyses are carried out between ADR vs. BTR, ADR vs. diary control and BTR vs. diary control, respectively.
Inclusion and exclusion criteria
Participants are eligible for inclusion if they meet the following criteria:
- Age +18 years
- Chronic, subjective tinnitus with a duration of ≥ 6 months
- At least mild tinnitus distress, corresponding to a score of ≥ 18 on THI (25)
- Signed informed consent
The study exclusion criteria are as follows:
- Objective tinnitus, either vascular or non-vascular in origin
- Acute inflammatory disease of the ear/ears
- Conductive hearing loss
- Blockage of the ear canal by e.g. cerumen, where removal is not desired by the individual
- Any current treatment for tinnitus
- Severe mental health issues (moderate/severe depression, bipolar disorder, psychosis, ADHD)
- Current use of a psychotropic drug for a mental health condition
- Substance abuse
- Previous or current neurofeedback treatment
- Previous or current neurological conditions (g. a history of seizures, brain tumor, hemorrhage/stroke).
Procedure/ Trial Protocol
Recruitment for the trial is done primarily through advertisements placed in local and regional media. In addition, a referral program is set up to entice ear-nose-throat physicians (ENT’s) to forward participants to the study. Interested individuals can read more about the study on a dedicated website, where they can also register their interest in trial participation. The website as well as the questionnaires are provided via the iterapi platform (26).
Insert here: (Figure 1: schedule of enrollment, interventions and assessments)
The enrollment period (t0, see figure 1 above) is a four-stage process. In the first stage, participants complete a battery of online screening questionnaires. These are mainly designed to sort individuals into principally eligible and non-eligible categories, based on the inclusion/exclusion criteria. Prior to taking the screening questionnaires, participants give informed consent for questionnaire completion only. For principally eligible individuals, the second stage in the screening process is a telephone interview. The purpose of this is to validate questionnaire responses and to assess motivation for participation in a trial. The third stage in t0 is a visit to the department of Psychology. The first part of this visit provides participants with detailed information about the study, about procedures and how to complete baseline assessments, including primary and secondary outcome measures on the secure online platform. Following this, participants sign the informed consent (see Additional file 4) and complete the State Anxiety Inventory online (STAI, 27). Participants then complete two attention tests, the Attention Network Task (ANT, 28) and the Sustained Attention to Response Task (SART, 29) respectively (in figure 1 referred to as cognitive assessment).
The fourth and final stage in t0 is a standard audiological examination at the ENT-clinic of the University Hospital of Giessen and Marburg. There are two main purposes of this visit: The first is to rule out possible non-subjective causes of tinnitus and second, to acquire audiometric data. At the clinic, participants first learn about the components of the audiological examination, before signing the informed consent (see Additional file 5). Subsequently, the audiological examination, in which the below information is acquired, takes place:
- Standard Audiogram (to 20.000 Hz)
- Tympanometry
- Otoacoustic emission (OAE)
- Tinnitus loudness discomfort level
- Tinnitus matching (loudness, pitch)
Participants who are included after the visit to the ENT clinic, take the baseline assessment in the pre-allocation (t1) stage. Upon completion of this assessment, participants move to the allocation (t1) stage and are randomly assigned to one of three arms in the study: Group 1, who receives alpha/delta neurofeedback training; Group 2, who receives beta/theta neurofeedback; Group 3, the diary control group.
The post-allocation (t2) intervention period extends over four weeks. Within this timeframe, ten neurofeedback training sessions are undertaken for the two neurofeedback groups, with a minimum of two and maximum of three sessions per week. Mid-point assessments are completed online after the first five of ten neurofeedback training sessions. For the diary control group, two face-to face visits to the department and two telephone calls are scheduled within the four-week period. Mid-point assessments are completed online after two weeks. In the post-allocation (t3) stage, participants complete the endpoint assessment online, prior to a visit to the department of Psychology, where they complete the ANT and SART (referred to as cognitive assessment II in figure 1).
The follow-up (t4) stage is three months after the post-allocation (t3) period. Here, participants complete follow-up assessments online and are invited to the department of psychology for a booster neurofeedback session. The booster sessions serves the practical goal of probing the participants about their experiences with neurofeedback. In addition, the visit is included to increase the likelihood of participant adherence to questionnaire completion.
Randomization and blinding
When, as in our study, specific prognostic factors may influence outcome, the best way to ensure a balanced distribution of participants across groups, is to use an allocation by minimization process (30). Accordingly, participants are stratified to one of the three study groups, based on their age and tinnitus severity (31). Age is included as a prognostic factor. This is based on the assumption that neurofeedback utilizes the brain’s ability to reorganize itself functionally and/or structurally (32, 33). With the current knowledge, evidence favors a hypothesis, that younger individuals compared to older, have greater neuroplastic potential (34). This is partially because of the association between increasing age, functional and structural brain deterioration (35, 36). However, it would be a falsehood to believe, that this deterioration affects all individuals to the same degree. With increased age come large individual differences in brain health (34). These are apparent already, when people are in their mid-fifties (35). Taken together, these observations have led to the following stratification categories for age in the present study: (1) ≤ 55 years; (2) > 55 ≤ 64; (3) > 64 ≤ 69; (4) > 69. Given the conservative inclusion criterion on the primary outcome variable ( 18 ≥THI score pre-intervention ≤ 100), stratifying participants into sub-groups based on the tinnitus distress classification by McCombe, Baguley (37), is considered a sensible option. This ensures a balanced distribution of the range of THI scores across the three groups. Importantly, this reduces the risk of heterogeneity of variance affecting the between-group outcome analysis. Thus, participants are stratified to one of four levels of tinnitus distress: 1) Mild (THI 18-36); 2) moderate (THI 38 – 56); 3) severe (THI 58-76); 4) catastrophic (THI 78-100).
Allocating participants to one of the three groups is done by a researcher independent of the study. This individual, trained in the minimization program, Minim (38), will subsequently inform the research team about group allocation of each participant. A group allocation list is kept in a separate and password secured excel file on the local drive of the independent researchers' PC.
To ensure blinding of trial participants, two identical neurofeedback training protocols were developed. First, information about the course of the training is kept identical for participants in both groups. Second, the position of the training electrodes is the same, i.e. positions FC1, FC2, F3 and F4 of the 10/10 International system are used in training. Third, the duration and intensity of training is identical for both groups. That is, all participants undergo ten training sessions, each consisting of initially four, progressing to five training blocks per session. Fifth, the neurofeedback training stimulus used, i.e. a fish swimming vertically across the screen, is the same for both groups. In other words, all variables are kept constant, bar the trained frequency bands, the information of which is not accessible to participants.
Interventions
Neurofeedback training groups
The neurofeedback training sessions were designed identically for the ADR and the BTR training. Both trainings will be applied unidirectionally, with the aim to a) decrease the ADR by an increase of alpha (8-12 Hz) and/or a reduction of delta (2–4 Hz) and b) decrease the BTR by inhibiting theta (4–8 Hz) and/or reinforcing beta (13–20 Hz) activity bilaterally over the fronto-central cortex (FC1, FC2, F3 and F4), respectively.
During each NF-training session, EEG, electromyography (EMG) as well as vertical and horizontal electrooculogram (EOG) are recorded from Ag/AgCl sintered ring electrodes with impedances kept below 5 kOhm using a 13-channel DC-amplifier (THERA PRAX ®MOBILE, Neurocare GmbH, Ilmenau, Germany; > 10 GOhm input impedance) and a sampling rate of 256 Hz. EEG electrodes are placed at F3, F4, FC1 and FC2 in accordance with the 10/10 electrode placement system (39) using an elastic electrode cap (Easy Cap, Germany). Reference and ground electrodes are attached to right and left mastoid, respectively. The electrode montage is identical to previous neurofeedback studies in tinnitus (14, 40) in order to guarantee comparability with former research. EMG is recorded with two electrodes placed at the upper descending part of the M. trapezius. To reduce artifacts produced by eye movements, a real-time EOG is recorded simultaneously, using four electrodes (two electrodes at external canthi, and two electrodes at infra‐ and supraorbital sides). Eye movements are removed from the feedback signal during the training using an online ocular correction as described by Schlegelmilch, Markert (41). In addition, all signals acquired during neurofeedback training are stored as raw signals for later offline processing without any calibration or filtering.
Online-processing in real time comprises a 50 Hz notch filter applied to EEG signals before direct feedback. Depending on the respective training group, BTR [theta(μV2/Hz)‐beta(μV2/Hz)/theta(μV2/Hz) + beta(μV2/Hz)] or ADR [delta(μV2/Hz)‐alpha(μV2/Hz)/delta(μV2/Hz) + alpha(μV2/Hz)] were extracted with a short‐time‐Fourier transformed moving average across the four training electrodes and fed back to the patients` monitor, using graphical objects (e.g. fish, moon etc.): horizontal movements of the object from the left to the right with constant speed represents the temporal proceeding of the trial (i.e., sampling rate), while vertical movements of the object, indicates the targeted changes in the feedback parameter (ADR or BTR), either by moving up (targeted change) or down (non-targeted change). Successful change in cortical activity (i.e., keeping the ADR or BTR above the training threshold for at least 250 msec.) is rewarded with the symbol of a sun after each trial, as the only performance-dependent reinforcement. The neurofeedback training protocols are largely in line with the successful tau (8-12 Hz; identical to our alpha frequency band)- to-delta-ratio protocol used previously (13). It comprises 10 sessions (4 runs à 10 trials; net training about 30 min) over the course of 4 weeks. One trial consists of 30 seconds of active neurofeedback training, followed by a 10 second inter trial interval, in which the last 2 seconds are used to determine the current baseline for both trainings. Therefore, each trial during training is individually baseline-corrected and the feedback object always starts moving from the same position on the monitor (Please see https://youtu.be/0ZD_mUFixHs for an illustration of the individual trial run). The scale of the feedback monitor is set to 2μV. For the first training trial in the first session, the training threshold is set to 20%, corresponding to reward reinforcement when the ADR or BTR is greater than 0,2 μV for at least 250 msec. If a reinforcement rate of at least 70% is reached for a training block, the threshold is increased by 3%. After blocks with less than 30% successful trials, the threshold is lowered by 2%. The following training session is started with the thresholds of the last block of the preceding session. In order to enhance transfer effects, we also included gradually more transfer trials in an additional block at the end of a session, starting from session 6 onwards (see figure 2). During transfer runs, patients do not receive any continuous feedback, but receive reinforcement for successful trials (reward symbol; please see https://youtu.be/shcR8Ilq3mo for an example of a transfer trial). The percentage of transfer trials increases gradually with session (20%, 40%, 60%. 80% and finally 100% transfer trials in within the final block of a session, respectively). The transfer block is implemented to facilitate the transfer of acquired changes into daily life when no feedback of physiological signals can be provided. Transfer trials, the use of which is reviewed comprehensively in (22), have successfully been employed in other frequency-based NF-protocols (42). To further facilitate transfer, participants are instructed to retrieve their neurofeedback experiences by designing personalized cues (i.e., printed graphics representing the mental strategy used during the neurofeedback training) and to use those cues both during within-session transfer trials and during daily life. Compliance is verified by questioning the participants after each session in a session evaluation, whether they have used the transfer cards over the intervention period.
Insert here: (Figure 2: the study flow divided into sessions, training blocks and trials)
Prior to the first training block in the first session, the training sequence on the monitor is shown to the participant. On the trainee monitor, the fish swims horizontally from left to right on the screen, in a sequence lasting 30 seconds. This is followed by a 10 second pause, before the process is repeated. During this presentation, the therapist tells the participants, that their task is to try to move the fish above the black reference line, in the middle of the screen. No further instructions on how to accomplish this is given. This mirrors the instructions given in (13), thus ensuring greater comparability between results. The addition to these instructions is that during the breaks between training blocks, the therapist discusses briefly with the participant, the types of mental activities they develop and employ for achieving the training goals.
Diary control group
In the diary control group, Group 3, an eight-item diary is completed each evening for seven consecutive days, in week one and week three of the four-week intervention period. The items are purposefully designed to make participants reflect more positively about their experience with tinnitus (e.g. The tinnitus didn't disturb me today). As adherence to diary completion is a concern, participants have two supportive face-to-face meetings. In these, participants are informed about the content and structure of the diary and instructed on its completion. In addition, there is an element of psychoeducation on tinnitus etiology, the content of which is based mainly on recognized self-help sources (43, 44). In addition, two follow-up telephone calls are held throughout the four-week period. The telephone follow-up calls are supportive in nature and give participants an opportunity to share their experience with the diary exercise.
Adherence to the interventions
To maximize treatment adherence in clinical trials in general, several suggestions have been proposed (e.g. 45, 46, 47). Of these suggestions, particularly three are relevant to our study. First, participants will find it relatively easy to complete questionnaires in the study, since all primary and secondary outcome measures are collected via an easy to navigate online platform. For people without access to the internet, pen and paper versions of questionnaires can be collected at the department or sent via mail upon specific request. Second, for participants in both neurofeedback groups and in the diary control group, there is frequent contact with members of the research team throughout the intervention period. Third, participants in the diary control group are given an incentive, a self-help book (43), upon completion of all outcome measures.
Measures
Primary outcome measures
There are two primary outcome measures in the study. First, change in tinnitus distress following neurofeedback training is measured with the Tinnitus Handicap Inventory (THI; 25). The THI is a 25-item self-report instrument. Each item is rated on a three-point Likert scale with responses “Yes” = 4 points, “Sometimes” = 2 points and “No” = 0 points, thus yielding a total score between 0 and 100. The instrument has three subscales, assessing functional, emotional and catastrophic reactions to tinnitus respectively. The overall test-retest reliability of the scale is 0.92 (48). It has been translated and validated into a German version (49, 50), which is used in the present study.
To assess the efficacy of neurofeedback training in reducing the perceived intensity, i.e. the sound percept of tinnitus, the Tinnitus Magnitude Index (TMI) is used (51). It is a three-item measure, designed to assess the individual perception of tinnitus sound intensity, without overlapping significantly with cognitive, behavioral and emotional reactions to tinnitus. The internal consistency of the scale is excellent with Cronbach's α = 0.86 and it has satisfactory discriminant validity (correlation of r=.62 with the THI). One obstacle in the development of the TMI, noted by Schmidt, Kerns (51), is the differential scaling of its three items. This has, according to the authors, the potential to increase measurement error. Schmidt, Kerns (51) suggested converting all three items to a standardized scale. Therefore, in the present study, the response scales for all three items are standardized to a range from 0 to 100.
Secondary outcome measures
To detect responsiveness to the intervention, the German version of the Tinnitus Functional Index (TFI; 52, 53) is used in the present study. It assesses treatment related changes in different areas of functionality, e.g. sleep or sense of control. It consists of 25 items, with 23 responses being rated on a Likert scale from 0 to 10 and two items between 0 to 100. When scoring the TFI, the two items ranging from 0 to 100 are divided by 10, thus yielding a total TFI score of the 25 items between 0 and 100. There are eight subscales associated with the TFI (intrusiveness, sense of control, cognitive interference, sleep, auditory difficulties, relaxation, quality of life, and emotional distress). Overall, the TFI has a good test-retest reliability of 0.78, convergent (r = 0.86 with the Tinnitus Handicap Inventory) and discriminant validity (r = 0.56 with Beck Depression Inventory Primary Care). (52).
To assess changes in sleep quality following neurofeedback training, the Insomnia Severity Index (ISI; 54) is used. The ISI consists of seven items, with Likert responses ranging from 0 to 4. Thus, a person can score between 0 to 28 on the total scale, with higher values indicating greater sleep disturbances. The ISI has a good internal consistency (Cronbach’s α = 0.91, 54) and convergent validity (r = 0.80 with the Pittsburg Quality Sleep Index, 55).
Changes in depressive symptoms following neurofeedback are assessed with the Personal Health Questionnaire-9 (PHQ-9; 56), a nine-item self-report instrument with Likert scale responses ranging from 0 to 3. The total score of the scale thus ranges from 0 to 27. The internal consistency of the PHQ-9 is good (Cronbach’s α = 0.89) and a test-retest reliability of r = 0.84 has been evidenced (57). Moreover, the PHQ-9 has strong convergent validity with other scores of depression, e.g. Beck Depression Inventory, r = 0.73 or General Health Questionnaire-12, r = 0.59 (58).
Further secondary exploratory research questions of the current trial focus on the moderating role of pre-treatment expectancy, perceived treatment credibility and somatic self-efficacy on treatment outcome. These moderator analyses are assessed with the following instruments: an adapted form of the treatment credibility and expectancy questionnaire (59); a translated to German version of the somatic self-efficacy questionnaire (60); an expectancy questionnaire, developed specifically for this study.
Training Outcomes
In order to keep track of individual learning curves during training, we will analyze ADR and BTR for each participant within and across sessions. However, session 1 will be discarded, because it is assumed that participants still have to habituate to the setting. Monitoring and encouraging learning is, one the one hand, done by the trainers. On a patient monitor, the trainer can keep track of training progression, as the percentage decrease in the BTR or ADR, compared to baseline for each training block separately.
On the other hand, the total number of successful trials per training block is used to guide, either upwards or downward adjustment of training thresholds. Dependent measures include the mean training level over training blocks within a session (%), the best run of each session (maximum training level), and the total number of obtained rewards per session. Furthermore, electrophysiological training raw data gathered with the THERA PRAX ®MOBILE described above will be analyzed offline with Brain Vision Analyzer v2.0 (Brain Products GmbH, Germany).
Preprocessing: Data is first band-pass filtered with Butterworth zero-phase filters between 0.1 Hz and 80 Hz with slopes of 24 dB/octave at the low and 48 dB/octave at the high cutoffs. In order to eliminate possible line noise, data is further filtered with a band-rejection filter with a central frequency of 50 Hz, a bandwidth of 1 Hz, and a slope of 24 dB/octave. For ocular correction we will use the algorithm of Gratton, Coles (61) as implemented in Brain Vision Analyzer software. Using a semi-automatic raw data inspection procedure, the recorded data from the training blocks will be screened for artifacts. Planned criteria for artifact screening will be as follows: maximal voltage step of 50 μV/ms, maximal amplitude of±100 μV, maximum allowed difference of 150 μV in each segment, values greater than 200 μV per 200 ms interval, activity below 0.5 μV in a 50-msec. period as criteria. Before and after detected artifacts, 100 msec. of data will be removed. A thorough visual inspection is performed in order to remove possibly remaining artifacts (i.e., muscle movements and short drifts or jumps over single or multiple electrodes) from the signal.
EEG Analysis: The artifact-free training data will be segmented into non-overlapping 2-s epochs and submitted to a Fast Fourier Transformation (FFT) with a 10% Hamming window. The resulting data are averaged in the frequency domains (delta 2-4 Hz; theta 4–8 Hz; alpha 8–12 Hz and beta 13–20 Hz) for feedback versus transfer trials for each individual session and each individual patient across the four training electrodes. Absolute power values (μV2) for each frequency band (i.e. delta, theta, alpha and beta), Theta/beta, delta/alpha power ratios as well as relative power (%) are calculated.
Data management and monitoring
Upon registration on the web platform, participants are assigned a computer-generated study code, which will follow them throughout the trial. All collected data are stored pseudonymized in locked cabinets or as computer files. The pseudonymization allocation list is kept separate from the pseudonymized data and deleted upon completion of the trial. Participants are informed about this. Only people who have signed a confidentiality agreement, and who are part of the research team, have access to the data collected in the study.
The study’s assessment questionnaires are answered online. All necessary precautions have been taken to ensure data protection and security. All data exchanged between participants and the online system are encrypted prior to transmission and storage. The online system is managed and hosted by the IT-department of Linköping University (Sweden). No entries in the form of video or sound recordings, which could make it possible for third parties to identify the participants, are registered. Participants can request their data deleted at any time upon stating their assigned code, until the pseudonymization allocation list is deleted. Lastly, data collected during the study are kept for ten years before being deleted.
Adverse events
Neurofeedback is generally considered safe and involving no risk to participants (62). This is perhaps best reflected in the fact, that it has been used extensively in studies involving children with ADHD. Thus, participants in our trial should not experience more serious side-effects than perhaps mild headaches, stemming mainly from the prolonged period of sustaining attention during training. Nevertheless, adverse effects of neurofeedback training will be routinely monitored as part of the ongoing dialogue between data collectors and trial participants. In the unlikely event, that a trial participant complains about serious adverse effects of training, data collectors will, as a first step, reduce the frequency of training sessions in the week of the complaint, and if inadequate, reduce the number of training segments per training session, until the trial participant no longer complains about adverse effects.
Ethics and dissemination
Should any future modification to the protocol occur, which changes the study objective, design or procedures, they will be decided upon by the study’s principal investigator. Substantial protocol amendments are submitted to and approved by the responsible ethics committee, prior to implementation.
Dissemination of results is expected to follow completion of data collection, scheduled to last until late spring of 2020. Results of primary outcome measures are reported and disseminated, regardless of the direction and magnitude of effect(s). No restrictions are imposed on which results can be disseminated from neither trial sponsor nor other interested parties.
Statistical analysis
Data will be analyzed primarily in the intention-to-treat (mITT) population. Supportive analyses are planned in the per-protocol (PP) population. mITT comprises all randomized patients, while PP analysis assesses mITT patients who do not meet any of the following criteria: violation of inclusion and/or exclusion criteria, major deviations from the visit schedule, and poor compliance during feedback sessions. For all outcome measures we will probe the longitudinal course across all assessments using a linear mixed model for repeated measures (MMRM; see 63). The MMRM model includes fixed effects for group (ADR vs. TBR vs. diary control), time and group-by-time interaction and a random intercept for subject specific effects using maximum likelihood estimation, adding sex, age, baseline tinnitus severity scores and patients’ expectations as covariates. The error variance-covariance matrices for the repeated factor will be specified in accordance to the data. The pattern of missing data is assumed to be random.