Protocol and umbrella repository for the Tinnitus Detection (TIDE) project.

TIDE is a multicentre, prospective, case–control study identifying and validating neurophysiological biomarkers of chronic constant tinnitus across seven international sites. This repository holds the study description, measurement protocol and shared documentation. Paradigm implementations and analysis code live in separate repositories — see Components.

Vanneste S, Yasoda-Mohan A, Chen F, et al. Objective data-driven personalised approach to diagnosis of chronic tinnitus: the Tinnitus Detection (TIDE) project — protocol for the identification and validation of a biomarker for tinnitus. BMJ Open 2026;16:e112788. doi:10.1136/bmjopen-2025-112788

Trial registration: NCT06520865 Status: recruitment from 1 August 2024, scheduled to complete April 2026; study end 31 December 2029.

Tinnitus is diagnosed by patient report. There is no established objective, quantitative biomarker for diagnosis or for characterising the condition. That absence sustains a trial-and-error approach to management, complicates differential diagnosis where psychiatric or non-auditory factors shape symptom expression, and leaves treatment trials dependent on subjective endpoints with inconsistent minimal clinically important differences.

TIDE tests two candidate mechanisms drawn from contemporary models of tinnitus, using standardised paradigms applied identically across sites.

Primary. Determine whether EEG/MEG event-related potential markers derived from (1) a cortical gap prepulse inhibition of the acoustic startle (GPIAS) paradigm and (2) an omission auditory oddball paradigm can differentiate individuals with chronic tinnitus from matched controls.

Secondary.

• Test whether these markers are associated with tinnitus intensity and severity (loudness matching, questionnaire-based severity)
• Assess reliability and reproducibility across sites, and within-site repeatability where applicable
• Evaluate the influence of covariates — hearing loss, hyperacusis, age, sex

Multicentre, prospective, case–control, experimental and patient-oriented.

560 participants: 280 chronic constant tinnitus cases, 280 non-tinnitus controls. Target 40 cases and 40 controls per site. Cases and controls group-matched for sex, age and hearing loss post-recruitment.

Sample size. Target 40 cases and 40 controls per site across seven sites, giving 280 per group. The target was set by site-level feasibility rather than by an a priori power calculation, as no prior effect-size estimates existed for these paradigms. With 280 per group, a two-group comparison at α = 0.05 (two-sided) has 80% power to detect d ≈ 0.24 (Cohen's f ≈ 0.12). The study is not longitudinal and no attrition is anticipated. Prespecified confounders: age, sex, hearing loss and hyperacusis.

Analysis is additionally supported by Winfried Schlee at the Institute of Information and Process Management, Eastern Switzerland University of Applied Sciences, St Gallen.

Approved by the ethics committee or IRB at each of the seven sites, compliant with US and European data protection regulation. Informed consent is obtained from all participants prior to participation, in accordance with the Declaration of Helsinki.

Inclusion

• Cases: chronic tinnitus for at least 6 months by history, constant, subjective, non-pulsatile, classified as primary tinnitus
• Controls: no tinnitus experienced for longer than 2 minutes in one instance
• Age 18–70
• Able to understand, consent and participate
• MoCA ≥ 26
• Hyperacusis Questionnaire (HQ) ≤ 27
• Speaks Dutch, English or German (screening instruments are not validated in other languages)
• If receiving tinnitus treatment, the regimen must be stable for ≥ 6 months (no initiation or major change); treatment history recorded

Exclusion

1. Objective tinnitus; heartbeat-synchronous tinnitus as primary concern
2. Otosclerosis; acoustic neuroma or other ear disorders with fluctuating hearing
3. Acute ear, nose and throat infection (otitis media, otitis externa, sinusitis)
4. Ménière's disease or similar syndromes
5. Vestibular migraine
6. Serious internal, neurological or psychiatric conditions
7. Epilepsy or other CNS disorders (brain tumour, encephalitis)
8. Clinically relevant drug, medication or alcohol abuse up to 2 weeks before study start
9. Severe hearing loss — PTA ≥ 71 dB HL in the better ear (mean of 0.5, 1, 2, 4 kHz)
10. Inability to communicate properly during the study
11. One deaf ear

Participants with red flags for secondary or retrocochlear pathology (unilateral tinnitus, focal neurological signs, vestibular symptoms, unexplained asymmetric sensorineural hearing loss, disproportionate speech discrimination deficits) are referred for standard-of-care evaluation including MRI of the internal auditory meati / cerebellopontine angle, and enrolled only if no secondary cause is found.

Total duration approximately 3 hours.

Order of the GPIAS and oddball paradigms is randomised between subjects, separately for each site.

Delivered online through REDCap where possible.

Plus additional items on pain and physical discomfort. Individual sites may administer further instruments beyond this common battery.

Pure-tone audiometry. Air-conduction thresholds in a sound-treated booth on a clinical audiometer calibrated to site standard, following otoscopy to confirm a clear ear canal. Thresholds at 0.5, 1, 2, 3, 4, 6, 8, 10 and 12.5 kHz, with bone conduction where indicated, per British Society of Audiology procedures using a standard ascending method (10 dB down / 5 dB up), threshold defined as the lowest level eliciting responses on at least 50% of ascending trials. Extended high-frequency audiometry to 16 kHz at sites so equipped. Masking applied where interaural attenuation or air–bone gap criteria indicate.

Tinnitus pitch. Paired-comparison procedure. A 500 Hz then a 4000 Hz pulsed tone; the participant selects which more closely matches the tinnitus pitch. Comparison then proceeds to 2000/8000 Hz or 250/1000 Hz accordingly, converging in half-octave steps. Repeated three times and averaged.

Loudness matching. Ascending method of limits from 10 dB above threshold, comparing tinnitus loudness to a 1000 Hz tone in 2 dB steps until bracketed. Repeated five times and averaged. Performed in the tinnitus ear, or the ear with the louder percept; if equal, the right ear. Unilateral tinnitus is analysed contralaterally to the tinnitus ear; bilateral tinnitus contralaterally to the worse ear.

Quality assurance. All sites follow a harmonised audiology Scope of Practice covering equipment calibration checks, standardised instructions, threshold criteria, masking rules and data-entry conventions. Site-specific deviations are documented; outcome variables and scoring definitions are held constant across centres.

64-channel EEG at all sites. Tübingen additionally attempts simultaneous MEG using low-interference Etymotic ER2 earbuds; where simultaneous acquisition is not achievable, EEG only is collected.

Recordings are made in a quiet, dimmed room with the participant seated upright.

Resting state — 5 minutes, eyes open.

Hypothesis. Individuals with tinnitus show impaired inhibition of cortical evoked responses to sound pulses when preceded by a silent gap embedded at a carrier frequency close to the tinnitus percept, relative to non-tinnitus individuals. Inhibition with a broadband carrier is expected to be similar between groups.

Unlike classical GPIAS, which indexes the motor startle reflex, this implementation measures the cortical response. GPIAS is regulated at the level of the auditory cortex, and cortical responses may capture inhibition more accurately than motor reflex measures.

Stimuli. Delivered binaurally through Etymotic ER2 ear inserts; background SPL below 35 dB(A). Sound is calibrated directly from the ear insert against a calibration microphone and compared with a high-frequency head calibrator.

• Pulse: broadband, 20 ms, 90 dB, no rise or fall time
• Carriers: continuous broadband noise (BBN), or Butterworth-filtered 1 kHz-wide narrowband noise centred at 3 kHz or 8 kHz (48 dB/octave), all at 60 dB
• Gap: 50 ms, 2 ms rise and fall ramp, down to background noise floor, preceding the pulse by 240 ms ISI
• Trial types: pulse only (PO), gap + pulse (GP), gap only (GO)

Structure. PO, GP and GO are presented 100 times in trios of randomised order within each carrier condition, 50 random trials per block. The three carrier sessions (BBN, 3 kHz, 8 kHz) are presented in randomised order.

Before and after the carrier sessions, an input/output block presents 25 PO trials at randomised intensities (70, 75, 80, 85, 90 dB) in a 60 dB BBN carrier — 50 trials in total — to characterise the ERP input/output growth function and habituation to pulses.

Outcome. Peak amplitude and latency of the N1 ERP.

Participants fixate a dark + on a white background throughout. Total GPIAS duration approximately 1 hour, with intercom contact after each session.

Hypothesis. Individuals with tinnitus show a stronger electrophysiological response to mismatch between prediction and perception (MMN), and increased amplitude with delayed latency of the late positive evoked response (P300), relative to controls. MMN is proposed as a marker for tinnitus loudness, P300 for tinnitus presence.

Procedure follows Wacongne and colleagues.

Stimuli. Two five-element sequences of 50 ms elements, 7 ms rise and fall, 250 ms ISI, normalised to equal root-mean-square power:

• Sequence X — five 500 Hz tones
• Sequence Y — four 500 Hz tones and one white noise burst

Presentation level. Set individually to the participant's most comfortable loudness, established with a 1–7 rating scale in 5 dB steps, followed by a loudness match between the tones and the noise burst. This controls for the possibility that ERP differences between standards and deviants reflect subjective loudness rather than deviance, and accounts for hearing loss in perceptual rather than physical terms.

Protocols. Three, presented in randomised order:

1. xxxxx — Sequence X standard, Sequence Y first deviant, omission second deviant
2. xxxxY — Sequence Y standard, Sequence X first deviant, omission second deviant
3. Omission-only — Sequence Omission, 100% expected

Each protocol comprises 5 blocks of 125 trials. Each block opens with 25 trials of the standard to establish the global rule; of the following 100 trials, standards occur with 75% probability, deviants 15% and omission sequences 10%. Across the five blocks of a protocol: 500 standards, 75 deviants, 50 omission sequences. The omission-only protocol presents 125 Sequence Omission trials separately.

Random intertrial jitter 1300–1500 ms. Participants attend to the stimuli with eyes fixed on a point ahead, with breaks every two blocks and between protocols.

EEG and MEG are non-invasive and serious adverse events are not expected. Anticipated events include transient scalp redness or mild skin irritation from the cap and electrolyte, transient discomfort from sound stimulation including startle or hyperacusis-related discomfort, fatigue or mild headache from prolonged testing, transient dizziness or nausea, and where applicable anxiety or claustrophobia during MEG.

Mitigations: skin-friendly procedures with minimal abrasion, disinfected electrodes, hypoallergenic electrolyte where needed, monitoring throughout, calibrated and conservative sound levels that can be reduced or the session paused or terminated at the participant's request, and regular breaks. Testing stops immediately if a participant reports unacceptable discomfort. All adverse events are recorded with severity and relatedness and reported per local ethics/IRB procedures.

Pilot cGPIAS testing indicated that 90 dB pulses are tolerated by individuals with and without tinnitus (~n=20), giving ERP amplitudes similar to 95 dB pulses, and that 60 dB carriers are as effective as 65 dB and more comfortable (Shabestari et al., JARO. 2025. 26(5):515–529. doi: 10.1007/s10162-025-00999-w). Hence, TIDE tested 90 dB pulses in a 60 dB carrier.

All data are managed under unique study codes. The key associating codes with participant identities is stored in a protected file with access restricted to designated members of the research team at each site. Paper documentation is held in locked cabinets in secured offices; electronic data on a secure, password-protected server. Participant names are not disclosed; analyses use de-identified coded data only, and only summary statistics are reported.

Variables collected for analysis:

1. Demographics — age, sex/gender, race/ethnicity where available, study site, assessment language
2. Clinical — case/control status, tinnitus duration and characteristics from ESIT-SQ, treatment history and stability, MoCA, HQ, THI, TFI, HADS, PSQ-30, VAS ratings
3. Audiological — pure-tone thresholds including high frequencies where possible, derived hearing-loss indices, tinnitus pitch and loudness matching
4. Neurophysiological — outcomes derived from EEG/MEG

No participant-level data are held in any TIDE repository. Data are read from paths outside the repository tree; a .gitignore and a pre-commit hook enforce this. Identifier crosswalks linking TIDE study codes to those of any other study are treated as the most sensitive file class and never enter a repository. See CONTRIBUTING.md.

Any public release of data follows TIDE-DMS-001, the consortium's de-identification standard, which requires a documented disclosure-risk assessment before deposit. Removal of an identifier column does not constitute de-identification.

Paradigm repositories are released as their corresponding manuscripts are published, each archived with its own DOI.

See CONTRIBUTING.md. Before your first commit:

chmod +x .githooks/pre-commit
git config core.hooksPath .githooks

Rainwater Charitable Foundation (#18340). CRC receives additional funding from the American Tinnitus Association.

If you use material from this repository, please cite the protocol:

@article{Vanneste2026TIDE,
  author  = {Vanneste, Sven and Yasoda-Mohan, Anusha and Chen, Feifan and others},
  title   = {Objective data-driven personalised approach to diagnosis of chronic
             tinnitus: the Tinnitus Detection (TIDE) project — protocol for the
             identification and validation of a biomarker for tinnitus},
  journal = {BMJ Open},
  year    = {2026},
  volume  = {16},
  pages   = {e112788},
  doi     = {10.1136/bmjopen-2025-112788}
}

Christopher Cederroth, University Clinic of Tübingen — christopher.cederroth[@]uni-tuebingen.de Sven Vanneste, Trinity College Dublin — sven.vanneste[@]tcd.ie