About Tinnitus Evidence

In the interest of scientific transparency: our research process does not systematically search every scientific database or every non-English publication. Databases and sources we do not currently search directly include EMBASE, Scopus, Web of Science, PsycINFO and CINAHL; some international trial registries (such as ChiCTR) that lack public interfaces; non-English literature (reached mainly through English-language systematic reviews); and conference abstracts or preprints as a systematic source. Important research from these channels usually reaches the sources we do monitor — for example through PubMed indexing at publication — but sometimes later than specialists would see it. Ongoing weekly monitoring, periodic independent audits, retraction/correction checks and evidence updates are used to reduce these limitations over time, and our audit records are public.

Scored from study quality, size, randomization, blinding, sham/placebo control, independent replication, clinical significance, follow-up length, safety, conflicts of interest and regulatory evidence. The reasoning behind every score is stored and shown on the treatment page, and score changes are logged in each treatment's change history with the reason.

Every treatment page carries an Evidence Profile: evidence quality (from the 1–5 evidence score), replication, loudness evidence and distress evidence (the structured none/limited/moderate/strong ratings), safety (a reviewed tolerability rating), availability (derived from regulatory status and access), and independence — who produced the evidence: primarily independent researchers, a mix, or primarily the treatment's own sponsor. Replication uses five honest categories: none yet · same group/sponsor only (repeats by the same lab or manufacturer are not independent replication) · limited independent · strong independent · and conflicting results, shown as its own amber-striped state when independent attempts disagree — disagreement is different from both "weak" and "none," and is never forced onto the positive scale. The evidence-quality bar has five segments — one per point of the 1–5 score, so 4/5 and 5/5 are visually distinct. The other bars show qualitative categories — Strong fills 4 segments, Moderate 3, Limited 2, Weak 1, None 0 — and a hatched bar means not yet assessed, because unknown is not the same as weak. We deliberately show no percentages and no single "effectiveness score": tinnitus studies measure different outcomes in different populations, and one universal number would be scientifically misleading. Changes to these values go through the same held-for-review process as rankings — they never change automatically.

Results on this site frequently cite standard questionnaires, and it matters what they do and do not measure. The Tinnitus Functional Index (TFI; Meikle 2012, Ear & Hearing) scores tinnitus impact 0–100; a drop of about 13 points is the commonly used threshold for meaningful improvement. The Tinnitus Handicap Inventory (THI; Newman 1996) and the Tinnitus Questionnaire (TQ) are older impact/handicap measures with their own thresholds. All of these measure how much tinnitus affects a person — none of them measures how loud the tinnitus is, which is why a THI/TFI improvement is never presented here as evidence the sound got quieter. Loudness requires its own measures (psychoacoustic loudness matching, loudness ratings). The international COMiT'ID consensus (Hall 2018) formalized that different intervention types should measure different core outcome domains — the methodological basis for our refusal to reduce treatments to a single universal "effectiveness score."

The treatment pages ask how strong the evidence is for a treatment. The Claims section asks a different question: is a claim that people actually meet (a video, an advertisement, a headline, a viral post) accurate? Each article takes one widely shared claim, traces it to the original research, and gives it one of five evidence verdicts: Supported, Partly Supported, Misleading, Unsupported or Insufficient Evidence. What the verdicts mean.

• Verdicts describe the evidence, not the people making a claim. “Unsupported” does not necessarily mean disproven: it can mean a claim was tested without benefit or was never tested well enough to know.
• Every article keeps loudness, distress, functional impact, temporary relief, sustained improvement and cure apart, lists its primary sources, shows its “evidence through” and “last reviewed” dates, and states its limitations and conflicts of interest.
• We name a product, company or person only when a regulator, a published study or a documented fact-check supports the specific statement, and we do not accuse anyone of fraud.
• Claims verdicts never change a treatment’s evidence score, tier or ranking.
• When an article relies on a long guideline, it says which part we read. For the 2026 AAO-HNSF tinnitus guideline we reviewed only the published abstract of its executive summary, not the summary’s full text or the full guideline.
• The Claims articles are for educational and informational purposes only. They are not medical advice, and they have not been reviewed or approved by a physician or other qualified healthcare professional. Our checking of sources is not a substitute for medical review, and the articles do not replace professional medical evaluation, diagnosis or treatment.

Roughly 14% of adults report some tinnitus and about 2% report severe tinnitus (global meta-analysis: Jarach 2022, JAMA Neurology) — but "tinnitus" covers very different situations that can require very different evaluation. Pulsatile tinnitus (rhythmic, heartbeat-synchronized) can have physical vascular causes that need imaging to identify — some treatable, a few dangerous. Somatosensory tinnitus (modulated by jaw or neck movement) has formal diagnostic criteria (Michiels 2018) and its own evidence base. Tinnitus with sudden hearing loss is time-sensitive. That heterogeneity is why this site tags treatments by studied population, shows narrow-population warnings, and why the My Tinnitus Profile red-flag check points characteristics like pulsatile tinnitus, sudden hearing loss, one-sided tinnitus or neurological symptoms toward professional evaluation. The site cannot determine which subtype anyone has — only a clinical evaluation can.

Tinnitus Evidence is kept current by an automated pipeline that runs once a week. Being straight about what that means:

• Automated systems monitor trusted research sources (PubMed, ClinicalTrials.gov, FDA, NIH/NIDCD, Cochrane ENT, major journals, EMA and others), identify relevant new material, extract structured study information (design, sample size, endpoints, outcomes, funding), compare findings against our existing evidence record, and classify weekly updates. Every item links to its original source.
• Routine research updates — a new study report, a trial status change verified at the registry, a follow-up publication — publish to "This Week in Tinnitus Research" automatically, after automated verification of the source, identifiers, dates, and duplicate checks. These items report what happened; they do not change our conclusions.
• Anything that could change a conclusion is never automatic. A possible evidence-score, tier, ranking, loudness/distress-rating, safety or regulatory-status change is placed in a held queue with a full proposed-change report. The existing public assessment stands unchanged until that hold is resolved — one new paper is never allowed to silently rewrite a medical conclusion. Where a held item's underlying fact is verified, it may appear in the weekly feed marked "Impact on Tinnitus Evidence rating: Under evaluation."
• Company announcements are monitored for pipeline intelligence but always labeled "Company-reported development" and never treated as independent evidence. Regulatory statuses update only against authoritative records (FDA databases, ClinicalTrials.gov, EMA), never from press coverage.
• Missing information is never inferred — items say "Not reported" or "Unable to verify." If a week's scan finds nothing meaningful, the feed says exactly that rather than manufacturing news.
• Retractions and corrections are monitored monthly: every research paper cited on this site is re-checked against PubMed and Crossref for retraction, expression-of-concern and erratum notices. A detected notice is flagged visibly on the affected study record and routed to held review — cited research is never silently left standing after its source is withdrawn.

Automated analysis is a monitoring and triage tool — it does not replace medical judgment, and this site does not claim physician or clinician editorial review of weekly items. A complete audit trail (source, identifier, extraction, classification, and publish/hold decision) is kept for every automated update.

When the same development appears in many places we rank credibility roughly as: regulatory record / original trial → peer-reviewed original research → systematic review / meta-analysis → university or hospital publication → trial registry → professional medical organization → reputable medical journalism → company press release → general media → blogs and testimonials. A lower source never overrides stronger primary evidence, and marketing claims are always kept separate from research evidence.

We actively flag: marketing beyond the published evidence, manufacturer-sponsored studies, missing sham controls, self-report-only results, small samples, short follow-up, subgroup-only effects, high dropout, and press releases stronger than the papers behind them. We also avoid dismissing a treatment merely because it is new. Failed and negative trials are shown prominently — they are information, not embarrassments.

Every treatment page shows: last reviewed date, evidence-included-through date, number of studies, evidence score with reasoning, a confidence level, and its full change history. Facts we could not verify against a primary source are marked low-confidence rather than guessed.