Tinnitus treatments compared

A — Strong: Consistent findings from high-quality systematic reviews, clinical guidelines, or several well-conducted controlled trials.

Changes the relationship with the sound rather than the sound itself. The most consistently supported approach for tinnitus distress.

Evidence: Multiple systematic reviews and meta-analyses of randomised trials, including a network meta-analysis; named first-line by the 2026 Nature Reviews Disease Primers volume

Worth knowing: Does not reduce loudness. People sometimes read a strong evidence rating as meaning the tinnitus will get quieter; it means daily life gets easier.

B — Moderate: Reasonable supporting evidence, with limitations that matter — smaller trials, shorter follow-up, or inconsistency between studies.

CBT delivered as an online programme with a clinician guiding it: the best-supported approach for tinnitus distress, in a form that can reach people who are never offered it in person.

Evidence: A randomised trial of audiologist-guided internet CBT against weekly monitoring, and a 2025 meta-analysis of nine studies of internet- and mobile-delivered CBT

Worth knowing: The trial compared it with weekly monitoring, not with face-to-face CBT. Pooled across nine studies it improved the Tinnitus Functional Index by 12.48 points but not the Tinnitus Handicap Inventory, and in a wider e-health meta-analysis the studies at higher risk of bias produced the larger effects. What was tested was a guided programme, not an app that borrows the name.

B — Moderate: Reasonable supporting evidence, with limitations that matter — smaller trials, shorter follow-up, or inconsistency between studies.

The VA's stepped programme: an audiologist teaches therapeutic use of sound and a psychologist teaches coping skills derived from CBT. In the trial, five telephone appointments.

Evidence: A randomised trial of 205 people comparing the telephone version against a six-month wait-list

Worth knowing: One randomised trial, against a wait-list rather than an active treatment, so it shows the programme beats waiting, not that it matches CBT or sound therapy delivered on their own. It is those two, sequenced and delivered, rather than a new intervention.

B — Moderate: Reasonable supporting evidence, with limitations that matter — smaller trials, shorter follow-up, or inconsistency between studies.

Acceptance-based approaches. In one head-to-head trial, ACT outperformed tinnitus retraining therapy on distress.

Evidence: Systematic review and meta-analysis, plus a head-to-head randomised trial against TRT

Worth knowing: A smaller literature than CBT's: the number of high-quality randomised trials is still fairly small, and the head-to-head win over TRT comes from a single trial.

B — Moderate: Reasonable supporting evidence, with limitations that matter — smaller trials, shorter follow-up, or inconsistency between studies.

Counselling combined with low-level sound, aimed at habituation over months rather than weeks.

Evidence: Systematic reviews, with methodological limitations noted in the guideline literature

Worth knowing: A 2025 review of 15 randomised trials found TRT no better than any of seven comparators, including usual care and an abbreviated version of TRT itself — a real question about the one-to-two-year protocol. Trial quality is also uneven and protocols vary between clinics.

B — Moderate: Reasonable supporting evidence, with limitations that matter — smaller trials, shorter follow-up, or inconsistency between studies.

Background sound to reduce the contrast between tinnitus and silence. Widely used, low risk, easy to try.

Evidence: Guideline support and recent umbrella reviews; the Cochrane review found no data at all for its main comparisons

Worth knowing: The 2018 Cochrane review found no data at all for the comparisons that matter — sound devices against a waiting list, a placebo, or education with no device at all. Newer umbrella reviews do report consistent benefit, and guidelines support it, but the trial that would settle it has not been run. Evidence is strongest for coping and sleep, weakest for changing the tinnitus itself, and notched-sound variants have produced mixed results in larger analyses.

B — Moderate: Reasonable supporting evidence, with limitations that matter — smaller trials, shorter follow-up, or inconsistency between studies.

For the substantial share of people whose tinnitus comes with hearing loss, amplification is often the single most effective option available.

Evidence: Multicentre clinical studies and guideline support; fewer large randomised trials than the clinical consensus implies

Worth knowing: The Cochrane review of hearing aids for tinnitus found one randomised trial, with 91 participants, and concluded there was no evidence to support or refute their use as a routine tinnitus intervention. The modern support is stronger and comes from pooled syntheses rather than a definitive trial — a 2026 umbrella review lists them among the approaches that consistently help, and the UNITI trial found them one of only two interventions with large effects alone. Only relevant where hearing loss is present, and benefit varies considerably between individuals.

B — Moderate: Reasonable supporting evidence, with limitations that matter — smaller trials, shorter follow-up, or inconsistency between studies.

In people implanted for severe hearing loss, tinnitus in the implanted ear often reduces substantially.

Evidence: Systematic reviews, mostly of before-and-after studies: in single-sided deafness, every study in a 2022 review found a significant reduction, and a 2024 meta-analysis of 28 studies and 853 people with sensorineural hearing loss found Tinnitus Handicap Inventory scores fell by 14.02 points on average

Worth knowing: Applies only to people who already meet the criteria for implantation. It is not a tinnitus treatment on its own terms. Most of the studies compare the same people before and after implantation rather than against an untreated group, many are small, and the 2024 meta-analysis found the effect may diminish over time.

B — Moderate: Reasonable supporting evidence, with limitations that matter — smaller trials, shorter follow-up, or inconsistency between studies.

Pairs sound with mild tongue stimulation. Its FDA De Novo authorisation rests on a single-arm pivotal trial and supporting data; no sham-controlled trial exists.

Evidence: One single-arm pivotal trial (TENT-A3) comparing each participant's bimodal weeks with their own earlier sound-only weeks, two randomised trials comparing stimulation settings with each other, and uncontrolled clinic chart reviews

Worth knowing: No trial has a sham arm, and the pivotal trial's fixed sound-then-bimodal order cannot separate the tongue stimulation from time, expectation or order effects. Its advantage over sound alone appeared in people still moderately bothered after six weeks of sound, not in the full cohort. Every Lenire study so far was either funded by the manufacturer or has an author with a declared tie to it.

C — Limited: Small studies, inconsistent results, or methodological weaknesses that make the findings hard to rely on.

Sensors show muscle tension, breathing, heart rate and other stress signals back in real time, so a person can learn to lower them. Needs equipment and a trained clinician.

Evidence: A randomised trial of 130 patients against a wait-list, testing biofeedback inside a cognitive-behavioural package, and a 431-patient study against routine care

Worth knowing: Neither substantial study isolates the biofeedback: the randomised trial tested a cognitive-behavioural package with biofeedback inside it against a wait-list, and the larger study compared a programme with routine care rather than with the same attention without sensors. Guidelines name CBT, not biofeedback.

C — Limited: Small studies, inconsistent results, or methodological weaknesses that make the findings hard to rely on.

Hypnosis or self-hypnosis aimed at easing the reaction to tinnitus. It looks effective against a waiting list and no better than a single counselling session.

Evidence: Small, mostly older controlled trials; the one randomised comparison against counselling, in 1996, found no difference on any quantitative measure

Worth knowing: The striking results come from comparisons against a waiting list or a brief tone, and from uncontrolled before-and-after figures. Against one counselling session, three sessions of hypnotherapy were no better on any quantitative measure, and the trial's authors could not say whether the one difference, a general sense of improvement, was more than extra attention from the therapist.

C — Limited: Small studies, inconsistent results, or methodological weaknesses that make the findings hard to rely on.

Music you enjoy with a narrow band around your own tinnitus pitch removed, listened to for a couple of hours a day over months.

Evidence: Randomised trials and three meta-analyses; most comparisons against the same music without the notch find no difference

Worth knowing: The comparison that tests the notch itself, notched music against ordinary music, mostly finds nothing: a 104-patient double-blind trial and a 2024 meta-analysis found no difference, and the 2025 meta-analysis that did has a confidence interval reaching almost to zero (−17.01 to −0.23 on the THI). The surviving signal is on loudness rather than distress, and the active ingredient may be the music.

C — Limited: Small studies, inconsistent results, or methodological weaknesses that make the findings hard to rely on.

Magnetic pulses aimed at brain regions associated with tinnitus.

Evidence: Meta-analyses with inconsistent findings; the largest and newest is null

Worth knowing: A 2026 meta-analysis of 8 randomised trials found a pooled effect of SMD -0.15 (95% CI -0.37 to 0.07) with zero heterogeneity — eight trials agreeing there is nothing to find. What earlier analyses found is better described as short-lived than small: a 2025 meta-analysis of 16 trials and 1,105 patients found a benefit at one month and none at six, with no effect on two of its four measures. A treatment whose effect has gone by half a year is a different proposition for a condition measured in years, and that is why this sits at Limited rather than Evidence against.

C — Limited: Small studies, inconsistent results, or methodological weaknesses that make the findings hard to rely on.

A weak constant current passed between electrodes on the scalp. Cheaper and more portable than TMS, and easier to sham convincingly.

Evidence: A 2022 meta-analysis of 14 controlled trials and 1,031 participants; a 2026 meta-analysis found the same effect size with a confidence interval crossing zero

Worth knowing: The pooled effect is small (SMD −0.35 on loudness), only stimulation over the left temporoparietal area produced it, and anxiety and depression did not move. A 2026 meta-analysis of an overlapping set of trials found almost the same estimate and could not rule out no effect. The trials used supervised protocols; a home device is not the intervention that was studied.

C — Limited: Small studies, inconsistent results, or methodological weaknesses that make the findings hard to rely on.

Training aimed at altering the brain activity patterns associated with tinnitus.

Evidence: A 24-participant study with no control group, a randomised trial of 94 patients against a minimal intervention, and a 43-person randomised trial of fMRI neurofeedback against group CBT

Worth knowing: Small, and pointing in different directions. The 24-participant study whose distress scores stayed lower for six months had no control group, and the randomised trial of 94 patients found no distress advantage over a minimal intervention that involved no neurofeedback, though both kinds of neurofeedback reduced rated tinnitus intensity. In the 2024 fMRI trial that beat CBT, the neurofeedback arm had 15 individual sessions against 10 group sessions of CBT, and the treatment needs an MRI scanner.

C — Limited: Small studies, inconsistent results, or methodological weaknesses that make the findings hard to rely on.

One of the most commonly tried alternative approaches for tinnitus.

Evidence: Many trials, systematically reviewed, with persistent quality and blinding problems

Worth knowing: Despite dozens of trials, methodological weaknesses mean the literature has not settled the question either way.

C — Limited: Small studies, inconsistent results, or methodological weaknesses that make the findings hard to rely on.

An injection of local anaesthetic near the sympathetic nerves in the neck, aimed at cochlear blood flow and arousal.

Evidence: Eleven randomised trials pooled in a 2026 meta-analysis, testing the block in combination rather than alone

Worth knowing: The pooled improvement in tinnitus handicap was 5.73 points, below the 7-point threshold used to define a responder — and the reported confidence intervals are far narrower than tinnitus trials normally produce.

C — Limited: Small studies, inconsistent results, or methodological weaknesses that make the findings hard to rely on.

Manual therapy, stretching and trigger-point work on the neck, for tinnitus that changes with neck or jaw movement.

Evidence: One delayed-start randomised trial of 38 patients with neck-related tinnitus, and a 2026 systematic review of 13 studies that found limited or no evidence on tinnitus outcomes

Worth knowing: In the trial, 53% improved substantially straight after six weeks of treatment, but only 24% had kept that improvement six weeks later, while the neck gains held. The wider review rated 10 of its 13 studies at high risk of bias. It is only relevant where the tinnitus has a somatic pattern.

C — Limited: Small studies, inconsistent results, or methodological weaknesses that make the findings hard to rely on.

Better trial data than most tinnitus supplements, particularly where poor sleep is part of the picture.

Evidence: Randomised trials, including a comparison against sertraline

Worth knowing: Benefit appears tied to improving sleep rather than acting on tinnitus directly.

C — Limited: Small studies, inconsistent results, or methodological weaknesses that make the findings hard to rely on.

Widely sold for tinnitus, often together. The research is small: an uncontrolled magnesium study, and zinc trials that found no benefit over placebo.

Evidence: One small single-arm, open-label magnesium study (19 of 26 completed), and a 2016 Cochrane review of three placebo-controlled zinc trials with 209 participants that found no evidence of benefit

Worth knowing: The evidence is considerably weaker than the marketing implies, and does not support routine supplementation. The magnesium study had no placebo group, and none of the zinc trials found a significant difference from placebo. Whether zinc helps people who are deficient is still open, since the placebo-controlled trials were not designed to answer it; correcting a documented deficiency is a separate, ordinary medical question.

D — Preliminary: Early-stage or exploratory research, often without controlled human trials yet.

Pairs nerve stimulation with tones, on the theory that this drives targeted plasticity.

Evidence: Early-stage trials and mechanistic work

Worth knowing: Still exploratory as a standalone treatment. Most of the practical development has gone into bimodal devices instead.

D — Preliminary: Early-stage or exploratory research, often without controlled human trials yet.

Seeds or beads taped to points on the outer ear for continuous pressure, a needle-free form of auricular acupressure. Studied as an add-on to conventional treatment, not a replacement.

Evidence: No randomised trial: a 100-patient retrospective comparison of ear acupressure plus methylcobalamin tablets against the tablets alone, and a longitudinal observational study; a meta-analysis of acupuncture, the parent technique, found no significant effect on its primary outcome

Worth knowing: The best comparison is retrospective: patients were not randomised, the groups were formed from the treatment each had already received, and there was no sham, so attention and expectation are part of what was measured. A meta-analysis of eight acupuncture trials found no significant effect on its primary outcome. What was studied was structured, practitioner-guided acupressure, not ear seeds bought online and self-applied.

E — Insufficient: Too little evidence exists to draw any meaningful conclusion in either direction.

Spinal manipulation aimed at tinnitus with a neck (cervicogenic) component. The mechanism it targets is real; the direct evidence for chiropractic is thin.

Evidence: Case studies and case series; no randomised trial of chiropractic care itself

Worth knowing: The one randomised trial nearby tested a physical therapy protocol, not chiropractic care as commonly practised. A meta-analysis found a small association between chiropractic care and cervical artery dissection (odds ratio 1.74) on very low quality evidence, and found no evidence for causation. Only plausible where the tinnitus changes with neck or jaw movement.

E — Insufficient: Too little evidence exists to draw any meaningful conclusion in either direction.

Cannabis and CBD products, often marketed with an implied benefit for tinnitus that no controlled trial has tested.

Evidence: No controlled human trial; a cross-sectional study of US adults (NHANES) and preclinical receptor research

Worth knowing: The largest dataset points the other way: in a US population survey, high-volume cannabis users had roughly double the odds of tinnitus (odds ratio 2.05). A survey of that kind shows an association, not a cause, and receptor research suggests compounds acting on CB1, as THC does, may worsen tinnitus while the more promising CB2 work is preclinical. Nothing supports starting cannabis or CBD to treat tinnitus.

F — Evidence against: Available evidence suggests no benefit, or that potential harms outweigh benefits.

A low-intensity laser aimed at or near the ear, on the theory that it improves blood flow and cell function in the inner ear.

Evidence: A 2022 systematic review of ten randomised trials, eight of which found it ineffective for chronic tinnitus; reviews that also counted non-randomised studies read more warmly

Worth knowing: The warmer reviews got there by counting weaker designs. The studies used ten different wavelengths and sessions from nine seconds to thirty minutes, so no clinic's protocol can claim the literature's support, and even the positive 2013 trial reported a non-response rate of 57% straight after treatment. It is not dangerous; the case against it is cost and expectation.

F — Evidence against: Available evidence suggests no benefit, or that potential harms outweigh benefits.

Devices that apply pulsed electromagnetic fields near the head or ear. Marketed for many conditions, and a different technology from TMS.

Evidence: Two small double-blind, placebo-controlled trials pointing in opposite directions: a 1993 trial of 58 people, positive on a one-week, four-category self-rating, and a 2004 pilot of 37 adults, negative on the Tinnitus Handicap Inventory, tinnitus ratings and hearing thresholds

Worth knowing: The verdict weighs two small trials of different devices and follows the better-measured one. The 2004 pilot treated people for a month and found no significant change on the Tinnitus Handicap Inventory, tinnitus ratings or hearing thresholds, in the real group or the placebo group. The 1993 trial found 45% improved against 9% on placebo, but after one week, on a single four-way self-rating, and counting only those who completed it. The newer 2024 report of improvement had no external control group and no sham stimulation.

F — Evidence against: Available evidence suggests no benefit, or that potential harms outweigh benefits.

A histamine-acting drug developed for Ménière's disease and widely prescribed for tinnitus in parts of Europe. It is not licensed for tinnitus in the UK.

Evidence: A Cochrane review of five placebo-controlled trials (303 to 305 participants), and a NICE evidence review

Worth knowing: Neither review found a benefit over placebo. Cochrane reports an absence of evidence of any effect on tinnitus, with the risk of bias unclear in every trial, and NICE found no clinical difference, with some possible harm in people who did not have Ménière's disease. The case for it is stronger alongside a Ménière's diagnosis than for tinnitus alone.

F — Evidence against: Available evidence suggests no benefit, or that potential harms outweigh benefits.

A corticosteroid, usually dexamethasone, injected through the eardrum into the middle ear. Standard care in sudden hearing loss; a separate question for chronic tinnitus.

Evidence: A meta-analysis of four double-blind, placebo-controlled trials, and a separate randomised trial of 36 patients; the positive figures come from uncontrolled studies

Worth knowing: Uncontrolled before-and-after studies report tinnitus improving in 75% of patients, but the four placebo-controlled trials pooled in the meta-analysis found no evidence of improvement over placebo. That is about chronic tinnitus without sudden hearing loss or Ménière's disease, where it means a procedure through the eardrum for a benefit the controlled trials did not find.

F — Evidence against: Available evidence suggests no benefit, or that potential harms outweigh benefits.

Breathing oxygen in a pressurised chamber. It has a narrow, time-limited role in sudden hearing loss and none established in chronic tinnitus.

Evidence: A Cochrane review of seven randomised trials and 392 participants, small and generally of poor quality

Worth knowing: The Cochrane review found no evidence of a beneficial effect on chronic tinnitus and does not recommend it for that purpose; its signal was in acute sudden hearing loss, and even there the effect on tinnitus could not be assessed. A course means repeated sessions in a pressurised chamber, and middle ear barotrauma is the main side effect: about 10% of 2,610 patients in a five-year French series.

F — Evidence against: Available evidence suggests no benefit, or that potential harms outweigh benefits.

The most widely recommended supplement for tinnitus.

Evidence: Randomised controlled trials, including large well-designed negative trials

Worth knowing: The largest and best-designed trials found no benefit over placebo. Included here precisely because it is so often recommended — and because a 2026 Korean national guideline now says it may be considered, against the two guidelines below that recommend against it.