Hearing

The single largest modifiable midlife dementia risk factor isn't a drug or a diet — it's the slow loss of hearing most adults dismiss as cosmetic. Nobody has yet proved that treating it protects the brain, but the case for getting your hearing measured no longer rests on that question.

Hearing loss has been quietly reclassified over the last decade. It is no longer treated as an isolated sensory deficit; it's a marker that tracks vascular and metabolic health elsewhere in the body, a measurable correlate of accelerated brain atrophy, and the largest single modifiable risk factor for dementia in midlife, tied with high LDL cholesterol — each accounting for an estimated 7% of global dementia cases.[1] Whether treating it prevents dementia is a separate and less settled question — the only randomised trial was null on its primary endpoint, and that endpoint was three-year cognitive decline, not dementia — but the case for treating hearing loss stands on its own regardless. Meanwhile a parallel diagnostic shift means the standard audiogram most clinics still use stops at 8 kHz and misses damage above it entirely.

Why hearing is on the longevity short list

Three independent signals converge — all observational, so read them as strong, consistent associations rather than proof of cause (Moderate):

  1. Dementia. The 2024 Lancet Commission ranks hearing loss as the largest modifiable midlife dementia risk factor, with a population attributable fraction of ~7%.[2] It tops the list because it is common, not because it is individually dangerous: the Commission pairs a modest relative risk of 1.4 (95% CI 1.0–1.9) with a 59% prevalence. The evidence that treating hearing loss lowers that risk is thinner than the ranking implies — the one randomised trial was null on its primary endpoint, and the supporting dementia signal is observational (see below).
  2. All-cause mortality. A 2025 meta-analysis of 36 cohort studies and 6.4 million people finds adults with hearing loss die at a modestly higher rate — about 21% higher all-cause mortality (hazard ratio 1.21, 95% CI 1.13–1.31), roughly 22% higher cardiovascular mortality (HR 1.22, 1.12–1.33) and 11% higher cancer mortality (HR 1.11, 1.02–1.22). The bracketed range is the 95% confidence interval — where the true effect most plausibly lies. Random error is the least of the uncertainty here, though: the 36 cohorts disagree sharply with one another, and the pooled figure is carried by the Asian and North American studies — across the five European cohorts the association is flat (HR 1.03, 95% CI 0.98–1.08, with no disagreement between them at all). Audiometrically measured loss carries a nominally stronger signal — about 28% higher mortality (HR 1.28, 1.10–1.49) versus about 13% for self-reported loss (HR 1.13, 1.06–1.20) — which fits the idea that an audiogram picks up sub-clinical physiology people don't notice. But the two intervals overlap, the review reports no formal test separating them, and the audiometric subgroup is the more variable of the two, so treat the gap as suggestive rather than established.[3]
  3. Regular hearing aid use is associated with lower long-term mortality. Among the 1,863 adults with audiometric hearing loss inside a National Health and Nutrition Examination Survey (NHANES) sample of 9,885, followed for a median 10.4 years, regular users died at about three-quarters the rate of never-users (adjusted hazard ratio 0.76, 95% CI 0.60–0.95). The word regular is load-bearing: occasional users showed no benefit (HR 0.93, 0.70–1.24), and pooling occasional users in with regular ones left the association no longer significant once medical history was accounted for (HR 0.83, 0.67–1.03). Income, education and insurance were adjusted for, but this remains an observational signal that cannot rule out healthier, more adherent people being likelier to wear aids — the authors say as much themselves.[4]

The stria vascularis — the cochlea's dense capillary bed — depends on continuous blood flow to sustain the ionic gradient that hair cells run on, and disrupted flow, hypoxia and inflammation in that bed are among the better-characterised mechanisms of age-related hearing loss.[5] That shared vascular biology is the usual explanation for why hearing loss travels with diabetes and cardiovascular disease. How far it makes hearing loss a whole-body prognostic marker is unsettled: in the German Gutenberg Health Study — 8,886 adults aged 35–74, all tested by audiometry — people with hearing loss died at more than five times the crude rate of those without, but the gap vanished once age, sex and socioeconomic status were accounted for (fully adjusted hazard ratio 1.01, 95% CI 0.72–1.40), and the authors read hearing loss as an early marker of vascular dysfunction rather than an independent predictor of death. Follow-up there was short — about seven years and 273 deaths — so one cohort does not overturn the pooled estimate above; it does suggest much of the mortality signal is a proxy for age and disadvantage.[6]

Do hearing aids prevent dementia? A null trial and a promising subgroup

This is the question the field most wants to answer, and the honest answer is "probably helps, not proven" (Moderate — one null RCT with a positive subgroup, plus consistent observational data). Be precise about what each study is:

Trial / cohortDesignn / populationKey finding
ACHIEVE (2023)Randomised controlled trial — the only one with cognitive decline as its primary endpoint977 older adults, from two cohorts: an at-risk group from the Atherosclerosis Risk in Communities (ARIC) study and healthier de-novo volunteersNull on its primary endpoint. Hearing intervention did not slow cognitive decline overall versus health education at 3 years. In the higher-risk ARIC subgroup, decline was ~48% slower — a prespecified stratified analysis with a significant interaction (p = 0.010), but still a 238-person stratum of a trial that was null overall, so hypothesis-generating rather than established.[7]
ASPREE (2026)Target-trial emulation — observational, not a trial, despite the name2,777 Australians aged 70+ reporting hearing problems (self-reported, used as a proxy for moderate hearing impairment — not audiometry), dementia-free at baseline; across the analysis's imputations a median of 664 of them received a new hearing-aid prescription7-year dementia risk 5.0% with hearing aids vs 7.5% without — about a third lower in relative terms, or ~2.5 fewer dementia diagnoses per 100 people over 7 years (risk ratio 0.67, 95% CI 0.37–0.97); cognitive impairment 42.4% → 36.1%, though that contrast does not clear the null (RR 0.85, 95% CI 0.70–1.00); and dementia risk fell stepwise with how often the aids were actually worn (7.3% never → 6.8% rarely/sometimes → 4.8% often/always — a separate and imprecise contrast, RR 0.67, 95% CI 0.29–1.06). The authors grade this as Class III evidence and say they cannot rule out residual confounding; a correction to the paper was published in Neurology on 28 April 2026.[8][9]

So the evidence is one null RCT whose apparent benefit sits in its higher-risk participants — a later analysis of the same trial, stratifying after the fact by predicted risk, put 3-year cognitive decline 61.6% slower (95% CI 33.7%–94.1%) in the top quartile[10] — plus a consistent observational signal with a dose-response, now pooled across 46 studies and 231,565 people (about 16% lower risk of cognitive impairment, RR 0.84, 95% CI 0.80–0.88).[11] That combination is a good reason to treat hearing loss — which is worth treating on its own merits regardless — but it is not proof that hearing aids prevent dementia, and anyone telling you the question is settled is overselling it.

Two further caveats from the ASPREE analysis:

  • The signal is concentrated at crossing the clinical threshold into dementia. Mean cognition scores among survivors barely moved (0.03 SDs, 95% CI −0.14 to 0.21). In this dataset hearing aids track the diagnosis rather than the test scores — but read that as provisional: it rests on few dementia events in an observational analysis, and it is not what the one randomised trial measured.
  • Use matters more than purchase. Buying a hearing aid that lives in a drawer is not the intervention. The ASPREE emulation graded use in times per month, not hours per day — never, occasional (≤3 times a month), regular (>3 times a month) — and its top-versus-never comparison is imprecise on its own. A 2026 pooled analysis of 61,089 hearing-impaired adults across seven cohorts sharpens the point: the lower dementia risk shows up only in people who report the aid actually improved their hearing (about 14% lower risk, HR 0.86, 95% CI 0.80–0.93), and vanishes in those who report it did not (HR 0.98, 0.90–1.07).[12]

Three mechanisms are proposed for how auditory deprivation might accelerate dementia. None is established: a 2026 editorial by four hearing-and-dementia epidemiologists argues the association is still "poorly understood", with measurement error, shared causes, reverse causation and selection bias each capable of generating it on its own.[13] Genetic evidence does not settle it either — a Mendelian randomisation study, which tests causation by using inherited genetic variants as a natural experiment, found no causal signal running from hearing loss to Alzheimer's disease specifically, though the associations it did find were mediated by loneliness, low mood and medial temporal lobe volume.[14] A 2025 review reached the same verdict: the evidence for a direct causal link is not conclusive, and some genetic studies suggest hearing loss may be an early feature of Alzheimer's disease rather than a cause of it.[15]

  1. Cognitive load. A degraded auditory signal forces the brain to spend executive resources decoding speech, depleting the working memory and consolidation capacity available for everything else.
  2. Brain atrophy. In the Baltimore Longitudinal Study of Aging, adults with hearing impairment lost whole-brain and right temporal-lobe volume faster than normal-hearing peers over about six years,[16] and poorer hearing measured in midlife (mean age 54) predicted steeper temporal-lobe, hippocampal and entorhinal decline nearly two decades later.[17] Both are observational and come from the same cohort, so they are a consistent association rather than two independent replications.
  3. Psychosocial pathway. Untreated hearing loss drives social withdrawal, and social isolation raises dementia risk by about 60% (relative risk 1.6) in the same Lancet Commission model that ranks hearing loss — a 5% population attributable fraction in its own right.[18] Poor social relationships are also associated with about 29% higher coronary heart disease risk and about 32% higher stroke risk.[19] The first leg of this pathway has randomised support: in ACHIEVE, hearing intervention slowed the loss of social network members and, in that trial, reduced loneliness over three years — though these were exploratory outcomes and the network effect was about one extra person.[20] See Purpose.

A Dartmouth-led study with McGill's Douglas Research Institute, funded by a $6 million National Institute on Aging grant in 2025, will track amyloid and tau across three timepoints over five years in about 70 adults with mild cognitive impairment to ask whether the relationship is "bottom-up" (sensory deprivation accelerates amyloid/tau accumulation) or "top-down" (early Alzheimer's pathology in the central auditory cortex manifests as hearing complaints).[21] No results have been reported yet; both directions may operate at once, and either way the practical implication is the same: treat the hearing loss.

The diagnostic shift: PTA4 is no longer enough

The standard four-frequency pure-tone average (PTA4: 0.5, 1, 2, 4 kHz) used in most primary-care hearing screens is a real but blunt predictor of the symptom most adults actually complain about — the inability to follow conversations in a noisy restaurant. Testing above 8 kHz, which PTA4 ignores entirely, appears to capture noticeably more of it.

The 2026 STOP (Swedish Tinnitus Outreach Project) cohort analysis (526 ears across 263 tinnitus-free adults aged 20–84) is the sharpest data on this so far — though it is a single-author, non-peer-reviewed preprint, cross-sectional, in a highly educated Swedish sample tested at a single noise level, so read the figures below as provisional rather than settled (Weak-preliminary):[22]

Diagnostic markerWhat it measuresVariance in speech-in-noise it explains
PTA4 (0.5–4 kHz)Mid-frequency thresholds~19–25%
EHF — extended high frequency (10–16 kHz)Basal cochlear hair cell integrity~34–36%
ABR (auditory brainstem response) Wave I amplitudeCochlear synaptic densityNegligible — no significant association, in a sub-sample of about 140 ears
ABR Wave I latencyAuditory nerve conduction speedModest on its own (~12%), but age-dependent: across the observed range of latencies a modelled 75-year-old's word score falls about 25 percentage points, to 57%, while a 45-year-old's barely moves

The widely quoted "64% versus 16%" pair measures something else entirely: it is how much of each test's own variance is explained by age — extended high-frequency thresholds are by far the most age-sensitive audiometric measure in this cohort — not the share of speech-in-noise performance either test accounts for. On the preprint's own figures no measure it tested explains more than about a third of why people differ in noise, and its best models leave roughly two-thirds unexplained.

Two practical takeaways:

  • If you struggle in noise but pass a standard audiogram, ask for extended high-frequency (EHF) audiometry — it tests above the standard 8 kHz ceiling, typically out to 16 or 20 kHz, and picks up the basal-cochlea damage the standard test cannot see. The clinical evidence behind it is thin: in a retrospective series of 40 symptomatic adults with normal conventional audiograms, 25 turned out to have hearing loss above 8 kHz, and the authors say plainly that no evidence-based algorithm for EHF audiometry yet exists.[23] The American Academy of Otolaryngology–Head and Neck Surgery Foundation (AAO-HNSF) 2024 age-related hearing loss guideline does not mention extended high-frequency testing at all, so expect to ask for it rather than be offered it.
  • Cochlear synaptopathy — the much-discussed "hidden hearing loss" attributed to lost synapses between hair cells and auditory nerve fibers — is contested, not settled. In the STOP cohort, ABR Wave I amplitude, the usual proxy for synapse loss, showed no relationship to speech scores, while Wave I latency did, which the preprint's author reads as conduction slowing rather than synapse loss — while noting explicitly that the finding "requires validation". Other groups disagree: a 2025 cross-species study with histologically confirmed neural degeneration found that degeneration does contribute to speech-in-noise difficulty and listening effort in midlife,[24] and a 2026 study of 60 adults with normal audiograms found neither synaptopathy nor extended high-frequency loss independently associated with difficulty in noise.[25] Part of the problem is measurement: Wave I amplitude is a weak, noisy proxy, so a null result for it is not a null result for synapse loss.

The metabolic comorbidity nobody screens for

Hearing loss is a heavily underdiagnosed comorbidity of type 2 diabetes (Strong as an association; the direction of cause is presumed microvascular). In NHANES audiometric data on 5,140 adults aged 20 to 69, hearing impairment was roughly twice as prevalent in adults with diabetes as in those without, across both low/mid and high frequencies, and the association held after adjustment for age, noise exposure, ototoxic medication use, and smoking — on self-reported diabetes that did not separate type 1 from type 2.[26] A 2026 systematic review of 29 studies found that about one in four adults with diabetes has moderate-or-worse hearing loss (pooled prevalence 24%, 95% CI 19–30%, from 23 of those studies), and across the 11 studies with a control group the odds of hearing loss were roughly 2.4 times higher than in controls (odds ratio 2.41, 95% CI 1.62–3.60). The gradient by age is the striking part: odds were tripled in adults under 60 and not significantly raised at all in adults 60 and over — an early complication, not a late one. Heterogeneity is high, small-study effects are present, and every included study is observational, so read the direction rather than the decimals.[27]

The presumed mechanism is microvascular and inflammatory — high blood sugar degrading the stria vascularis (the cochlea's blood supply) and the synapse between inner hair cells and the auditory nerve — though the mechanistic literature is still inconsistent and rests heavily on animal models.[28] What is measurable in people is that diabetes tracks with weaker otoacoustic emissions (the faint sounds a healthy cochlea produces on its own) and slower auditory-brainstem conduction, neither of which a standard audiogram looks at — and that hearing impairment also predicts incident diabetes, so the association runs both ways.[29] Prediabetes already carries about a 30% higher rate of hearing loss, so anyone with prediabetes, T2D, or metabolic syndrome should treat unexplained difficulty in noise as a possible early signal — not a normal aging variant.[30]

The exposome: noise + air pollution + ototoxic chemicals

Three environmental hazards combine — and they synergize.

Recreational and occupational noise

The pathophysiology is strictly dose-dependent (Strong — the noise dose-response is well established), and the decibel scale is logarithmic, so safe exposure time collapses fast. These are WHO's own weekly allowances, with WHO's own example sounds:[31]

Sound levelWHO's example soundSafe exposure per week
80 dBDoorbell40 hours
85 dBHeavy traffic, inside a car12 h 30 min
90 dBShouted conversation4 hours
95 dBMotorcycle1 h 15 min
100 dBHair dryer20 minutes

A hair dryer is a 100 dB sound, not an 80 dB one: the loudest appliance in an ordinary bathroom spends a whole week's allowance in twenty minutes.

WHO estimates that more than 1 billion young people aged 12–35 may be at risk of hearing loss from unsafe listening through earbuds and at venues.[32] That headline rests on a modelled range — 0.67 to 1.35 billion — built from 33 studies of listening habits, and it counts people exposed to unsafe listening, not people already losing their hearing.[33] WHO's 2022 standard for venues and events recommends a 100 dB average ceiling, live sound monitoring on calibrated equipment, and hearing protection made available to audiences.[34] These are recommendations, not law; WHO's most recent safe-listening standard is the May 2025 one covering video gameplay and esports.

The air-pollution synergy is severe

A 2025 prospective cohort of 1,179 oil workers in Hebei, followed from 2017–2019 baselines to 2023 (Moderate — a single occupational cohort), found that high cumulative noise exposure (≥85 dB(A)-year) alone more than doubled the risk of occupational hearing loss (RR 2.36). Air quality made that worse, but how much worse depends on which model you read. In the primary fully adjusted analysis the two hazards together added about 8% on top of their separate effects (interaction RR 1.08, 95% CI 1.03–1.13), and the equivalent term for nitrogen dioxide (NO₂) was similar in size but no longer statistically significant once the model was fully adjusted. A post-hoc regrouping of the pollutant data produced the much larger figures usually quoted from this paper — nearly six-fold for the worst air-quality group (RR 5.77, 95% CI 3.22–10.34) and roughly four-fold for the highest-NO₂ group at the 80 dB(A)-year noise level (RR 4.29, 95% CI 1.88–9.83). Read those as exploratory: in the same cohort coarse particulates (PM₁₀) ran the other way, fine particulates traced a U-shaped curve, and sulphur dioxide one that rose and then fell.[35] The proposed mechanism is systemic oxidative stress and vascular inflammation stripping the cochlea of its defenses against acoustic trauma.

UK Biobank data (n = 79,277) extends this to tinnitus, though weakly. Each interquartile-range increase in a composite air-pollution score was associated with 6% higher odds of current tinnitus before adjustment, falling to about 3% after full adjustment for workplace noise, hearing difficulty and hearing-aid use — and the association with constant tinnitus faded out altogether. Adults who sat in both the most-polluted fifth and the top tier of genetic risk (measured by a polygenic risk score, a tally of many small-effect gene variants) had 55% higher odds of constant tinnitus than adults low on both. That is a joint association rather than a demonstrated interaction: the authors found very limited multiplicative interaction between pollution and any variable they tested, the analysis is cross-sectional, and their own prospective follow-up was largely null, so treat it as hypothesis-generating.[36] See Environmental toxins for the broader air-pollution story.

Ototoxic chemicals — a silent third channel

A 2024 review of the US Agency for Toxic Substances and Disease Registry's toxicological profiles gathered ototoxicity information on 62 substances and found strong evidence for 26 of them, 13 of which occupational health agencies have already formally designated ototoxic. Its own headline is how thin the evidence still is for the rest, and it found no usable evidence on tinnitus for any substance (Moderate).[37] The CDC's National Institute for Occupational Safety and Health (NIOSH) estimates that about 22 million US workers — roughly 13% of the civilian workforce — are exposed to ototoxic chemicals each year, and about 12 million to both loud noise and ototoxic chemicals. Many of these substances are breathed in, swallowed with contaminated food or drink, or absorbed through the skin, so acoustic hearing protection does nothing about them. NIOSH lists these groups as occupational hazards that compound noise damage:[38]

CategoryCommon examplesWhere you'd encounter them
Industrial solventsToluene, styrene, xylene, trichloroethylenePaints, adhesives, degreasers, refinishing, hobbies
Heavy metalsLead, mercury, organic tin compoundsOld paint and plumbing, contaminated water and fish, antifouling paints and PVC stabilisers. Cadmium, arsenic and manganese are suspected but not established
AsphyxiantsCarbon monoxide, hydrogen cyanide, tobacco smokeEngine exhaust, poorly ventilated garages, smoking and second-hand smoke
NitrilesAcrylonitrile, 3-butenenitrilePlastics, rubber and acrylic-fibre manufacture
PesticidesOrganophosphates and othersAgriculture, forestry, spraying work
Aminoglycoside antibioticsGentamicin, streptomycinTreatment for serious infections; monitor especially with kidney disease
Loop diureticsFurosemideHypertension/heart failure; risk compounds with other ototoxic drugs
OTC analgesicsRegular NSAIDs and acetaminophen; high-dose salicylatesAcute salicylate tinnitus reverses on stopping. The chronic signal is different and small: in 55,850 women followed for two decades, more than six years of regular NSAID or acetaminophen use tracked about 10% higher risk of persistent hearing loss, while duration of aspirin use tracked none.[39]
Platinum chemotherapyCisplatin, carboplatinThe drug class NIOSH singles out; permanent and dose-cumulative, with hearing loss in half or more of treated patients

For most healthy adults the relevant hazards are solvents (DIY refinishing without ventilation), CO from engines in garages, and being on multiple ototoxic medications simultaneously — flag this with a pharmacist if you're taking three or more.

Sleep, the cochlear clock, and reactive oxygen species

The evidence here is Weak — a plausible mechanism plus one observational dose-response curve. The cochlea has its own peripheral circadian clock, synchronized to the brain's master clock (the suprachiasmatic nucleus), that governs local sensitivity to noise via brain-derived neurotrophic factor and glucocorticoid (stress-hormone) signalling.[40] Sound transduction generates reactive oxygen species, and oxidative stress is one of the main mechanisms of noise-induced hearing loss. The tempting next step — that sleep is the window in which the cochlea clears them — is an extrapolation from glymphatic-clearance work in the brain; the cochlear circadian literature does not show it.

A NHANES analysis of 4,883 adults aged 20–69 found a U-shaped relationship between sleep duration and hearing thresholds, with the best thresholds at about 8 hours and worse at both ends — but only in men and in older adults. In women, young adults and middle-aged adults there was no relationship at all.[41] The study is cross-sectional, so it captures an association at one moment rather than sleep changing hearing over time, and it measured nothing about reactive oxygen species or hair cells. Sleep well for the reasons that are already established, and treat the Sleep prescription as a plausible bonus for hearing rather than a proven one.

Diet — modest but real signals

The systemic vascular and oxidative-stress mechanisms make the dietary signal mostly the same one as for cardiovascular disease and dementia: a Mediterranean / plant-rich pattern (Weak-to-Moderate — the evidence is entirely observational).

  • A 2025 systematic review and meta-analysis of 33 observational studies, 21 of them cross-sectional, reported inverse associations between hearing loss and eight dietary exposures: vitamin B2, β-carotene, carotenoids, β-cryptoxanthin, fat, protein, fibre and fish.[42] Read the size, not the list: the associations are small and mostly turn up in subgroups of otherwise null pooled analyses — the whole vitamin-B pooling was flat, with only the four vitamin-B2 studies significant; the whole carotenoid pooling was non-significant, with only β-carotene reaching significance; and fish rests on four studies at the edge of significance. What did not show up matters for the ACEMg bullet two below: minerals, carbohydrates, and vitamins A, C and E were all null.
  • The Chinese Longitudinal Healthy Longevity Survey (3,839 older adults, mean age 79) found a higher plant-based dietary diversity score tracking lower incident hearing loss over a mean 4.2 years — about 22% lower risk in the most varied diets (HR 0.78, 95% CI 0.64–0.96), on a linear dose-response.[43] Total, animal-based and protein-based diversity scores showed nothing. Two caveats do a lot of work here: hearing loss was self-reported on a questionnaire rather than measured by audiometry, and the cohort averaged 79 years old, so this says little about a 50-year-old.
  • ACEMg (vitamins A, C, E + magnesium) is the most-marketed targeted formula for cochlear oxidative stress, and it does not have the evidence its marketing implies. The widely circulated figure — otoacoustic-emission scores (a measure of outer-hair-cell function, not of hearing thresholds) unchanged or improved in 75.3% of users, against 26.8% of the untreated group — comes from a retrospective chart review with a historical, non-randomised comparison group, published in an integrative-medicine journal by authors tied to the product.[44] A retrospective cohort compared against historical controls is the design that manufactures effects out of selection bias, and the authors' own conclusion concedes that alternative explanations cannot be excluded. Two randomised, placebo-controlled, double-blind trials of ACEMg exist, and both were null. In the first, 51 patients were enrolled around cochlear-implant surgery to see whether the supplement protected the residual hearing in the operated ear, and among the 49 analysed, hearing loss at 500 Hz three months on was 4.15 dB lower with ACEMg than with placebo — a gap entirely compatible with chance (95% CI −12.95 to 4.65, p = 0.35). The full paper states that no superiority over placebo could be demonstrated, though the abstract's conclusion is more upbeat. The trial stopped early when the supply of study drug broke down, it has never been replicated, and the 8 dB figure sometimes quoted from it comes from a per-protocol re-analysis that was null as well.[45] That trial was about protecting residual hearing during surgery, not about age-related loss. The second gave the same combination (with β-carotene as the vitamin A) to normal-hearing adults before loud music; it did not prevent the temporary threshold shift the music caused, and transient tinnitus was reported more often in the supplement group.[46] So the honest statement is narrower than "no trial exists": both randomised tests of ACEMg in humans found nothing. The meta-analysis two bullets above found vitamins A, C and E and minerals all null for hearing risk too. Treat ACEMg as unproven.

The honest read: there is no specific "ear vitamin." The protective dietary signal is the same Mediterranean and MIND (Mediterranean-DASH Intervention for Neurodegenerative Delay) pattern that protects the heart and brain, mediated by the same cochlear microvasculature.

Modern hearing aids and filtered earplugs

The technology has changed faster than most people realize.

Hearing aids in 2026 are dual-chip devices: a conventional acoustics processor paired with a deep neural network coprocessor trained on millions of hours of real-world acoustic environments. Modern algorithms perform real-time beamforming and speech isolation — separating a target voice from competing multi-talker babble — rather than the indiscriminate noise suppression of earlier generations. The most-quoted number, up to 28% better speech intelligibility in difficult listening, comes from one manufacturer's launch material for its neural-network directional system; it is a ceiling rather than a typical result, it describes one product line rather than hearing aids as a class, and it has no independent replication (Weak).[47] What independent evidence supports is more modest and more useful: a 2026 Cochrane review of 16 randomised trials in 2,261 adults found a large benefit on hearing-specific quality of life and on listening ability, at moderate certainty — and no demonstrated benefit for working memory, so the tidy story about amplification freeing up cognitive capacity is still a hypothesis.[48]

Over-the-counter hearing aids have been legal in the US since 17 October 2022: adults aged 18 and over with perceived mild-to-moderate hearing loss can buy them without a prescription, a medical exam or a professional fitting, and the devices must let the buyer set them up themselves.[49] Self-fitting works better than most people expect. In a randomised trial, adults who set up their own over-the-counter device with remote support did as well at six weeks as adults whose identical device was fitted by an audiologist to prescriptive targets with real-ear verification.[50] A 2026 systematic review of 15 studies in 774 adults found no difference on any pooled outcome — hearing-aid benefit, spatial hearing, satisfaction or speech in noise — though the certainty is low and most trials ran only a few weeks.[51] Both come from the same research group, and the trial device is made by hearX, a company two of the authors report payments from. The practical read for mild-to-moderate loss: an over-the-counter device is a reasonable first move rather than a compromise. See a clinician first instead if hearing dropped suddenly, is much worse in one ear, or comes with pain, discharge, ringing in one ear only, or vertigo.

Filtered earplugs — Loop and similar — are passive devices, and that is the point. A precision acoustic channel and a filter lower the whole spectrum roughly evenly, so music and speech stay intelligible instead of muffled; there are no electronics, no battery and no app, and models with switchable modes change them mechanically.[52] Manufacturers quote attenuation in the high teens to mid-twenties of decibels on the European SNR scale; the US NRR figure printed on the same product is much lower — Loop's Experience is 18 dB SNR but 7 dB NRR — so compare like with like. For concerts, factory floors, motorcycles, and other situational noise exposure, a pair of well-chosen filtered earplugs is the most cost-effective hearing-protection upgrade a healthy adult can make, and the only kind people will actually wear consistently. Electronic (level-dependent) hearing protection, which amplifies quiet sound and clamps loud transients, is a separate and more expensive category, worth considering mainly for shooting and prolonged machinery work. Cearvol, sometimes listed alongside Loop, is not an earplug brand at all: it sells over-the-counter hearing aids.

Aural rehabilitation: the brain has to relearn the signal

A common point of failure (Moderate): a person gets fitted with hearing aids, finds the amplified soundscape exhausting and unfamiliar, and stops wearing them within months. Most people do need weeks to months before amplified sound stops feeling strange. How much of that is genuine central reorganisation is less settled than the phrase implies: a 2025 preregistered systematic review of 25 controlled studies in new hearing-aid users found acclimatisation in some users and some outcomes but not others — of the 18 studies measuring speech recognition, 10 showed it and 7 showed none — and where it appeared the gains were modest.[53] The one variable that consistently predicted it was consistent daily wear. As audiologists put it: you hear with your brain, not your ears.

Auditory-training apps exist and people enjoy them, but be honest about what they buy. The reference systematic review found that training reliably improves performance on the trained task, generalises weakly and inconsistently to real speech understanding, and is not robust enough to guide clinical practice.[54] A 2025 meta-analysis of seven randomised trials found a signal on global cognition but nothing on working memory or attention, with heterogeneity high enough that the pooled numbers barely mean anything.[55] No trial has tested whether training makes people keep wearing their devices. If you want one anyway, check that it is built for adults adapting to hearing devices rather than for children or cochlear-implant users, and expect it to sharpen the drill more than the restaurant.

Gene therapy and the regenerative pipeline

Until recently, mammalian sensory hair cells and auditory neurons were considered permanent once destroyed. That changed on 23 April 2026, when the FDA granted accelerated approval to Regeneron's Otarmeni (lunsotogene parvec-cwha) — the first gene therapy for OTOF-related deafness. The licence covers children and adults with severe-to-profound and profound sensorineural hearing loss — any frequency worse than 90 decibels hearing level (dB HL), the scale audiograms are plotted on — from molecularly confirmed biallelic variants in the OTOF gene — which encodes otoferlin, a protein the inner-ear synapse needs to release its neurotransmitter — and only where outer hair cell function is preserved and the ear has not already received a cochlear implant. The label lists procedure-related complications including vertigo, tinnitus, cerebrospinal fluid leak, facial nerve paralysis, taste changes, meningitis, infection and labyrinthitis.[56] It is a dual-vector adeno-associated virus (AAV) therapy: the otoferlin gene is too large for one AAV, so it is split across two. Approval rested on Otarmeni's own CHORD trial — single-arm and uncontrolled. Its first 12 children were published under the drug's development name, DB-OTO, and 9 of the 12 reached the trial's primary hearing target (average thresholds of 70 dB HL or better) at 24 weeks;[57] by the time of the FDA's review the trial had treated 24 children aged 10 months to 16 years, with hearing improved in 80% of the 20 who could be evaluated.[58] It is an accelerated approval, meaning the licence is provisional and continued marketing depends on confirming durability and effects on speech development. The decision came 61 days after filing, under the FDA's national priority voucher pathway. Every patient treated so far has been a child; the label states the product has not been studied in anyone 65 or older, and administration requires ear surgery plus a short course of systemic corticosteroids (Moderate, and only for this narrow indication).[59]

The longest durability data on OTOF gene therapy come from a separate product and trial on the same target — a Chinese multicentre single-arm study of AAV1-hOTOF in 42 participants aged 0.8–32.3 years (not the Otarmeni/CHORD cohort). A single intracochlear injection recovered hearing in 90%, with auditory-brainstem-response thresholds improving from worse than 97 dB at baseline to 42 dB at 2.5 years and behavioural thresholds to 37 dB. Children aged 0.5–18 responded better than adults. No dose-limiting toxicities occurred, though grade 3 adverse events included decreased neutrophil count.[60]

Otarmeni reaches only OTOF-related deafness, which the FDA puts at 2–8% of inherited, non-syndromic cases — and genetic causes account for only about half of congenital hearing loss in the first place, so this is a small slice of a small slice. It matters mostly as proof of concept, and the pipeline behind it has already moved on. Gene therapy for GJB2 — the most common genetic cause of congenital severe-to-profound hearing loss in many populations[61] — has already reached patients: a Phase 1/2 trial of a single intracochlear injection in young children has been recruiting since May 2026.[62] It needs only one vector; the two-vector trick is specific to the oversized otoferlin gene rather than a platform being carried across. Rinri Therapeutics' Rincell-1, allogeneic cells aimed at regenerating auditory neurons, is a 20-adult safety study in people who already qualify for a cochlear implant, with the cells injected during implant surgery; the registry listed it as registered but not yet recruiting as of August 2026, with results not expected before 2028.[63] The regenerative era for hearing has started; it doesn't yet reach the average healthy adult with age-related decline, but the pipeline is now real.

Screening: what the guidelines actually say

The AAO-HNSF 2024 Clinical Practice Guideline on age-related hearing loss is more assertive than what most primary-care offices currently practice. It is scoped to adults aged 50 and over, and it grades every one of its 11 key action statements, so the weight behind each line below is not the same. These are the statements that bear on this article, not all eleven:[64]

  • Screen for hearing loss at the time of a healthcare encounter from age 50 — a recommendation rather than a strong one, and the guideline itself says there is no consensus on how often to screen.
  • Look in the ear before blaming age — otoscopy for wax impaction, infection or other abnormality whenever a screen suggests loss.
  • Obtain an audiogram whenever a screen suggests hearing loss — one of only three strong recommendations. The guideline sets no deadline for it; a separate 2021 AAO-HNSF quality-measure set proposed a 4-week window.
  • Escalate the red flags — markedly asymmetric loss, conductive or mixed loss, or poor word recognition all need evaluation for something other than ageing.
  • Offer appropriately fitted amplification, and refer for a cochlear-implant assessment when someone has well-fitted aids and still cannot understand speech. These are the other two strong recommendations.
  • Counsel the patient and their family or care partner on how hearing loss affects communication, safety, function, cognition and quality of life, and on communication strategies and assistive listening devices.
  • Check within a year whether communication goals were met and hearing-related quality of life improved — for everyone with hearing loss, not only device users. Retesting hearing at least every 3 years in anyone with known loss or a new concern is offered as an option, the guideline's weakest tier.

Two things the guideline does not say, though both are widely attributed to it: it recommends nothing before age 50, and it does not mention extended high-frequency audiometry anywhere.

The US Preventive Services Task Force (USPSTF) does not agree, and has not moved since: for asymptomatic adults 50 and over it finds the evidence insufficient to weigh the benefits and harms of screening for hearing loss — a grade I statement, which is a refusal to conclude either way rather than a verdict against screening, unchanged from 2012 and still current.[65] The disagreement is narrower than it looks. The Task Force statement explicitly excludes anyone "reporting signs and symptoms of hearing loss", and tells clinicians to use their judgement for patients who have raised a concern. So if you are struggling in restaurants, nothing here is in tension: get tested. What is unsettled is whether testing people who have noticed nothing does more good than harm — not what to do once loss is found, which is where the guideline's strongest evidence sits.

A practical hearing protocol for healthy midlife adults

  1. Get a baseline audiogram in your forties — this one is our suggestion, not a guideline recommendation: the AAO-HNSF guideline starts at 50, and the USPSTF finds the evidence for screening asymptomatic adults insufficient either way. The case for a baseline is simply that a personal reference point makes later change legible instead of a guess. Repeat every 3–5 years; sooner if anything changes.
  2. If you struggle in noisy environments but pass standard audiometry, request EHF audiometry. It tests above the standard 8 kHz ceiling, typically out to 16 or 20 kHz, and picks up basal-cochlea damage. Speech-in-noise testing is also worth asking for.
  3. Treat any confirmed hearing loss promptly. It is worth treating for its own sake; whether it protects against dementia is not yet proven. The device that lives in a drawer doesn't count.
  4. Use filtered earplugs at concerts, sporting events, motorcycles, leaf blowers, table saws. A 100 dB venue is hazardous after 20 minutes; a pair of filtered earplugs preserves the music while protecting the cochlea.
  5. Cap personal-audio listening at the 80 dB / 40 hours per week envelope — Mode 1 of the 2019 safe-listening standard from WHO and the International Telecommunication Union; the setting for children is 75 dB over the same 40 hours.[66] Most phones can show weekly listening dose — watch it.
  6. Mind the synergy: don't combine high noise with poor air quality (airport runways, heavy traffic, refinery work). The one occupational cohort that measured both found the combination consistently worse than noise alone, though how much worse depends heavily on how the pollutant data are grouped.
  7. Audit ototoxic exposures — DIY solvents, garage CO, multiple ototoxic prescriptions. Ventilate, monitor, talk to a pharmacist.
  8. Treat metabolic health as ear health. Adults with diabetes carry roughly twice the odds of hearing impairment, worse long-term blood-sugar control tracks worse hearing thresholds, and the loss shows up earlier in life than most people expect. Whether tightening glucose control preserves hearing has not been shown in any trial — treat the ear as one more reason to manage the metabolic disease, not as a promised return on it.
  9. Sleep close to 8 hours. One cross-sectional survey put the best hearing thresholds near 8 hours, with worse at both ends — but only in men and older adults. That is thin evidence, and the mechanism usually cited for it, overnight antioxidant clearance in the cochlea, has never been demonstrated. Sleep well for the reasons that are established, and count your ears as a possible bonus.
  10. Diet pattern, not supplements. Mediterranean / MIND, with adequate fish for omega-3s, is what the observational data supports. No supplement has randomised evidence that it protects age-related hearing — both randomised ACEMg trials were null — a 51-patient study of residual hearing during cochlear-implant surgery that stopped early, and a small study of music-induced temporary threshold shift in normal-hearing adults.[67]
  11. If you've been fitted with hearing aids, commit to the rehabilitation period. Daily wear is the one variable that consistently predicts adaptation. A training app is optional: it reliably improves the task it drills and not much beyond it.

What's overhyped or wrong

  • "Hearing loss is just a normal part of aging." It is the largest modifiable midlife dementia risk factor, and in most cohorts adults with it die at modestly higher rates. Whether treating it bends either curve is unproven — the one randomised trial was null on its primary endpoint, and the mortality evidence is entirely observational. Treat it because untreated hearing loss costs you conversation, work and social life today; the longevity upside is a plausible bonus, not a promise.
  • "Hearing aids are vanity / for old people." The devices are now small, recharge in hours, and isolate speech in noise. What they do not yet have is proof of cognitive protection: the one randomised trial was null overall, and slowed decline only in its higher-risk stratum. Calling them vanity is still the most expensive piece of received wisdom in the field.
  • "A standard audiogram is sufficient screening." In the one cohort that has compared them head to head, extended high-frequency audiometry explained about a third of the variance in speech-in-noise performance against roughly a fifth for the standard PTA4 — better, but far from the whole story: two-thirds of the variance is explained by neither, and the evidence is a single non-peer-reviewed preprint. Anyone reporting noisy-restaurant trouble despite "normal" hearing can reasonably ask for EHF.
  • "Cochlear synaptopathy is the main cause of hidden hearing loss." Contested rather than resolved. The 2026 STOP preprint found that Wave I latency (nerve conduction speed) tracked speech-in-noise scores while Wave I amplitude (the usual synapse-count proxy) did not — but a 2025 cross-species study with histologically confirmed neural degeneration found that degeneration does contribute,[68] and a 2026 study found neither synaptopathy nor extended high-frequency loss independently associated with difficulty in noise.[69] Anyone stating this one confidently, in either direction, is ahead of the data.
  • "Stem cells / gene therapy will fix age-related hearing loss soon." The 2026 FDA approval is for one specific genetic deafness — OTOF-related loss, in ears with preserved outer hair cell function that have not already been implanted. Acquired age-related hearing loss is the much harder problem; meaningful adult therapeutics are years away. Don't wait.
  • "Loud workouts / earbuds at the gym are fine." Twenty minutes at 100 dB spends the entire WHO-recommended allowance at that level — and the allowance is weekly, not per session, so every loud gym visit draws on the same budget.
  • "Hearing supplements." Audifort, ACEMg and the rest rest on press releases, retrospective chart reviews and historical-control comparisons. Both randomised tests that exist were null: 51 patients given ACEMg or placebo around cochlear implantation, stopped early and never replicated (4.15 dB less hearing loss with ACEMg, 95% CI −12.95 to 4.65), and normal-hearing adults dosed before loud music, with no effect on temporary threshold shift.[70] No oral supplement has been shown in a randomised trial to preserve human hearing.

Further reading

  • Livingston G et al. Dementia prevention, intervention, and care: 2024 report of the Lancet Standing Commission. Lancet 2024.[71]
  • Lin FR et al. Hearing intervention versus health education control to reduce cognitive decline in older adults with hearing loss in the USA (ACHIEVE): a multicentre, randomised controlled trial. Lancet 2023.[72]
  • Cribb L et al. Treating hearing loss with hearing aids for the prevention of cognitive decline and dementia (ASPREE target-trial emulation). Neurology 2026.[73]
  • Jia J et al. Hearing loss and its association with all-cause and cause-specific mortality: a meta-analysis of cohort studies. PLoS One 2025 — 36 cohorts, 6.36 million participants; heterogeneity is extreme.[74]
  • Choi JS et al. Association between hearing aid use and mortality in adults with hearing loss in the USA: a mortality follow-up study of a cross-sectional cohort. Lancet Healthy Longev 2024 — a correction notice was published for this paper (Lancet Healthy Longev 2024;5:e96).[75]
  • Extended High-Frequency Hearing Loss, Not Cochlear Synaptopathy, Predicts Speech Recognition in a Population Cohort. medRxiv 2026.[76]
  • Saade M et al. Utility of Extended High-Frequency Audiograms in Clinical Practice. Laryngoscope 2024 — retrospective series of 40 symptomatic adults with normal standard audiograms; Level of Evidence 4.[77]
  • Bainbridge KE et al. Diabetes and hearing impairment in the United States: audiometric evidence from NHANES 1999–2004. Ann Intern Med 2008.[78]
  • Nisar M et al. Hearing Loss in Adults With Diabetes and Prediabetes: A Systematic Review and Meta-Analysis. Diabetes Metab Res Rev 2026 — 29 studies; supersedes Horikawa 2013.[79]
  • Li Z et al. Air pollution × occupational noise synergy on hearing — oil workers cohort. BMC Public Health 2025.[80]
  • Yang D et al. Effects of Composite Air Pollution and Genetic Susceptibility on Tinnitus Risk. J Otolaryngol Head Neck Surg 2026.[81]
  • Sleep duration and hearing loss — NHANES U-shaped relationship. Front Neurosci 2025.[82]
  • Cochlear circadian clock and noise modulation — review. Front Neurosci 2025.[83]
  • Lu W et al. Dietary nutrients & hearing loss — systematic review and meta-analysis. Front Nutr 2025.[84]
  • Du Q et al. Plant-based dietary diversity & hearing loss — CLHLS. Front Nutr 2025.[85]
  • Le Prell CG et al. Dietary supplement comprised of β-carotene, vitamin C, vitamin E, and magnesium: failure to prevent music-induced temporary threshold shift. Audiol Neurotol Extra 2016 — the second randomised ACEMg trial, null (added 2026-10-19).[86]
  • Seifer BS et al. Impact of the ACEMg Biomedicine on Sensorineural Hearing Loss and Auditory Function: Analysis of Real-World Clinical Data. Glob Adv Integr Med Health 2026 — retrospective, historical controls, product-linked authors.[87]
  • Ototoxic environmental pollutants — comprehensive review. Toxics 2024.[88]
  • CDC NIOSH science bulletin — chemicals and hearing loss. 2026.[89]
  • WHO — global standard for safe listening at venues and events. 2022.[90]
  • WHO-ITU — safe listening devices and systems (80 dB for adults, 75 dB for children, 40 hours per week). 2019.[91]
  • WHO — deafness and hearing loss: safe listening. The Q&A that carries the weekly exposure table. 2026.[92]
  • Dillard LK et al. Prevalence and global estimates of unsafe listening practices in adolescents and young adults: a systematic review and meta-analysis. BMJ Glob Health 2022 — the research behind WHO's "over 1 billion" figure.[93]
  • FDA — first-ever gene therapy approval for genetic hearing loss (Otarmeni). 2026.[94]
  • Valayannopoulos V et al. DB-OTO Gene Therapy for Inherited Deafness (CHORD). N Engl J Med 2026.[95]
  • Jiang L et al. Multicentre gene therapy for OTOF-related deafness followed up to 2.5 years. Nature 2026 — AAV1-hOTOF, 42 participants, 90% hearing recovery (distinct from Otarmeni).[96]
  • Tsai Do BS et al. Clinical Practice Guideline: Age-Related Hearing Loss. Otolaryngol Head Neck Surg 2024 — 11 key action statements, scoped to adults aged 50 and over.[97]
  • Levett BA et al. Hearing impairment and dementia: cause, catalyst or consequence? J Neurol 2025 — a correction was published in October 2025.[98]

— § —