Hearing Loss and Dementia Risk: What the Research Actually Shows

Hearing loss is consistently associated with a higher risk of dementia. It has never been shown to cause it. The only randomized trial built to test whether treating hearing loss slows cognitive decline found no benefit across its full study population over three years, and a benefit only inside one higher-risk subgroup.

That distinction is the whole article. The gap between "associated with," "causes," and "treating it prevents" is where most of the marketing for over-the-counter hearing aids now lives. We read the trial papers and the Lancet Commission report, not the press releases. Here is how strong the case actually is.

The short answer

• The association is large and replicated. In the best-known cohort, moderate hearing loss carried roughly three times the dementia risk of normal hearing.
• The causal evidence is thin. Three competing explanations fit the same data, and only one says hearing loss harms the brain.
• The one randomized trial was null overall. Across 977 older adults followed three years, treating hearing loss made no measurable difference to cognitive decline.
• It was positive in one prespecified subgroup. People already at elevated risk declined roughly 48% more slowly. That is a hypothesis, not a proof.
• "Largest modifiable risk factor" is a population statistic. It is not your personal risk reduction.
• Most people with hearing loss never develop dementia. Hearing loss is common and dementia has many causes. This is not a sentence.

The observational link between hearing loss and cognitive decline

The most-cited numbers come from a single prospective cohort. Lin and colleagues, publishing in Archives of Neurology in 2011, followed 639 dementia-free adults in the Baltimore Longitudinal Study of Aging who had audiometric testing between 1990 and 1994. Over a median 11.9 years, 58 developed dementia. Compared with normal hearing, the adjusted hazard ratio for all-cause dementia was 1.89 for mild hearing loss, 3.00 for moderate and 4.94 for severe. Those are the familiar "two, three and five times" figures, usually quoted without qualifiers.

• They are associations, not effects. The models adjusted for age, sex, race, education, diabetes, smoking and hypertension. Adjustment reduces confounding; it cannot remove what was never measured.
• The severe-loss figure rests on six people. Only six participants had severe hearing loss, which is why that confidence interval ran from 1.09 to 22.40. A hazard ratio of 4.94 carrying that much uncertainty is a direction, not a measurement.
• Most participants never developed dementia. Fifty-eight cases among 639 people over roughly twelve years. The National Institute on Deafness and Other Communication Disorders puts hearing loss at about one in three US adults aged 65 to 74, and difficulty hearing at nearly half of those 75 and older. The great majority never develop dementia.

The one randomized trial, and what it actually found

Everything above is observational. Only one randomized trial has tested what people actually want to know: if you treat the hearing loss, does cognition hold up better?

That trial is ACHIEVE, published in The Lancet in July 2023. It enrolled 977 adults aged 70 to 84 with untreated hearing loss and no substantial cognitive impairment at four US sites, assigning 490 to a best-practice hearing intervention of audiological care plus hearing aids and 487 to a health education control, and followed them three years.

The primary result was null. Global cognition declined 0.200 standard deviations in the hearing group and 0.202 in the control group. The between-group difference was 0.002 SD, 95% CI -0.077 to 0.081, p equal to 0.96. That is as close to nothing as a trial result lands.

Now the finding that made the headlines, in its proper order. ACHIEVE recruited two populations. 238 participants, 24.4%, came from the Atherosclerosis Risk in Communities study: older, with more cardiovascular risk factors, lower baseline cognition and faster decline during the trial. The other 739 were healthier de novo volunteers. A prespecified sensitivity analysis found the intervention's effect differed significantly between the two cohorts, with a p value for interaction of 0.010. Within the ARIC group, the intervention produced a 0.191 SD difference in three-year cognitive change, 95% CI 0.022 to 0.360, p equal to 0.027, reported as roughly a 48% reduction in the rate of decline. Within the de novo group, nothing: -0.061 SD, 95% CI -0.151 to 0.028, p equal to 0.18.

A 2025 secondary analysis in Alzheimer's and Dementia ranked participants by a risk model built in 2,692 ARIC volunteers. Among the trial's top quartile of predicted risk, three-year decline was 61.6% slower with the hearing intervention, 95% CI 33.7% to 94.1%.

Read that pattern carefully. A subgroup finding inside a null trial is a hypothesis. This one is prespecified, plausible, and visible under two risk definitions. It is also unmasked by design, driven by 238 people, and the sort of result that fails to replicate often enough that trialists call it hypothesis-generating. The investigators said longer follow-up may be needed to detect any effect in the lower-risk group, and that follow-up is ongoing. What the trial supports today is narrow: hearing intervention may slow cognitive change in people already at increased risk, and does nothing measurable over three years for those who are not.

The "largest modifiable dementia risk factor" claim, explained

The phrase traces to the standing Lancet Commission on dementia prevention, intervention and care, whose 2024 report in The Lancet lists 14 modifiable risk factors and estimates that 45% of dementia cases worldwide could in theory be prevented or delayed if all 14 were eliminated. Hearing loss carries a population attributable fraction of 7%, tied with high midlife LDL cholesterol as the largest single share. Newly added vision loss gets 2%.

A population attributable fraction is a modelling number. It estimates the share of cases in an entire population that would not occur if that risk factor were removed from it, assuming the association is genuinely causal and the factor can actually be eliminated. Both assumptions do heavy lifting.

What a PAF is not is your personal risk reduction. Audiologists reviewing the updated figures in Canadian Audiologist in 2024 said it directly: the PAF cannot be used to estimate individual risk. So "hearing aids cut your dementia risk by 7%" is not a cautious restatement of the Commission's finding. It is a different claim, it is unsupported, and it is the most common error in this category's advertising.

Three explanations for the link, only one of which is "hearing loss damages the brain"

Reviews of the mechanistic literature, including a 2026 review in Frontiers in Dementia, group the candidate pathways as follows.

• Cognitive load and sensory deprivation. Degraded sound takes more mental effort to decode, diverting resources from memory and other tasks, and reduced auditory input may drive structural change over time. Researchers call this the "outside-in" account.
• The social isolation pathway. Hearing loss is associated with withdrawal from conversation, less social engagement and depression, each of which independently raises dementia risk. The Lancet Commission counts social isolation and depression as separate risk factors in its own model, so part of hearing loss's apparent contribution may simply be travelling through them.
• Reverse causation and shared cause. Early neurodegeneration can affect central auditory processing, making hearing difficulty an early symptom rather than a cause. Shared causes are equally plausible: the same reviews point to vascular disease and genetic factors such as APOE variants that could damage ear and brain alike.

A fourth problem is not a mechanism at all. It is measurement. Most cognitive tests are verbal and given out loud, so someone who cannot hear the tester may score worse without being more impaired. A 2023 study in the Journal of Alzheimer's Disease simulated hearing loss in normal-hearing listeners, found immediate verbal recall declined, and concluded that auditory factors confound cognitive assessment and lead to underestimation of cognitive function. Reviews note the counterweight: a 2026 review in Audiology Research points out that the association also turns up on predominantly non-auditory measures, including verbal fluency tasks with minimal perceptual load, so test bias alone is unlikely to explain it.

Only the first explanation implies that treating hearing loss should protect the brain. The rest fit the observational data equally well, which is why the randomized trial matters more than any cohort study.

The hearing-aid studies are hard to trust, for a specific reason

People who seek out, buy and keep wearing hearing aids are not a random sample. They tend to have more income, more education, more engagement with health care and fewer competing illnesses, and every one of those traits independently predicts lower dementia risk. So when a study finds hearing aid users declining more slowly, some unknown share of that is the hearing aid and some unknown share is the kind of person who owns one.

That caps what the literature can prove. A systematic review and meta-analysis by Yeo and colleagues in JAMA Neurology in 2023 included 31 studies covering 137,484 participants, 19 of which entered the quantitative analyses. Across eight studies and 126,903 people, users of hearing aids or cochlear implants had a 19% lower hazard of long-term cognitive decline than people with uncorrected hearing loss, a hazard ratio of 0.81, 95% CI 0.76 to 0.87. Short-term cognitive test scores improved about 3% across 11 studies of 568 participants. The authors flagged the constraint themselves: the evidence was mostly observational, and residual confounding is likely.

The field has also been burned in public. In April 2023 The Lancet Public Health published a UK Biobank analysis reporting that hearing aid users carried dementia risk similar to people without hearing loss. The journal retracted it that December: an error in the output settings of the authors' statistical code had switched the hearing-aid-user and non-user groups, rendering the conclusions false and misleading. It surfaced only because outside scientists tried to reproduce the result and could not.

A more basic question comes first: whether a device improves your hearing in the situations that actually bother you. We looked at whether OTC hearing aids actually work separately. The dementia question is downstream of that one.

A sensible plan if you have hearing loss

• Find out what your hearing actually is. Start by learning how to test your hearing at home, so you are working from more than a guess. A home screening is a starting point, not a diagnosis.
• See a clinician for red flags first. Hearing loss that comes on suddenly, affects one ear only, or arrives with pain, drainage, dizziness or new ringing should be assessed promptly rather than self-treated.
• Treat it for the benefits with the most direct evidence. ACHIEVE's secondary analyses are the useful part. In JAMA Internal Medicine in 2025, the hearing group retained about one more person in their social network over three years, a difference of 1.05, 95% CI 0.01 to 2.09, plus small improvements in loneliness the authors cautioned may not be clinically meaningful. A Lancet Public Health analysis in 2025 reported 27% fewer falls, a mean difference of -0.54 falls, 95% CI -0.77 to -0.31, which its authors framed as exploratory. In JAMA Network Open in 2024, the same trial found no effect on health-related quality of life. These researchers publish their nulls.
• Treat the cognitive angle as a possible bonus, not the reason. If you match the higher-risk group ACHIEVE identified, the case is stronger. It is still not a promise, and no device has been shown to prevent dementia.
• Keep hearing in proportion. It is one of 14 factors on the Commission's list, alongside blood pressure, LDL cholesterol, activity, smoking and social contact. Readers ask about supplements for brain health in the same breath; that is a separate and generally weaker body of evidence.
• Take memory concerns to a doctor. If you or your family have noticed changes in memory, that belongs in a conversation with a clinician.

Our take

The honest formulation is simple, and nobody selling hardware will give it to you. Treating hearing loss has clear, proven benefits for communication, measured if modest gains in social connection, plus credible but exploratory evidence on falls. There is a plausible but unproven case that it also protects cognition, strongest for people already at elevated risk.

That is a good enough reason to treat hearing loss. It is not a reason to panic about dementia. If the honest version disappoints next to the advertising, that is the point: the advertising is ahead of the evidence.

If you have decided to act, the hardware questions come next. Start with our complete OTC hearing aid guide.

Does hearing loss cause dementia?

No study has shown that it does. Hearing loss is consistently associated with higher dementia risk in observational cohorts, but reverse causation and shared causes such as vascular disease remain live explanations, and the only randomized trial of treating hearing loss found no effect on cognitive decline in its full study population.

Do hearing aids reduce dementia risk?

Not proven. The ACHIEVE trial found no difference in three-year cognitive decline overall, a between-group difference of 0.002 SD with a p value of 0.96, though a prespecified analysis showed roughly a 48% slower decline in the higher-risk ARIC subgroup. Observational studies suggest benefit but are confounded by who buys and wears hearing aids.

How much does hearing loss increase dementia risk?

In the Baltimore Longitudinal Study of Aging, published in Archives of Neurology in 2011, adjusted hazard ratios for incident dementia were 1.89 for mild, 3.00 for moderate and 4.94 for severe hearing loss compared with normal hearing. Those are observational associations rather than measured effects, and the severe-loss estimate rested on only six participants.

Should I get my hearing checked if I am worried about my memory?

Yes, and raise both with your doctor. Because most cognitive tests are spoken aloud, untreated hearing loss can make test performance look worse than a person's actual cognition, so having hearing assessed makes any memory evaluation more accurate.