Hearing Loss and Dementia: One of the Largest Modifiable Risk Factors

Hearing loss is one of the largest single modifiable risk factors for dementia — but not for the reason most headlines suggest. In the 2024 Lancet Commission’s accounting, midlife hearing loss carries a population attributable fraction of 7%, tied with high LDL cholesterol as the biggest of 14 modifiable factors. Yet the one large randomised trial of hearing aids found no cognitive benefit in its overall study population. It did find a substantial effect in a prespecified higher-risk subgroup. The gap between that population statistic and that trial result is the entire story, and popular coverage almost always gets it wrong.

What the Lancet Commission Actually Says

The 2024 report of the Lancet standing Commission on dementia prevention, intervention, and care (Livingston et al., Lancet 2024;404:572–628) is the current edition and the reference document for this field. It expanded the list from 12 modifiable risk factors to 14, adding high LDL cholesterol and untreated vision loss, and concluded that 45% of dementia cases worldwide are potentially preventable or could be delayed if all 14 were eliminated.

In the Commission’s published breakdown of individual population attributable fractions, hearing loss sits at 7% — level with high LDL cholesterol and ahead of everything else. For comparison, the rest of the list runs: less education 5%, social isolation 5%, depression 3%, traumatic brain injury 3%, air pollution 3%, physical inactivity 2%, diabetes 2%, smoking 2%, hypertension 2%, visual loss 2%, obesity 1%, and excessive alcohol 1%.

Two things about that 7% deserve emphasis.

• A population attributable fraction is a modelled ceiling, not a measured benefit. It answers a hypothetical: if this risk factor vanished from the population entirely, and if the observed association is fully causal, how much dementia would disappear? Both conditions are assumptions.
• Hearing loss scores highly partly because it is so common. A modest relative risk applied to a very prevalent exposure produces a large attributable fraction. The World Health Organization projects that nearly 2.5 billion people will have some degree of hearing loss by 2050, with over 700 million needing rehabilitation. Prevalence does a lot of the work in that 7%.

What the One Big Trial Found — and What It Did Not

ACHIEVE is the only large randomised test of whether treating hearing loss slows cognitive decline — and the study on this topic most often misreported.

The design

ACHIEVE (Lin et al., Lancet 2023;402:786–797) was a multicentre, parallel-group, unmasked randomised controlled trial across four US community sites. It enrolled 977 adults aged 70–84 with untreated hearing loss and no substantial cognitive impairment, randomising them 1:1 to a hearing intervention (audiological counselling plus hearing aids) or an active control of individual health-education sessions on chronic disease prevention. The primary endpoint was 3-year change in a global cognition standardised factor score, analysed by intention to treat.

Critically, participants came from two very different populations: 238 (24%) from the long-running Atherosclerosis Risk in Communities (ARIC) observational cohort, and 739 (76%) healthy de novo community volunteers. The ARIC participants were older, had more risk factors for cognitive decline, and lower baseline cognitive scores.

The primary result was null

In the primary analysis combining both cohorts, 3-year cognitive change was not significantly different between groups: −0.200 SD (95% CI −0.256 to −0.144) in the hearing intervention arm versus −0.202 SD (95% CI −0.258 to −0.145) in the control arm. The between-group difference was 0.002 SD (95% CI −0.077 to 0.081), p=0.96. That is about as flat a result as a trial can produce.

Say it plainly: hearing aids did not slow cognitive decline in the overall trial population. Any headline claiming ACHIEVE showed hearing aids prevent dementia is wrong twice over — the primary result was null, and the trial measured cognitive change over three years, not dementia diagnoses.

The higher-risk subgroup told a different story

A prespecified sensitivity analysis found the intervention’s effect differed significantly between the two source populations (p for interaction = 0.010). In the ARIC cohort (n=238), the hearing intervention was associated with a 48% reduction in 3-year cognitive change relative to control — a difference of 0.191 SD (95% CI 0.022 to 0.360), p=0.027. In the healthier de novo cohort (n=739), there was no effect: −0.061 SD (95% CI −0.151 to 0.028), p=0.18.

The trial’s own interpretation is appropriately hedged: hearing intervention might reduce cognitive change over three years in populations at increased risk, but not in those at decreased risk. That is a hypothesis worth testing next, not a settled finding.

ACHIEVE was funded by the US National Institutes of Health. Its authors disclose extensive industry ties: the senior author reports consulting fees from Frequency Therapeutics and Apple and directs a research centre part-funded by a donation from Cochlear, and Sonova/Phonak donated the hearing technologies used in the trial. Those ties do not invalidate a null result, but they are worth knowing.

Where the 7% Figure Comes From: The Observational Evidence

The Commission’s estimate rests on cohort studies, not trials. The most-cited is Lin and colleagues in Archives of Neurology (2011;68:214–220): a prospective study of 639 dementia-free adults in the Baltimore Longitudinal Study of Aging who underwent audiometric testing, followed for a median 11.9 years, with 58 incident dementia cases.

Adjusted for age, sex, race, education, diabetes, smoking, and hypertension, risk rose log-linearly with baseline hearing loss severity — a hazard ratio of 1.27 per 10 dB of loss (95% CI 1.06 to 1.50). Against normal hearing, the hazard ratios for all-cause dementia were 1.89 (95% CI 1.00 to 3.58) for mild loss, 3.00 (95% CI 1.43 to 6.30) for moderate loss, and 4.94 (95% CI 1.09 to 22.40) for severe loss.

Notice how wide those confidence intervals get; the severe-loss estimate rests on just six people. And the authors closed by asking whether hearing loss is a modifiable risk factor or simply an early marker of dementia already underway — a question their design could not answer, and one still unresolved.

Three Plausible Mechanisms

Mechanistic plausibility is part of why this association is taken seriously rather than dismissed as confounding. Three explanations are usually proposed. None is established, and they are not mutually exclusive.

Cognitive load

When the signal arriving at the ear is degraded, the brain works harder to reconstruct it. Resources that would otherwise go to encoding what was said and holding it in working memory get spent decoding the acoustics instead. Over years, the theory goes, that chronic reallocation of effort has a cost.

Reduced social engagement

People who struggle to follow conversation in restaurants, at family gatherings and on the phone tend to do less of it. That matters because social isolation is itself on the Commission’s list, carrying its own 5% attributable fraction. Hearing loss may therefore act partly through a second, already-recognised risk factor rather than a direct neural pathway.

Brain structural change

In the neuroimaging substudy of the Baltimore Longitudinal Study of Aging, Lin and colleagues in NeuroImage (2014;90:84–92) tracked 126 adults aged 56–86 for a mean of 6.4 years after baseline MRI. The 51 participants with hearing impairment showed accelerated volume decline in the whole brain and in right temporal lobe regions — superior, middle and inferior temporal gyri and parahippocampus — compared with the 75 with normal hearing (p<0.05), robust to adjustment for demographic and cardiovascular factors.

That is a small sample, and atrophy in auditory-adjacent regions could reflect reduced input rather than a dementia process. Suggestive, not decisive.

Limitations Worth Sitting With

On this topic the honest summary is more useful than the confident one.

• The primary trial result was null. ACHIEVE is the highest-quality causal evidence available, and in its total cohort the answer was no effect. Everything more optimistic comes from subgroups or from observational data.
• Subgroup findings are fragile by nature. The ARIC result was prespecified rather than fished for, which counts in its favour, and the interaction test was significant. But it rests on 238 people, its confidence interval (0.022 to 0.360) nearly touches zero, and subgroup effects that look convincing in one trial frequently fail to replicate.
• Reverse causation is a live possibility. Early neurodegeneration can impair the central processing of sound years before diagnosis. Some of the association between poor hearing and later dementia may be dementia causing measured hearing difficulty rather than the reverse.
• Residual confounding is unresolved. Cohort studies adjust only for what they measured, and cardiovascular health, noise-exposed occupations, socioeconomic position and healthcare access all track with both hearing and cognition.
• ACHIEVE was unmasked. Participants knew which arm they were in — you cannot blind someone to a hearing aid. That opens the door to expectation effects in both directions.
• The literature has already produced one high-profile error. A widely covered 2023 UK Biobank analysis of 437,704 people reporting that hearing aid use was associated with lower dementia risk was retracted by The Lancet Public Health in January 2024 after two prior corrections. Its headline figures are still circulating. If you see them cited, that is why they should be ignored.

So What Should You Actually Do?

What makes this topic unusual among longevity interventions: the case for acting does not depend on the dementia question being settled. Treating hearing loss has its own well-established payoff.

The Cochrane review of hearing aids for mild to moderate hearing loss in adults (Ferguson et al., 2017, CD012023) pooled five RCTs with 825 participants and found a large benefit on hearing-specific quality of life — a mean difference of −26.47 on the Hearing Handicap Inventory for the Elderly (95% CI −42.16 to −10.77, three studies, 722 participants, moderate-quality evidence) — plus improvements in general health-related quality of life (SMD −0.38, 95% CI −0.55 to −0.21) and listening ability (SMD −1.88, 95% CI −3.24 to −0.52). Hearing aids reliably do the thing they are actually for.

1. Get an actual audiogram, not a self-assessment. Hearing loss is gradual enough that people adapt without noticing. The Commission places hearing loss in the midlife window, so a baseline test in your forties or fifties is reasonable rather than premature.
2. Treat identified loss rather than deferring it. Typical delays between noticing difficulty and seeking help run to several years. Whatever the dementia evidence eventually shows, those are years of degraded communication for no benefit.
3. Protect the hearing you have. Noise-induced loss is largely preventable and largely permanent. WHO estimates over a billion young adults are at risk from unsafe listening practices. Earplugs at concerts and on power tools, and lower headphone volumes, are the cheapest intervention on this entire page.
4. Do not treat a hearing aid as a dementia drug. Expecting cognitive protection you were never promised is how people end up disappointed with a device that is working exactly as intended.
5. Address the other 13 factors too. Blood pressure, LDL cholesterol, physical activity, smoking and social connection sit on the same list, and several have far stronger interventional evidence behind them.

The Bottom Line

Hearing loss is a genuinely underrated longevity topic. It is common, it is measurable, it is treatable, and a major commission ranks it at the top of the modifiable dementia risk list. What it is not — yet — is a proven causal lever. The observational associations are consistent and dose-dependent but cannot rule out reverse causation or residual confounding, and the single large randomised trial designed to settle the question came back flat in its primary analysis.

The defensible position is the boring one. Get your hearing checked, treat loss when it is found, and protect your ears from noise — those things demonstrably improve how you hear and how you live. If the ARIC subgroup signal proves real, you will have been doing the right thing anyway.

This is educational content summarising published research, not medical advice. Hearing changes and cognitive concerns both warrant a proper assessment — see an audiologist or your doctor rather than self-diagnosing from an article.

---

Want more science-backed breakdowns like this one? Subscribe to The Wonder Drop on YouTube for new videos every week.