In older adults who already have mild cognitive impairment, being fitted with hearing aids did not significantly reduce the chance of progressing to dementia-level impairment over two years. That is the primary result of a randomised trial published on 28 September in JAMA Neurology — and it is the second large trial of hearing aids for cognition to come back null on its main endpoint.
The trial, known as CHOICE, was run in Shanghai at three hospital otology clinics and three community health stations. Investigators screened 21,908 people aged 60 and over to find 703 who met two criteria at once: mild cognitive impairment (MCI), and moderate-to-severe hearing loss untreated with hearing aids in the previous year. Participants were assigned at random to either free behind-the-ear hearing aids fitted by audiologists, with training and adjustments at each visit, or a hearing care and education programme. Everyone was assessed at baseline and at 6, 12 and 24 months. Of the 703, 610 (86.8%) completed the final visit.
What the main result was, and what a null result means here
Progression to dementia-level impairment — a Clinical Dementia Rating global score of 1 or more, up from 0.5 at entry — occurred in 3.09% of the hearing-aid group and 4.74% of the education group. The risk difference was −2.07 percentage points, with a 95% confidence interval running from −4.74 to 0.61. The relative risk was 0.59 (95% CI 0.25 to 1.38; p = .23).
That interval is wide enough to contain both a substantial benefit and a small increase in risk, and the authors do not pretend otherwise. Among their listed limitations: fewer participants progressed to dementia-level impairment than the sample-size calculation had assumed, “resulting in underpowered analyses for the primary outcome.” In other words, the trial could not settle the question it was built around.
The number that will travel further than the main result
One prespecified secondary outcome did separate the two groups, and it is the figure most likely to circulate. The share of participants whose CDR global score moved back to 0 — scored as cognitively unimpaired at 24 months — was 15.34% in the hearing-aid group versus 2.36% in the education group (risk difference 12.00 percentage points, 95% CI 8.79 to 15.21; relative risk 5.94, 95% CI 2.46 to 14.37).
The trial’s own discussion is a good deal more restrained than that ratio suggests. Prespecified or not, the secondary outcomes were treated as exploratory: “this exploratory finding warrants cautious interpretation and further investigation,” the authors write. There are concrete reasons for the caution. Returning from a CDR of 0.5 to 0 is reversion, and reversion is common in MCI — the paper notes rates reported across studies ranging from 2% to 59% over follow-up periods of one to 17 years. The same instrument was used to define who entered the trial and to measure what happened afterwards, which the authors flag as a possible source of evaluation bias. And the CDR 0.5 entry criterion itself was not prespecified in the trial registration, an omission the authors describe as potentially undermining transparency about the workflow.
The devices did work as devices
None of this disputes that hearing aids did their own job. Mean pure-tone average in the intervention group went from 52.65 dB HL at baseline to 32.18 dB HL at 24 months with aids in place. Average daily wearing time was 5.9 hours — lower than the 7.2 to 8.1 hours reported in the earlier American trial, which the authors attribute partly to memory decline affecting consistent wear. Five participants in the education group obtained hearing aids independently, and 20 non-study-related serious adverse events were reported, ten in each group, with no study-related adverse events.
This is not the first null primary endpoint
The pattern is consistent with ACHIEVE, the American trial published in The Lancet in 2023. It randomised 977 adults aged 70 to 84 who had untreated hearing loss but no substantial cognitive impairment. Over three years, global cognition changed by essentially the same amount in both arms: a between-group difference of 0.002 SD (95% CI −0.077 to 0.081; p = 0.96).
A prespecified sensitivity analysis did detect an interaction between the trial’s two recruitment cohorts (p = 0.010). In the older, higher-risk group drawn from an existing heart-health cohort — 238 people — hearing intervention was associated with a 48% reduction in three-year cognitive decline. In the 739 healthier volunteers recruited fresh, there was no effect. That subgroup finding, rather than ACHIEVE’s headline, is where hopes for a cognitive benefit have rested; CHOICE, testing a population that already had MCI, did not reinforce it on its own primary measure.
Why hearing loss stays on the prevention list anyway
None of this removes hearing loss from the dementia-prevention conversation. A 2024 systematic review and meta-analysis in The Lancet Healthy Longevity collected reported population attributable fractions for 61 modifiable risk factors and meta-analysed 12 of them across 48 studies. Weighted for overlap between risk factors, hearing loss came in at 7.2% (95% CI 5.2 to 9.7) — behind low education (9.3%) and physical inactivity (7.3%), just ahead of hypertension (7.1%). The authors also note that heterogeneity for hearing loss was total (I² = 100%) and that publication bias was significant for the unweighted estimates of most factors, hearing loss included. Reported fractions were higher in low-income and middle-income countries than in high-income ones for both education and hearing loss.
What the two trials have shifted is narrower than “hearing aids protect the brain.” Treating hearing loss has not yet been shown to slow cognitive decline over two or three years in the populations tested — one group of cognitively healthy older adults, and now one group of East Asian adults with MCI. The authors of the CHOICE paper argue that the wider case for treating hearing loss, and for widening access to hearing care, does not depend on the cognitive result.
Where that leaves the underlying problem
The scale is not in dispute. The World Health Organization estimates that more than 5% of the world’s population — about 430 million people — need rehabilitation for disabling hearing loss, with nearly 2.5 billion people projected to have some degree of hearing loss by 2050. Among people over 60, more than 25% are affected by disabling hearing loss. WHO lists “increased risk of developing dementia and accelerated cognitive decline when hearing loss remains unaddressed” among the consequences of unaddressed hearing loss.
In the United States, the National Institute on Deafness and Other Communication Disorders puts the figure at roughly 15% of adults aged 18 and over — about 37.5 million people — reporting some trouble hearing, with about one in three people aged 65 to 74 and nearly half of those over 75 affected. The same agency is blunt that scientists do not yet know how to prevent age-related hearing loss, though the noise-induced portion is preventable.
For anyone who suspects a problem, the practical routes are settled. The FDA established a category of over-the-counter hearing aids for adults in 2022, available without a hearing test or a clinician visit for mild to moderate loss; more complex or more severe loss is a matter for an audiologist or an ear, nose and throat specialist, as is tried-and-failed use of an over-the-counter device. The NIDCD also publishes a short screening questionnaire: if you answer “yes” to two or more of its items, or “sometimes” to three or more, it suggests considering a hearing check.
Cognitive change is not something to self-assess from a news article, in either direction. This piece reports what a trial and federal and international agencies published; it is not medical advice, and questions about your own hearing or memory belong with a healthcare professional. For how we handle that boundary, see our medical disclaimer.
Sitting alongside this is the question of what the social and cognitive side of later life looks like — the territory our recent piece on social connection and capability in older age mapped, and the same broad family of concerns behind our archive look at conditions that tend to go unnoticed.