Do Hearing Aids Cut Cognitive Decline by 50%? The ACHIEVE Trial in Full Context

The inventory of a famous headline: 977 volunteers; ages 70 to 84; untreated mild-to-moderate hearing loss; no substantial cognitive impairment. One intervention kit—hearing aids, fitting, counseling, related support. One control kit—health-education sessions. And one number that escaped the paper and went everywhere: cognitive decline cut by almost 50 percent.

Itemized more carefully, the trial contains something else: a primary analysis of the full population that was null; a prespecified higher-risk subgroup where the intervention appeared to slow decline; a later secondary analysis finding an even larger effect among those at highest predicted risk; and—separately, robustly, across the whole trial—a large and durable improvement in hearing-related communication.

The question matters because hearing loss is common, treatable, and entangled with isolation. The most defensible current answer: the 50-percent headline belongs to a subgroup, not to every older adult with hearing loss, while the communication benefit reached across participants. Both facts are true. Neither cancels the other.

What ACHIEVE randomized

ACHIEVE enrolled 977 adults aged 70 to 84 with untreated mild-to-moderate hearing loss and without substantial cognitive impairment. Participants received either a best-practice hearing intervention - hearing aids, fitting, counseling, and related support - or a health-education program. Investigators followed cognition for three years.

Participants came from two populations. Some were already part of the long-running Atherosclerosis Risk in Communities, or ARIC, cohort and had more cardiovascular and cognitive risk factors. Others were newly recruited community volunteers who were generally healthier. That difference became central to the result.

The primary result in all 977 participants was null

Across the complete trial population, three-year change in global cognition did not differ between hearing intervention and health education. The estimated difference was 0.002 standard-deviation units, with a 95% confidence interval from -0.077 to 0.081 and a p value of 0.96. In plain language, the trial did not show that hearing intervention slowed cognitive decline for the full group.

This is the prespecified primary analysis, and it belongs in any accurate summary. Reporting only the favorable subgroup converts a heterogeneous trial into a universal claim.

Where the 48% figure came from

The trial also prespecified an analysis by recruitment source. In the higher-risk ARIC subgroup, hearing intervention was associated with 48% less cognitive decline over three years relative to health education. In the healthier de novo group, there was no cognitive benefit. The interaction between cohort and intervention was statistically significant.

A prespecified subgroup with a significant interaction is more credible than an arbitrary post hoc slice. It can identify a population in which an intervention may be especially useful. But subgroup estimates are less precise, involve fewer participants, and need replication. The correct conclusion is that the intervention may slow decline among older adults at higher baseline risk, not that hearing aids reduce decline by 48% in everyone.

A later risk-model analysis sharpened the same pattern

In 2025, ACHIEVE investigators published a secondary analysis that predicted each participant's baseline risk of cognitive decline. Among the highest-risk quarter, hearing intervention was associated with 61.6% slower three-year decline, with a 95% confidence interval from 33.7% to 94.1%. No cognitive benefit was found in lower-risk quartiles.

This strengthens the hypothesis that baseline risk modifies the effect. It does not turn a secondary analysis into a new primary trial. Prediction models can discover clinically useful heterogeneity, but the threshold and effect size should be tested prospectively in another randomized population.

Observational evidence is supportive but cannot settle causality

Large cohort studies often find that hearing-aid users have lower rates of cognitive decline or dementia than nonusers with hearing loss. A systematic review and meta-analysis also found associations favoring hearing restoration. These data are consistent with a protective effect, but users and nonusers differ in healthcare access, education, income, social engagement, disease burden, and willingness to seek treatment.

A UK Biobank analysis designed to emulate a hypothetical intervention illustrates the problem. The apparent direction and size of the hearing-aid association changed substantially after accounting for healthcare use and the timing of hearing-loss diagnosis. The authors concluded that residual confounding remained a serious concern. Randomization is valuable precisely because statistical adjustment cannot guarantee comparable groups.

Possible mechanisms remain hypotheses

Hearing loss could influence cognition through greater cognitive load during listening, reduced social engagement, shared vascular or neurodegenerative causes, changes in brain structure, or combinations of these pathways. Hearing treatment might reduce listening effort and help people remain connected. Alternatively, hearing loss may partly mark underlying processes that an aid cannot reverse.

These mechanisms are plausible, not mutually exclusive, and not proven by association alone. ACHIEVE suggests that treating hearing loss may matter most when cognitive risk is already elevated, but it does not identify a single causal pathway.

The clearest randomized benefit was communication

Focusing on the cognitive nuance should not obscure what hearing care demonstrably achieved. In a prespecified secondary analysis, self-reported communicative function improved within six months and remained better through three years. At year three, the between-group difference in change on the Hearing Handicap Inventory for the Elderly screening scale was 9.5 points in favor of hearing intervention.

A separate analysis did not find an overall improvement in broad physical or mental health-related quality-of-life scores. Again, outcomes must remain specific. Hearing aids improved hearing-related communication; they were not a universal quality-of-life intervention and have not been established as a dementia-prevention drug.

The useful conclusion is more personal than the viral one

Untreated hearing loss deserves evaluation because hearing and communication are important now, independent of any future cognitive effect. For an older adult at elevated risk of cognitive decline, ACHIEVE provides a credible reason to study hearing intervention as one component of risk reduction. It does not justify promising a 48% benefit to a healthy low-risk person or using hearing aids as a substitute for broader medical, sensory, cardiovascular, and social care.

The accurate headline is: a hearing intervention did not slow cognitive decline across the full ACHIEVE trial, but it did in a prespecified higher-risk subgroup, while communication improved robustly across participants. That version is less absolute and more useful. Educational, not medical advice.