Ryan NguyenCan treating hearing loss reduce the chance that an older adult with mild cognitive impairment progresses to dementia-level impairment? The CHOICE trial tested that question directly. Its primary result was not statistically significant.
There was also a favorable secondary finding for cognitive improvement. Those results can coexist. Reporting them accurately requires keeping the trial’s main question separate from the other outcomes it measured.
For people following brain-aging research, this is a useful example of why a study cannot be reduced to “worked” or “failed.” It can narrow one claim while raising another question worth investigating.
What was the CHOICE hearing-aid trial?
Published September 28 in JAMA Neurology, CHOICE randomized 703 adults in Shanghai with moderate to severe hearing loss and mild cognitive impairment to hearing-aid intervention or hearing-care education. Participants were followed for 24 months. The main endpoint was conversion to dementia-level impairment, defined using the Clinical Dementia Rating, or CDR. [1]
Conversion rates were 3.09% with intervention and 4.74% with education; the difference was not statistically significant (P=.23). A prespecified secondary outcome favored intervention: cognitive improvement to CDR 0 occurred in 15.34% versus 2.36%. [1]
The primary outcome therefore does not establish dementia prevention in this population. The secondary result supports further investigation, while requiring its own interpretation.
Why the primary and secondary outcomes matter
A trial’s primary endpoint defines the central question it is designed to answer. Specifying it in advance reduces the temptation to choose whichever result looks best after the data arrive.
Secondary outcomes can still be informative, especially when prespecified. They may capture a different aspect of a response, help explain the main outcome, or identify a question for another trial. They should not silently replace the original endpoint in the headline.
In this case, avoiding progression to a threshold and improving on a rating scale are different outcomes. A participant can improve without the study demonstrating a statistically reliable reduction in progression across the full group.
The appropriate article should make both results visible early. Readers should not have to reach the limitations section to discover that an encouraging outcome was secondary or that the main question remained unresolved.
Does a nonsignificant result prove hearing aids have no effect?
No. Statistical nonsignificance means the study did not establish the effect under its analysis; it does not demonstrate that the true effect is exactly zero.
The distinction is especially important when relatively few participants experience the primary event. A study can have hundreds of participants yet still provide limited precision for an outcome that occurs infrequently. The relevant information includes the number of events and the uncertainty around the estimate, rather than sample size alone.
Conversely, a numerical difference should not be advertised as an established benefit simply because it points in a favorable direction. The uncertainty is part of the finding.
For a reader, the most defensible conclusion is limited: CHOICE did not demonstrate a statistically significant reduction in dementia-level progression over two years among the participants it enrolled. That leaves room for further research without turning uncertainty into proof of benefit.
How CHOICE compares with the ACHIEVE trial
The earlier ACHIEVE trial studied a different population and endpoint. It enrolled older adults without substantial cognitive impairment and assessed cognitive change over three years. The overall primary analysis did not show a significant intervention effect, while a prespecified analysis in the higher-risk ARIC subgroup found approximately 48% less cognitive decline. [2]
That subgroup result is often repeated without its context. It should not be generalized to every person with hearing loss or treated as equivalent to preventing a dementia diagnosis.
Comparing the trials requires attention to baseline cognition, background risk, follow-up time, and the outcome measured. One study’s cognitive-change result cannot answer another study’s progression question automatically.
The comparison is still useful. It suggests that future research should identify which populations are most likely to benefit and which outcomes an intervention can realistically change. A targeted hypothesis is more testable than a universal claim that hearing treatment prevents dementia.
Why hearing and cognitive assessment are difficult to separate
Hearing is relevant both to everyday communication and to how some cognitive tasks are administered. That creates a methodological challenge: an improved test result may reflect several changes, including better access to instructions or greater ease of participation.
This is not a reason to dismiss an improvement. It is a reason to examine how it was measured. Investigators can strengthen the interpretation by using assessments designed to limit hearing-related measurement bias and by including outcomes that reflect daily function.
Longer follow-up can also help distinguish a temporary change in assessment performance from a sustained change in cognitive trajectory. Where possible, outcome assessors should be protected from knowing which intervention a participant received.
These are considerations for interpreting and designing trials. They do not establish which mechanism produced CHOICE’s secondary finding. A careful account should leave that question open rather than invent an explanation the experiment did not resolve.
What should readers take from the trial?
The dementia result should not be used as a reason to ignore hearing difficulties. The question of whether an intervention reduces dementia progression is distinct from whether it helps a person hear and communicate. Those benefits should be evaluated on their own terms.
At the same time, a device or service should not be sold with a stronger brain-health promise than the evidence supports. Someone considering hearing care deserves a clear account of what it is intended to improve and what remains uncertain.
For future dementia-prevention research, the next steps include longer follow-up, enough primary events to estimate effects precisely, and replication of the secondary cognitive finding. Trials should also establish whether changes translate into meaningful daily benefits.
CHOICE adds a substantial randomized dataset to that effort. Its contribution is most useful when the nonsignificant primary outcome and favorable secondary result are reported together, without allowing either to erase the other.
Sources
- Chen Y, Shi H, Xiang M, et al.; CHOICE Collaborative Research Group. Hearing Intervention for Older Adults With Mild Cognitive Impairment: The CHOICE Randomized Clinical Trial. JAMA Neurology. Published September 28, 2026. https://doi.org/10.1001/jamaneurol.2026.3395.
- Lin FR, Pike JR, Albert MS, et al.; ACHIEVE Collaborative Research Group. 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. The Lancet. 2023;402:786–797. https://doi.org/10.1016/S0140-6736(23)01406-X.
