Older adults who received the shingles vaccine Shingrix had a 24% lower risk of dementia over four years, according to a large analysis — the latest in a string of studies hinting at a link that none has proven.
Researchers led by Brown University analysed Medicare claims and health records from 509,926 adults aged 66+ admitted to skilled nursing facilities between 2017 and 2022. Only 8,843 had been vaccinated. The findings appeared in Annals of Internal Medicine.
Why anyone thinks a shingles vaccine might affect dementia
The hypothesis is not arbitrary. Shingles is caused by reactivation of the varicella-zoster virus, which establishes lifelong latency in nerve cells after chickenpox and can re-emerge decades later when immune surveillance weakens.
Two mechanisms have been proposed. Reactivation causes inflammation in nervous tissue, and repeated episodes might contribute cumulatively to neurodegeneration — particularly since the virus can affect cerebral blood vessels. Alternatively, the vaccine’s effect may be non-specific: vaccines can produce broad immune changes beyond their target, and a generally better-regulated immune system in later life might influence dementia risk through inflammation.
Neither is established, but both are biologically coherent, which is why the repeated observational signal has attracted serious attention rather than dismissal.
What the study did differently
The researchers used target trial emulation — a method designed to mimic a randomised trial using real-world data.
The approach forces analysts to specify in advance what trial they would have run: eligibility criteria, the moment of treatment assignment, follow-up rules and the outcome. They then construct that trial from observational records, which avoids several errors that plague conventional observational analysis — particularly biases arising from how time is handled around the moment of treatment.
It is a genuine methodological improvement. It does not create randomisation where none exists.
They estimated the vaccine might prevent roughly one dementia case per 17 people vaccinated.
Why that number should be treated cautiously
One case prevented per 17 vaccinated would be an extraordinary effect — comparable to or better than almost any intervention in medicine for a disease with no effective prevention.
Effects of that magnitude appearing in observational data, for an intervention not designed to produce them, are more often explained by confounding than by biology. The size of the estimate is itself a reason for scepticism rather than excitement.
The confounding problem, stated plainly
The authors are explicit that the study cannot establish cause and effect. Vaccinated participants were slightly younger and healthier, and although the researchers adjusted for such differences, those adjustments “did not fully explain the association.”
The deeper issue is healthy vaccinee bias. People who receive optional preventive vaccines differ systematically from those who do not, in ways records capture poorly. They attend appointments, follow medical advice, have transport and support, and are frequently more cognitively intact at the time of vaccination — which matters enormously when the outcome is dementia.
Early cognitive decline reduces the likelihood of obtaining an elective vaccine years before diagnosis. The association could therefore run backwards: not vaccination preventing dementia, but incipient dementia preventing vaccination.
The unusual population
One feature deserves particular attention. All participants were admitted to skilled nursing facilities, and only 1.7% had been vaccinated.
That is a low vaccination rate in a frail population, which means the vaccinated group is a small and highly selected minority. Whatever distinguished those 8,843 people — better baseline health, more engaged families, better healthcare access — is precisely what statistical adjustment struggles to capture.
Why it keeps coming up
Several independent studies using different populations and methods have found similar associations, and consistency across designs is generally reassuring.
It is less reassuring here, because the same confounding applies to all of them. Healthy vaccinee bias operates identically in every observational dataset, so replication demonstrates the bias is consistent rather than that the effect is real.
What would settle it
Confirming a genuine protective effect would require randomised clinical trials. Randomisation is the only method that balances unmeasured differences, including the ones nobody thought to record.
Such a trial faces real obstacles: dementia takes years to develop, requiring long follow-up in older participants, and withholding a recommended vaccine from a control group raises ethical questions. Design workarounds exist — comparing vaccination timing rather than vaccination versus none — but none is straightforward.
The practical takeaway is unchanged
Shingrix, a recombinant zoster vaccine introduced in 2017, is the only shingles vaccine on the market and is already recommended for older adults to prevent shingles.
Shingles itself is worth preventing: the rash is painful, and a substantial minority develop postherpetic neuralgia — nerve pain persisting months or years after the rash resolves, and notoriously difficult to treat.
So the recommendation does not change. Get the vaccine to prevent shingles, which it demonstrably does. A dementia benefit would be a remarkable bonus, and remains a promising lead rather than proof. This summarises observational research and is not medical advice.