In 2013, the Welsh National Health Service did something that, three decades from now, may turn out to be one of the most useful accidents in dementia research. To roll out the new live shingles vaccine without overwhelming GP clinics, the NHS picked a hard birthday cutoff: anyone who was 79 on September 1, 2013, was eligible for one year. Anyone born a week earlier, already 80, would never become eligible. Anyone born a week later would have to wait. The cutoff was administrative, and the two cohorts on either side of it came out near-identical on the covariates the researchers later checked.
A decade later, a research team led by Markus Eyting realized what that cutoff actually was: a quasi-randomized trial of a few hundred thousand nearly-identical 80-year-olds, half of them vaccinated against shingles and half not. The team ran the numbers. Seven years out, the people who had been just barely eligible were 3.5 percentage points less likely to be diagnosed with dementia, a roughly 20% relative reduction. And here is the part you should hold onto, because it reframes the whole story: a separate Oxford analysis published a few months later found the same-sized dementia signal for GSK’s new RSV vaccine, Arexvy, which targets a respiratory virus that has nothing to do with shingles. The two GSK vaccines have one obvious thing in common, and it isn’t the virus they fight. It’s the adjuvant they share.
Before we get to that, a word about why the Welsh paper matters in its own right. Vaccinated populations have always been suspect in observational dementia work, because the kind of person who shows up for an optional shot at 79 is also the kind of person who eats vegetables and remembers to walk the dog. Researchers call this the healthy-vaccinee bias, and it can manufacture a vaccine “benefit” out of nothing more than the difference between joggers and shut-ins. The Welsh cutoff blew a hole in that explanation. The two groups on either side of September 2, 1933 were not joggers versus shut-ins. They were people born a week apart. In the Nature paper, the vaccine moved dementia diagnoses and didn’t appear to move other major outcomes the authors examined, which is the pattern you’d expect from something acting on the brain specifically rather than on overall frailty.
The signal has become hard to dismiss. The much more interesting question is what produces it.
The first explanation everyone reached for was elegant and intuitive. Varicella zoster virus, the agent that gives children chickenpox and reactivates in old age as shingles, lives in the nervous system. Each reactivation drives a wave of neuroinflammation. Suppress the virus, suppress the inflammation, slow the cognitive decay. The hypothesis fit neatly into a broader story about neurotropic herpesviruses and aging brains, the same family of arguments that links HSV-1 to Alzheimer’s pathology. Two more big analyses seemed to back it up. A 2024 Nature Medicine paper from the Oxford group led by Maxime Taquet used U.S. electronic health records to show that GSK’s recombinant vaccine, Shingrix, was associated with a 17% lower six-year dementia risk than the older live shingles vaccine. This month, a target trial emulation of more than 500,000 newly-admitted nursing-home residents, led by Brown’s Kaley Hayes and funded by GSK, reported a 24% lower four-year dementia risk in the vaccinated group, with roughly one prevented dementia diagnosis for every seventeen vaccinated residents. A separate Stanford-led paper in Cell reported that even patients who already had dementia died of it less often if they had received the vaccine.
Four studies, four different designs, four different populations, all pointing the same way. By the standards of dementia research, that is an avalanche.
Then comes the result that quietly rearranges everything. In June 2025 the Oxford team published a follow-up in npj Vaccines comparing more than 430,000 U.S. adults who received either a shingles vaccine or Arexvy against people who got an ordinary flu shot. Arexvy and Shingrix share a proprietary adjuvant, AS01, a saponin-and-lipid cocktail that ramps up the immune response. The RSV vaccine, which targets a respiratory virus with no herpesvirus-style latency-in-nerves story behind it, was associated with a 29% reduction in dementia diagnoses over the following 18 months. The effect was essentially indistinguishable from the shingles vaccine’s effect in the same dataset.
If that comparison holds up, the headline is not “Shingrix prevents dementia.” It is “AS01 may be doing something to aging immunity.” A protective signal that travels with the adjuvant rather than the antigen points away from killing the target virus and toward whatever the adjuvant does to the immune system on its way to building antibodies against the antigen of the day. That is a very different drug story than “stop the shingles, save the brain.” It is a story about an off-the-shelf saponin formulation that may, as a side effect, retune the aging immune system in ways that are downstream-protective for the brain.
It is also, conveniently, a story about a GSK adjuvant. The Brown nursing-home paper that produced this week’s MedPage coverage ran on GSK money; the Oxford recombinant and AS01 analyses ran on Wellcome Trust and academic funding, which makes them more useful, not less. The IDWeek 2025 presentation that claimed Shingrix also cuts heart attacks, strokes, and overall mortality is the kind of everything-bagel benefit list that should trigger any epidemiologist’s healthy-user reflex, and it has not been peer-reviewed. None of this means the dementia signal is fake. It does mean the studies most likely to ask the cleanest mechanistic question, the one that would distinguish virus from adjuvant, are not the ones GSK is paying for, and the marketing arc around Shingrix has been climbing for years from “you don’t want shingles” toward something that sounds a great deal like “you might lose your mind without it.”
The right read here is not that the effect is illusory. The Welsh natural experiment is one of the cleanest pieces of population-scale evidence we have on any vaccine’s secondary effects, and it ran on NIH money, with no competing interests, on a vaccine (Merck’s Zostavax) that GSK no longer sells. The right read is that the cause is probably not what the press releases say it is.
Two things to watch next. The first is whether anyone runs a real randomized trial. A clean RCT of AS01-containing versus AS01-free vaccines, with cognitive outcomes as a pre-registered endpoint, would settle the adjuvant-versus-virus question in a way no amount of electronic-health-records cleverness can. The second is what GSK and the FDA do with the implication. If the adjuvant is doing the work, the right product is not a shingles vaccine. It is whatever drug the adjuvant happens to be riding on.
Sources
- Nature, Eyting et al. (2025): A natural experiment on the effect of herpes zoster vaccination on dementia
- npj Vaccines, Taquet et al. (2025): Lower risk of dementia with AS01-adjuvanted vaccination against shingles and RSV
- Nature Medicine, Taquet et al. (2024): The recombinant shingles vaccine is associated with lower risk of dementia
- Cell, Geldsetzer et al. (2025): Shingles vaccination at different stages of the dementia disease course
- MedPage Today, June 15, 2026: Risk for Dementia After Shingles Vaccine in High-Risk Group
- CIDRAP: IDWeek 2025 Shingrix preprint on dementia, heart disease, and mortality
- News-Medical: Shingles and RSV vaccines with AS01 adjuvant reduce dementia risk
- PubMed record: Eyting et al., natural experiment on herpes zoster vaccination and dementia
- Eric Topol on the shingles vaccine and dementia evidence