GSK Bankrolled the Study That Turned Its Shingles Vaccine Into a Dementia Headline

At the foot of the paper, below the tables and the confidence intervals, in the small type where journals keep the parts nobody puts in a headline, sits the sentence that explains the headline. The authors of the study now generating headlines that a shingles shot fends off dementia reported receiving funding from GlaxoSmithKline, the company that makes the shingles shot. GSK, the disclosure adds, had no control over the study’s design, its analysis, or the decision to publish. That last clause is standard, and it is doing a lot of work. The problem was never that GSK cooked the numbers. It is that the company paid to ask a question whose most flattering answer sells more vaccine, using a design that cannot separate that answer from a duller one.

The study itself is careful, by the standards of its kind. A team led by Kaley Hayes at Brown University, working with the University of Delaware and the Providence VA, pulled Medicare claims and health records for 509,926 adults aged 66 and older who entered skilled nursing facilities between 2017 and 2022, and, writing in the Annals of Internal Medicine, asked a clean question: among people admitted to a nursing home, did the ones who got Shingrix within a year go on to be diagnosed with dementia less often than the ones who didn’t? Over four years, dementia appeared in 18.8 percent of the vaccinated and 24.6 percent of the unvaccinated. Run 18.8 against 24.6 as a relative reduction and you get the “24 percent lower risk” in every headline; stated as raw incidence it is a 5.8-point gap. It works out to roughly one dementia case avoided for every seventeen people vaccinated, which, if it were causal, would be a bigger effect than almost anything else on the table.

If it were causal. Here is the number the press releases left out: of those half-million nursing-home residents, 8,843 got the vaccine. Fewer than two in a hundred. Think about who those two are. In a population of frail, mostly cognitively declining elderly people, the resident who receives an elective shingles shot within twelve months of admission is not a random draw. He is the one with a doctor paying attention, a family advocating, the physical reserve to be offered a vaccine at all, and, often, a brain that is not yet failing. The authors know this. They report that the vaccinated were “slightly younger and healthier,” they adjusted for what they could measure, and they concede the adjustment “did not fully explain the association.” Every observational study of a self-selected treatment says a version of that sentence, and it is always the most honest line in the paper. You cannot adjust for the vigor you never wrote down.

The study GSK didn’t pay for

So the interesting question is not whether the shot might protect the aging brain. There is a genuine biological case that it does, and it does not rest on this study. It rests on a piece of work GSK did not pay for. In Wales, eligibility for the shingles program was set by an accident of the calendar: people born before 2 September 1933 were shut out for life, while those born a week later were offered the vaccine. Markus Eyting and Pascal Geldsetzer at Stanford turned that arbitrary line into a natural experiment, comparing roughly 280,000 older adults who differed by little except which side of a birthday they landed on, and found those offered the vaccine were about 20 percent less likely to develop dementia over the seven years that followed. Because the birthday, not the patient’s health-consciousness, decided who got vaccinated, that design does what a claims database cannot: it addresses the healthy-vaccinee problem at its source. The Stanford team designed the confound out. The GSK-funded study could only adjust for it, and conceded the adjustment fell short.

There is a second detail the celebratory coverage tends to skip. The vaccine in that Welsh natural experiment was the old live-virus shot, Zostavax, the product Merck pulled from the U.S. market in 2020. The vaccine in GSK’s study, and the one in every “get the shot to save your brain” segment this month, is Shingrix, the recombinant successor GSK sells. The strongest causal evidence we have is for a vaccine that is no longer on American shelves. The evidence for Shingrix specifically is thinner and softer: an Oxford analysis in Nature Medicine comparing recombinant to live recipients, a hypothesis that the effect belongs to the AS01 adjuvant rather than to shingles at all, and now a nursing-home database with GSK’s name in the acknowledgments.

There is even a tell buried in the broader literature that ought to slow everyone down. A 2025 meta-analysis in Age and Ageing pooled twenty-one studies covering more than 104 million people and found that essentially every adult vaccine travels with less dementia: shingles, yes, but also influenza, pneumococcus, and the tetanus-diphtheria-pertussis shot. One reading is that immune activation broadly protects the brain, a serious and testable idea. The duller reading is the one manufacturers never lead with: people who keep up with their vaccines are, as a class, people who keep up with everything, and it is the keeping-up the databases are measuring.

None of this makes Shingrix a bad vaccine at its actual job, which is preventing a painful, sometimes disabling rash. It makes the dementia claim a marketing frontier rather than an established fact, and the timing is worth sitting with. Shingrix is GSK’s best-selling product, roughly £3.6 billion in 2025, a franchise the company still expects to grow. It is also a franchise running into the wall every mature product hits: GSK has told investors its growth is slowing as it reaches the harder-to-activate unvaccinated. A vaccine sold as protection against shingles has a natural ceiling, set by how many people fear shingles. A vaccine sold as protection against dementia does not.

No one has run the study that would actually settle it, the one that randomly assigns people to Shingrix or not and counts the dementias years later, and until someone does, the honest verdict is that a good vaccine may or may not carry a valuable side benefit. Kaley Hayes and her co-authors wrote that down plainly, in the disclosures and the limitations, the way careful researchers do. The company that funded them, and the market that priced them, will report a cleaner number.

Sources

1. Brown University – news release on Hayes et al., Annals of Internal Medicine (2026)
2. ScienceDaily – “Shingles vaccine linked to 24% lower dementia risk,” release on the Brown/Annals study (Aug 2026)
3. Nature – Eyting, Geldsetzer et al., “A natural experiment on the effect of herpes zoster vaccination on dementia” (2025)
4. Age and Ageing – Maggi et al., “Association between vaccinations and risk of dementia: a systematic review and meta-analysis” (2025)
5. Nature Medicine – Taquet et al., “The recombinant shingles vaccine is associated with lower risk of dementia” (2024)
6. npj Vaccines – “Lower risk of dementia with AS01-adjuvanted vaccination against shingles and RSV” (2025)
7. Drugs.com – Zostavax (zoster vaccine live) FDA approval and U.S. discontinuation history
8. GSK – 2025 full-year results and long-term outlook (Shingrix sales)
9. BioPharma Dive – GSK vaccine outlook and Shingrix market slowdown