A Welsh birthday cutoff gave the best evidence yet that a shingles jab may delay dementia

People who received the live shingles vaccine were about a fifth less likely to be diagnosed with dementia over the following seven years, according to a Welsh study that exploited an accident of policy to get closer to cause and effect than any previous work [s1]. It is the strongest evidence to date that a vaccine might delay or prevent dementia, and it is still a single striking result that tested the older live vaccine, not the newer one most people now receive.

Why the birthday cutoff matters

The study's power comes from how Wales rolled out the vaccine. Eligibility for the live herpes zoster vaccine was set by a person's exact date of birth: those born before 2 September 1933 were ineligible and stayed ineligible for life, while those born on or after that date became eligible [s1]. Someone born a week before the cutoff and someone born a week after are, on average, indistinguishable in health, wealth and behaviour, but one group could get the vaccine and the other could not. That is a natural experiment, and it lets researchers use a regression discontinuity design that mimics the random assignment of a trial without running one [s1].

The eligibility rule worked as a switch. Among people who were merely a week too old, just 0.01% ever received the vaccine; among those a week younger, 47.2% did [s1]. That sharp jump, shown above, is the engine of the analysis: because the two groups differ so much in vaccination but so little in everything else, differences in their later dementia rates can be attributed to the vaccine with far more confidence than an ordinary observational comparison allows.

What it found

Over a seven-year follow-up, receiving the vaccine reduced the probability of a new dementia diagnosis by 3.5 percentage points (95% confidence interval 0.6 to 7.1, P = 0.019), which corresponds to a 20.0% relative reduction (95% confidence interval 6.5 to 33.4) [s1]. The protective effect was stronger in women than in men [s1]. The authors then repeated the analysis in a different population, the combined population of England and Wales, using death-certificate data and an outcome, deaths with dementia as the primary cause, that does not depend on a timely diagnosis, and the finding held [s1].

The rationale is not far-fetched. Several neurotropic herpesviruses have been implicated in dementia, so a vaccine that suppresses reactivation of the virus that causes shingles could plausibly matter, and vaccines are increasingly recognised to have off-target immune effects beyond the specific infection they prevent [s1]. But plausibility is not proof, and the study does not establish a mechanism.

How strong is it, really

Two things make this more convincing than the usual "vaccinated people are healthier" correlation. The birthday cutoff breaks the main confounder, because whether you fall a week either side of it has nothing to do with how health-conscious you are; and the result reproduces in a second dataset with a harder outcome [s1]. That is a well-designed piece of causal inference.

Two things temper it. It is one natural experiment in one country, and a regression discontinuity estimates the effect only for people near the cutoff, here, those around age 80, so it does not automatically generalise to younger adults. And it tested the live-attenuated vaccine, which has largely been replaced in the UK and US by a recombinant vaccine, discussed in our explainer on Shingrix eligibility and efficacy.

That last point is where separate, weaker evidence lines up in the same direction. A large US cohort study of the recombinant vaccine, covering 4,502,678 people, found dementia incidence of 99.1 cases per 10,000 person-years in the fully vaccinated versus 135.0 in the unvaccinated, and a hazard ratio of 0.68 for two doses [s2]. That is an observational study, open to the confounding the Welsh design avoids, but it suggests the signal is not unique to the older vaccine [s2]. The shingles vaccines have also been linked to reduced cardiovascular events, part of a broader pattern of apparent off-target benefits that is still being sorted out.

What it means for a reader

No health authority recommends the shingles vaccine to prevent dementia, and this study does not change that; the vaccine's licensed job is preventing shingles and its complications. What the Welsh result adds is the best causal-leaning evidence so far that a widely used vaccine might also delay dementia, a possibility worth taking seriously precisely because dementia has so few modifiable risk factors with strong evidence. The appropriate next step is a randomised trial designed to test the dementia question directly, rather than a change in practice built on one natural experiment. For now, the honest reading is: a genuinely strong study, a plausible effect, and not yet a reason on its own to get vaccinated.