The Prevention Nobody Planned

Not metaphorically. If you had chickenpox as a kid — and if you were born much before the mid-nineties, you almost certainly did — the varicella zoster virus did not leave when the spots faded. It withdrew into your dorsal root ganglia, the little bundles of nerve cell bodies tucked alongside your spinal cord, and went quiet. It has been there ever since. Decades. It is there right now, doing nothing in particular, waiting for your immune system to get old enough to stop paying attention.

When it wakes up, we call it shingles, and it is famously unpleasant: a burning rash that traces a single nerve's territory around your torso, the pain drawing you an accurate map of your own wiring.

So we built a vaccine for it. Shingrix, licensed in 2017 — a recombinant protein plus an adjuvant, roughly 90% effective against the rash. Ordinary public health. Prevent painful thing. Bill Medicare. Move on.

This week, a team at Brown reported that the same shot appears to cut dementia diagnoses by about a quarter.

Kaley Hayes and colleagues at Brown's School of Public Health, publishing in Annals of Internal Medicine on August 25, worked through Medicare claims and health records for 509,926 people aged 66 and older who entered skilled nursing facilities between 2017 and 2022. They used a method called target trial emulation — a way of coaxing observational data into approximating the randomized trial nobody ran. Over four years, dementia showed up in 24.6% of the unvaccinated and 18.8% of the vaccinated. Hayes puts the useful version plainly: roughly one in seventeen dementia cases potentially prevented.

The tempting framing here — the one most of the coverage will reach for — is happy accident. Bonus. Side effect. We aimed at a rash and hit Alzheimer's, lucky us.

That framing is wrong, and it's worth being precise about why.

Because this is not the first time. In 2025 a group led by Markus Eyting published a genuinely elegant piece of work in Nature, exploiting an accident of Welsh policy: eligibility for the older shingles vaccine was set by exact date of birth. Born before 2 September 1933, ineligible for life. Born on or after, eligible. Uptake jumped from 0.01% to 47.2% across a one-week gap in birthdays — which is about as close to a randomized trial as history ever hands you for free. Dementia diagnoses fell on the eligible side of the line.

Different vaccine — Zostavax was a live attenuated virus, Shingrix a recombinant protein with a synthetic adjuvant. Different country, different health system, different decade. And the same answer falls out.

But the two are not a matched pair, and it matters which one is carrying the weight. The Welsh study is the strong one, and its strength is structural: nobody chooses which side of a birthday to be born on. That severs the link between getting vaccinated and being the sort of person who gets vaccinated — the healthy-vaccinee problem, which haunts every study of this shape. The Brown study cannot do that. It adjusts, and adjustment is not randomization. What Brown has instead is scale, the current vaccine, and a population the Welsh experiment says nothing about.

So this isn't two independent legs under a table. It's one quasi-experiment with real causal traction, plus a large confirmatory study vulnerable to precisely the bias the first one rules out. That's weaker than the happy-accident framing implies, and still strong enough for the thing that matters: our model of dementia is missing a term.

The rest of the honest limits are worth stating. GSK, which makes Shingrix, funded the work — disclosed, with no control over design or analysis. And nobody has the mechanism. Maybe decades of low-grade viral reactivation quietly inflame the brain. Maybe the adjuvant does something general and useful to an aging immune system. Maybe some of it is confounding that survived the math. The researchers say clinical trials are needed, and they are right.

But here is the number that stopped me, and it isn't in the headline. Of those 509,926 nursing home residents — the exact population most likely to get shingles and most likely to get dementia — 8,843 were vaccinated. Under two percent.

That scarcity is the scandal, and it is also the caveat, which is an uncomfortable shape to hold in one hand. When fewer than one in fifty residents gets a shot, the ones who do are not a random fiftieth. They are the people with an involved family, an attentive prescriber, enough remaining vitality that the trip seemed worth making. Every one of those independently predicts keeping your mind longer. The adjustment tries to subtract them; it cannot promise it succeeded. And the implicit offer inside one-in-seventeen — vaccinate the other 98% and get the same result — is shakiest exactly where the number is most quotable.

Still. We went looking for a way to stop a rash. What we may have found instead is that a passenger we have all been carrying since kindergarten has been doing something to us the entire time, and that switching it off — for reasons nobody planned and nobody yet understands — might let some number of people keep their own minds a little longer.

And notice what it took to find out. The most credible causal evidence we have that a licensed, cheap, already-stocked vaccine may delay dementia exists because a Welsh administrator once drew a line through a birthday in 1933. Nobody designed that experiment. Nobody would fund it: a decade-long prevention trial, on a drug already approved for something else, aimed at an indication with no patent runway and no sponsor waiting at the end. The question wasn't asked. It was tripped over.

That is the part that generalizes past shingles. We tend to picture the frontier of medicine as the place where the hard questions are being worked on. The truer picture is that the questions being worked on are the ones somebody can afford to ask — and an effect falling outside the indication a molecule was licensed for gets found by accident, or doesn't get found. Useful things don't hide, exactly. They get filed under the wrong heading and wait to see whether anyone is reading carefully.

One in seventeen. It isn't a cure, and the number will move. It's a shot most people over sixty can already get, and mostly don't.