A Nature paper published on April 2, 2025 used a real-world quirk to get close to a randomized result. Wales rolled out its shingles vaccine with a hard birthdate cutoff: people born on or after September 2, 1933 were eligible, those born just before were not. People on either side of that line are otherwise alike, so comparing them (a regression-discontinuity design) sidesteps the usual problem that people who get vaccinated tend to be healthier to begin with.
A natural experiment can strengthen a causal case without becoming a clinical trial. Our shingles-vaccine evidence review walks through that distinction, and the study-reading primer is the practical framework for interpreting the next headline.
The finding: over 7 years, vaccination was associated with a 20% relative reduction in new dementia diagnoses, which works out to about 3.5 fewer cases per 100 people in absolute terms. The effect was stronger in women.
Why this matters for a carrier: a later genotype-stratified analysis found carriers appeared to benefit too. And this is a cheap, low-regret, already-recommended vaccine, so the downside of acting on it is small.
The honest caveat: this is still a natural experiment, not a randomized dementia-prevention trial. The design is unusually strong for observational work, but it cannot fully close the gap to proof.
We unpack it in the deep dive on the shingles vaccine and dementia risk, and on reading a study like a skeptic for why this design is convincing. Primary source: Nature, 2025. This is education, not medical advice.