This is the foundation, and it is the best supported thing here. Self controlled case series designs, which use each patient as their own comparison and therefore neuter the healthy user problem entirely, consistently show a sharp spike in myocardial infarction and stroke in the weeks following a zoster episode. The risk decays over the following months. The biology is unremarkable once you say it out loud. Acute viral infection produces inflammation, inflammation destabilizes plaque, and zoster in particular is a vasculopathy that infects arterial walls directly.

Zoster is not behaving strangely. It is behaving like influenza, which does the same thing, and like any acute inflammatory insult in a patient who already has plaque.

The meta-analysis presented at the European Society of Cardiology last August pooled 19 studies and reported vaccine effectiveness of 18 percent in adults 18 to 49 and 16 percent in adults 50 and over. The absolute figures are the ones worth carrying around. Somewhere between 1.2 and 2.2 fewer cardiovascular events per 1,000 person years.

That is a real effect of a plausible size. It is roughly what you would predict if preventing zoster prevents the post zoster risk spike, which is exactly the mechanism claim in card one. The pieces fit.

Eight of the nine analyzable studies were observational. The presenting author works in Global Medical Affairs at GSK, which manufactures Shingrix, and said plainly that further research is needed before anyone attributes the association to the vaccine. When the manufacturer is the one applying the brakes, take the brakes seriously.

TriNetX, 123,411 vaccinated patients matched to an equal number of unvaccinated ones, all with atherosclerotic cardiovascular disease, followed from one month to one year after vaccination. Reported reductions of 32 percent for MI, 25 percent for stroke, 25 percent for heart failure, 46 percent for the composite.

The effect size is roughly triple what the pooled meta-analysis found, in a population where the pooled estimate should be more reliable, not less. The authors acknowledged healthy user bias and adjusted for socioeconomic and behavioral proxies in the claims data. Those proxies are weak. Nobody codes flossing.

This is a conference poster. It has not been peer reviewed and it has not been published. The press release, the Medscape writeup, and every downstream article are all describing the same unrefereed abstract.

This comparison came from the lead author and traveled through the coverage unchallenged. It equates a relative risk reduction from a single unadjudicated observational cohort with the effect of smoking cessation, which is supported by decades of cohort data, dose response gradients, and mechanistic work across essentially every organ system.

The 66 percent all cause mortality reduction sitting in the same dataset is the reason to decline the comparison. Any analysis that shows an intervention preventing death from every cause at once is reporting the health of the people who chose it, not the effect of the thing they chose. That confound does not politely restrict itself to the endpoints you are less interested in.

Wales set eligibility for the zoster vaccine by exact date of birth. Anyone born before 2 September 1933 was ineligible and stayed ineligible for life. Anyone born on or after that date got a year of eligibility. Vaccine uptake went from 0.01 percent among people one week too old to 47.2 percent among people one week younger. Nothing else about those two groups differs. Birthday functioned as a coin flip.

Eyting and colleagues ran a regression discontinuity on that cutoff and reported a 3.5 percentage point absolute reduction in new dementia diagnoses over seven years, a 20 percent relative reduction, published in Nature in April 2025. The effect was stronger in women.

The design is the point. This is the objection that sinks the cardiac poster, tested directly and failing. The investigators checked whether the vaccine affected other common causes of death and morbidity, and it did not. They checked whether vaccinated people went on to take up other preventive measures, and they did not. The benefit did not show up everywhere at once. It showed up in shingles and in dementia, which is what a real effect looks like.

Replications followed in Australia, Canada, England, and New Zealand, all using the same trick of a birthdate eligibility cutoff. A December 2025 follow up in Cell extended the finding to mild cognitive impairment and to dementia deaths among people already diagnosed.

The regression discontinuity estimates are anchored on people aged roughly 79 to 80, which is where the eligibility cutoff sat. The paper says plainly it cannot estimate the effect at other ages.