The longevity secret hiding in plain sight

ELI5/TLDR

A retired tech guy who runs a longevity channel argues that vaccines are quietly the most powerful longevity drug we have, and we missed it for decades because nobody was looking at the right data. Recent studies that match a million vaccinated people against a million similar unvaccinated people show that flu shots cut Alzheimer’s risk by 20-40%, the shingles shot cuts heart attacks and strokes by 46% in the year after the jab, and overall biological aging slows by about 19%. The catch is that none of this could be seen until big medical databases and machine learning came along. The rest of the episode is a tour through why six different groups, from worried mothers to anti-vaccine lawyers, all think they understand the elephant.

The Full Story

The host, Chris (no last name given on the channel, but he’s the “Viva Longevity” guy), opens by acknowledging that his last episode set off a comment war. He had said two things people did not want to hear: that trust in science is the biggest predictor of population health after poverty, and that vaccines look like one of the strongest known drivers of longevity. This episode is his attempt to make the data case without sounding like, in one viewer’s words, “a know-it-all.”

The headline studies

The piece of evidence Chris keeps coming back to is a 2022 study from the University of Texas led by neurologist Paul Schultz. The team pulled records on roughly a million seniors who never got the flu vaccine and matched each one against a similar senior who did. Same age, same sex, same area, same chronic conditions. This trick is called propensity matching. The idea is to fake a randomized trial after the fact by finding pairs of people who differ only in the thing you care about, in this case the shot.

The result, in Schultz’s own words: three flu shots over three years cut the risk of Alzheimer’s by about 20%, six shots cut it by about 40%. Translated into bodies, in a cohort of 900,000 people, 80,000 of the unvaccinated got Alzheimer’s compared to 50,000 of the vaccinated. Thirty thousand people who did not lose themselves to a disease that erases people one brain cell at a time.

A few more recent studies in the same vein:

A Stanford team noticed that in Wales, people born just before a certain date were ineligible for the shingles vaccine and people born just after were eligible. That cutoff acted like a natural experiment, the closest thing to randomization you can get without running a trial. The eligible group had 20% lower dementia.
Dr. Robert Nuin at UC Riverside used a 117-million-patient database called TriNetX to compare 247,000 vaccinated and unvaccinated people with partially clogged arteries. In the first year after the shingles shot: 46% fewer heart attacks and strokes, 66% fewer deaths from any cause.
USC’s gerontology school looked at frozen blood samples from 4,000 retired seniors and found shingles-vaccinated people had 19% lower markers of biological aging. For an American with a life expectancy of 79, that translates to roughly 18 extra months. For someone in a high-trust-in-science zip code where life expectancy is closer to 85, more like three years.
Same Texas team showed that the higher-dose flu shot for over-65s cut Alzheimer’s incidence by 54% over three years.

The pattern is consistent across multiple vaccines (flu, TDAP, shingles, pneumonia) and multiple end conditions (Alzheimer’s, heart disease, biological aging). That kind of repetition is what epidemiologists look for. One study is interesting. Five studies pointing the same way is a signal.

Why nobody saw this until now

Chris’s analogy for why epidemiology was historically clunky is the early search engine wars. Yahoo, Excite, and WebCrawler hired editors who hand-curated results. Google hired thousands of math PhDs who let the data speak. The “letting the data speak” part is what biomedical informatics now does to medical records. You stop testing one hypothesis at a time and instead let machine learning surface unexpected patterns.

Epidemiology start with certain like hypothesis and test that hypothesis using statistic. In biomedical informatics, what we do is that like use machine learning model first and let the data speak first.

That is the data scientist on Schultz’s team explaining how the flu-vaccine-meets-Alzheimer’s connection got noticed in the first place. It wasn’t a hunch. The model surfaced flu vaccines as suspiciously associated with lower Alzheimer’s onset, and the team then went and tested it formally. She literally calls it “AI’s idea.”

This matters for understanding the rest of the episode, because Chris’s argument is that almost everyone arguing about vaccines, on every side, is working with information that pre-dates this kind of analysis.

The mechanism, sort of

Why would a flu vaccine prevent Alzheimer’s, or a shingles vaccine prevent heart attacks? The honest answer is no one fully knows yet. Two clues:

The first comes from a Dutch researcher in Guinea-Bissau named Christine Stabell Benn, who has been tracking 200,000 people for 25 years. She noticed that vaccines often improve overall survival far beyond what they should from preventing the targeted disease. The hypothesis is that live vaccines (especially) train the entire immune system, like a tennis player’s first practice match against a beginner before facing a pro.

The second comes from Bali Pulendran at Stanford, who has been mapping what the yellow fever vaccine, an extraordinarily potent live vaccine, actually does to the immune system. Whatever it does, it does for years.

The candidate mechanism Chris keeps mentioning is reduced inflammation. Most of the diseases of aging, from Alzheimer’s to heart disease, have inflammation as a core ingredient. Anything that lowers chronic inflammation, including stopping the small fires lit by repeated infections, plausibly slows aging.

The bulk of the episode is Chris walking through six groups who all touch a different part of the elephant.

Worried mothers are not crazy or uneducated. They are evolved pattern matchers being asked to make a decision in an environment full of noise. The expert he leans on here is Heidi Larson of the Vaccine Confidence Project, who points out that scientists speak elite language and big institutions have lost public trust. When a child is diagnosed with autism after a vaccine, the human brain wires causality into the sequence even when none exists.

Doctors and nurses see the cost of preventable disease up close. Chris quotes a pediatric nurse, Dani Fritz of the Kid Nurse blog, who started out vaccine-skeptical, went to study the harms for her thesis, and could not find substantiated evidence of them. Then she watched children die of preventable diseases in the ICU and changed sides.

Epidemiologists and data scientists see populations, not individuals. Chris draws a sharp distinction between a doctor (front-line, individual patients) and an epidemiologist (statistician of populations). He also distinguishes a master’s in public health (a few epidemiology classes) from a PhD epidemiologist with years of post-doc work, and notes how rarely the public ever sees the latter on TV.

Public health scientists see vaccination as a collective decision, not a personal one. The Australian HPV program is his prime example: girls and boys vaccinated in school since 2007, cervical cancer rates halved, on track for elimination by 2035. In the US, where 23% of parents refuse the HPV vaccine compared to 2% in Australia, 14,000 women get cervical cancer every year and 4,000 die. Another quarter-million have precancerous lesions removed.

The truly misinformed are everyone, including Chris, in the senses they don’t realize. He argues the core problem is that vaccines do not text you when they work. You never get a notification saying “your uncle was a hepatitis B carrier and just kissed your baby, but the birth dose has you covered.” You just end up with a healthy 40-year-old child decades later and forget why you bothered.

Anti-vaccine advocates as a profession. This is where Chris gets sharpest. He notes RFK Jr. earned $836,000 from his anti-vaccine nonprofit in 2022, that Jenny McCarthy charged $50-100k per speaking gig, and that Andrew Wakefield bought an estate in Austin off his speaking circuit. His argument: in every field there are people willing to deny science for money. The tell is that they don’t update when new evidence comes in.

He also goes after vaccine injury lawyer Aaron Siri, whose firm is funded by a US-only vaccine injury court that pays lawyers regardless of outcome. Chris’s accusation is that the legal incentive is structurally aligned with manufacturing fear, and that the court was designed in the 1980s before the modern data infrastructure existed.

Key Takeaways

Vaccines may be doing more than preventing the diseases they target. The new evidence suggests they reduce Alzheimer’s risk (20-40% for flu), heart attacks (46% for shingles), and biological aging (19% for shingles).
Propensity matching on huge databases is the new tool. When you have 100+ million patient records, you can build pseudo-randomized trials by matching vaccinated and unvaccinated people on dozens of variables. This is not the same kind of evidence as a randomized trial but it is much closer than anything we had before.
The mechanism is probably reduced chronic inflammation. Not proven, but the leading hypothesis. Live vaccines may train the broader immune system in ways inactivated ones don’t.
The high-dose flu shot for over-65s appears more protective than the standard dose for Alzheimer’s. Worth knowing if you’re advising older parents.
Trust in science is correlated with longevity at the zip-code level in the US. Chris’s example: his Silicon Valley area has the same life expectancy as Japan, partly because vaccination rates are 2.5x the national average.
The HPV vaccine is one of the most underused interventions in the US. Australia’s outcome data is the cleanest argument for it.

Claude’s Take

Chris is making a strong claim and the evidence backing it is real but it deserves a few caveats he glosses over.

The propensity-matched studies are the best you can do without running a randomized trial, but they are still observational. The healthy user effect, where people who get vaccinated are also more likely to exercise, eat well, see doctors regularly, and own gym memberships, is the standard worry. The Texas team tries to address this by comparing high-dose to standard-dose flu shots within the vaccinated population. That is genuinely clever and reduces the worry, but it doesn’t eliminate it. The Welsh shingles study using the eligibility cutoff is the strongest piece of evidence Chris cites, because it comes closest to true randomization.

The 46% heart attack reduction and 66% all-cause mortality reduction in the first year after the shingles vaccine is, frankly, almost too good. Numbers that big should always trigger a “what am I missing” reflex. To Chris’s credit, he quotes the data scientist saying the same thing: “When we see that kind of like big effect, our first reaction is not excitement, we rather we say we think like what are we missing.” That is the right instinct. Replication and time will tell whether the effect sizes hold up or shrink, as they often do.

The episode is also unmistakably a polemic. Chris’s framing of the six groups is sharp on the bad actors and warm on the worried mothers, which is probably the right pitch for the audience he’s trying to reach. But his attack on Aaron Siri and the vaccine injury court conflates several things. The court exists because vaccines, like any medical intervention, do occasionally cause harm. That doesn’t make every plaintiff lawyer a fear merchant, even if some clearly are.

The strongest part of the episode is the meta-argument: that data science is genuinely showing us things about long-term vaccine effects that we could not have known before. That is true, it’s underappreciated, and the implication, that the cost-benefit calculation on adult vaccines may be much better than current public discourse suggests, is worth taking seriously.

The weakest part is the assumption that everyone who disagrees with the new data is either ignorant, emotional, or grifting. There are legitimate reasons to be skeptical of observational studies showing dramatic effect sizes, and the response to that skepticism should be more data, not more confidence.

A 7. Solid evidence assembly, useful framework for understanding why people disagree, but Chris is so frustrated with the bad-faith side that he undersells the legitimate scientific uncertainty on the side he’s defending.