The 55% That Doesn't Exist: How a Modest Insurance Claims Study Became a Flu Shot Miracle
Inside the alchemy that turns fragile epidemiological signals into pharma-friendly headlines — and what the study’s blind spots reveal about the vaccine-centric paradigm itself
Something remarkable happened last week in the world of science communication. A study published in Neurology — a retrospective analysis of insurance billing codes — was transmuted, through the reliable machinery of press-release journalism, into front-page headlines declaring that a flu vaccine “decreases Alzheimer’s disease risk by 55 per cent.”
The Daily Mirror ran it. The Daily Express ran it. The New York Post ran it. GB News ran it. The Daily Record ran it. Google News surfaced these outlets at the top of its results, ensuring that hundreds of millions of people were exposed to a number — 55% — that does not appear anywhere in the actual study.
Let’s look at what the study actually found, what it cannot tell us, what the resulting media coverage reveals about the architecture of manufactured consent in public health, and what the study’s deepest blind spots tell us about the limits of the vaccine-centric paradigm itself.
What the Study Actually Did
Bukhbinder et al. (2026) used the IQVIA PharMetrics Plus database — a US commercial insurance claims dataset — to compare Alzheimer’s dementia diagnoses among adults 65 and older who received either the high-dose flu vaccine (Fluzone High-Dose) or a standard-dose inactivated influenza vaccine between 2016 and 2019. They followed participants for up to three years using a target trial emulation framework with inverse probability weighting.
The methodological scaffolding is competent. The study design is internally coherent. None of that is in question.
What is in question is what the study means — and how far from that meaning the public narrative has traveled.
The Numbers That Actually Exist in the Paper
In the per-protocol analysis — the one the authors themselves emphasize — the risk ratios for Alzheimer’s dementia after high-dose vs. standard-dose vaccination ranged from 0.78 at month 1 to 0.89 at month 25.
That’s an 11–22% relative risk reduction. Not 55%.
The absolute risk differences were tiny: ranging from 0.02 to 0.54 percentage points over the entire follow-up. The number needed to treat at month 25 was 185 — meaning you’d need to switch 185 people from standard-dose to high-dose flu shots to prevent one additional Alzheimer’s diagnosis code from appearing in an insurance database.
MedPage Today, to their credit, reported the finding accurately: “roughly 20% lower risk.” Co-author Paul Schulz himself described it as “about a 20% reduction.”
So where did 55% come from? It appears nowhere in the paper. Not in the primary analysis. Not in the secondary analyses. Not in the sensitivity analyses. It is a phantom number, born somewhere in the press-release-to-headline pipeline and now permanently lodged in the public record.
What does this actually mean for a real person?
Let’s put it in terms anyone can understand. At the point of maximum observed effect (month 25), the absolute risk difference between the two groups was 0.54 percentage points. That means out of every 1,000 people who got the high-dose shot instead of the standard dose, roughly 5 fewer would receive an Alzheimer’s diagnosis code in an insurance database over two years. The other 995 would experience no measurable difference whatsoever.
Flip it around: 185 elderly people would need to be injected with the high-dose formulation — with its four-fold antigen load and its aluminum adjuvant — to prevent a single additional billing code from appearing in a claims database. Not a confirmed Alzheimer’s diagnosis. Not a cognitive test result. Not a brain scan. A billing code, in a system the authors themselves acknowledge is rife with misclassification.
And that’s at the peak of the effect. In the early months, you’re looking at absolute risk differences of 0.02 percentage points — meaning out of every 10,000 people switched to the high-dose vaccine, roughly 2 would see any difference at all.
Now consider the denominator of risk. Every one of those 185 people (or 995 out of 1,000 who receive zero benefit) is still exposed to the known risks of the injection: the aluminum adjuvant, the inflammatory cascade it provokes, and the rare but real adverse events associated with influenza vaccination — from Guillain-Barré syndrome to the inflammatory responses documented in pregnant women and cardiovascular patients. The U.S. Vaccine Adverse Event Reporting System records more Guillain-Barré reports for influenza vaccines than for any other vaccine. The risk is small per individual. But when you multiply it across the millions of elderly people this study implicitly encourages to choose the high-dose shot specifically to prevent Alzheimer’s, you are exposing an enormous population to real immunological provocation for an effect so small it is indistinguishable from statistical noise in most of the sensitivity analyses.
This is the arithmetic the headlines erase. “55% risk reduction” sounds like a medical breakthrough. “5 fewer billing codes per 1,000 people, in a database riddled with misclassification, at the cost of injecting all 1,000” sounds like what it actually is: a finding that would not survive a serious benefit-risk analysis if the risks of the intervention were honestly accounted for.
No Placebo. No Unvaccinated Control. No Cognitive Testing.
This point cannot be stated clearly enough: there is no unvaccinated group in this study. Both arms received an aluminum-adjuvanted inactivated influenza vaccine. The comparison is exclusively between two doses of the same product category among people who chose to get vaccinated.
The authors frame this as a strength, arguing that it reduces healthy-vaccinee bias. And it does narrow certain confounding pathways. But it also means the study is structurally incapable of answering the question the headlines are implying — namely, whether flu shots prevent Alzheimer’s.
What the study can say: among vaccinated older adults, those who got the higher-dose formulation had slightly fewer dementia diagnosis codes in their insurance records.
What the study cannot say:
Whether either vaccine is neuroprotective compared to no vaccination
Whether either vaccine contributes to neurological harm
Whether the observed difference reflects genuine cognitive protection or differences in health-seeking behavior, diagnostic patterns, survival, or data quality
No participant was cognitively tested. No biomarkers were measured. No brains were imaged. The “Alzheimer’s dementia” outcome is an ICD billing code — the same code that the authors themselves acknowledge is systematically misclassified in claims data.
The Effect Vanishes Under Scrutiny
The authors, to their credit, ran extensive sensitivity analyses. Several of those analyses are devastating to the primary finding:
When they removed the nonspecific “senile” and “unspecified” dementia codes from the outcome definition — keeping only AD-specific ICD codes — the number of cases dropped by approximately 80%, and the entire effect disappeared. This means the finding depends on the loosest, most misclassification-prone diagnostic category in the claims database.
When they restricted the sample to statin-adherent patients (a proxy for overall health-consciousness and medication compliance), the effect disappeared. The authors attribute this to reduced statistical power, but it is equally consistent with healthy-user bias driving the primary result.
When they applied a one-month lag period to check for reverse causality and detection bias, the first seven months of apparent protection vanished. The effect only appeared from months 8 to 28, suggesting that the early signal in the primary analysis may reflect pre-existing undiagnosed dementia or differential surveillance rather than any vaccine effect.
When they excluded people who had received any flu vaccine in the prior two years, there was no significant difference between high-dose and standard-dose in the per-protocol analysis.
Any one of these findings might be attributable to power loss. Taken together, they describe an effect that is fragile, sensitive to analytic choices, and potentially explained by artifacts of the data rather than by biology.
The Aluminum Question Nobody Asked
Both the high-dose and standard-dose vaccines in this study contain aluminum adjuvants. The high-dose formulation (Fluzone High-Dose) contains four times the antigen of standard-dose vaccines. The adjuvant profiles and total aluminum loads differ between formulations.
The study has no way to measure whether either exposure is neurologically harmful, neutral, or protective in absolute terms. You are seeing a relative comparison within a narrow band of aluminum-adjuvanted vaccine exposure. The possibility that both arms experienced some degree of adjuvant-related neuroinflammatory burden — and that the observed difference reflects something other than “protection” — is entirely consistent with the data.
The paper does not discuss aluminum adjuvant exposure as a variable. It does not cite any of the published literature on aluminum neurotoxicity (Exley, Shaw, Gherardi, Crépeaux, and others). This is a notable omission, though a predictable one given the institutional consensus that aluminum adjuvants are inert. That consensus is itself contested, but that’s a separate discussion.
The structural point stands: an active-comparator design between two aluminum-containing products cannot distinguish neuroprotection from differential neurotoxicity. The data is silent on this question, and so is the paper.
The Headline Machine
Now look at what happened when this study entered the media ecosystem:
Notice the pattern. The outlets with the largest general audiences — tabloids, television news websites, aggregators — uniformly ran the fabricated number. The outlets read primarily by clinicians and researchers got it right or close to right.
This is not a failure of science communication. It is science communication working exactly as designed. The press release creates a simplified narrative. Newsrooms without subject-matter expertise repackage it as a headline optimized for clicks. The headline becomes the public reality. The actual paper — with its fragile effect sizes, its vanishing signals under sensitivity analysis, its complete absence of cognitive data — is read by almost no one.
Eric Topol, quoted in the CIDRAP piece, was the only expert voice in any of this coverage to offer honest context: the evidence here is “not nearly as strong” as the shingles vaccine natural experiments, and if the effect is real, it likely works through general immune stimulation rather than anything flu-specific. That single quote contains more analytical value than the combined output of dozens of newsrooms.
What This Is Really About
This study will now enter the permanent record as evidence that flu shots prevent Alzheimer’s. It will be cited in CDC promotional materials, in pharmacy advertisements, in public health campaigns targeting the elderly. The 55% figure will be repeated until it fossilizes into received wisdom, immune to correction.
The actual finding — a small, fragile, dose-dependent association between influenza vaccine formulation and dementia billing codes in a commercial insurance database, which vanishes under multiple sensitivity analyses and cannot be distinguished from confounding, misclassification, or differential survival — will be forgotten.
This is how the machine works. Not through conspiracy, but through the accumulated weight of institutional incentives:
Researchers are incentivized to publish and to frame findings in the most newsworthy terms possible
Journals are incentivized to accept papers that will generate citations and media attention
University press offices are incentivized to write releases that maximize coverage
Newsrooms are incentivized to write headlines that maximize engagement
Public health agencies are incentivized to amplify anything that supports existing vaccination recommendations
Pharmaceutical companies are incentivized by a market that responds to headlines, not methods sections
At no point in this chain is anyone incentivized to say: “This is a modest observational finding with significant limitations that should not change anyone’s clinical decision-making.”
And so nobody does.
The Wider Context: What the Flu Vaccine Evidence Actually Shows
This study does not exist in a vacuum. It arrives at a moment when the evidence base for influenza vaccination itself is under extraordinary strain.
Cochrane Collaboration reviews — the acknowledged gold standard of evidence-based medicine — have repeatedly found limited or no conclusive benefit for influenza vaccination in the very populations targeted by this study. For the elderly, Cochrane concluded that “the available evidence is of poor quality and provides no guidance regarding the safety, efficacy or effectiveness” of flu vaccination for those 65 and older — even after reviewing 75 studies spanning four decades. In healthy adults, a separate Cochrane analysis found only “modest” effects on symptoms, with “no evidence that they affect complications, such as pneumonia, or transmission.”
Real-world data have proven equally sobering. A 2024–2025 study of over 53,000 Cleveland Clinic employees found negative vaccine efficacy — the vaccinated group experienced a 27% higher flu infection rate than the unvaccinated, yielding a calculated vaccine effectiveness of -26.9%. While this was a preprint, it echoes a pattern observed as far back as the 2009 pandemic, when Canadians who had received the prior seasonal flu shot experienced higher rates of pandemic H1N1 illness.
The structural absurdity runs deeper still. As I documented in “WHO’s Latest Prescription for Failure,” the WHO selects three influenza strains by committee months before deployment, for a pathogen landscape comprising over 200 respiratory viruses — of which influenza A and B represent roughly 10%. The 2025–2026 season proved the point yet again: H3N2 subclade K emerged after vaccine compositions were already locked. Cochrane’s review of 75 studies spanning four decades of elderly vaccination could not produce clinical guidance. And the only validated safety study on inactivated flu vaccines in children under two is a single trial. This is the evidentiary foundation upon which the Bukhbinder study now asks us to build an entirely new therapeutic indication: Alzheimer’s prevention.
And yet here we are: a vaccine that cannot reliably prevent the disease it was designed for is now being promoted as preventing a completely different disease, on the basis of billing codes. The authors propose several hypothetical pathways: trained immunity, reduced neuroinflammation from fewer infections, modulation of inflammaging. These are interesting ideas, yet they are not evidence.
Meanwhile, the study’s authors — and the media covering them — make no mention of what does have clinical trial evidence for influenza prevention that outperforms flu shots: bovine colostrum. A landmark 2007 clinical trial (the San Valentino study) found colostrum supplementation was at least three times more effective than flu vaccination at preventing influenza episodes. Among high-risk cardiovascular patients, the colostrum group had zero hospitalizations during flu season, while the vaccinated group suffered multiple hospitalizations and one death. A 2010 follow-up in Panminerva Medica confirmed colostrum’s superiority, especially when combined with a probiotic immunomodulator.
These findings were published in peer-reviewed journals. They have never been refuted. They have simply been ignored — because a natural, unpatentable substance that outperforms a multi-billion-dollar pharmaceutical product is not a story that institutional medicine is structured to tell.
The irony is exquisite: the same scientific establishment that converts a fragile billing-code association into global headlines about flu shots preventing Alzheimer’s has, for nearly two decades, maintained near-total silence about direct clinical evidence that a safe, low-cost supplement prevents influenza more effectively than the vaccine. The asymmetry reveals what the system values. It is not evidence. It is narrative alignment.
The Deeper Question: Poisoned, Not Infected?
There is one more dimension to this story that the study’s institutional framing will never reach.
The authors hypothesize that the high-dose vaccine may reduce Alzheimer’s risk by better preventing influenza infection, thereby reducing the systemic inflammation that promotes neurodegeneration. This assumes a clean causal chain: virus enters body → inflammation damages brain → vaccine prevents virus → brain is protected. It is a tidy story. It is also an incomplete one.
What if the relationship between infection, inflammation, and neurodegeneration is not what it appears?
As I have explored at length in my “Poisoned, Not Infected” series, an emerging body of research is revealing that what we have historically attributed to viral pathogenesis may, in many cases, reflect the body’s response to toxic environmental exposures. When cells are damaged by chemical toxins — whether pharmaceutical drugs, pesticides, heavy metals, or adjuvants — they release extracellular vesicles (exosomes) that are functionally and morphologically indistinguishable from virus particles. These exosomes carry molecular signatures of injury: oxidative stress markers, inflammatory cytokines, apoptotic signals. They propagate distress from cell to cell, tissue to tissue. And in doing so, they can mimic the symptoms and even the diagnostic markers of infectious disease.
The implications for a study like Bukhbinder et al. are profound.
If both vaccine formulations introduce aluminum adjuvants into elderly patients — a population already burdened by decades of accumulated toxic exposures, compromised blood-brain barrier integrity, and immunosenescence — then the observed “dementia diagnoses” in both arms of the study may not be measuring viral susceptibility at all. They may be measuring the downstream consequences of cumulative toxic burden, of which the vaccine itself is one contributor among many.
The study compares two groups of people who have both been exposed to known neurotoxic adjuvants, at different doses, and observes a small difference in dementia billing codes. From within the vaccine-centric paradigm, this is interpreted as evidence that more antigen equals more protection. From within a toxicological framework, the same data could be interpreted entirely differently: different aluminum loads, different inflammatory cascades, different rates of exosome-mediated neuroinflammatory signaling — all producing slightly different rates of cognitive decline that have nothing to do with influenza prevention and everything to do with differential toxic burden.
This is not a fringe interpretation. It is what the data structurally permits. The study cannot distinguish between these explanations because it was not designed to ask these questions. It was designed within a paradigm that assumes vaccines are inert except for their intended immune effect, that aluminum adjuvants are biologically irrelevant, and that any observed benefit must flow from infection prevention. These are assumptions, not findings.
The deeper truth the Poisoned Not Infected framework illuminates is this: when we measure “disease” using insurance billing codes, when we define “protection” as fewer diagnostic labels appearing in a claims database, when we attribute causality to one variable (vaccine dose) while ignoring the toxic properties of the intervention itself — we are not doing science. We are performing a ritual of confirmation within a closed system of belief.
The body does not operate according to the logic of claims databases. It responds to the totality of what enters it — the antigens, the adjuvants, the preservatives, the cumulative environmental load, the nutritional substrate, the stress milieu. A study that measures none of these things and claims to have discovered that a flu shot prevents Alzheimer’s is not illuminating biology. It is generating a marketing asset.
And the headlines prove it.
The Bottom Line
The Bukhbinder et al. study is a competently executed pharmacoepidemiological analysis that found a small, fragile association in claims data. It is a reasonable contribution to the hypothesis-generating literature. It does not demonstrate that flu shots prevent Alzheimer’s. It does not demonstrate that high-dose flu shots protect the brain. It cannot distinguish neuroprotection from differential neurotoxicity, confounding, misclassification, or healthy-user bias.
The 55% risk reduction now circulating in global headlines is a fabrication — a number that exists in no table, no figure, and no supplementary appendix of the published paper.
The gap between what this study shows and what hundreds of millions of people now believe it shows is not an accident. It is the predictable output of a system that converts scientific uncertainty into commercial certainty, one headline at a time.
And the questions this system will never ask — about the neurotoxicity of the interventions themselves, about the natural alternatives that outperform them, about the possibility that we are measuring poisoning rather than protection — remain, as always, outside the frame.
It is up to us to put them back in.
For over 100 potential adverse effects of the influenza vaccine documented in the peer-reviewed literature, visit the GreenMedInfo influenza vaccine research database. For research on over 221 evidence-based natural approaches to influenza prevention, mitigation, and treatment, visit the GreenMedInfo influenza databasse
To me it says, "Don't worry, we're still hard at work trying to kill you and reduce the population."
Thank you for this brilliant analysis! If only we could get all pro-vaxxers to read this so they would see how all their “science“ is actually manipulated to provide misleading and even false proclamations!