How might vaccines Protect against Dementia?

By /
The mechanism question: is the shingles vaccine protecting your brain from the virus, the adjuvant, or something else?
The Vaccine–Brain Connection · Post 2 of 5

The mechanism question: is the shingles vaccine protecting your brain from the virus, the adjuvant, or something else entirely?

Post 1 made the case that shingles vaccination is consistently linked to lower dementia risk. What it couldn’t answer — because the literature can’t either, not yet — is why. There are three serious explanations on the table. Only one has a piece of evidence that’s genuinely hard to argue with.

Understanding mechanism matters more here than it usually does. If the protection works mainly by suppressing VZV reactivation, then what matters most is the durability of vaccine immunity. If it’s the AS01 adjuvant doing something directly to neuroinflammation, that reshapes which vaccines we should be paying attention to. And if it’s largely a statistical artefact of how dementia gets diagnosed, and who tends to get vaccinated, the clinical picture looks quite different.

There are currently three serious explanations circulating. They’re not mutually exclusive, but they predict different things, draw on different evidence, and point toward different conclusions. Rather than letting the most recent paper set the terms of the debate, I want to look at all three side by side.

Three narratives, one consistent signal

Before getting into the competing explanations, it’s worth being clear about what we’re actually trying to explain. Across multiple studies, including the Welsh and Australian natural experiments, shingles vaccination is consistently linked to lower dementia risk in older adults. This was first shown with Zostavax, a live-attenuated vaccine with no adjuvant beyond the live virus itself. Shingrix, the recombinant vaccine, seems to do better. That gap between the two is itself a clue, and different explanations interpret it differently.

Narrative 1
It’s the virus

The oldest and most direct explanation: VZV reactivation causes neurological damage, and vaccinating against reactivation prevents it.

VZV is neurotropic. It establishes lifelong latency in sensory ganglia and can reactivate as cellular immunity wanes with age. The consequences range from the familiar painful rash of shingles to less visible neurological injury: vasculopathy, encephalitis, and neuroinflammatory cascades linked to dementia pathology. Cell culture and animal model work, including Cairns, Itzhaki and Kaplan (2022) specifically on VZV’s potential to reactivate latent HSV-1 and trigger Alzheimer-like pathological features, gives this a biologically coherent foundation. There’s also the hypothesis that subclinical reactivation — without any rash — may contribute chronic low-grade neurological insult over years, never announcing itself clinically. That remains a hypothesis: Polisky et al. note explicitly that their study only captured clinically manifested shingles.

Most important evidence for this narrative Polisky et al., Nature Medicine (2025): 100M+ US EHRs, 2007–2023. Recurrent HZ episodes associated with modestly higher dementia risk (~7–9%) — dose-response consistent with viral mechanism. Both Zostavax and Shingrix outperform PPSV23 active comparator. Notably: among Zostavax recipients over 15 years, dementia protection waned in a similar pattern to anti-HZ protection (Pearson r = 0.59). This correlation is derived from serially autocorrelated time series and should be treated as suggestive rather than definitive — but the directional coherence over 15 years is notable.
Narrative 2
It’s the adjuvant

This one gained significant traction in 2025. It centres on the AS01 adjuvant system in Shingrix and in Arexvy, GSK’s RSV vaccine. AS01 contains monophosphoryl lipid A (MPL), which activates toll-like receptor 4 (TLR4), and QS-21. Together they trigger a cytokine cascade that ends in interferon-gamma (IFN-γ) production. Animal model data suggest TLR4 stimulation by MPL improves Alzheimer’s pathology in mice, and that IFN-γ may slow amyloid plaque deposition. The pathway is biologically coherent, though it’s worth noting that mouse AD models have a poor track record of translating to humans.

The key epidemiological argument comes from Taquet et al. in npj Vaccines (2025): both AS01-containing vaccines were linked to reduced dementia risk within 18 months, with no significant difference between them. If the effect were purely about preventing VZV reactivation, the RSV vaccine shouldn’t show a similar signal. The shared component is AS01.

Important caveat on this evidence The RSV vaccine code in EHR data doesn’t reliably distinguish between Arexvy (AS01-containing) and Abrysvo (non-AS01). The authors estimate roughly 76% Arexvy after exclusions and assumptions, so some dilution of adjuvant specificity is likely. The inference “therefore AS01” is weaker than a clean head-to-head comparison would support. A methodological critique was published by Williams et al. in npj Vaccines (December 2025), with a reply from the original authors.
Narrative 3
It’s a general vaccine effect

The most sceptical of the three. Williams and colleagues (December 2025) argue the evidence doesn’t exclude a much simpler explanation: that vaccination in older adults reduces apparent dementia risk through mechanisms unrelated to any specific vaccine or adjuvant.

The argument runs like this: vaccination prevents severe acute illness and hospitalisation, which is a well-recognised trigger for rapid cognitive decline in people with pre-existing subclinical cognitive impairment. Someone with undiagnosed mild cognitive impairment who has a bad infection may get a dementia diagnosis that would otherwise have come later. Since the endpoint in these studies is dementia diagnosis rather than underlying neuropathology, healthcare utilisation and diagnostic timing are plausible contributors, at least for shorter time horizons. Multiple vaccine types show associations with reduced dementia incidence regardless of adjuvant content.

Note on the Williams et al. critique The methodological points are legitimate and deserve engagement on their merits. One specific concern: the influenza vaccine comparator used in Taquet’s AS01 analysis may itself have some protective association with dementia in the literature, which would attenuate rather than eliminate any true adjuvant signal. These are appropriate design-level critiques of the observational evidence, not claims that the association doesn’t exist.

Weighing the narratives: a structured assessment

The debate between these three positions has played out through published papers, critiques, and replies, without anyone really stepping back to apply the same evidentiary bar to all three simultaneously. Existing frameworks weren’t built for this: Bradford Hill addresses a single putative cause, GRADE rates intervention evidence, and the IOM/WHO causality frameworks were designed for adverse event assessment. So here’s my attempt to do that comparison directly, using five criteria applied identically across all three narratives.

Mechanistic plausibility assessment
Five criteria applied identically to each narrative. Ratings reflect the current evidence base — not the inherent plausibility of the underlying biology.
Criterion Narrative 1
VZV reactivation		 Narrative 2
AS01 adjuvant		 Narrative 3
General vaccine effect
Biological plausibilityCredible pathway with experimental support? Strong
VZV neurotropism well-established. Cairns/Itzhaki cell models link VZV to Alzheimer-like pathological features. Plausible subclinical reactivation hypothesis.
 Partial
TLR4/IFN-γ/amyloid pathway coherent in animal models. Mouse AD models have poor translational track record.
 Partial
Acute illness → cognitive decompensation mechanism biologically plausible and clinically observed. Non-specific rather than vaccine-specific.
Timescale coherenceMechanism operates on the observed timescale? Mixed
Short-end challenge: protective signal within 12–18 months is not straightforward to attribute solely to accumulated shingles prevention. Long-end: 15-year waning shows a similar directional pattern to anti-VZV protection (r = 0.59), though this correlation is derived from serially autocorrelated time series.
 Partial
Short-term immune modulation explains early signal plausibly. But the adjuvant narrative cannot easily explain the 15-year waning correlation. No saturation effect predicted by immune-modulation pathway.
 Partial
Explains short-horizon effects well (prevented hospitalisations → delayed diagnoses). Does not explain why dementia protection wanes in a similar pattern to anti-VZV protection specifically.
Biological gradientDose-response relationship predicted and observed? Strong
Recurrent HZ → modestly higher dementia risk (~7–9%). Two doses of Shingrix predict lower risk than one. Shingrix outperforms Zostavax. Consistent gradient across multiple metrics.
 Weak
No additive effect observed when both AS01 vaccines administered (Taquet 2025). Authors invoke saturation effect — plausible but post-hoc. Difficult to predict from immune-modulation pathway.
 Neutral
No clear dose-response prediction. Gradient data does not specifically support or refute this narrative.
Mechanistic exclusivityEvidence supporting this narrative, hard to explain under competitors? Strong
The 15-year waning pattern (dementia protection declining alongside anti-VZV protection) is the most distinctive finding for this narrative. The Pearson r = 0.59 correlation is derived from serially autocorrelated time series and should be treated as suggestive rather than definitive — but a confounding process mimicking it tightly over 15 years would be an unusually high bar.
 Partial
RSV vaccine showing similar protection to Shingrix is hard to explain under a pure VZV-specific mechanism. But vaccine misclassification (~24% may be Abrysvo/non-AS01) dilutes the inference. Active debate ongoing.
 Weak
Multiple vaccine types showing associations is consistent with, but does not specifically require, a general vaccine mechanism. Does not explain waning correlation or dose-response patterns.
Evidential directnessHuman experimental > human observational > animal > theoretical Strong
Multiple converging lines of direct human evidence: natural experiments (Wales, Australia), large EHR studies, dose-response, 15-year waning data. Animal/cell model support adds mechanistic coherence.
 Weak
Primary mechanistic evidence (TLR4/amyloid) is animal model only. Human evidence inferred from indirect observational comparisons complicated by vaccine misclassification. No direct human mechanistic evidence.
 Partial
Methodological critique has force at the design level. But diagnostic endpoint concern applies to all three narratives — it cannot selectively explain VZV-specific patterns in the data.
Overall weight of evidence Most support. Waning pattern is suggestive and directionally coherent, with important methodological caveats. Dose-response patterns consistent throughout. Not fully coherent at short timescales — mechanisms likely coexist. Genuine argument from RSV vaccine finding. Primary mechanistic evidence remains animal model only. Misclassification and lack of dose-response are concerns. Methodologically legitimate. Does not account for the specificity of waning patterns or dose-response. Applies to all three narratives equally.

This assessment applies the same evidentiary bar to each narrative. Ratings reflect current evidence, not biological plausibility in the abstract. Mechanisms are likely not mutually exclusive — the table identifies which has the strongest current support, not which is the sole explanation.

Shareable graphic · The waning correlation
The Vaccine–Brain Connection · Post 2 · diplomaticimmunity.blog
A key data point in the mechanistic debate — with important caveats
Polisky et al., Nature Medicine 2025 · 100M+ US health records · 15-year follow-up of Zostavax recipients
Years 1–5
Zostavax provides substantial protection against both shingles and dementia. Anti-VZV immunity is high.
Years 5–10
Zostavax protection against shingles begins to wane — a well-documented finding. Dementia protection wanes in a similar pattern.
Years 10–15
Anti-VZV protection largely lost. Dementia protection is largely lost over the same period. Pearson r = 0.59 between the two waning curves — suggestive, but derived from serially autocorrelated data.
As Zostavax lost its ability to prevent shingles, dementia protection waned — over the same 15-year period.

This pattern is consistent with what the VZV-reactivation hypothesis predicts. The Pearson r = 0.59 correlation should be treated as suggestive rather than definitive — it is derived from serially autocorrelated time series, and shared estimation machinery can induce correlated trends. That said, a confounding process replicating this directional coherence over 15 years would need to track anti-VZV protection unusually tightly.

Polisky et al. Nature Medicine 2025 · doi:10.1038/s41591-025-03972-5 Helen Petousis-Harris · diplomaticimmunity.blog

Where does this leave us?

Applying these criteria to the available evidence, the VZV narrative currently has the most going for it. The waning pattern is suggestive and directionally coherent, though it carries methodological caveats that prevent treating it as definitive. The dose-response relationships — more reactivation episodes associated with more dementia, more vaccine doses associated with less — fit a viral mechanism in ways that aren’t easily explained by adjuvant effects or general vaccine effects alone.

The AS01 hypothesis remains genuinely interesting and shouldn’t be written off. The RSV vaccine finding is surprising under a pure VZV mechanism, and the short-term protection timescale argument is a legitimate challenge. But the mechanistic evidence for AS01’s direct neuroprotective effects in humans is still indirect, and the vaccine misclassification issue in the main supporting study warrants caution.

The general vaccine effect narrative raises a real methodological concern. Multiple vaccines do seem to reduce apparent dementia risk through non-specific pathways. But it doesn’t explain the specific patterns of protection, dose-response, and 15-year waning that make the shingles data distinctive.

The most defensible position right now: these mechanisms probably aren’t mutually exclusive. VZV reactivation is almost certainly part of the story. Whether AS01 provides additional neuroprotective benefit on top of that, through direct immune modulation rather than antigen-specific protection, is a genuinely open question the current evidence can’t resolve.

Designing a study that could answer it — comparing AS01-adjuvanted versus non-adjuvanted formulations of the same antigen in a natural experiment or randomised framework — remains one of the most important outstanding questions in this field.

What this means for clinical conversations

The good news is that the mechanistic debate doesn’t need to be resolved before it has clinical relevance. If protection is primarily through VZV suppression, Shingrix’s superior durability over Zostavax is a strong argument for the recombinant vaccine on brain protection grounds, not just shingles prevention. If AS01 does contribute independently, that strengthens the case further. And if general vaccine effects are part of the picture, the already robust shingles evidence becomes easier to act on, not harder.

The evidence we have now is sufficient to inform clinical conversations. The mechanistic debate matters — not least because this field is moving fast and the next few studies could be decisive — but it’s a reason to stay engaged with the science, not to wait before acting on what we already know.

What remains unresolved No study has yet directly isolated the adjuvant effect from the antigen-specific effect in humans. The ideal design would compare an AS01-adjuvanted formulation against an identical antigen formulation without AS01. That study doesn’t exist in the published literature for any of the relevant vaccines. Until it does, the AS01 hypothesis stays in the “interesting and plausible, but not established” column.
Shareable graphic · Three narratives, one effect
The Vaccine–Brain Connection · Post 2 · diplomaticimmunity.blog
Why does the shingles vaccine protect against dementia?
Three competing explanations — not mutually exclusive, but with different levels of current evidential support
Narrative 1
VZV reactivation
Vaccine prevents shingles virus from damaging the brain. 15-year waning of dementia protection tracks loss of anti-VZV immunity. Dose-response present throughout.
Most current support
Narrative 2
AS01 adjuvant effect
Shingrix’s adjuvant system activates immune pathways that reduce neuroinflammation independently of VZV. RSV vaccine (same adjuvant) shows similar signal.
Interesting, not yet established
Narrative 3
General vaccine effect
Any vaccination reduces severe illness and hospitalisation, delaying dementia diagnosis. Diagnostic endpoint rather than neuropathological endpoint may amplify signal.
Methodologically valid, insufficient

Mechanisms are likely not mutually exclusive. The honest verdict: VZV suppression is almost certainly contributing; whether AS01 provides additional benefit is an open question; general vaccine effects are real but the specificity of the waning pattern and dose-response data is not easily explained by non-specific mechanisms alone.

Polisky et al. Nat Med 2025; Taquet et al. npj Vaccines 2025; Williams et al. npj Vaccines 2025 Helen Petousis-Harris · diplomaticimmunity.blog

Primary sources

  1. Eyting M et al. A natural experiment on the effect of herpes zoster vaccination on dementia. Nature. 2025;641:438–446. doi:10.1038/s41586-025-08800-x
  2. Xie M, Eyting M et al. The effect of shingles vaccination at different stages of the dementia disease course. Cell. 2025;188(25):7049–7064. doi:10.1016/j.cell.2025.11.007
  3. Taquet M, Dercon Q, Todd JA, Harrison PJ. The recombinant shingles vaccine is associated with lower risk of dementia. Nature Medicine. 2024;30:2777–2781. doi:10.1038/s41591-024-03201-5
  4. Taquet M, Todd JA, Harrison PJ. Lower risk of dementia with AS01-adjuvanted vaccination against shingles and respiratory syncytial virus infections. npj Vaccines. 2025;10:130. doi:10.1038/s41541-025-01172-3
  5. Williams SE, Luisi K, Liang C, Cane A, Begier E. Methodological issues in Taquet et al.’s analysis preclude any conclusions regarding AS01 adjuvant’s specific role in dementia prevention. npj Vaccines. 2025;10:255. doi:10.1038/s41541-025-01309-4
  6. Taquet M, Todd JA, Harrison PJ. Reply to Williams et al. npj Vaccines. 2025;10:256. doi:10.1038/s41541-025-01312-9
  7. Polisky V et al. Varicella-zoster virus reactivation and the risk of dementia. Nature Medicine. 2025;31:4172–4179. doi:10.1038/s41591-025-03972-5
  8. Pomirchy M et al. Herpes zoster vaccination and dementia occurrence. JAMA. 2025;333(23):2083–2092. doi:10.1001/jama.2025.5013
  9. Cairns DM, Itzhaki RF, Kaplan DL. Potential involvement of varicella zoster virus in Alzheimer’s disease via reactivation of quiescent herpes simplex virus type 1. Journal of Alzheimer’s Disease. 2022;88:1189–1200.
  10. Michaud JP et al. Toll-like receptor 4 stimulation with the detoxified ligand monophosphoryl lipid A improves Alzheimer’s disease-related pathology. PNAS. 2013. doi:10.1073/pnas.1215165110
  11. Chakrabarty P et al. IFN-gamma promotes complement expression and attenuates amyloid plaque deposition. J Immunol. 2010;184:5333–5343.
  12. Ogunjimi B et al. Varicella zoster virus and the central nervous system. Nature Reviews Microbiology. 2026. doi:10.1038/s41579-026-01289-9

About this series

The Vaccine–Brain Connection examines what the evidence actually shows about vaccines and dementia risk — and why that evidence isn’t reaching the people who need it.

  • Post 1: The evidence — what do we actually know?
  • Post 2: The mechanism — virus, adjuvant, or something else?
  • Post 3: The women’s story — why the strongest finding is being ignored
  • Post 4: The regulatory paradox — why nobody’s allowed to tell you
  • Post 5: The counter-narrative gap — and who fills it
Infographic — Why does the shingles vaccine protect the brain?
The Vaccine–Brain Connection · Post 2 of 5
Why does the shingles vaccine protect against dementia?
Three competing explanations — and the evidence that separates them
Narrative 1
It’s the virus
VZV reactivation causes neurological damage. The vaccine prevents reactivation — and the brain injury that follows. Dose-response patterns and 15-year waning data both support this.
Most current support
Narrative 2
It’s the adjuvant
Shingrix’s AS01 adjuvant activates immune pathways (TLR4 → IFN-γ) that may reduce neuroinflammation independently of VZV. The RSV vaccine — same adjuvant — shows a similar signal.
Interesting · not established
Narrative 3
It’s a general effect
Any vaccination reduces severe illness and hospitalisation, delaying dementia diagnosis. The endpoint is diagnosis, not neuropathology — which may amplify the apparent signal.
Valid · insufficient
A key finding — with caveats
Dementia protection waned in a similar pattern to anti-VZV immunity — over 15 years
As Zostavax lost its ability to prevent shingles, it lost its ability to protect against dementia — over the same 15-year period. This pattern is consistent with the VZV hypothesis, though the correlation is derived from serially autocorrelated time series and should be treated as suggestive rather than definitive.

Polisky et al., Nature Medicine 2025 · 100M+ US health records
Zostavax recipients · protection over time
0 5 yr 10 yr 15 yr
Anti-VZV immunity
Dementia protection
r = 0.59
Correlation between the two waning curves · suggestive, not definitive
These mechanisms are probably not mutually exclusive. VZV suppression is almost certainly part of the story. Whether AS01 provides additional neuroprotective benefit through direct immune modulation is a genuinely open question — and an important one.
diplomaticimmunity.blog · Helen Petousis-Harris
Polisky et al. Nat Med 2025; Taquet et al. npj Vaccines 2025

Discover more from

Subscribe to get the latest posts sent to your email.

Related Posts

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top

Discover more from

Subscribe now to keep reading and get access to the full archive.

Continue reading