Predicting Alzheimers & Dementia (and minimizing risk)

John_Hemming · May 15, 2026, 3:06pm

All of this confused me until I looked at the paper:

Lithium replacement therapy

The observation that Li is sequestered by amyloid deposits in MCI and AD prompted a search for therapeutic Li salts with reduced amyloid binding. We reasoned that the electrostatic interaction of the Li ion with Aβ deposits would be a function of the ionization capacity of the salt, and that Li salts with reduced ionization might show reduced amyloid sequestration. To assess ionization directly, we measured the conductivity of 16 lithium salts. Inorganic Li salts, including the clinical standard lithium carbonate (Li2CO3, hereafter LiC), showed significantly elevated conductivity, indicative of increased ionization, relative to organic Li salts (P = 8 × 10−4; Fig. 5a and Extended Data Fig. 7a). Of the organic Li salts, lithium orotate (C5H3LiN2O4, hereafter LiO) showed the lowest conductance across a broad Li concentration range (Fig. 5a and Extended Data Fig. 7a) and was therefore selected for further comparison with the clinical standard LiC.

L_H · May 16, 2026, 6:43am

I thought id ask Claude for a summary of the evidence for/against rapamycin in prevention of Alzheimer’s…

Rapamycin and Alzheimer’s — Evidence Review

The Evidence Hierarchy — Honest Framing Upfront

No completed human RCT for AD prevention exists. What exists is:

Strong mechanistic rationale
Compelling animal data (among the most replicated in neurodegeneration research)
Emerging human observational and early trial data
Indirect human evidence from mTOR pathway genetics

The animal-to-human translation gap is real but narrower here than in many areas, because the pathological mechanisms targeted (tau, amyloid, neuroinflammation) are conserved and the drug’s human PK is well understood.

1. Animal Evidence — The Core Case

Tau pathology:

The most replicated finding in rapamycin neurodegeneration research. Across multiple independent labs and multiple tauopathy mouse models:

Rapamycin reduces soluble hyperphosphorylated tau, reduces neurofibrillary tangle burden, and preserves cognitive function in PS19 (P301S tau), rTg4510, and 3xTg-AD mice
Mechanism: autophagy-mediated clearance of soluble/oligomeric tau before aggregation — not clearance of mature NFTs (which are largely autophagy-resistant)
The temporal window matters critically — rapamycin is effective when started before or during early tangle formation; less effective after extensive NFT deposition. This is the central argument for early/preventive use
Caccamo et al. (2010, Journal of Biological Chemistry) — landmark paper showing rapamycin reduces tau pathology and cognitive deficit in 3xTg-AD mice; reduction in p62/SQSTM1 confirmed autophagy as mechanism
Majumder et al. (2011) — showed both tau and amyloid pathology reduced with rapamycin in same model

Amyloid pathology:

Rapamycin reduces Aβ42 levels and plaque burden in APP/PS1 and 3xTg-AD models
Mechanism is dual: autophagy-mediated APP intermediate clearance + BACE1 suppression via 4EBP1 (the translation suppression effect discussed earlier — your elevated Cmax this cycle is specifically relevant here)
Spilman et al. (2010, PLoS ONE) — rapamycin begun at 9 months in PDAPP mice reduced Aβ levels and reversed cognitive impairment. Notably used a late-intervention design, strengthening translational relevance

Cognitive restoration in aged non-transgenic mice:

Arguably more relevant for prevention than transgenic AD models:

Harrison et al. (2009, Nature) — the landmark ITP study showing rapamycin extends lifespan in aged mice (started at 20 months equivalent). Cognitive benefits were a secondary finding
Halloran et al. (2012) — rapamycin reversed age-related cognitive decline in normal aged mice without AD transgenes — relevant to the question of whether it preserves cognition through general neuronal autophagy rather than specifically anti-amyloid mechanisms

2. The mTOR-AD Connection — Pathological Evidence

Independent of rapamycin studies, the mTOR pathway is directly implicated in AD pathology in human post-mortem data:

mTORC1 is hyperactivated in AD brains — multiple post-mortem studies show elevated p-S6K1, p-4EBP1, p-S6 in hippocampal and cortical neurons of AD patients vs controls
This hyperactivation correlates with tangle density and cognitive decline severity, not just presence of pathology
mTOR hyperactivation impairs autophagy → accumulation of tau and Aβ → further mTOR activation (tau oligomers activate mTOR) — a feed-forward pathological loop
Cai et al. showed that mTOR activity inversely correlates with autophagy flux markers in human AD hippocampus — the pathway is causally implicated, not merely associated

This is important: it means rapamycin is targeting a mechanism that is demonstrably dysregulated in human AD pathology, not merely a theoretical target extrapolated from mouse models.

3. Genetics — Indirect Human Evidence

Several lines of genetic evidence converge on mTOR pathway:

APOE4 and mTOR:

APOE4 (strongest genetic AD risk factor) is associated with enhanced mTOR signalling in neurons
APOE4 astrocytes show impaired autophagy vs APOE3 — partly mTOR-mediated
Rapamycin in APOE4 neuronal models partially corrects the autophagy deficit

PTEN and TSC mutations:

Loss-of-function PTEN mutations (mTOR activating) are associated with neurodegeneration phenotypes
TSC1/2 mutations (mTOR hyperactivation) produce severe neurological disease — Tuberous Sclerosis — with tau and autophagy pathology, partially responsive to rapamycin/everolimus in human studies

mTOR SNPs:

Several GWAS studies identify mTOR pathway genes in AD risk loci — though these are not yet cleanly mechanistically interpreted

4. Human Observational Evidence

Transplant recipients on sirolimus/everolimus:

The most direct human data available. Immunosuppressed transplant patients on mTOR inhibitors provide a natural experiment:

Multiple studies show transplant patients on mTOR inhibitors (vs calcineurin inhibitors) have lower rates of cognitive decline in long-term follow-up
Confounded by indication, comorbidity, and concurrent immunosuppression — but directionally consistent across multiple cohorts
Notably: these are therapeutic (higher) doses than longevity protocols — suggesting even with significant systemic immunosuppression the CNS signal is detectable

Metformin/berberine AMPK data:

Indirect but relevant — AMPK activators (mechanistically overlapping with rapamycin via TSC2) show protective associations with dementia in diabetic cohorts (the metformin-AD epidemiology literature). This supports the mTOR/AMPK pathway hypothesis in humans without being rapamycin-specific.

5. Emerging Human Trial Data

PEARL Trial (rapamycin for AD prevention):

Launched ~2022-2023, University of Texas Health Science Center
Targeting cognitively normal adults with AD biomarker risk (amyloid PET, CSF tau)
Rapamycin vs placebo, primary endpoints including CSF biomarkers and cognitive measures
Results not yet published as of mid-2025 — this will be the most important data point in the field when it emerges

REACH (Rapamycin Efficacy in Alzheimer’s and Cognitive Health):

Separate trial at multiple US sites
Similar design targeting early/pre-symptomatic population
Also ongoing

Everolimus trials:

Several small trials using everolimus (oral rapamycin analogue, more consistent bioavailability) in MCI and early AD
Small n, mixed results on cognitive endpoints — but biomarker data (CSF Aβ42/40, p-tau) showing trends toward improvement in some cohorts

The fact that multiple independent research groups have launched human trials reflects the strength of the preclinical signal — this is not a fringe hypothesis.

6. Neuroinflammation Arm — The Underappreciated Mechanism

Beyond tau and amyloid directly:

Microglial mTOR hyperactivation in aged and AD brains drives a chronic neuroinflammatory phenotype — elevated IL-1β, TNF-α, complement activation, excessive synaptic pruning
Rapamycin in aged mice normalises microglial morphology and cytokine secretion — the immune rejuvenation effect
This is potentially as important as the direct neuronal autophagy effect — chronic neuroinflammation is now considered a primary driver of AD progression rather than a secondary phenomenon
The Treg expansion discussed earlier (rapamycin’s paradoxical immunostimulatory effect on regulatory T cells) also dampens neuroinflammation peripherally, with CNS consequences via reduced peripheral immune cell infiltration

7. Limitations and Honest Caveats

Concern	Severity	Context
No completed human prevention RCT	High	PEARL/REACH results awaited
Mouse AD models poorly translate historically	Moderate	But mTOR mechanism is conserved; post-mortem human data supports pathway
Optimal dosing/timing for CNS unknown	High	Longevity protocols extrapolated from PK, not CNS pharmacodynamic endpoints
Rapamycin may acutely suppress neurogenesis	Moderate	mTOR required for hippocampal neurogenesis — pulse dosing likely mitigates vs chronic
Metabolic side effects at higher doses	Low at your dosing	Insulin resistance risk is real at therapeutic doses; less relevant at longevity doses
APOE4 genotype modifies response	Unknown	You haven’t mentioned your APOE status — this is worth knowing

Your APOE Status — Worth Flagging

The rs429358 SNP you sought to look up via impute.me in April is the defining APOE4 variant. Your AD risk profile and the magnitude of benefit from mTOR inhibition are both substantially modified by APOE genotype:

APOE4/4: highest AD risk, strongest argument for rapamycin neuroprotection
APOE3/4: approximately 3x population risk, meaningful benefit plausible
APOE3/3: population risk, benefit still present but less compelling urgency
APOE2/3: below population risk

If you got your rs429358 result, it directly stratifies how aggressive your rapamycin protocol is justified neurologically.

Bottom Line Assessment

The case for rapamycin in AD prevention is the strongest in longevity pharmacology after lifespan extension data — arguably stronger than the case for any other current intervention because:

The target pathway (mTOR hyperactivation) is demonstrably dysregulated in human AD pathology
The mechanism (tau/amyloid autophagy clearance, microglial normalisation) is pathologically specific, not just generic anti-aging
Animal data is unusually well-replicated across independent labs and multiple model systems
The temporal logic is sound — preventive use targeting soluble tau/Aβ before aggregation is mechanistically coherent and the window you’re currently in (pre-symptomatic) is exactly the right one
Human trials are now running — the field has moved from hypothesis to clinical testing

The honest position: compelling preclinical and mechanistic evidence, awaiting definitive human trial confirmation. For someone already on a biweekly rapamycin protocol, the neurological rationale is among the strongest justifications for continuing it.

L_H · May 16, 2026, 1:17pm

I had another look at the Lancet report. Its interesting what’s not included (yet):

Factors Missing from Lancet 2024 Dementia Prevention Model

Missing Factor	Estimated Impact	Why Absent
Shingles vaccine (Shingrix)	15–25%	Too recent; Welsh natural experiment data 2023
Rapamycin/mTOR inhibition	15–25%	No human AD trial data yet
GLP-1 agonists (semaglutide)	10–20%	Emerging; EVOKE trial ongoing
Viral burden (HSV-1/VZV/CMV)	10–20%	Mechanistically contested, no intervention trial
Sleep architecture quality	15–20%	Duration is in model; staging/slow-wave quality is not
Hormonal timing (HRT/testosterone)	8–15%	Critical timing window hypothesis not yet modelled
PDE5 inhibitors (tadalafil)	12–18%	Retrospective data only
TRE/intermittent fasting	10–15%	Insufficient RCT data in humans
Spermidine	10–15%	Small trials only
Urolithin A/mitophagy	8–12%	Too early stage
Exogenous ketones/C8	8–12%	Symptomatic data only
Sauna (3x+/wk)	8–12%	Observational only; Laukkanen Finnish data
Oral/systemic inflammation (periodontal)	8–10%	P. gingivalis data emerging
Eudaimonic purpose/meaning	5–10%	Hard to operationalise
Gut microbiome diversity	Unknown	No clean intervention data

For late onset Alzheimer’s I’m encouraged to think that the modifiable % could be much higher than 60%. I appreciate we’re guessing here, and these other interventions will certainly not be neatly additive, But the sheer number of possibles with decently encouraging evidence would have me guessing at 90%+ modifiability for a motivated individual.

adssx · May 16, 2026, 6:19pm

It’s funny because yesterday I asked ChatGPT 5.5 Thinking Extended the exact same question and it said it wouldn’t go beyond 55%. Which LLM did you use?

L_H · May 16, 2026, 6:22pm

Claude 4.7 it said 90%+ probable

Maybe AI really does just tell you what you want hear!

u

adssx · May 17, 2026, 5:23am

Ah OK, I understand the discrepancy. You just asked Claude to add potential factors. It listed random BS ones. This is not the same as asking Claude to read the current report, the “Potential risk factors considered with insufficient evidence to include in the model,” and estimate them. Those factors are:

Sleep — short and long sleep duration, shift work, sleep apnoea, and benzodiazepine use. Evidence on short sleep is suggestive but confounded by reverse causation; long sleep is judged not a risk factor (the prodrome causes it). The Commission could not make a recommendation.
Diet — Mediterranean, DASH, and MIND diets; ultraprocessed foods; omega-3s; the gut microbiome.
Dietary interventions — multivitamins, cocoa flavanols, anthocyanins, and MIND-diet counselling trials, all with small/null effects.
Infections and systemic inflammation — hospital-treated infections, inflammatory markers (CRP, IL-6), COVID-19, vaccines, NSAIDs, and antibiotics like minocycline.
Dental disease — including periodontal disease (treated as a subsection of infections/inflammation).
Bipolar disorder
Psychotic disorders, including schizophrenia (and very-late-onset schizophrenia-like psychosis)
Anxiety
Post-traumatic stress disorder
Menopause and hormone replacement therapy
Multimorbidity and frailty

Only the above are likely to join the list of modifiable risk factors within the next decade. Claude Opus 4.7 Adaptive’s conclusion:

A reasonable speculative range for “what would the headline number be if the next Commission accepted all of these”: 48–58% modifiable, with the lower bound being the more honest estimate once overlap with the existing 14 factors is taken seriously.

Read the conversation: Claude

L_H · May 17, 2026, 6:46am

Yes, i asked Claude to read the report but specifically NOT limit itself to the factors included in the Lancet report.

I’m not sure why we would limit this question to the Lancet report. The pharmaceutical sector is certainly not, Billions of dollars are being invested in areas factors not included in the Lancet report. So it would seem logical/obvious that there are many factors outside the limited remit of the Lancet report.

Essentially the Lancet Commission’s approach is conservative because the report is explicitly designed to inform policy and public health action at scale.

It’s not designed to identify the true modifiable vs non-modifiable risk factors of dementia. It wants to identify the low hanging fruit for policy makers.

Your Claude answer seems to be answering the wrong question if you’re trying to understand what the modifiable percentage of dementia will be in the future.

This is the part I most strongly disagree with. We have so many avenues of attack for solving this problem - and with AI assistance i expect huge progress.

adssx · May 17, 2026, 7:04am

Because those are the avenues with the most evidence as of today. Whereas tadalafil or Urolithin A or sauna are BS for instance.

John_Hemming · May 17, 2026, 7:59am

Noting the recent report of an association between lower 25OHD levels and higher dementia (ie vitamin D earlier in life prevents dementia) I thought I would ask a question of chatGPT:

So the best summary is:

Sunnier / higher-UV countries tend to have lower age-adjusted dementia incidence in ecological analyses, but the association is vulnerable to confounding and diagnostic bias. The most plausible biological link is vitamin D plus circadian, sleep, inflammatory and vascular pathways, but causality is not established.

The safest interpretation is: low sunlight exposure is a plausible dementia-risk marker, not yet a proven country-level causal factor.

L_H · May 17, 2026, 8:32am

I think you may have misunderstood what the Lancet Paper is trying to do.

Dismissing the broader avenues (“clues”)(neuroinflammation, microbiome, protein clearance pathways, mitophagy, autophagy, immune modulation etc.) as BS just because the Lancet Report (a deliberately conservative policy document) didn’t include them yet is exactly the naive take I’m pushing back on.

Those broader areas have real mechanistic depth and serious money behind them for a reason. The Lancet identifies today’s public policy low-hanging fruit. It doesn’t try to define the true ceiling of what’s modifiable. It tries to definethe conservative, policy-grade lower bound.

So if the question is, what will the next Lacet Commission say meets this criteria for public policy, sure we should he pessimistic. But as to what is and will be truly modifiable that’s a whole different question.

And I’d bet a lot higher. The remaining 55% isn’t a brick wall, its not “non-modifiable”, it’s a research agenda. Is Urolithin A BS? or does it modify risk of dementia? We dont have hard RCT evidence yet, but I’d guess that its mitophagy affect does modify risk to some degree. But that’s not the point, the point is that mitophagy and brain mitochindrial health is one of many,many avenues/“clues” we have for improving ageing brain function and reducing risk of dementia. So the idea that we have “no clues” and will reach a low ceiling on what’s modifiable seems obviously wrong. Especially in an age of Alphafold and AI driven biology.

adssx · May 17, 2026, 10:41am

I said BS because most of those listed in your post were BS or promising but unproven. Others are promising and are included in the Lancet report (e.g. vaccines).

In any case the discussion was about “modifiable risk factors” and this is what the Lancet paper is about. They’re not conservative, they’re just looking at the real evidence beyond some hypotheses.

So my point is that in 2035 I don’t expect the scientific consensus (that the Lancet represents) to be that modifiable risk factors for dementia are more than 60%. I wish it could be the case but it’s unlikely. I would be willing to bet on that (it’s a safe bet for me because it’s like an insurance against science being slow ). Of course I would prefer to be wrong that we have to be realistic.

By the way, this article is about Alzheimer’s and dementia: for other forms of dementia such as Parkinson’s disease dementia, the modifiable risk factors are closer to 0%.

L_H · May 17, 2026, 2:06pm

I was just making the point that the 14 modifiable risk factors listed in the Lancet paper are those with robust enough evidence to drive policy recommendations. That’s what the Lancet paper “is about” it restricts itself to where there is "Population Attributable Fraction ", trying to represent a scientific consensus of where there’s robust enough evidence to act at a population/policy level. That’s very different from trying to reach a scientific consensus as to biologically modifiable vs mon-modifiable risk factors.

The part we really disagree on though is whether we have “no clue” as to the other modifiable risk factors and therefore whether the modifiable risk factors % will reach a ceiling:

When it seems to me we have a huge number of clues and other factors to explore. A good example is nutrigenomics. I have PEMT TT so my liver doesnt synthesise phosphatidylcholine well and so i rely more on dietary choline (and supplements) than typical. Is that a modifiable risk factor for dementia? phophatidylcholine deficiency is very likely to be a big risk factor for dementia. And its easily modified for me - I eat eggs and take choline tablets. By 2035 I would certainly assume we’ll have much greater understanding around this specific area. But will it ever feature in the Lancet report? I doubt it - because it wont make sense at a population level. “Lack of Nutrigenomics” might feature, but i doubt even that.

I like your bet - its safe for me too because if you’re right probably neither of us will remember the bet!

adssx · May 17, 2026, 2:29pm

If your example was correct then for sure it would appear in the Lancet report because it would be some kind of deficiency. But it’s also possible that in 2035 we’ll learn that choline supplementation or whatever other intervention increases the risk of dementia.

We should create a prediction market

starspawn0 · May 17, 2026, 3:09pm

I asked GPT-5.5-thinking about epidemiological evidence for the effect of sustained ibuprofen use on Alzheimer’s risk reduction:

Its conclusion:

The epidemiologic case for sustained ibuprofen is moderate as an association and weak as causal evidence. The strongest ibuprofen-specific evidence is the large VA observational study showing a >40% lower Alzheimer’s risk after >5 years of ibuprofen use, but that result could be biased. Broader cohort/meta-analytic evidence supports a possible long-duration NSAID association, but not cleanly for ibuprofen alone.

Given the known harms of chronic ibuprofen—higher risks of heart attack, stroke, gastrointestinal bleeding/ulcers, kidney problems, and greater risk with long-term or high-dose use—it should not be used for Alzheimer’s prevention outside medical supervision or a trial.

The “known harms” mostly apply to higher-dose ibuprofen use. I think 100 to 200 mg per day, which would be low-dose, might be safe – especially the 100 mg range. And earlier in GPT-5.5-thinking’s analysis it is written:

Shorter use was not protective, and cumulative dose was not associated with lower risk, suggesting that duration rather than intensity drove the association.

So, you may not get any additional benefit from high-dose anyways, if there is any benefit at all.

…

I will say that I have noticed that low-dose ibuprofen improves my sleep quality and duration, which may be part of the benefit against Alzheimer’s risk (if there is any benefit at all). Acetaminophen does also, but it requires much higher dose to see much benefit. In fact, if I take 2 or 3 extra-strength Tylenol (which is a very bad thing to do, as it probably powerfully harms the liver; but I have done it), my sleep is usually very good.

L_H · May 17, 2026, 5:30pm

I love that idea - and for science more broadly. It allows market economics to highlight areas of most promising research, and it allows impoverished researchers an opportunity to bet on (or against as a hedge) themselves and their research.

I used the choline/PEMT example precisely because I don’t think the Lancet commission would include it - even if the proof was there. Choline supplementation at population level may indeed not be productive (the association seems to be u-shaped) but may be strongly protective if an individual’s underlying genes cause a deficiency.

The Lancet commission would struggle to include choline in its list even if the evidence becomes very high for those with PEMT TT - Because the inclusion of a risk factor that depends heavily on individual genomics would require a fundamental shift in how the Commission calculates “Population Attributable Fraction”.

The one intervention I’d love to have a bet on as being effective though, is low dose lithium. It seems like the effect could be large. The associations from natural water supply, the hypothesized mechanisms and RCT pilot studies all point in the right direction and suggest a big potential affect size. Lithium alone could get us above your prediction - but I dread to imagine the policy debate about adding it to tap water!

desertshores · May 17, 2026, 5:47pm

NSAIDs appear on the Beers Criteria for potentially inappropriate medications in older adults.

Certainly, ibuprofen is not on my list for Alzheimer’s prevention.

GI bleeding—the risk rises steeply with age, and the elderly are more likely to bleed silently until they’re significantly anemic or hemodynamically unstable. Concurrent aspirin, anticoagulants, SSRIs, or corticosteroids compound this.

Cardiovascular — increased MI, stroke, and heart failure exacerbation. Ibuprofen also blunts the antiplatelet effect of low-dose aspirin if taken close together.

Hypertension — NSAIDs can raise BP and antagonize antihypertensive medications.

RapAdmin · May 18, 2026, 3:07am

Researchers have shown for the first time that malfunctioning mitochondria — the cell’s energy generators — may directly cause cognitive decline in neurodegenerative diseases. By creating a new tool that temporarily boosts mitochondrial activity in the brain, scientists restored memory performance in mouse models of dementia. The discovery hints that energy failure inside neurons could happen before brain cells die, potentially offering a new target for future Alzheimer’s treatments.

L_H · May 18, 2026, 7:17am

I really like your list, And thought Id put it through Claude for a ranking and estimate of impact on dementia risk:
DEMENTIA RISK REDUCTION — RANKED INTERVENTIONS
RRR = estimated relative risk reduction · [cond] = conditional on subgroup · [e4] = APOE4-amplified
Evidence: A = RCT/strong prospective · B = cohort/large observational · C = small trials/mechanistic

Intervention Tier Est. RRR Evidence Key notes

  ── TIER 1 — Highest impact, strongest evidence ──

1 HSV antivirals (if HSV+) [cond] 2 Sauna 4–7x/week 3 Vigorous cardio (Zone 4–5) + strength 4 Hearing loss treatment [cond] 5 Alcohol abstinence [e4-amplified] 6 HRT — women, post-menopause [cond] 7 GLP-1 agonists (semaglutide) 8 Sleep optimisation (7–8hr, N3) 9 Diet: keto-Mediterranean intersection 10 Blood pressure control (midlife) 11 Sildenafil / tadalafil (PDE5i) ── TIER 2 — Moderate-high 12 Low-dose lithium 13 Social engagement / cognitive reserve 14 B vitamins (methylated, high dose) [cond] 15 Air quality / PM2.5 avoidance 16 Omega-3 ≥2g/day [e4-amplified] 17 Green tea (3–4 cups EGCG) + coffee 18 Toxic metal elimination (Hg, Pb) [cond] 19 Homocysteine lowering 20 Metformin 21 HBOT (males only) [cond] 22 Testosterone optimisation (males) [cond] 23 Fasting / TRE (independent of diet) ── TIER 3 — Promising, emerging 24 Rapamycin (higher intermittent dose) 25 Spermidine (wheat germ / supplemental) 26 Plasmalogens 27 Intranasal insulin [cond] 28 Melatonin (physiological 0.3–0.5mg) 29 Pioglitazone [cond] 30 MCT C10 / exogenous ketones [e4 note] 31 CoQ10 (ubiquinol) + GG [e4-amplified] 32 NO synthesis pathway (nitrates, 33 40Hz multi-sensory stimulation 34 Lion’s mane (Hericium erinaceus) 35 NMN / NR (NAD+ precursors) 36 Curcumin (Longvida / Theracurmin) 37 Urolithin A (supplemental) 38 Berberine 39 Trehalose 40 Selenium + magnesium L-threonate 1 60–90% B+ Tzeng cohort; Itzhaki mechanistic. HSV1+ only.
1 ~65% B+ Laukkanen/Kuopio prospective. HR ~0.35 vs. 1x/wk.
1 35–45% A– VO2max top quintile ~35% RRR. Strength adds ~10% independently.
1 19–48% A– Largest PAF in Lancet 2020 (~8%). ACHIEVE RCT: 48% slower decline.
1 40–50% B+ Multiplicative interaction in e4. ~15–20% general population.
1 30–50% B+ Bioidentical estradiol, within 5yr of menopause.
1 40–70% B+ Cleveland Clinic 1M+ retrospective. RCTs ongoing.
1 25–35% A– Glymphatic amyloid clearance. U-shaped curve: <6hr and >9hr both elevated.
1 35–53% B MIND diet 53% RRR. Keto adds neuronal ketone substrate.
1 20–30% A SPRINT-MIND RCT. Target <120 systolic.
1 40–69% B Cleveland Clinic 7.2M-person database. cGMP → NO → cerebrovascular.
impact, reasonable evidence ──
2 25–35% B Kessing Danish cohort; Forlenza MCI RCTs. GSK-3β → tau + amyloid. BDNF, autophagy.
2 20–25% B+ Lancet 2020 factor. Bilingualism, new skills, music.
2 30–40% A– VITACOG: 53% reduced brain atrophy. MTHFR variants need 5–10x standard dose.
2 25–35% B ~30–40% elevated risk in high-exposure areas. HEPA filtration.
2 20–30% B e4: re-esterified TG + phospholipid form. Ethyl ester absorption impaired.
2 25–35% B EGCG: amyloid inhibition, autophagy. Coffee ~27% RRR (meta-analyses).
2 25–40% B Test first. Near-zero if unexposed. Chelate only if burden confirmed.
2 20–30% B+ Target Hcy <10 µmol/L. Delivered via B vitamins (#14).
2 20–25% B– AMPK → mTOR suppression. TAME trial ongoing. May blunt exercise adaptation.
2 20–30% B– Efrati 60-session protocol. ↑CBF, ↓senescent cells. Male-specific.
2 15–30% B– Male HRT equivalent. Androgen receptors neuroprotective.
2 15–20% B– Autophagy window distinct from dietary ketosis. Circadian-aligned TRE.
evidence ──
3 20–40% C+ Animal evidence strong. e4: mTOR hyperactivation documented. PEARL trial ongoing.
3 15–25% C+ Kiechl/Graz cohort. EP300 inhibition → autophagy. Wirth RCT: memory improvement.
3 15–25% C+ Temporal cortex depletion precedes AD symptoms. Yanagisawa human trials.
3 15–25% C+ Bypasses BBB and peripheral IR. High-value if insulin resistance present.
3 15–20% C+ Anti-Aβ fibrillogenesis, metal chelation, glymphatic support.
3 15–20% C+ TOMORROW trial genotype-stratified subgroup. PPARγ → brain insulin resistance.
3 10–20% C+ C10 → PGC-1α. Note: MCT response attenuated in e4.
3 10–20% C+ e4 Complex I/III impairment. GG as isoprenoid precursor.
oral biome, MB) 3 10–20% C+ Dietary nitrates + avoid antiseptic mouthwash. MB: mitochondrial electron carrier.
+ intranasal PBM 3 10–20% C Tsai/MIT gamma entrainment. Cognito Therapeutics trials. PBM: cytochrome c oxidase.
3 10–20% C+ NGF + BDNF. Only dietary agent with adult neurogenesis mechanism. Mori MCI RCT.
3 10–20% C+ SIRT1/3 → amyloid + tau biology. NAD+ decline drives multiple AD mechanisms.
3 10–15% C+ UCLA/Small PET: reduced brain amyloid + tau. Standard form not absorbed.
3 10–15% C+ Specific mitophagy. Gut conversion variable — Timeline Mitopure 500–1000mg bypasses this.
3 10–15% C+ AMPK, mitophagy, gut-brain axis, anti-amyloid aggregation.
3 10–15% C TFEB autophagy. Bioavailability caveat: gut trehalase degrades oral form.
3 8–12% C+ Se: GPx4 (ferroptosis). Mg-L-threonate: BBB-crossing form, ↑brain Mg2+.

L_H · May 18, 2026, 7:22am

I then asked Claude to estimate my dementia risk with some or all of these interventions at age 80, 90 and 100

DEMENTIA RISK PROJECTION
Male · APOE ε3/ε3 · Healthy baseline at 53 · UK population
Cumulative incidence if surviving to each age

Stack Age 80 Age 90 Age 100
────────────────────────────────────────────────────────
Baseline (no interventions) ~11% ~26% ~45%
Top 10 interventions ~6% ~13% ~23%
Top 20 interventions ~4% ~10% ~17%
Full stack (all 40) ~3% ~8% ~13%

Sub-additive pathway modelling · 25–35% publication bias discount applied
Realistic RRR: Top 10 ~50% · Top 20 ~62% · Full stack ~70%
Irreducible floor ~2–3% regardless of intervention intensity

Evidence is very far from perfect for all of these - but Im inclined to think this is also very far from BS especially in aggregate.
In aggregate, error bars should be reduced. And the cost side for most of these is relatively trivial (financial, time and side effect risk). I do 26/39 already and that covers most bases. It seems a sensible approach while we wait for Ai-led biology to solve this.

One extra point on rapamycin dosing - I recently tested a higher dose and hit the v. small mouth ulcer level at 9mg. Claude argues that i should use this dose quarterly because of its increased BBB penetration. Claude is guessing at serum sirolimus and cns penetration - but the logic is directionally sound. BBB penetration does seem to non-linearly correlate with serum levels. I’m inclined to go with a quarterly 9mg (normal dose 6mg) for brain-autophagy benefits - any thoughts?

AustraliaLongevity · May 18, 2026, 9:42am

Taken together, our data demonstrated that RJ could reduce cholesterol levels, regulate Aβ levels and enhance neuronal metabolic activities in AD rabbits, providing preclinical evidence that RJ treatment has the potential to protect neurons and prevent AD.