She had undergone genetic testing and learned that she carries one copy of APOE4, the most common genetic risk factor for late-onset Alzheimer’s disease. She wanted to be proactive about prevention.

Her primary care physician had started her on what she called a low-dose “GLP3” drug to lower inflammation and reduce her future risk of dementia.

But her metabolic profile was already favorable. Her BMI was below 25. Her hemoglobin A1c was 4.6%. Her high-sensitivity C-reactive protein, or hsCRP, was undetectable. She did not have diabetes, obesity, or an obvious metabolic abnormality.

Her question was simple:

Is this actually good for my brain?

GLP-1–based drugs have substantial benefits when they treat a biological problem that is present. In obesity, diabetes, insulin resistance, and cardiovascular disease, there is a measurable target.

But what happens when the target is missing?

For a lean, normoglycemic APOE4 carrier, we have no evidence that GLP-1 therapy prevents dementia. At the same time, these drugs suppress appetite and energy intake, cause gastrointestinal side effects, and can reduce both lean mass and fat.

That changes the risk-benefit calculation.

First: What Drug Are We Talking About?

There is no standard medication class called “GLP3.”

Patients may mean a GLP-1 receptor agonist such as semaglutide, a dual GIP/GLP-1 agonist such as tirzepatide, or one of the newer multi-receptor agonists.

The exact molecule, dose, formulation, and source matter, especially with compounded products. But even with an FDA-approved product, the more important question is the indication.

What Problem Are We Treating?

The major GLP-1 trials were conducted in people with identifiable metabolic or cardiovascular risk.

The SELECT trial enrolled 17,604 adults age 45 or older with BMI ≥27 and established cardiovascular disease, but without diabetes. Semaglutide reduced major cardiovascular events in this high-risk population. [1]

STEP 1 enrolled adults with BMI ≥30, or BMI ≥27 with at least one weight-related condition, and demonstrated substantial weight loss with semaglutide. [2]

These are important results, but neither trial studied healthy, normal-weight adults taking GLP-1 therapy for longevity or dementia prevention.

The evidence supports a narrower conclusion:

GLP-1 drugs improve important outcomes when there is metabolic or cardiovascular pathology to treat.

They should not automatically be extrapolated into longevity drugs for metabolically healthy people.

Inflammation Is Not One Biological Process

The argument that GLP-1 drugs might prevent dementia because they “lower inflammation” needs more precision.

Inflammation is not a single biological state. Acute immune responses, autoimmune inflammation, infection-related inflammation, neuroinflammation, and the chronic low-grade inflammation associated with obesity and insulin resistance involve overlapping but distinct pathways.

The inflammatory state most closely linked to the metabolic effects of GLP-1 therapy is metabolic inflammation, sometimes called metaflammation: chronic, low-grade immune activation associated with excess adiposity, insulin resistance, type 2 diabetes, and metabolic syndrome.

In that setting, dysfunctional adipose tissue, altered adipokine signaling, and immune-cell activation contribute to insulin resistance and cardiometabolic disease. Reducing excess adiposity and improving insulin sensitivity can therefore reduce a real inflammatory burden. [10]

But that is different from assuming that any reduction in systemic inflammation is beneficial to the brain.

Neuroinflammation in Alzheimer’s disease is not synonymous with metabolic inflammation, and circulating CRP is not a direct measure of inflammatory activity in the brain.

That distinction is especially relevant in someone who is lean, normoglycemic, and has undetectable hsCRP.

EVOKE Helps Separate Systemic Inflammation From Brain Benefit

The EVOKE and EVOKE+ trials are informative because participants were selected for early symptomatic Alzheimer’s disease—not for obesity, diabetes, insulin resistance, or cardiovascular disease.

Participants were 55 to 85 years old with mild cognitive impairment or mild dementia due to Alzheimer’s disease, objective cognitive impairment, and confirmed amyloid pathology. Mean BMI was about 25.7 kg/m², nearly half had a normal BMI, and only a minority had obesity or type 2 diabetes. [3]

This was very different from the populations in SELECT and STEP 1.

EVOKE and EVOKE+ randomized 3,808 participants to oral semaglutide or placebo. Semaglutide did not slow decline on the primary clinical endpoint, the Clinical Dementia Rating–Sum of Boxes. Figure 1 [3]

Yet semaglutide lowered hsCRP, Figure 2.

That combination is important:

Systemic inflammation decreased, but clinical Alzheimer’s progression did not.

This does not mean inflammation is unimportant in Alzheimer’s disease. Nor does it exclude the possibility that treating obesity, insulin resistance, diabetes, or vascular disease earlier in life could indirectly reduce dementia risk.

But it shows that lowering a systemic inflammatory marker with a GLP-1 receptor agonist, in a population not selected for major metabolic disease, was not sufficient to produce cognitive benefit.

For a lean patient whose hsCRP is already undetectable, lowering CRP further has no demonstrated meaning for brain protection yet.

But the more relevant question is:

What inflammatory process is present, and is that process actually being treated?

The Lean-Mass Problem

For a lean APOE4 carrier, the concern is not simply weight loss. It is catabolism.

GLP-1 drugs reduce appetite and energy intake. In a person with obesity, that can be beneficial because much of the lost weight is excess fat. But the weight lost is not entirely fat.

In the STEP 1 body-composition substudy, semaglutide was associated with reductions in both fat and lean mass. DXA-measured lean mass is not identical to skeletal muscle—it also includes water, organs, connective tissue, and glycogen-associated water—but the finding remains relevant when there is little excess weight to lose.

In the SURMOUNT-1 DXA substudy, approximately 75% of weight lost with tirzepatide was fat mass and about 25% was lean mass, Figure 3. [4]

For someone with obesity, the net change may still be favorable.

For someone who is already lean, there is less excess adipose tissue to remove and less margin for avoidable loss of lean tissue.

Regaining Lost Muscle May Be More Difficult With Aging

A loss of lean mass with aging is problematic. A 2025 study by Jiang and colleagues provides important insights on why it is difficult to regain lean mass as we age. The investigators used UM-HET3 mice, the genetically heterogeneous strain used extensively in the NIA Interventions Testing Program. Importantly, these mice were first made obese and metabolically unhealthy with a high-fat diet. [5]

Continuous liraglutide improved weight and metabolic health. Again, the drug was acting in a system with clear metabolic pathology to correct.

Repeated treatment and withdrawal produced a different pattern. Fat mass returned more readily than lean mass, while visceral adiposity and hyperleptinemia increased and some metabolic benefits disappeared.

This was not a lifespan or cognition experiment, and mouse findings should not be assumed to occur in humans. But it illustrates an important concept:

Weight regain after stopping GLP1s is not necessarily muscle recovery.

That may matter more with aging, when skeletal muscle is already harder to rebuild because of anabolic resistance and progressive loss of muscle mass and strength.

Returning to the same number on the scale may not mean returning to the same body composition.

Why APOE4 and Catabolism Matter

APOE4 affects lipid transport, glucose metabolism, inflammation, and brain energy metabolism. This makes weight loss more complicated than simply reducing fat mass.

Adipose tissue and skeletal muscle are also endocrine organs. They release adipokines and myokines that participate in signaling between peripheral tissues and the brain and can influence metabolism, vascular function, inflammation, and neuronal health.

When obesity, insulin resistance, diabetes, or metabolic syndrome are present, reducing dysfunctional adipose tissue may improve metabolic inflammation and vascular risk enough to outweigh some loss of lean mass.

But the balance may be different in a lean APOE4 carrier. Loss of adipose tissue and muscle can also reduce metabolic reserve and alter adipokine and myokine signaling.

In our review of APOE alleles and diet, Caleb Finch and I discussed evidence that APOE4 alters brain energy metabolism and that the cognitive consequences of weight loss may differ by genotype. We highlighted Look AHEAD, where favorable cognitive effects of intensive weight loss were seen primarily in APOE4 non-carriers. [6]

These findings do not prove that GLP-1–induced weight loss harms APOE4 carriers. They suggest a more nuanced principle:

When metabolic disease is present, correcting it may provide substantial brain benefit. When metabolic health is already favorable, loss of muscle, adipose-derived signals, and metabolic reserve may carry greater importance.

Side Effects Matter

The most common adverse effects of GLP-1–based therapies are gastrointestinal.

Jalleh and colleagues reviewed nausea, vomiting, diarrhea, constipation, delayed gastric emptying, and related gastrointestinal effects of GLP-1 receptor agonists in The Lancet Gastroenterology & Hepatology. [7]

A large systematic review and meta-analysis of randomized trials also found increased risks of cholelithiasis and probably gastroesophageal reflux disease. [8]

For a patient with obesity, diabetes, or substantial cardiovascular risk, these adverse effects may be an acceptable tradeoff for demonstrated benefit.

For a healthy lean person, the calculation is different.

Reduced appetite may also reduce total energy, protein, and micronutrient intake. In someone with little excess weight to lose, that reinforces the concern about catabolism and preservation of lean tissue.

Does “Microdosing” Solve the Problem?

An increasingly common proposal is to use a very small dose of a GLP-1 drug—often called microdosing—with the hope of obtaining metabolic or anti-inflammatory effects while avoiding substantial weight loss or side effects.

The problem is that this strategy has not been validated for dementia prevention.

Komé, Chandran, Buse, Klein and colleagues discussed fractional semaglutide dosing in Diabetes Care in 2025, largely in the context of individualized diabetes care, tolerability, and cost. It was not evidence that microdosed semaglutide prevents Alzheimer’s disease or extends lifespan. [9]

The major clinical benefits of GLP-1 drugs were demonstrated at specific studied doses in defined disease populations.

Lower doses may reduce adverse effects, but they may also reduce efficacy. A patient may still experience early satiety, nausea, constipation, reflux, or reduced food intake at a dose that has never been shown to reproduce the clinical benefits demonstrated in outcome trials.

So microdosing introduces a different uncertainty:

A person may receive enough drug to alter normal physiology without receiving a dose that has demonstrated clinical benefit.

When GLP-1 Therapy Is Indicated, Protect Muscle

None of this argues against GLP-1 therapy when there is an appropriate indication.

If a patient has obesity, diabetes, insulin resistance, cardiovascular disease, or another established indication, the potential benefits may substantially outweigh the risks.

But muscle preservation deserves particular attention whenever these drugs are used in older adults or anyone vulnerable to age-related muscle loss.

Aging itself is associated with progressive loss of muscle mass and strength and reduced anabolic responsiveness. Adding pharmacologic weight loss to that process makes muscle an important treatment outcome.

Resistance training should be incorporated into the therapeutic plan, together with adequate protein and total energy intake. In older adults, monitoring weight alone is insufficient. Strength, functional performance, dietary intake, and, where appropriate, body composition should also be followed. [11]

The pharmaceutical pipeline is moving in the same direction. Investigational drugs targeting myostatin and activin signaling are being combined with GLP-1–based therapies to preserve lean mass while maintaining fat loss. Bimagrumab has been studied with semaglutide, and apitegromab with tirzepatide, with early trials showing improved preservation of lean tissue. [12,13]

These approaches remain experimental, but they reinforce an important shift:

The goal should not simply be maximum weight loss. It should be loss of excess fat while preserving functional muscle.

What We Still Do Not Know

The uncertainty cuts in both directions.

EVOKE and EVOKE+ studied people with symptomatic Alzheimer’s disease. They cannot tell us whether treatment decades earlier might alter dementia risk.

APOE4 carriers who also have obesity, diabetes, insulin resistance, metabolic syndrome, or substantial vascular disease may benefit when those conditions are effectively treated.

Conversely, we do not have randomized trials showing that GLP-1 therapy causes cognitive harm in lean APOE4 carriers.

The concern about catabolism is therefore biologically plausible, not clinically proven.

A definitive prevention trial would require metabolically healthy participants, APOE genotyping, careful measurement of muscle and body composition, nutritional status, inflammatory phenotype, drug exposure, and long-term cognitive outcomes.

That study is not likely to be done, given the complexity of long-term cognitive outcome trials.

For preventive pharmacology in someone who is currently healthy, absence of demonstrated harm should not be mistaken for evidence of benefit.

What I Would Tell This Patient

I would tell her that GLP-1 drugs are effective medications when there is a clear metabolic or vascular indication.

But we currently do not have evidence that either standard-dose or microdose GLP-1 therapy prevents dementia in a lean, normoglycemic APOE4 carrier with undetectable hsCRP and no clear evidence of metabolic or vascular inflammation.

I would return to one question:

What abnormality are we treating?

If the answer is obesity, diabetes, insulin resistance, metabolic syndrome, cardiovascular disease, fatty liver disease, or another established indication, there is a biological target and an evidence base. Treat it—and, particularly with aging, actively protect muscle.

If the answer is simply APOE4 carrier status and dementia prevention, the evidence is not there.

For APOE4 carriers seeking prevention, we already have measurable targets: blood pressure, apoB and LDL cholesterol, glycemic health, fitness, muscle strength, sleep apnea, smoking, diet quality, hearing, and other vascular and metabolic risk factors.

The goal should be to identify abnormalities and correct them—not assume that altering normal metabolism is beneficial simply because the same drug helps people with metabolic disease.

Take-Home Messages

• GLP-1 drugs have proven benefits when there is metabolic or cardiovascular disease to treat.

• They are not proven longevity or dementia-prevention drugs in healthy, lean adults.

• Inflammation is heterogeneous. Metabolic inflammation associated with obesity, insulin resistance, diabetes, and metabolic syndrome is not the same as neuroinflammation.

• In EVOKE/EVOKE+, semaglutide lowered systemic inflammation in an Alzheimer’s population not selected for obesity or diabetes, yet did not slow cognitive decline.

• GLP-1–associated weight loss includes lean-mass loss, which matters increasingly with aging.

• In APOE4 carriers, treat harmful metabolic disease when it is present, but do not assume that loss of muscle, adipose tissue, and metabolic reserve is inherently brain-protective when metabolic health is already favorable.

• Withdrawal deserves attention because fat may return more readily than lean tissue.

• Microdosing has not been validated for dementia prevention and may produce pharmacologic effects without proven clinical benefit.

• When GLP-1 therapy is indicated, particularly during aging, resistance training, adequate nutrition, and monitoring of muscle and function should be part of treatment.

• The central clinical question remains: What biological abnormality are we treating?

References

1. Lincoff AM, Brown-Frandsen K, Colhoun HM, et al. Semaglutide and cardiovascular outcomes in obesity without diabetes. New England Journal of Medicine. 2023.

2. Wilding JPH, et al. Once-weekly semaglutide in adults with overweight or obesity. New England Journal of Medicine. 2021.

3. Johannsen P, Cummings J, et al. Efficacy and safety of oral semaglutide in early-stage symptomatic Alzheimer’s disease: EVOKE and EVOKE+. Lancet. 2026.

4. Look M, Dunn JP, Kushner RF, et al. Body composition changes during weight reduction with tirzepatide in the SURMOUNT-1 study of adults with obesity or overweight. Diabetes, Obesity and Metabolism. 2025;27(5):2720-2729. doi:10.1111/dom.16275.

5. Jiang N, Yin J, Lawrence N, et al. Repeated withdrawal of a GLP-1 receptor agonist induces hyperleptinemia and deteriorates metabolic health in obese aging UM-HET3 mice. Aging Cell. 2025.

6. Yassine HN, Finch CE. APOE alleles and diet in brain aging and Alzheimer’s disease. Frontiers in Aging Neuroscience. 2020;12:150.

7. Jalleh RJ, Rayner CK, Hausken T, Jones KL, Camilleri M, Horowitz M. Gastrointestinal effects of GLP-1 receptor agonists: mechanisms, management, and future directions. Lancet Gastroenterology & Hepatology. 2024;9:957–964.

8. Chiang C-H, et al. Glucagon-like peptide-1 receptor agonists and gastrointestinal adverse events: a systematic review and meta-analysis. Gastroenterology. 2025.

9. Komé AM, Chandran MM, Tungate Lopez SS, Buse JB, Klein KR. One size does not fit all: understanding microdosing semaglutide for diabetes in multidose pens. Diabetes Care. 2025;48:e25–e27.

10. Tilg H, Ianiro G, Gasbarrini A, et al. Adipokines: masterminds of metabolic inflammation. Nature Reviews Immunology. 2025.

11. McKendry J, Coletta G, Nunes EA, Lim C, Phillips SM. Mitigating disuse-induced skeletal muscle atrophy in ageing: resistance exercise as a critical countermeasure. Experimental Physiology. 2024.

12. Heymsfield SB, Aronne LJ, Montgomery P, et al. Bimagrumab plus semaglutide alone or in combination for the treatment of obesity: a randomized phase 2 trial. Nature Medicine. 2026.

13. Pratley RE, Denham DS, Trivedi R, et al. Apitegromab for lean mass preservation during tirzepatide-induced weight loss: a randomized, double-blind, placebo-controlled phase 2 trial. Nature Medicine. 2026.

Disclosure: This post contains partially AI-assisted text. All references were reviewed for accuracy.

Amyloid plaque burden imaged using PET shows that it can accumulate years before symptoms. Tau PET, by contrast, tends to align more closely with where the brain is failing and how cognition is declining. This observation has shaped the field. Based on findings that tau burden and distribution are more closely associated with cognitive impairment and disease stage than amyloid plaque burden [1], why don’t we directly target tau?

That is the assumption the ceperognastat trial forces us to examine.

Ceperognastat is an oral drug designed to inhibit O-GlcNAcase, also called OGA. OGA is an enzyme that removes a small sugar-like chemical tag, called O-GlcNAc, from proteins inside cells. The idea was that blocking OGA would keep more O-GlcNAc on tau, reduce tau phosphorylation and aggregation, and slow Alzheimer’s disease progression. [2]

The rationale was plausible. But plausibility is not proof.

PROSPECT-ALZ

PROSPECT-ALZ was a large, expensive Phase 2 program. It enrolled 327 participants with early symptomatic Alzheimer’s disease, ran at 72 sites in 5 countries, used tau PET screening/enrichment, MRI, plasma biomarkers, repeated clinical assessments, randomized to placebo, low-dose ceperognastat, or high-dose ceperognastat, and followed participants for up to 76–124 weeks, with an extension period.

The trial did not show clinical benefit. In the primary outcome population, the low-dose group had no clinically meaningful improvement over placebo, and the high-dose group showed 32% greater clinical progression than placebo on the Integrated Alzheimer’s Disease Rating Scale at 100 weeks (Figure 1).

Some biomarker measures moved in directions that could have appeared favorable, including less whole-brain volume loss and a smaller increase in tau PET signal in one brain region Figure 2. [3]

This is the central lesson: a biomarker can track disease without proving that modifying it will improve cognition.

The Correlation–Causation Problem

The strongest argument for tau as a therapeutic target comes from human observational data. Tau PET burden and distribution correlate more closely with cognitive impairment and clinical stage than amyloid plaque burden. That makes tau PET valuable for staging disease, selecting participants for trials, and estimating risk of progression. [1]

But correlation is not causation.

A marker of disease severity is not automatically the mechanism that must be modified to improve the disease. In Alzheimer’s disease, tau pathology could play several roles. It may directly injure neurons. It may amplify stress initiated by amyloid, inflammation, vascular injury, or endolysosomal dysfunction. It may reflect a cellular response to injury. Or it may mark neurons that are already failing.

These possibilities lead to different therapeutic predictions.

If tau is a direct and reversible driver of neuronal death, lowering tau should help preserve neurons and cognition. If tau is downstream of a broader cellular failure, lowering tau may improve tau biomarkers without restoring the injured system that produced the tau pathology in the first place.

Ceperognastat does not prove that tau is irrelevant. It shows that changing tau-related biology through broad OGA inhibition did not produce clinical benefit in early symptomatic Alzheimer’s disease. [3]

Genetic Tauopathies Are Not the Same as Alzheimer’s Disease

There are diseases where tau has a stronger causal role.

MAPT is the gene that encodes the tau protein. In some inherited frontotemporal dementias and related primary tauopathies, mutations in MAPT alter tau itself. Some mutations change the tau protein sequence. Others alter tau splicing, which changes the balance of tau isoforms. These changes can make tau more prone to misfolding, aggregation, or toxic dysfunction. In these diseases, tau biology is closer to the root cause of neurodegeneration. [4]

Alzheimer’s disease is different.

Most patients with late-onset Alzheimer’s disease do not have MAPT mutations. AD is not usually a disease of genetically overexpressed tau. It is a disease in which tau pathology emerges in the setting of amyloid, endolysosomal stress, immune activation, vascular injury, metabolic dysfunction, synaptic failure, and aging.

This distinction matters because many preclinical tau models rely on high expression of mutant human tau. One common model, rTg4510, overexpresses human tau carrying the P301L mutation. P301L means that one amino acid in tau — proline at position 301 — is replaced by leucine. This mutation pushes the system toward tau pathology. It is useful for testing whether a drug changes tau aggregation, but it does not reproduce the full biology of sporadic Alzheimer’s disease. [4,5]

A mutant tau overexpression model asks one question:

Can this drug change tau aggregation biology?

A human Alzheimer’s trial asks a broader question:

Can this drug preserve neurons, synapses, networks, and cognition in a multifactorial disease of aging?

Those are not the same question.

Tau Production, Tau Aggregation, and Tau Spread Are Different Targets

One reason tau drug development is difficult is that “targeting tau” can mean several different things.

Tau production means how much tau protein a cell makes. The MAPT gene is copied into MAPT mRNA, and that RNA message is used to produce tau protein.

Tau aggregation means tau proteins begin to misfold and clump together into abnormal species, including oligomers, fibrils, and eventually neurofibrillary tangles.

Tau spread refers to the observation that tau pathology appears to move through connected brain networks as Alzheimer’s disease progresses.

A drug can target one step without fixing the others.

An antisense oligonucleotide, or ASO, is a short synthetic strand of nucleic acid designed to bind a specific RNA message. A tau ASO binds MAPT mRNA, the message used to make tau protein, and promotes its degradation or blocks its translation. In plain English, a tau ASO reduces tau production. [8,9]

OGA inhibitors do something different. They do not reduce tau production. They try to change tau’s chemical state after tau has already been made by increasing O-GlcNAcylation, with the hope that tau becomes less likely to phosphorylate, aggregate, or spread. [2]

This distinction matters. A tau ASO tests whether reducing total tau production helps. An OGA inhibitor tests whether changing tau modification helps. These are related but not identical hypotheses.

Why the Ceperognastat Rationale Looked Appealing

The rationale for ceperognastat followed a clear biological sequence.

Tau is regulated by post-translational modifications. One of these is O-GlcNAcylation, the addition of a small O-GlcNAc tag to serine or threonine residues on proteins. OGA removes this tag. Inhibiting OGA increases O-GlcNAcylation. [2,6]

Because tau phosphorylation and O-GlcNAcylation can interact, OGA inhibition was proposed as a way to shift tau away from pathological phosphorylation and aggregation. In animal tauopathy models, OGA inhibition increased tau O-GlcNAcylation and reduced some measures of tau pathology. [2,4,5]

Ceperognastat was then optimized as a potent, orally available, CNS-penetrant OGA inhibitor. Phase 1 studies showed high brain OGA occupancy in humans. OGA occupancy means the percentage of OGA enzyme molecules bound by the drug. High occupancy shows that the drug entered the brain and engaged its target. In ceperognastat studies, brain OGA occupancy above 95% was reported. [2]

But there is a gap between target engagement and disease modification.

Disease modification, explained in greater detail before, means changing the underlying biology of the disease in a way that slows future decline. In Alzheimer’s disease, a disease-modifying therapy should ultimately preserve cognition, daily function, synapses, neurons, or brain networks over time.

Showing that a drug reaches the brain and binds OGA is important. It does not prove that the intervention improves neuronal function. It does not prove that the dose is safe for synapses. It does not prove that modifying tau-related chemistry will alter the clinical trajectory of Alzheimer’s disease.

This is where the evidentiary foundation appears limited.

The Preclinical Data Answered a Narrow Question

The preclinical work showed that OGA inhibition could reduce tau pathology in tauopathy mouse models. That is meaningful, but narrow.

In rTg4510 mice, chronic OGA inhibition reduced aggregated tau, several phosphorylated tau species in insoluble brain fractions, and CSF total tau. These are pharmacodynamic effects on tau biology (Figure 4). But rTg4510 mice overexpress mutant human tau. They do not reproduce the full biology of late-onset Alzheimer’s disease. [4]. And the investigators missed behavioral readouts.

This is a recurring problem in Alzheimer’s drug development. Animal models are often interpreted beyond the question they were designed to answer. A model can be useful for testing target biology and still be limited for predicting clinical benefit.

The relevant limitation is not that the tau mouse data were wrong. The limitation is that they did not establish that broad OGA inhibition would preserve synaptic function and cognition in cell and animal models before they went to human AD.

That should have been a higher bar before moving into a large and expensive clinical program. The unfortunate implications of null trials are decreased investment.

OGA Is Not a Tau-Specific Target

The failure of ceperognastat may also reflect the biology of the target.

OGA does not act only on tau. O-GlcNAcylation regulates many nuclear and cytoplasmic proteins involved in metabolism, stress responses, cell signaling, survival, protein turnover, transcription, and synaptic function. [2,6]

This matters because cognition depends on synaptic physiology, not only on pathology markers.

If a drug reduces a tau signal but disrupts synaptic transmission or plasticity, the net clinical effect may be neutral or harmful. The phase 2 trial raises this concern because high-dose ceperognastat was associated with worse clinical progression despite some biomarker changes that did not look unfavorable. [3]

A later brain-slice study came too late for the Lilly-sponsored trial. Investigators tested three structurally distinct OGA inhibitors — ceperognastat, ASN90, and MK8719 — in adult mouse hippocampal slices. All three suppressed short-term synaptic plasticity and long-term potentiation, or LTP (Figure 5). LTP is one of the experimental measures scientists use to study how synapses strengthen during learning. The same study also reported altered synaptic protein markers, including PSD-95 and synaptophysin 1. [7]

This experiment does not prove that synaptic toxicity caused the clinical outcome in PROSPECT-ALZ. But it identifies exactly the kind of functional risk that should be assessed early in CNS drug development, before moving to expensive Phase II or Phase III trials.

For Alzheimer’s disease, the key preclinical question should not be limited to:

Does the drug lower tau?

It should also include:

Does the drug preserve synaptic transmission, plasticity, and network function?

What the Trial Actually Tells Us

The PROSPECT-ALZ trial should not be interpreted as proof that tau is irrelevant.

It tells us something more specific: broad OGA inhibition with ceperognastat did not slow clinical progression in early symptomatic Alzheimer’s disease, and the high dose was associated with worse clinical outcomes. [3]

It also highlights a biomarker–clinical disconnect. Some imaging outcomes appeared to move in directions that might otherwise be interpreted as favorable, while cognition and function did not improve. [1,3]

That is important because Alzheimer’s trials increasingly rely on biomarkers for enrollment, staging, target engagement, and early decision-making. Biomarkers are essential. But they are not complete substitutes for clinical outcomes.

A biomarker can identify disease without proving that it is the best therapeutic lever. Tau PET may be a strong marker of neurodegeneration and clinical stage. That does not mean every tau-lowering or tau-modifying strategy will be beneficial.

What About Tau ASOs?

Tau ASOs should be considered separately from ceperognastat.

Unlike OGA inhibitors, tau ASOs target MAPT mRNA to reduce tau production more directly. This makes them a cleaner test of the tau-lowering hypothesis than broad O-GlcNAc manipulation. [8,9]

The leading example is diranersen, also known as BIIB080, an investigational tau-targeting ASO. Early clinical studies showed reductions in CSF tau measures and suggested effects on tau PET in smaller datasets. [8] In 2026, Biogen reported phase 2 CELIA results showing robust reductions in CSF total tau and tau PET pathology, with prespecified cognitive signals. However, the study did not meet its primary endpoint assessing dose response, and the strongest clinical signals were not simply dose-dependent. [9]

This is encouraging, but it still requires careful interpretation.

If tau ASOs show reproducible clinical benefit in larger trials, that would support tau as a therapeutically relevant pathway in selected Alzheimer’s populations or disease stages. If they lower tau biomarkers without preserving cognition, that would strengthen the concern that tau may often be a marker or amplifier rather than the primary reversible cause of neuronal loss.

The key question for tau ASOs is therefore not only whether tau can be lowered.

The question is whether lowering tau preserves synapses, neurons, networks, and cognition.

A Challenge to the Linear Amyloid Cascade

Ceperognastat also fits into a broader problem with the linear amyloid cascade model.

The classic sequence is familiar: amyloid accumulates, tau spreads, neurons die, cognition declines. This model explains important parts of Alzheimer’s disease biology, especially in familial AD and in amyloid-positive disease trajectories. It has also supported amyloid-lowering therapies that modestly slow decline.

But late-onset Alzheimer’s disease is not a pure amyloid disease and not a pure tauopathy. I discussed problems with the linear amyloid cascade in a previous post.

Briefly, late-onset AD genetics implicate endolysosomal trafficking, immune activation, lipid metabolism, vascular injury, mitochondrial stress, proteostasis, and synaptic vulnerability. In this broader framework, amyloid and tau remain important. They may be toxic. They may amplify injury. They may stage disease. But they may not always be the root cause.

This is where targeting endolysosomal and immune pathways becomes relevant. If upstream cellular clearance and immune pathways are failing, then amyloid and tau may partly reflect the visible output of a deeper biological process.

Lowering one output may not restore the system.

The Uncertainty We Need to Acknowledge

The field should be careful here.

We do not yet have definitive proof that lowering tau improves cognition in Alzheimer’s disease. Ceperognastat modified tau-related biology but did not improve clinical outcomes. [3] Diranersen has reported encouraging phase 2 biomarker and clinical signals, but it has not yet provided replicated registrational evidence that tau lowering slows AD progression. [9]. Cell or Animal models that genetically overexpress tau (or amyloid) create environments that do not mimic human AD pathology and complicate distinguishing causation from association. The effect of targeting excessive tau expression in animal models may not translate to a mild or moderate increase in tau pathology in chronic diseases like aging.

This uncertainty does not mean tau is unimportant. It means the therapeutic role of tau in AD remains unresolved.

Tau may be causal in some contexts, amplifying in others, and downstream in others. The task is to identify when tau is a driver, when it is a marker, and when lowering it meaningfully preserves brain function.

Take-Home Messages

1. Tau PET is a strong disease marker, but not yet a validated therapeutic surrogate.
Tau correlates with cognition better than amyloid, but this does not prove that lowering tau without understanding the mechanism of action will improve cognition.

2. Correlation is not causation.
Tau may drive neuronal injury, amplify other disease processes, or mark neurons already under stress. These possibilities require different treatments.

3. Alzheimer’s disease is not the same as a genetic tauopathy.
MAPT mutation diseases show that tau can cause neurodegeneration, but late-onset AD usually does not arise from MAPT mutations or tau overexpression.

4. Tau production, aggregation, and spread are different therapeutic targets.
Tau ASOs reduce tau production by targeting MAPT mRNA. OGA inhibitors try to change tau modification after tau has already been made.

5. Target engagement is not disease modification.
High OGA occupancy showed that ceperognastat reached and bound its target in the brain. It did not prove that synapses, neurons, or cognition would be protected.

6. OGA inhibition is not tau-specific.
OGA affects O-GlcNAcylation on many proteins involved in metabolism, signaling, stress responses, and synaptic function.

7. Synaptic function should be part of preclinical screening.
For Alzheimer’s drugs, moving a biomarker is not enough. A therapy should preserve or improve synaptic transmission, plasticity, and network function before advancing to large trials.

8. Tau ASOs remain important, but they still must prove clinical benefit.
They are a cleaner test of tau lowering than OGA inhibition, but the key endpoint remains cognition and function, not tau reduction alone.

9. Alzheimer’s disease likely requires systems-level thinking.
Amyloid and tau are important, but late-onset AD also involves endolysosomal, immune, vascular, metabolic, and synaptic biology.

10. The goal is not to treat the biomarker.
The goal is to preserve the living system the biomarker is supposed to represent: neurons, synapses, networks, and cognition.

References

1. Rabinovici GD, Grill JD. Ceperognastat in Alzheimer disease: lessons from a negative clinical trial. JAMA. 2026.

2. Kielbasa W, Goldsmith P, Donnelly KB, et al. Discovery and clinical translation of ceperognastat, an O-GlcNAcase inhibitor, for the treatment of Alzheimer’s disease. Alzheimer’s & Dementia: Translational Research & Clinical Interventions. 2024.

3. Fleisher AS, Munsie L, Mancini M, et al. Ceperognastat in early symptomatic Alzheimer disease: a randomized clinical trial. JAMA. 2026.

4. Hastings NB, Wang X, Song L, et al. Inhibition of O-GlcNAcase leads to elevation of O-GlcNAc tau and reduction of tauopathy and cerebrospinal fluid tau in rTg4510 mice. Molecular Neurodegeneration. 2017;12:39.

5. Graham DL, Gray AJ, Joyce JA, et al. Increased O-GlcNAcylation reduces pathological tau without affecting its normal phosphorylation in a mouse model of tauopathy. Neuropharmacology. 2014;79:307–313.

6. Pratt MR, Vocadlo DJ. Understanding and exploiting the roles of O-GlcNAc in neurodegenerative diseases. Journal of Biological Chemistry. 2023;299:105411.

7. Meade J, Mesa H, Alamgir S, Bieniecka I, Liu L, Zhang Q. Synaptic toxicity of OGA inhibitors and the failure of ceperognastat. The Journal of Prevention of Alzheimer’s Disease. 2026;13:100456.

8. Mummery CJ, Börjesson-Hanson A, Blackburn DJ, et al. Tau-targeting antisense oligonucleotide MAPTRx in mild Alzheimer’s disease: a phase 1b randomized, placebo-controlled trial. Nature Medicine. 2023;29:1437–1447.

9. Biogen. Phase 2 CELIA results for diranersen (BIIB080) in early Alzheimer’s disease. 2026.

APOE4 is not simply a “bad gene,” and APOE2 is not simply a “good gene.” Many APOE4 carriers do not develop dementia, and some APOE2 carriers have a lipid handling disorder.

APOE alleles reflect biological tradeoffs. A variant that helps in one environment, tissue, or life stage may harm in another.

That is why the question sounds easier than it is:

If APOE2 protects against Alzheimer’s disease, why don’t we give APOE2 to APOE4 carriers?

The answer is that APOE2 is not a vitamin, supplement, or simple replacement part. APOE is a lipid transport protein, receptor ligand, immune modulator, and brain-cell signaling molecule. Its effects depend on where it is made, which cell makes it, how it is lipidated, when it is delivered, and what disease stage is already present.

Not a Good or a Bad Gene Copy

APOE is a lipid transport gene with three common alleles: APOE2, APOE3, and APOE4. In many large datasets, APOE3 is the most common allele, roughly 75–80%, while APOE4 is about 10–15% and APOE2 about 5–10%, with variation by ancestry and population. In genotype terms, APOE3/3 is the most common, often present in about two-thirds of people [1].

APOE4 increases late-life Alzheimer’s risk, but human data suggest it may have carried advantages in environments where infection, undernutrition, and reproduction shaped survival more strongly than late-life dementia.

In the Tsimane, a natural fertility population with high infectious burden, a study of 795 women found that APOE4 carriers had about 0.3 to 0.5 more children than APOE3 homozygotes, while APOE4 homozygotes had about 1.4 to 2.1 more children [2]. In a Brazilian shantytown cohort of 72 children followed from birth, APOE4 carriers appeared more resilient to the long-term cognitive effects of heavy early-childhood diarrhea [3]. Several studies also describe amyloid-beta as an antimicrobial-like peptide that can inhibit microbial growth through membrane disruption or microbial entrapment [4].

These studies do not prove that APOE4 is beneficial. They support a more careful idea: what helps early-life survival may not help the aging brain.

APOE2 also has tradeoffs. APOE2 is associated with lower Alzheimer’s risk and, in some studies, longevity. Key to how it works, APOE2 binds poorly to the LDL receptor. This property may explain why about 5–10% of APOE2 homozygotes develop type III hyperlipoproteinemia, with accumulation of triglyceride-rich remnant particles [4]. APOE2 has also been associated with increased risk of cerebral amyloid angiopathy and selected neurological disorders [5].

So the goal is not to replace a “bad” allele with a “good” allele. The goal is to identify which APOE2 functions protect the aging brain — lipidation, amyloid handling, synaptic support, vascular effects, or microglial signaling — and reproduce those functions without creating APOE2-related complications.

What APOE2 Does Differently from APOE4

APOE is best known as a lipid carrier. In the brain, APOE helps move cholesterol and phospholipids between cells. This is essential because neurons depend on glial cells to help maintain membranes, synapses, repair programs, and lipid balance.

APOE2 differs from APOE3 and APOE4 at key amino acid positions. One important consequence is weaker binding to the LDL receptor (LRP1). In some lipid-handling contexts, APOE2 can therefore behave like a partial loss-of-function variant.

But in the brain, APOE2-containing lipoprotein particles may be more lipidated than APOE3 or APOE4 particles. More lipidated particles may better support lipid transport, synaptic maintenance, amyloid plaque seeding, and amyloid clearance [4].

But there is the APOE2 paradox:

In the periphery, APOE2 can impair remnant clearance. In the brain, APOE2 may create a more protective lipidated particle.

That is why therapeutic design has to be precise.

APOE2-Associated Longevity Involves Lipid Handling, Not Just Alzheimer’s Pathology

APOE2’s link to longevity may not simply reflect lower Alzheimer’s pathology. In an analysis of NACC clinical records, carrying APOE4 had a similar effect on survival independent of AD pathology. Similarly, APOE2 had a protective effect on survival independent of amyloid and tau pathology-Figure 2 [6]. I have discussed before why amyloid and tau do not fully explain the clinical syndrome of AD, and here is where lipids become interesting.

In the same eLife study, APOE2-targeted replacement mice lived longer in a model that did not overexpress mutant APP or tau and did not show Alzheimer-like amyloid or tau deposition. In 118 mice, median survival was 911 days for APOE2, 825 days for APOE3, 753 days for APOE4, and 738 days for Apoe knockout [6].

This same study linked APOE2 to preserved activity and lipid metabolism during aging. The authors reported associations among activity, apoE protein, HDL-cholesterol, and triglycerides, and suggested that higher HDL-cholesterol and triglyceride levels regulated by apoE2 might contribute to APOE2-mediated longevity [6]. These findings may offer interesting insights into how Obicetrapib may affect AD biology.

This matters because APOE2 may influence aging through lipid transport and brain-body metabolism, not only through amyloid and tau. An APOE2 therapy should therefore ask not just whether plaques go down, but whether lipid handling, vascular integrity, brain metabolism, learning, and resilience move in the right direction.

The Brain Has Its Own APOE System

One major reason we cannot simply give APOE2 by mouth or IV is that the body and brain have largely separate APOE pools.

A classic liver transplantation study showed that after liver transplant, the APOE phenotype in plasma changed almost completely to match the donor liver. But the APOE phenotype in cerebrospinal fluid did not change [7].

Liver APOE does not simply become brain APOE.

The liver supplies most plasma APOE. The brain makes APOE locally, mainly through astrocytes, with contributions from microglia, vascular mural cells, choroid plexus, and stressed neurons [4].

So oral APOE would likely be digested like other proteins. IV APOE may mostly act in the periphery, where it could affect triglycerides, remnant lipoproteins, vascular biology, or inflammation. But it may not meaningfully change APOE biology inside the brain.

The therapeutic problem is not just what to deliver.

It is where to deliver it, and whether APOE is properly lipidated.

AAV Gene Therapy: Silence and Replace

One promising approach is CNS gene therapy.

In APOE4 carriers, the ideal intervention may not be adding APOE2 alone. It may require two steps:

Silence APOE4. Replace with APOE2.

This makes biological sense because APOE4 may not simply represent loss of protection. APOE4 may also have toxic gain-of-function effects involving lipidation, amyloid aggregation, amyloid clearance, synaptic injury, mitochondrial function, blood-brain barrier integrity, and neuroinflammation.

A recent AAV “silence-and-replace” strategy used microRNAs to suppress endogenous APOE4 while adding a modified APOE2 gene designed to resist that silencing [8].

The idea is elegant: reduce the high-risk APOE4 signal and shift the local CNS APOE environment toward APOE2.

But delivery remains difficult. In mice, APOE AAV studies often bypass the blood-brain barrier by direct stereotaxic brain injection. In the APOE4 silence-and-replace study, investigators drilled burr holes and injected AAV.S2 vectors bilaterally into the hippocampus. That is very different from giving a drug by mouth or IV, and it highlights the central challenge for human translation: how to deliver APOE2 safely, broadly, and to the right CNS cells without direct regional injections into every vulnerable brain area.

Many AAV vectors preferentially transduce neurons. Yet most endogenous brain APOE is made by astrocytes. That mismatch matters. If a therapy causes neurons to make APOE2, is that equivalent to astrocytes or microglia making APOE2?

Likely not.

Cell Type Matters

This may be the most important concept.

APOE2 made by a neuron may not be the same as APOE2 made by an astrocyte or microglia.

Astrocyte-derived APOE participates in lipid transport, synaptic maintenance, cholesterol handling, and neuronal support.

Microglial APOE may be more closely linked to immune activation, plaque-associated responses, TREM2 biology, phagocytosis, damaged-neuron clearance, and disease-associated microglial states.

Cell type also affects APOE processing. APOE made in different cells may differ in lipidation, secretion, receptor interactions, and post-translational modification [8].

So simply measuring “more APOE2 in the brain” may not tell us whether the right biology has been restored.

The future may require cell-specific APOE2 expression.

Current Clinical Translation

The leading clinical example is LX1001, an investigational AAVrh.10-based gene therapy designed to deliver human APOE2 into the central nervous system of APOE4 homozygotes with Alzheimer’s disease [9].

The Phase 1/2 study tests one-time intrathecal administration of AAVrh.10hAPOE2 in APOE4 homozygotes with mild cognitive impairment or dementia due to Alzheimer’s disease. A long-term follow-up study is also underway [9,10].

Intrathecal delivery means placing the therapy into the cerebrospinal fluid, the fluid that surrounds the brain and spinal cord. This brings an AAV vector closer to the CNS than an IV infusion, but it does not guarantee that the right brain cells will express APOE2. After intrathecal delivery, APOE2 may be expressed by cells reached from the CSF space — including meningeal, ependymal, choroid plexus, spinal, or some brain parenchymal cells depending on the vector. The key question is whether this reproduces the biology that matters most for Alzheimer’s disease: APOE made by astrocytes and microglia, properly lipidated, secreted at the right level, and active in the brain regions most vulnerable to disease

Interim topline results reported 15 dosed participants, dose-dependent APOE2 detection in CSF, general tolerability, and favorable movement in some amyloid and tau biomarkers [11].

These findings support continued study, but they are not proof of clinical efficacy.

The core translational questions remain:

Can APOE2 be delivered safely to the CNS?
Can APOE2 be expressed durably at meaningful levels?
Does it improve amyloid, tau, lipid, vascular, or inflammatory biology?
Does it avoid vascular, inflammatory, lipid, lysosomal, or immune toxicity?
Does it slow cognitive and functional decline?

What Would Success Look Like?

The benchmark for success cannot be “we detected APOE2.” That is target engagement, not disease modification.

A successful APOE2 therapy would need several layers of evidence.

First: safety. No unacceptable inflammation, immune response, edema, hemorrhage, vascular injury, or accelerated neurodegeneration.

Second: target engagement. APOE2 should appear in CSF or other CNS-linked measures in a dose-responsive and durable way.

Third: biological correction. The therapy should improve APOE lipidation, lipid transport, amyloid handling, tau-related injury, synaptic biology, glial function, or vascular integrity.

Fourth: clinical effect. It should slow worsening on cognition, function, MRI atrophy, tau PET, amyloid PET, CSF biomarkers, or plasma biomarkers.

The hardest endpoint is the one patients care about most:

Do people think, remember, function, and live independently longer?

That is the real bar.

The Liposome Cautionary Tale

A recent liposome-based APOE2 study is important because it complicates the simple story.

The study used brain-targeting liposomes to deliver APOE2 plasmid material in an amyloid mouse model. APOE2 delivery produced some potentially favorable transcriptional signals related to neuronal development and synaptic pathways [12].

But it also increased soluble amyloid-beta, including oligomeric forms, and worsened neurite dystrophy [12].

That distinction matters.

Lowering plaque burden is not automatically the same as lowering toxicity. Some soluble amyloid species may be more synaptotoxic than deposited plaque.

So “APOE2 reduces plaques” is not enough.

We need to ask:

What happens to soluble amyloid?
What happens to synapses?
What happens to neurites?
What happens to microglia?
What happens over time?

Prevention Is Not Treatment

A person born with APOE2 has decades of APOE2 biology. Their brain develops and ages with different lipidation, receptor binding, amyloid handling, synaptic support, vascular biology, and immune tone.

That is not the same as giving APOE2 at age 70 after amyloid plaques, tau pathology, vascular injury, or neuroinflammation are already present.

APOE2 may be most powerful as preventive biology. It may reduce amyloid accumulation, improve lipid transport, support synapses, or shape microglial responses before pathology is advanced.

But once amyloid burden is high, adding APOE2 could have different effects. It may mobilize amyloid into soluble species, alter glial lipid handling, or interact with tau and vascular disease in ways we do not yet understand.

This does not mean APOE2 therapy cannot work in treatment.

It means disease stage matters.

One APOE2 strategy may not fit young APOE4 carriers, preclinical amyloid-positive carriers, mild cognitive impairment, established dementia, vascular comorbidity, amyloid-dominant disease, or tau-dominant disease.

The Ethical Question

APOE2 gene therapy also raises some ethical issues.

AAV-based gene therapy is designed to be long-lasting. That makes the risk-benefit calculation different from taking a pill that can be stopped.

The prevention question is harder. What level of risk would justify CNS gene therapy in a cognitively normal APOE4 homozygote? Amyloid positivity? Tau positivity? Plasma p-tau? Family history? Age?

Equity also matters. Genetic testing, biomarker screening, lumbar puncture, gene therapy delivery, and long-term monitoring may not be equally accessible.

The ethical standard should be high:

Clear risk disclosure, careful genetic counseling, long-term follow-up, transparent uncertainty, fair access, and a strict distinction between prevention and treatment.

So Why Aren’t We Giving APOE2 to APOE4 Carriers?

Because APOE2 is not simply good, and APOE4 is not simply bad.

We are not replacing a missing vitamin.

We are trying to reprogram a complex lipid-immune-neurodegenerative pathway inside the brain.

APOE genotypes teach us that Alzheimer’s risk is modifiable. It points toward protective mechanisms involving lipidation, receptor binding, amyloid handling, synaptic support, vascular biology, microglial function, and lipid metabolism.

But APOE2 is not a simple replacement part for APOE4.

The future is likely precision APOE biology: the right APOE function, in the right brain compartment, in the right cell type, at the right dose, and at the right disease stage. This is where downstream mechanisms become interesting. Can we lipidate APOE3 and APOE4 to behave like APOE2? Can we block APOE4-accentuated neuroinflammation, like APOE2?

APOE2 may or may not be the answer by itself.

It will offer us important clues.

Key Takeaways

• APOE4 and APOE2 are not simply bad and good alleles. They likely reflect biological tradeoffs.

• APOE2 is associated with lower Alzheimer’s risk and longevity, but it is not universally benign.

• APOE2 longevity may involve lipid metabolism, not only amyloid and tau.

• Peripheral APOE and brain APOE are largely separate pools, and oral or IV APOE2 is unlikely to reproduce brain APOE2 biology.

• Cell type matters. Astrocyte, microglial, and neuronal APOE2 may not have the same effects.

• AAV silence-and-replace strategies are promising, but delivery and cell targeting remain major challenges.

• Reducing plaques is not enough if soluble amyloid oligomers or neurite injury increase.

• Prevention and treatment may require different APOE2 strategies.

• The future is likely precision APOE biology, not simply “give APOE2.”

References

1. Farrer LA, Cupples LA, Haines JL, et al. Effects of age, sex, and ethnicity on the association between apolipoprotein E genotype and Alzheimer disease. JAMA. 1997;278(16):1349–1356.

2. Trumble BC, Stieglitz J, Blackwell AD, et al. Apolipoprotein-ε4 is associated with higher fecundity in a natural fertility population. Science Advances. 2023.

3. Oria RB, Patrick PD, Zhang H, et al. APOE4 protects the cognitive development in children with heavy diarrhea burdens in Northeast Brazil. Pediatric Research. 2005.

4. Moir RD, Lathe R, Tanzi RE. The antimicrobial protection hypothesis of Alzheimer’s disease. Alzheimer’s & Dementia. 2018.

5. Li Z, Shue F, Zhao N, Shinohara M, Bu G. APOE2: protective mechanism and therapeutic implications for Alzheimer’s disease. Molecular Neurodegeneration. 2020;15:63.

6. Shinohara M, Kanekiyo T, Tachibana M, et al. APOE2 is associated with longevity independent of Alzheimer’s disease. eLife. 2020;9:e62199.

7. Linton MF, Gish R, Hubl ST, et al. Phenotypes of apolipoprotein B and apolipoprotein E after liver transplantation. Journal of Clinical Investigation. 1991;88(1):270–281.

8. Karan KR, Hackett NR, Crystal RG. Adeno-associated virus-mediated central nervous system gene transfer to suppress Alzheimer’s disease high-risk APOE4 variant and replace with protective APOE2. Human Gene Therapy. 2026.

9. ClinicalTrials.gov. Gene Therapy for APOE4 Homozygote of Alzheimer’s Disease. NCT03634007.

10. ClinicalTrials.gov. Long-Term Follow-up of Gene Therapy for APOE4 Homozygote Alzheimer’s Disease. NCT05400330.

11. Johnson K, et al. Topline results from Phase 1/2 AAV gene therapy (LX1001) in APOE4/4 homozygotes with Alzheimer’s disease. Alzheimer’s & Dementia. 2025.

12. Wang N, Parsons TM, Ren Y, et al. Brain-targeting liposome-based APOE2 gene delivery exacerbates soluble amyloid-β accumulation in AppNL-G-F mice. Heliyon. 2024;10:e39607.

She was not trying to become a bodybuilder. She was scared of something very real: losing muscle mass as she aged.

She told me she had been aiming for 100 grams of protein per day. She had a relatively small muscle build, but since she started using protein bars — including Maui and David bars — a DEXA scan showed that she had gained muscle mass. She was excited. Understandably so.

Then she asked me the more important question.

Is this actually good for my health?

She admitted that maintaining that level of protein intake was hard. The bars helped her reach the number. But she wanted to know whether chasing 100 grams of protein every day was the right goal for aging.

That is the question this post is really about.

Protein matters. Muscle matters. But aging well is not just about pushing protein intake higher until the number looks right. It is about preserving strength, mobility, metabolic health, gut health, resilience, and independence.

The protein bar may help someone hit a target.

But the bigger question is whether the entire dietary pattern helps the body age well.

Among those in the image above, who will age better?

The answer is not simply either/or. Strength training matters. Protein matters. But the protein craze often overlooks the bigger question and ends up nutrient-poor.

The Protein Halo

Protein has become the “super nutrient.”

It is added to bars, yogurts, cereals, smoothies, cookies, chips, coffee drinks, and pancake mixes. Once a package says “20 grams of protein,” many people assume it is healthy.

That is the protein halo.

But a high-protein cookie is still a cookie. A protein bar can be useful in a pinch, but it is not the same as a meal built from beans, lentils, yogurt, tofu, fish, eggs, vegetables, nuts, seeds, and whole grains.

The danger is not protein itself.

The danger is using protein as a shortcut for health.

Do We Really Need 100 Grams a Day?

The woman at the conference had absorbed one of the most common messages in wellness culture on social media: aim for 100 grams of protein per day.

But that number is not magic.

Protein needs depend on body size, age, health, appetite, physical activity, total calories, and training goals. The current U.S. Dietary Guidelines for Americans, 2025–2030, now list a protein target of 1.2 to 1.6 grams per kilogram of body weight per day, adjusting as needed for individual caloric requirements. That is a major shift from the older adult RDA of 0.8 grams per kilogram per day, which was designed mainly to prevent deficiency rather than optimize muscle or metabolic function. [1,2].

For a 60 kg person, the current U.S. guideline range of 1.2 to 1.6 g/kg/day translates to about 72 to 96 grams per day. So 100 grams is not wildly outside the new federal target, but it is still not a magic number. For a smaller person, it may be at the high end or above it; for a larger or very active person, it may be reasonable.

Her DEXA scan showed more muscle mass, and that is encouraging. But a DEXA scan does not answer every health question. It does not tell us whether the diet improved insulin sensitivity, vascular health, gut microbial diversity, inflammation, appetite regulation, or long-term resilience.

So the question is not only:

Did the protein target increase lean mass?

It is also:

What did it take to get there, and what did it replace?

If reaching 100 grams requires multiple protein bars, fewer legumes, fewer whole grains, fewer fruits and vegetables, and less dietary variety, then the number may be distracting us from the larger goal.

A protein bar may deliver 20 grams of protein.

But it may deliver little fiber, plant diversity, healthy fat, slow carbohydrate, or microbial fuel.

Whole foods deliver protein inside a broader matrix.

That matrix matters.

The Real Tension: Muscle Versus Longevity Signals

The protein debate in aging has a real tension.

On one side, older adults are often told to eat more protein to preserve muscle and prevent frailty. Some studies support this. In the Health ABC Study, older adults with higher protein intake lost less lean mass over three years than those with lower intake [3].

On the other side, aging biology raises a different concern.

In The Hallmarks of Protein and Amino Acid Restriction in Aging and Longevity, Bailey Knopf and Dudley Lamming review evidence that protein restriction, and restriction of specific amino acids such as methionine, isoleucine, and valine, can improve metabolic health, reduce frailty, extend healthspan, and lengthen lifespan in several model organisms [4].

So which is it?

Do we need more protein for muscle, or less protein for longevity?

The answer is that protein should not be treated as a single lever. Protein affects muscle, but amino acids also act as nutrient signals. They influence pathways that regulate growth, repair, metabolism, inflammation, cellular cleanup, mitochondrial function, and aging.

Protein is not just building material.

Protein is also information that affects metabolism.

That means more is not automatically better. It depends on the person, the dose, the source, the exercise pattern, and the whole diet.

Protein Restriction Biology: Why the Science Is Interesting

The science of protein restriction is not about starving the body of protein.

It is about how the body responds when amino acid availability changes while basic nutritional needs are still met.

Knopf and Lamming describe several hallmarks of protein restriction that overlap with major pathways in aging biology.

One is nutrient sensing. Amino acids regulate mTORC1, a growth-related pathway that helps drive protein synthesis and cell growth when nutrients are abundant. That is useful for building muscle after resistance exercise. But chronically high nutrient signaling may also reduce cellular cleanup processes such as autophagy, which help remove damaged proteins and organelles.

Protein restriction can also activate GCN2, a stress-responsive pathway that helps cells adapt to amino acid scarcity. This is part of a broader shift away from constant growth and toward maintenance, repair, and metabolic adaptation.

Another important signal is FGF21, a hormone induced by protein restriction and some amino acid restrictions. In animal and human studies, FGF21 is linked to changes in energy expenditure, glucose metabolism, lipid handling, and insulin sensitivity.

Protein restriction also intersects with mitochondrial function, cellular senescence, and inflammation. In several models, lower protein or selective amino acid restriction improves metabolic health and reduces markers of aging-related dysfunction. Restriction of specific amino acids, especially methionine and some branched-chain amino acids such as isoleucine and valine, can reproduce some of the benefits of overall protein restriction.

This does not mean people should randomly restrict amino acids.

It means the biology of protein is not simple.

Protein is a muscle nutrient, but amino acids are also signals that tell the body whether to grow, repair, conserve, recycle, or adapt.

Healthspan Is Not Just Muscle Mass

The “more protein” argument often assumes that preserving muscle mass is the same as preventing frailty.

It is not that simple.

Muscle matters, but aging well is about function: strength, balance, walking speed, glucose control, recovery from stress, immune resilience, and independence.

Knopf and Lamming point out that higher protein can promote muscle growth, especially with resistance exercise, but protein restriction can blunt age-related increases in frailty in animal studies. In some late-life models, protein restriction reduced lean mass while also reducing or slowing further declines in frailty [4].

That sounds paradoxical only if muscle size is the only outcome.

The better goal is not maximum muscle at any cost.

The better goal is durable function.

Resilience: The Missing Word

A useful concept in the review is resilience.

Resilience means the ability to bounce back from stress: infection, injury, fasting, heat, cold, hospitalization, inflammation, or metabolic strain.

In humans, loss of resilience contributes to chronic disease, multimorbidity, disability, and death. In animal studies, protein restriction has improved some forms of resilience during aging, including responses to starvation, bacterial infection, and heat stress in fruit flies [4].

We do not yet know how much of this translates to humans.

But the concept is important.

Aging nutrition should not only ask:

How much muscle can I build?

It should also ask:

Can this body recover, adapt, resist disease, and remain independent?

Protein may be part of that answer.

But it is not the whole answer.

Okinawa: A Low-Protein Diet Was Not a Frailty Diet

Okinawa is important because it challenges the assumption that aging well requires protein-maxing.

The traditional Okinawan diet was not built around protein bars, powders, or large portions of animal protein. It was largely plant-based, rich in complex carbohydrates, and centered on foods such as sweet potatoes, vegetables, legumes, and soy. Knopf and Lamming note that the traditional Okinawan diet has been estimated at 9% protein, similar to some protein-restriction laboratory diets, and about 80% of calories from plant sources [4].

This does not prove that low protein caused Okinawan longevity.

The Okinawan pattern also included lower calorie intake, daily movement, social connection, and a broader cultural environment that supported aging.

But that is exactly the point.

Okinawa was not a protein-maximizing culture. It was a dietary and lifestyle pattern. The population became famous for longevity, not frailty.

That should make us cautious about reducing aging nutrition to a single protein target.

The Mediterranean Diet: Muscle Without Protein Obsession

The Mediterranean diet tells a similar story in a more contemporary human study.

Mazza and colleagues studied 528 adults aged 50 and older living in southern Italy and examined dietary patterns in relation to handgrip strength, muscle mass, and sarcopenia [5].

The average participant was not protein deficient. Mean protein intake was about 1.0 g/kg/day.

The researchers identified several dietary patterns. The Mediterranean pattern was characterized by higher intake of carbohydrates, plant proteins, and fiber. People with higher adherence consumed more legumes, cereals, vegetables, and fruit, and lower amounts of meat, fish, and eggs.

This pattern was associated with better handgrip strength and lower prevalence of sarcopenia. Low adherence to the Mediterranean pattern was associated with higher odds of low muscle strength and sarcopenia [5].

This matters because the analyses adjusted for daily protein intake per kilogram of body weight.

So the result was not simply “more protein equals better muscle.”

It was a pattern effect.

The authors make the key point directly: the Mediterranean diet is primarily plant-based, about 20% lower in total protein than a typical Western diet, and relies more on legumes, nuts, seeds, and whole grains as protein sources. They argue that protein quality may be more important than quantity in explaining the benefits of the Mediterranean diet for muscle mass [5].

This helps resolve the protein paradox.

A diet can support muscle without making protein the star.

It can provide adequate amino acids while also delivering fiber, unsaturated fats, polyphenols, minerals, antioxidants, and plant diversity.

The Mediterranean diet is not a low-muscle diet.

It is a food-pattern diet.

Why Protein Bars Miss the Point

A protein bar can deliver 20 grams of protein.

But it cannot deliver a dietary pattern.

It cannot reproduce a meal built from legumes, greens, whole grains, fruit, nuts, seeds, olive oil, yogurt, fish, tofu, herbs, and fermented foods. It cannot compress plant diversity, fiber architecture, resistant starches, polyphenols, minerals, food texture, satiety signals, and microbial substrates into a bar.

This matters for the gut microbiome.

The gut microbiome is shaped by repeated exposure to diverse plant foods. Fibers and polyphenols from legumes, vegetables, fruits, whole grains, nuts, and seeds feed microbial communities that produce metabolites such as short-chain fatty acids. These metabolites influence inflammation, glucose handling, gut barrier function, immune tone, and possibly aging biology.

Protein bars are built around density: more protein in less space.

Healthy aging may depend on the opposite: diversity, structure, and food complexity.

The real question is not, “Did I hit my protein target?”

It is:

What kind of internal ecosystem is this diet building?

The Real Deficiency Is Often Fiber

The modern protein craze has a hidden cost.

When protein goes up, something else often goes down.

For many people, that “something else” is carbohydrate. But carbohydrates are not one thing. Soda and lentils are both carbohydrates. White bread and barley are both carbohydrates. Candy and carrots both contain carbohydrates.

When people cut carbohydrates aggressively to make room for protein, they often cut the foods that feed the gut microbiome: beans, oats, whole grains, fruit, vegetables, potatoes, nuts, and seeds.

That is how a high-protein diet becomes a low-fiber diet.

Fiber supports bowel health, gut microbial diversity, short-chain fatty acid production, cholesterol metabolism, glucose control, satiety, and inflammation.

It is not glamorous. It is not a gym supplement. It does not look sexy on TikTok.

But it may be one of the most important missing nutrients in modern diets.

Muscle Still Needs Training

None of this means protein is irrelevant.

Protein supports muscle repair and adaptation. But resistance exercise is the main signal that tells the body to keep muscle.

Someone can eat a high-protein diet and still lose strength if they do not train. Someone can eat adequate protein from whole foods and preserve muscle if they lift, carry, climb, walk, and remain active.

For aging, muscle protection is a system:

Resistance training + adequate protein + enough calories + fiber-rich plants + sleep + metabolic health + recovery.

Protein is one part of that system.

It should not take over the whole conversation.

The Better Question

Instead of asking:

How do I get more protein?

Ask:

Am I getting enough protein from real foods while also getting enough fiber, plant diversity, healthy fats, and exercise?

That changes the meal.

Greek yogurt becomes Greek yogurt with berries, walnuts, and chia.
Eggs become eggs with greens, mushrooms, and whole-grain toast.
Fish becomes fish with lentils, roasted vegetables, and olive oil.
Tofu becomes tofu with vegetables, sesame, edamame, and brown rice.

Protein stays in the meal.

It just stops being the whole story.

Take-Home Messages

• Protein is essential, but more is not always better.

• A 100-gram target is not magic. Needs depend on body size, age, health, appetite, activity, and goals.

• Protein is both building material and biological signal. Amino acids influence mTOR, GCN2, FGF21, autophagy, mitochondria, senescence, and metabolism.

• The frailty argument is more complicated than “eat more protein.” Muscle mass matters, but aging well is about function and resilience.

• Okinawa and the Mediterranean diet challenge protein-maxing. These patterns are plant-rich, fiber-rich, and not built around protein density.

• Protein quality may matter more than protein quantity.

• The gut microbiome needs plant diversity. Fiber, polyphenols, and resistant starches cannot be fully compressed into a bar.

• Muscle needs training, not just protein. Resistance exercise is the main signal.

• The best target is protein adequacy inside a high-quality dietary pattern.

References

1. U.S. Department of Agriculture and U.S. Department of Health and Human Services. Dietary Guidelines for Americans, 2025–2030. 10th Edition. January 2026. Available at RealFood.gov.

2. Bauer J, Biolo G, Cederholm T, et al. Evidence-based recommendations for optimal dietary protein intake in older people: a position paper from the PROT-AGE Study Group. Journal of the American Medical Directors Association. 2013;14(8):542–559.

3. Houston DK, Nicklas BJ, Ding J, et al. Dietary protein intake is associated with lean mass change in older, community-dwelling adults: the Health ABC Study. American Journal of Clinical Nutrition. 2008;87(1):150–155.

4. Knopf BA, Lamming DW. The hallmarks of protein and amino acid restriction in aging and longevity. Cell Press Blue. 2026. doi:10.1016/j.cpblue.2026.100079.

5. Mazza E, Ferro Y, Maurotti S, et al. Association of dietary patterns with sarcopenia in adults aged 50 years and older. European Journal of Nutrition. 2024;63:1651–1662. doi:10.1007/s00394-024-03370-6.

An umbrella review is a review of reviews. Instead of asking what one study found, it asks what the field is showing across systematic reviews, meta-analyses, and the primary studies inside them.

Our question was simple:

Does physical activity appear to matter differently for people who carry APOE4?

The answer was encouraging, but not final.

The strongest favorable signals came from observational studies of younger or cognitively unimpaired adults. In several studies, physical activity was linked to better cognition, lower dementia risk, better brain imaging markers, or better vascular and metabolic health — sometimes more clearly in APOE4 carriers.

But clinical trials have been less convincing. Many enrolled older adults with mild cognitive impairment or dementia. APOE4 subgroups were often small. Interventions were short. Outcomes varied. Many trials asked whether exercise could improve cognition after disease was already underway.

That may be the wrong question.

Exercise may be most powerful before symptoms begin, when vascular, metabolic, inflammatory, and brain-resilience pathways are still modifiable.

Why APOE4 Carriers Should Care

If you carry APOE4, especially two copies, “exercise is good for you” is probably not motivating enough.

You already know that.

The better question is:

Can movement change the biology that makes APOE4 risky?

APOE4 is not only an Alzheimer’s risk gene. It is involved in lipid transport, cholesterol metabolism, vascular function, inflammation, glucose handling, and brain energy use. These are systems that exercise can influence.

Exercise does not erase genetic risk.

But it may improve the biological environment in which that risk is expressed.

For APOE4 carriers, physical activity may not be just about weight or fitness. It may be a way to train the vascular and metabolic systems that support the brain.

What the Positive Studies Suggest

Several studies help explain why physical activity remains interesting for APOE4 carriers.

In the CAIDE cohort, Rovio and colleagues followed adults from midlife into later life. Leisure-time physical activity at least twice per week was associated with lower dementia and Alzheimer’s risk, with more pronounced benefits reported among APOE4 carriers [2].

Krell-Roesch and colleagues found that inactive APOE4 carriers had increased risk of incident mild cognitive impairment compared with active non-carriers. Physical activity in midlife and late life was linked to lower MCI risk [3].

Shih and colleagues reported that APOE4 carriers with low physical activity had more than a threefold higher risk of dementia or cognitive impairment compared with highly active non-carriers. When low activity, APOE4, and diabetes occurred together, risk was nearly tenfold higher [4].

These studies lacked the sample size to directly test whether exercise works better for APOE4 carriers.

They show a simpler pattern: people with APOE4 who are inactive often have higher cognitive risk, so we need better trials to test whether exercise changes that risk.

That finding is important because it points to the system, not just the gene.

APOE4 risk may interact with diabetes, blood pressure, insulin resistance, lipids, inflammation, and vascular health. Exercise is one of the few interventions that touches many of these pathways at once.

Other studies point to brain markers. Physical activity or fitness in APOE4 carriers has been linked to lower amyloid burden, preserved hippocampal volume, better brain activation during memory tasks, and greater cerebral glucose uptake [5–8].

Figure 1. A hypothetical model of protective effects of physical activity (PA) on age-related pathology in APOE ε4 carriers across the lifespan, from Ali et al, 2026 (1).

None of this proves that exercise prevents Alzheimer’s disease in APOE4 carriers.

But together, the studies suggest something worth taking seriously:

Inactivity may be a particularly bad environment for an APOE4 brain.

Why Trials Have Been Disappointing

The trial evidence has not yet matched the observational signal.

That does not mean exercise does not work.

It may mean we have not tested it in the right way.

Many trials enrolled people who already had MCI or dementia. At that stage, the brain may already have substantial synaptic loss, hippocampal atrophy, vascular injury, amyloid, tau, and inflammation. Exercise may still help function, sleep, mood, mobility, falls, and vascular health. But it may be harder to show cognitive rescue.

Some trials were also too small to test APOE4 subgroups properly. Others used broad cognitive tests that may be too blunt. Some measured physical activity by self-report. Some did not prove that fitness actually improved.

A better trial would enroll low-active APOE4 carriers before dementia begins. It would measure movement objectively. It would test whether fitness improves. It would focus on outcomes that exercise is likely to change first: blood pressure, insulin resistance, lipids, inflammation, sleep, cerebral blood flow, hippocampal structure, and blood biomarkers such as p-tau217.

Dementia prevention trials are hard because dementia takes years to develop.

So we need smarter intermediate outcomes.

What Does Sedentary Mean?

Sedentary does not mean lazy.

Sedentary means spending too much waking time sitting, reclining, or lying down with very low energy use.

A practical definition is this:

You sit most of the day, get little intentional movement, and rarely reach moderate-intensity activity.

Many people under about 5,000 steps per day are functionally low-active. But step count does not capture everything. You can reach a decent number of steps and still sit for long blocks.

That matters.

My colleague at USC, David Raichlen, led an analysis of a large cohort that linked sedentary behaviors and dementia risk. This UK Biobank study used wrist accelerometers to measure sedentary behavior in nearly 50,000 older adults. Dementia risk began to rise after about 10 hours per day of sedentary time. Compared with about 9.3 hours per day, risk was higher at 10 hours, much higher at 12 hours, and highest at 15 hours [9].

This does not prove that sitting alone causes dementia.

But it tells us something practical:

The total time spent sedentary matters.

For people who work behind a computer, 10 hours can happen easily: eight hours of work, commuting, meals, and evening screen time.

So the first exercise prescription may not be “join a gym.”

It may be:

Stop letting the chair win the whole day.

The Benefit Is Not Linear

One of the most important ideas in exercise science is also one of the most encouraging:

The benefit is not linear.

Going from sedentary to somewhat active may matter more than going from active to perfectly optimized.

This is good news.

It means the first steps count. The first walk counts. The first week of interrupting long sitting counts. The goal is not to become an athlete. The goal is to stop giving the body a signal of stillness all day long.

For a sedentary person, the first meaningful intervention is often simple:

Move out of the sedentary category.

A person who goes from 3,000 to 6,000 steps per day may gain more biologically than a person who goes from 10,000 to 14,000 steps per day. A person who interrupts sitting every hour may change their daily metabolic pattern even without joining a gym.

This does not mean higher fitness is useless. Fitness matters. Strength matters. Vigorous activity can be valuable.

But the priority is not optimization.

The priority is restoring movement to the day.

A Gym Session Does Not Cancel a Sedentary Day

There is another trap.

Someone may exercise for 45 minutes in the morning and then sit behind a computer for 8 to 10 hours.

That is better than not exercising.

But it does not fully solve the problem.

We were not built to move once and then remain still all day. Human physiology expects repeated muscle contraction, posture change, walking, lifting, bending, carrying, and light movement across the day.

A workout is a dose.

A movement pattern is an environment.

For brain health, the environment may matter as much as the dose.

So the question is not only:

Did I exercise today?

It is also:

How long did I sit without interruption?
How often did I move during the day?
Did my muscles wake up after meals?
Did I spend the whole day in a chair?

This is especially relevant for computer work. A desk job can quietly create 10 hours of sedentary time: work, meetings, commuting, meals, and evening screens.

That is why the practical goal is not only more exercise.

It is less uninterrupted sitting.

Exercise Is Not Just Exercise

When people hear “exercise,” they picture a gym.

But physical activity is broader.

It includes walking to the store, gardening, climbing stairs, cleaning, carrying groceries, dancing, walking after meals, playing with grandchildren, standing from the floor, taking a sunset stroll, and breaking up long sitting time.

Structured exercise matters.

Habitual movement matters too.

The body does not care whether movement came from a gym or from daily lifestyle. It cares that the muscles worked, the heart responded, the blood moved, and the pattern repeated.

The best plan is usually both:

Daily movement plus structured training.

The First Target: Interrupt Sitting

If you are sedentary, the first goal is not VO2 max.

The first goal is to break up long sitting.

Try this:

Every 30 to 60 minutes, stand up and move for 2 to 5 minutes.

Walk around the room.
Climb one flight of stairs.
Do 10 body-weight squats.
Stretch your hips and calves.
Carry laundry.
Water plants.
Walk outside for sunlight.
Take a phone call standing or walking.

These small breaks look unimpressive.

But they repeatedly activate muscle, blood flow, glucose handling, and posture throughout the day.

A single workout is good. A day with repeated movement is better.

Ordinary Movement Counts

Physical activity does not have to look like exercise.

A good hike counts.

Walking the extra distance from the parking lot to the store counts.

Taking the stairs counts.

A morning stroll counts.

A sunset walk counts.

Stretching during the workday counts.

Gardening, cleaning, carrying groceries, walking the dog, dancing, and playing with grandchildren count.

These are not inferior forms of movement. They are the kind of movement human lifestyle used to contain by default.

The body does not care whether movement came from a gym, a trail, a sidewalk, a garden, or a grocery store parking lot.

It cares that the muscles worked, the heart responded, the blood moved, and the pattern repeated.

The goal is not to optimize a workout. The goal is to build a lifestyle the brain can live in.

Steps: Useful, Not Sacred

Steps are a simple way to measure daily movement.

You do not need a wearable, but a phone, watch, ring, or simple pedometer can help you see your baseline.

Measure one normal week.

If you average 3,000 steps per day, aim for 3,500 to 4,000.
If you average 5,000, aim for 6,000.
If you average 7,000, make it consistent.

The first goal is not 10,000 steps.

The first goal is more than your current normal.

For many adults, 7,000 to 10,000 steps per day is a reasonable long-term range. But the biggest gain may come from moving out of the lowest activity category.

A useful rule:

Increase gradually, then make it boringly consistent.

The Walk After Meals

One of the simplest habits is a 10-minute walk after meals.

This is especially useful after lunch or dinner, when blood sugar often rises, and people are likely to sit.

Walking after meals does not require a gym. It does not require special clothes. It does not require motivation for a full workout.

It just requires standing up and going outside.

For APOE4 carriers, this habit makes biological sense because exercise may act through vascular and metabolic pathways: glucose handling, insulin sensitivity, triglycerides, blood pressure, inflammation, and blood flow.

A walk after meals is small enough to repeat.

That is why it works.

The Morning or Sunset Stroll

Not every walk has to be a workout.

An early morning or sunset stroll can combine movement, light exposure, stress reduction, and rhythm.

Morning light helps anchor circadian timing. Evening walks can reduce stress and prevent the drift into hours of sitting after dinner.

These walks may not raise VO2 max dramatically.

That is not the point.

They help build the daily pattern.

And the pattern is the intervention.

Stretching Has a Role

Stretching is not a substitute for aerobic exercise or strength training.

But it helps people move.

Tight hips, stiff ankles, back pain, and poor balance all make activity harder. A few minutes of stretching during the workday can reduce stiffness and make the next walk more likely.

Try simple movements:

Calf stretch
Hip flexor stretch
Hamstring stretch
Chest opener
Thoracic rotation
Neck and shoulder mobility
Gentle yoga or Tai Chi

The goal is not flexibility for its own sake.

The goal is to keep the body available for movement.

Then Add Structured Exercise

Once daily movement is improving, add structured exercise.

Aim for at least 150 minutes per week of moderate aerobic activity, or 75 minutes per week of vigorous activity, plus two days per week of strength training [10].

Moderate means you can talk, but not sing.

Examples include brisk walking, cycling, swimming, hiking, dancing, rowing, or elliptical training.

The simplest plan:

30 minutes, 5 days per week.

If you are starting low, begin with 10 minutes and build.

A good hike once a week can be a powerful anchor. It adds longer aerobic work, hills, sunlight, balance, and often social connection.

That may be better than thinking of exercise only as something done indoors on a machine.

What About VO2 Max?

VO2 max measures how well your body uses oxygen during hard exercise. It is useful in research, but most people do not need a lab test.

You can track fitness in simpler ways:

Can you walk the same route faster?
Can you climb stairs with less effort?
Can you walk uphill without stopping?
Can you cover more distance in six minutes?
Is your resting heart rate lower than before?

Those are practical fitness markers.

If safe, add short intervals once or twice per week:

Warm up for 5 minutes.
Walk fast for 1 minute.
Walk easy for 2 minutes.
Repeat 4 to 6 times.
Cool down.

You do not need to suffer.

You need to progress.

If you have chest pain, severe shortness of breath, dizziness, heart disease, or uncontrolled blood pressure, talk with a clinician before vigorous exercise.

Strength and Balance

Strength training is not optional after midlife.

Muscle is a metabolic organ. It helps glucose control, balance, bone health, mobility, and independence.

Do strength training two days per week.

Include:

Squat or sit-to-stand
Hip hinge
Push
Pull
Carry
Core stability

Use body weight, resistance bands, machines, dumbbells, or household objects.

You do not need to become a bodybuilder.

You need enough strength to keep moving through daily activities.

For older adults, balance also matters. Falls can change the course of aging. Tai Chi, yoga, heel-to-toe walking, step-ups, single-leg stands, and supervised balance work all count [11].

How Do You Know If It Is Working?

Do not expect your memory test to change after 12 weeks.

Do not expect plasma Alzheimer’s biomarkers to clearly improve after a few months of exercise.

Blood biomarkers such as p-tau217 are promising research tools. Future trials should use them. But they are not yet practical personal feedback tools for exercise.

Look for nearer signals.

Your sitting time is lower.
Your steps are higher.
You walk after meals more often.
Your blood pressure improves.
Your resting heart rate falls.
Your walking pace improves.
Stairs feel easier.
Your sleep becomes more regular.
Your waist changes, even if weight does not.
Your triglycerides fall.
Your HDL rises.
Your glucose or A1c improves.
Your strength improves.
Your balance improves.
Your mood and energy improve.

For APOE4 carriers, LDL cholesterol may or may not fall with exercise alone. LDL is often influenced by genetics, diet, saturated fat intake, body weight, and medication.

That does not mean exercise failed.

It means exercise is one part of the system.

The first win is not a biomarker. The first win is changing the daily pattern.

Does Exercise Work Alone?

Exercise helps, but it does not live alone.

A person who exercises but sleeps poorly, eats a highly processed diet, has untreated sleep apnea, uncontrolled blood pressure, high stress, social isolation, or insulin resistance is still pushing the brain in the wrong direction.

This is not because people need a perfect protocol.

They do not.

It is because biology is a network.

Movement improves glucose handling, blood pressure, sleep, mood, inflammation, and muscle metabolism. Better sleep improves appetite, stress hormones, amyloid clearance, and immune regulation. A healthier diet supports the microbiome, lipids, and inflammation. Social connection lowers chronic stress signaling.

These are not separate boxes.

They reinforce each other.

Exercise may be the best first domino because it often makes the next healthy behavior easier.

Uncertainty

We should be honest.

We do not yet have definitive proof that exercise prevents dementia specifically in APOE4 carriers.

Most favorable evidence comes from observational studies. Active people may differ from inactive people in diet, education, sleep, income, medical care, vascular risk, and social connection.

The trials so far have not settled the question.

But the absence of definitive proof is not the same as evidence of no benefit.

Exercise is biologically plausible, low-cost, and strongly beneficial for systems that matter to the brain: blood vessels, metabolism, muscle, sleep, mood, inflammation, and independence.

For APOE4 carriers, that is enough reason to take it seriously while still asking for better trials.

Take-Home Messages

• Exercise for APOE4 is not just gym advice. The goal is to improve the vascular, metabolic, inflammatory, and brain-resilience systems that shape how genetic risk is expressed.

• Timing may matter. The most encouraging signals appear earlier in life, before dementia, especially in cognitively unimpaired adults.

• The benefit is not linear. Moving from sedentary to somewhat active may matter more than moving from active to highly optimized.

• Sitting is its own problem. A morning workout is helpful, but it does not fully erase 8 to 10 hours of uninterrupted sitting.

• We are built to move repeatedly. Walking, standing, stretching, carrying groceries, taking stairs, gardening, hiking, and short movement breaks all count.

• Steps are useful, not sacred. Measure your baseline if it helps, then increase gradually. The first goal is not 10,000 steps. It is more than your current normal.

• Look for near-term wins. Better blood pressure, sleep, glucose control, waist size, triglycerides, HDL, strength, balance, mood, and energy may appear long before any change in memory or Alzheimer’s biomarkers.

• Exercise works best as part of a lifestyle pattern. Diet, sleep, blood pressure control, metabolic health, stress reduction, social connection, and cognitive engagement reinforce each other.

• Do not overpromise. Exercise is not a guaranteed shield against Alzheimer’s disease.

• Do not underuse it either. For APOE4 carriers, becoming a person who moves throughout the day may be one of the most rational places to start.

Citation and Acknowledgment

Ali N, Chakbazof N, Ghasem Pour S, Contreras L, Estrada J, Alexander GE, Raichlen DA, Yassine HN. APOE ε4, physical activity, and the brain: a review of systematic reviews. Frontiers in Aging Neuroscience. 2026;18:1798639. doi:10.3389/fnagi.2026.1798639.

The work was supported in part by the National Institutes of Health/National Institute on Aging, the Alzheimer’s Drug Discovery Foundation, donations from the Vranos and Tiny Foundations and Lynne Nauss, and support to Gene E. Alexander from the Arizona Department of Health Services and the McKnight Brain Research Foundation.

The authors have no conflicts of interest.

References

1. Ali N, Chakbazof N, Ghasem Pour S, et al. APOE ε4, physical activity, and the brain: a review of systematic reviews. Frontiers in Aging Neuroscience. 2026.

2. Rovio S, et al. Leisure-time physical activity at midlife and the risk of dementia and Alzheimer’s disease. Lancet Neurology. 2005.

3. Krell-Roesch J, et al. Physical activity and incident cognitive impairment: the Mayo Clinic Study of Aging.

4. Shih IF, et al. Physical activity modifies the influence of APOE ε4 and type 2 diabetes on dementia and cognitive impairment. Alzheimer’s & Dementia. 2018.

5. Head D, et al. Exercise engagement as a moderator of APOE effects on amyloid deposition.

6. Smith JC, et al. Physical activity, APOE ε4, and brain activation during memory processing.

7. Woodard JL, et al. Physical activity, APOE ε4, hippocampal volume, and cognitive decline.

8. Deeny SP, et al. Exercise, APOE ε4, and cerebral glucose metabolism.

9. Raichlen DA, Aslan DH, Sayre MK, et al. Sedentary behavior and incident dementia among older adults. JAMA. 2023;330(10):934–940.

10. CDC. Adult physical activity recommendations.

11. CDC. Physical activity recommendations for older adults.

12. del Pozo Cruz B, Gallardo-Gómez D, del Pozo-Cruz J, Ding D. How many steps a day to reduce the risk of all-cause mortality? A dose-response meta-analysis. Journal of Internal Medicine. 2022;291(4):519–521.

13. Ekelund U, Tarp J, Steene-Johannessen J, et al. Dose-response associations between accelerometry measured physical activity and sedentary time and all-cause mortality. BMJ. 2019;366:l4570.

The short version: landing in the highest p-tau217 group is an important finding, and it’s also, like most things in medicine, probabilistic rather than deterministic. It meaningfully shifts the odds of decline. It does not function like a positive or negative diagnostic result, and it does not tell a person what their future holds. Most people in that highest group, in every dataset discussed below, did not go on to develop dementia within the years these studies were able to follow them. That distinction, informative versus determinative, is the throughline of everything that follows. (More on how to weigh that balance, with a real-world comparison, at the end.) Today, p-tau217’s clearest use is identifying good candidates for prevention trials, replacing expensive and complex PET scans; but it is not yet ready for clinical use, and current guidelines recommend against using it to screen healthy people outside of research.

What does p-tau217 actually measure?

p-tau217 is a small piece of tau protein that leaks out of brain cells into blood. Its release is triggered mostly by amyloid, the sticky protein plaques that build up early in Alzheimer’s disease. So for most of its range, a rising p-tau217 mainly means “more amyloid,” not “more tau tangles.” Only later in the disease does it start reflecting actual tau buildup too, the kind seen on a specialized brain scan called a tau PET. That scan lights up wherever tangled tau protein has physically accumulated inside neurons; tau doesn’t spread evenly, it tends to start in one small area and then move into connected regions over time, the way an outbreak spreads along a transit line rather than raining down on a whole city at once.

Three more things are worth knowing in plain terms. A Centiloid is just a 0-to-100-plus scale for how much amyloid shows up on a brain scan, built so different scanners and tracers can be compared fairly; 0 is a young, healthy brain, 100 is roughly what’s typical in mild-to-moderate Alzheimer’s dementia. Harmonizing means converting each lab’s raw number onto a shared scale (usually “how many standard deviations from average”), since different test brands don’t report the same raw numbers for the same biology. And logistic regression, a term that shows up later, is simply the standard math for turning a measurement into a probability, like “what’s the chance this person’s tau has started accelerating,” producing a curve rather than a hard yes/no line.

How results get labeled, and why that label isn’t the whole story

One widely used commercial test (Lumipulse, made by Fujirebio) sorts results into three bands:

Category Lumipulse p-tau217 (pg/mL) Interpretation Negative ≤ 0.18–0.20 Alzheimer’s-type pathology unlikely Indeterminate / gray zone 0.18–0.32 Borderline; further evaluation recommended Positive ≥ 0.32–0.33 High probability of Alzheimer’s-type pathology

What these numbers actually mean: these specific cutoffs were derived by comparing symptomatic memory-clinic patients with confirmed Alzheimer’s pathology (via PET scan or spinal fluid) against those without it. They mark where, in that particular symptomatic population, the test best tells the two groups apart, not a universal biological line in the sand. Other manufacturers (Eli Lilly, Roche Elecsys, C2N, and others) use different chemistry and validate their own, different pg/mL cutoffs. There is no single, agreed-upon p-tau217 cutoff across labs today. That lack of a uniform cutoff is one of the main reasons this test cannot yet support a specific clinical recommendation, and it’s worth remembering as a take-home point on its own.

Just as important regardless of which band a result falls in: no p-tau217 result should replace an actual cognitive evaluation. A “low” number doesn’t mean skip a checkup if something feels off, and a “high” number doesn’t mean a diagnosis has been made. The blood test is one input, not a verdict, and it hasn’t been validated for screening healthy people outside of research studies.

Research studies mostly sidestep these fixed bands. Instead they rank people against each other, low, intermediate, high, “very high,” based on their own study population. So “very high” in a study (roughly the top fifth of people, corresponding to substantial amyloid buildup on a brain scan) is a relative ranking within that study, not a number you could match against a personal lab report using the commercial bands above.

The JAMA study: what nearly 2,700 people showed

This study pooled six research groups (ADNI, the A4 and LEARN studies, Harvard Aging Brain Study, Wisconsin Registry for Alzheimer’s Prevention, and HABS-HD), for 2,684 cognitively healthy older adults followed a median of about 5.4 years, up to 13.5 years. Everyone had both a blood test and a brain amyloid scan at the start.

One term needs defining clearly: “progression.” Here it meant reaching a diagnosis of mild cognitive impairment (MCI), dementia, or two consecutive visits showing very mild impairment on a standard staging scale. A separate, second measure tracked gradual decline on cognitive testing, without requiring a formal diagnosis; it told the same story.

The results, in plain numbers:

• Low or intermediate p-tau217 (roughly under 25 Centiloids of amyloid): just a 12–15% chance of progressing over 5 years. Genuinely reassuring.

• High p-tau217: 24% over 5 years, 62% over 10.

• Very high p-tau217 (roughly the top fifth, over 60 Centiloids): 38% over 5 years, 78% over 10, about four times the lowest group’s risk.

Here’s the point worth sitting with: even in that highest-risk group, nearly two-thirds had not progressed after 5 years, and even at 10 years, about 1 in 5 still hadn’t. A high result shifts the odds substantially. It doesn’t seal anyone’s fate.

Some honest limits: these six groups were mostly white, well-educated research volunteers, not a random slice of the public, so results may not generalize perfectly. The four risk groups also overlap somewhat rather than being sharply distinct categories, deaths weren’t tracked as a competing outcome, and only 5% of people were followed past 10 years, so the longest-term numbers are least certain. Risk also ran a bit higher in men and older participants at the same p-tau217 level, a reminder that one number doesn’t mean the same thing for everyone.

A quick note on sources, because three different tools appear in this piece and they shouldn’t be treated as one. The JAMA study above used its own six research groups and its own definition of progression. A separate paper from the Mayo Clinic Study of Aging, cited by the JAMA authors, used a different measure (brain amyloid scans rather than blood tests) to estimate risk specifically of MCI. And the Mayo calculator used below is a third, distinct tool, built on blood p-tau217, projecting decades into the future using its own statistical model. All three are useful. None of their numbers should be lined up against each other as if they were measuring the exact same thing.

Seeing it with real amyloid numbers

A related analysis (from the A4 and LEARN studies) split people into thirds by p-tau217 and matched each third to actual brain amyloid levels: roughly 47, 61, and 84 Centiloids for the low, middle, and top thirds. The top third showed a clearly steeper rise in confirmed decline over six years than the bottom third, whose amyloid levels averaged only about 4 Centiloids.

A worked example: the Mayo calculator

The Mayo Clinic Study of Aging built a public tool (the Mayo CACR calculator) that turns an actual p-tau217 number, plus age, sex, and genetics, into a personalized curve. It’s a different, separate model from the JAMA study above, using its own long-term projections, so treat the two as complementary rather than directly interchangeable; its multi-decade curves are computed projections, not observed follow-up. Readers curious to see how the numbers shift for a different age, sex, or p-tau217 value can try it directly at the link above.

Running it for a 60-year-old woman carrying the APOE4 gene (more on that below), at three “positive-range” p-tau217 values:

By age 80 (20 years out) p-tau217 0.2 p-tau217 0.3 p-tau217 0.4 Chance of MCI 26% 34% 40% Chance of dementia 11% 17% 24%

Even within this one “positive” range, risk more than doubles. That’s the main lesson: a positive result isn’t one fixed risk level; it behaves like a sliding scale.

And here’s the projection worth remembering most: even at 0.4 pg/mL, the highest value tested here, the model estimates about three in four people (76%) may not develop dementia within 20 years. This example is not exactly representative; it is based on cognitively normal individuals with AD-level brain amyloid deposition (rare occurrence). This is also a simulation, not an observed outcome, so treat it as a projection rather than a fact about any real person’s future.

Picture 100 women exactly like this one, modeled forward: somewhere between 11 and 24 are projected to develop dementia by their early 80s, and the rest are projected not to, within that stretch of time, though real life includes things this kind of model can’t fully capture, other illnesses, earlier death from unrelated causes, or simply better or worse luck than average. A high biomarker is a real signal. It is not a countdown clock.

A useful way to weigh how important this all is: a classic analysis validating the Framingham risk model found men with the highest total cholesterol (280 mg/dL or above) were about twice as likely to have a coronary heart disease event as men with a more typical level (160–199 mg/dL). That twofold gap is in the same range as the two- to fourfold difference this JAMA study found between its highest and lowest p-tau217 groups. High cholesterol built an entire field of preventive cardiology on a risk difference of that size, and nobody dismisses it as minor. Yet no one calls a high cholesterol reading a diagnosis: most people with high cholesterol never have a heart attack. Both things are true of p-tau217 too. A high result is a real, medically meaningful risk signal, and it still isn’t a prediction about any one person’s future.

Genetics changes the math: the Brain study on APOE4

The APOE4 gene is the strongest common genetic risk factor for Alzheimer’s. A study in Brain, using two independent groups of people (one from ADNI, one from the A4-LEARN study), asked a sharp question: does carrying APOE4 change what a given p-tau217 number means?

The answer split in an interesting way. APOE4 did not make amyloid trigger more p-tau217 release. But it did make a given p-tau217 level more likely to trigger actual tau spread across the brain, and the effect grew with each copy of the gene. Put simply: carrying APOE4 seems to lower the bar for tau pathology to start actively spreading, at the very same blood test number.

Using statistical modeling, the researchers estimated the p-tau217 level marking a 50% chance of that tau-spread milestone: in one group, about 0.62 pg/mL for non-carriers, 0.34 for one copy of APOE4, 0.15 for two copies. A second, independent group confirmed the same pattern using a different lab test, though the specific numbers from different assay brands shouldn’t be compared to each other directly, they’re not on the same scale.

A separate comparison using the Mayo calculator points the same direction:

p-tau217 (pg/mL) Carrier: dementia risk by 80 Non-carrier Gap 0.2 11% 9% 2 points 0.3 17% 14% 3 points 0.4 24% 20% 4 points

The gap between carriers and non-carriers widens as p-tau217 rises. That’s consistent with the Brain study’s finding, though this specific comparison comes from a separate tool and isn’t direct proof of the same mechanism.

One nuance worth flagging: does this hold at every level, or only some? The Brain study’s effect was clearer in the group with more advanced disease and murkier in the earlier-stage group, so it’s not yet settled whether the genetic effect is constant from the very earliest biology all the way through, or whether it grows stronger over time. That distinction matters for anyone hoping to use a single p-tau217 cutoff for everyone: the same number may carry meaningfully different risk depending on genotype, and possibly depending on how far along someone already is.

And the thresholds identified for “tau starting to spread” are low in absolute terms, especially for APOE4 carriers, yet the resulting dementia risk, as shown above, still stays fairly modest for years afterward. That’s not a contradiction: crossing that biological threshold marks the quiet start of a process that typically takes additional years to turn into noticeable decline. Think of it as an early warning light coming on in a car with a lot of road left before the engine actually stalls.

Some real limits here too: a key brain region (the hippocampus) had to be excluded from the tau scans for technical reasons, the “spread” maps rely on estimated brain wiring rather than confirmed anatomy, and blood p-tau217 reflects tau leaking out of cells, not the actual tangles forming inside them. The two research groups also used different lab-test brands, so their exact pg/mL cutoff numbers, and the Mayo calculator’s numbers above, shouldn’t be treated as the same scale even where the figures happen to look similar.

What’s coming next: chasing a “goldilocks” window

That gap between early biology and later symptoms is exactly what the newest prevention trials are trying to exploit. The idea: treat people early enough that amyloid is clearly present (worth the risks of a drug) but before tau pathology has spread very far, since once tau is well underway, removing amyloid may not change the trajectory much anymore. Call it a goldilocks window, not too early, not too late.

Three trials are testing this now. TRAILBLAZER-ALZ 3 (donanemab) screened over 63,000 people with plasma p-tau217 to find about 2,200 eligible participants meant to have amyloid with only early tau. AHEAD 3-45 (lecanemab) splits people into two amyloid-level groups, an earlier group with moderate amyloid buildup and a more advanced group with higher amyloid, and uses blood tests mainly to cut down on unnecessary brain scans rather than to gatekeep tau status directly. PrevenTRON, a newly announced trial of a drug called trontinemab, plans to enroll around 1,600 people selected largely by an elevated blood result, with the main goal of measuring time until noticeable decline begins.

Here’s the catch, and it connects right back to the earlier point that p-tau217 partly reflects tau, not just amyloid: using it alone to find “amyloid present, tau still limited” people risks letting some people in who already have more tau than intended. TRAILBLAZER-ALZ 3’s own data show this plainly, roughly 1 in 6 to 1 in 4 of its p-tau217-selected participants already had elevated tau on a brain scan. That’s less a flaw than an unavoidable trade-off: scanning 63,000 brains isn’t realistic, so a blood test is used instead, at some cost to precision.

It’s also worth saying plainly that this whole “goldilocks” idea comes mostly from looking back at subgroups of older trials, not from a trial designed around it from the start. Ideas like that often look weaker, or disappear, once tested directly instead of spotted after the fact. These three trials are a real test of whether the idea holds up.

Research tool, not a clinical test

To repeat the core message: this blood test is not ready for ordering on a healthy person outside of research. Official guidelines recommend against it in routine care, and even the JAMA study’s own authors describe their findings as useful for research and trial design, “not yet precise enough to guide individual prognosis.”

What is a very high p-tau217 good for today? Mainly, identifying good candidates for prevention trials, exactly how it’s used in the trials described above. It is not a trigger for a specific medical decision, because there’s no approved treatment yet that a healthy person’s blood result would unlock. The general advice for someone with a high result, manage vascular risk factors, stay physically and mentally active, sleep well, avoid excess alcohol, is the same advice anyone at elevated risk should already be getting, biomarker or not. Being “eligible for a trial” and “needing a different care plan” are not the same thing, and it’s easy to blur the two. Larger, more representative studies, and clearer genetics- and lab-specific calibration, are still needed before that changes.

Take-home messages

• A very high p-tau217 result is a genuinely important finding. It meaningfully raises the odds of future cognitive decline in a healthy older adult, comparable in scale to how a marker like high cholesterol raises heart disease risk.

• Important is not the same as certain. Most people, even in the highest-risk group, do not progress within the years these studies could follow. A strong signal shifts the odds; it doesn’t decide any one person’s fate.

• There’s no universal cutoff yet. Different labs and assays define “positive” differently, largely because those cutoffs were built for diagnosing symptomatic patients, not screening healthy ones.

• A blood result never replaces a clinical evaluation, whatever the number says.

• Genetics changes what a given number means. Carrying APOE4 appears to shift the biology, so the same result may not carry the same meaning for everyone.

• Long-range projections are models, not facts. Multi-decade risk curves are useful for intuition, not for predicting any individual’s future.

• New prevention trials are testing an appealing but unproven idea: treating people early enough to matter, before too much tau has spread. Whether that strategy actually works is still an open question.

• This is a research tool, not a clinical test. Current guidelines recommend against using it to screen healthy people outside of research studies and trials.

References

1. Buckley RF, Townsend DL, Birkenbihl CJ, et al. Prognostic value of blood-based p-tau217 levels for progression to cognitive impairment. JAMA. Published online July 14, 2026. doi:10.1001/jama.2026.12556

2. Sperling RA, Donohue M, Rissman R, et al. Amyloid and tau prediction of cognitive and functional decline in unimpaired older individuals: longitudinal data from the A4 and LEARN studies. J Prev Alzheimers Dis. 2024;11(4):802-813.

3. Jack CR Jr, Hu M, Wiste HJ, et al. Lifetime and 10-year absolute risk of cognitive impairment in relation to amyloid PET severity: a retrospective, longitudinal cohort study. Lancet Neurol. 2025;24(12):1016-1025.

4. Mayo Clinic Study of Aging. CACR p-tau217 calculator (Fujirebio Lumipulse Research Use Only assay, version 2). Available at: rtools.mayo.edu/CACR. Accessed July 2026.

5. Steward A, Dewenter A, Hirsch F, et al. ApoE4 lowers the ptau217 threshold for tau aggregation and spread in an allele dose-dependent manner. Brain. 2026;149(6):1902-1914. doi:10.1093/brain/awaf463

6. Palmqvist S, Whitson HE, Allen LA, et al. Alzheimer’s Association Clinical Practice Guideline on the use of blood-based biomarkers in the diagnostic workup of suspected Alzheimer’s disease within specialized care settings. Alzheimers Dement. 2025;21(7):e70535.

7. Yaari R, Holdridge KC, Williamson M, et al. Donanemab in preclinical Alzheimer’s disease: screening and baseline data from TRAILBLAZER-ALZ 3. Alzheimers Dement. 2025;21(9):e70662. doi:10.1002/alz.70662

8. Rafii MS, Sperling RA, Donohue MC, et al. The AHEAD 3-45 study: design of a prevention trial for Alzheimer’s disease. Alzheimers Dement. 2023;19(4):1227-1233. doi:10.1002/alz.12748

9. Roche/Genentech. PrevenTRON: rationale and design of a Phase III trial of trontinemab in cognitively unimpaired individuals with biomarkers of Alzheimer’s disease at high risk of clinical decline. Presented at the Alzheimer’s Association International Conference (AAIC), London, July 2026.

10. Gander J, Sui X, Hazlett LJ, Cai B, Hébert JR, Blair SN. Factors related to coronary heart disease risk among men: validation of the Framingham Risk Score. Prev Chronic Dis. 2014;11:E140. doi:10.5888/pcd11.140045

Finding the right population

The first problem for lifestyle trials is deciding who to enroll. Enroll people already impaired, and you’re fighting neurodegeneration built up over a decade. Enroll young and healthy people, and you may need twenty years before enough develop dementia to measure anything. FINGER recruited adults aged 60–77 with an elevated CAIDE dementia risk score (vascular risk factor score) and cognition at or just below what’s expected for their age [1]. U.S. POINTER cast a wider net: about 2,100 older adults, 78% with a family history of memory loss and 30% carrying the APOE-e4 gene copy, recruited to include groups historically underrepresented in dementia research [2]. MIND enrolled people with a family history of dementia, a BMI over 25, and — this is worth being precise about — a MIND diet screener score of 8 or less out of 14, meaning eligibility required eating poorly by that specific measure [3]. That’s a genuinely hard recruitment target, and credit is due for the effort: of roughly 1,900 people screened, more than 500 were excluded for already eating too well on the MIND scale. This was not a casual “self-reported suboptimal diet” — investigators actively went looking for a diet-deficient population.

There’s a further wrinkle worth naming, though: advantaged and diet-poor are not the same thing. FINGER and POINTER recruited educated, engaged older adults with reasonable baseline access to exercise, social stimulation, and health care — capable of sustaining a multi-year program. MIND, despite deliberately selecting for a poor baseline diet, still enrolled a demographically similar cohort on nearly every other axis: 87.7% white and an average of 16.9 years of education, recruited through two academic medical centers. A low MIND-diet score and genuine socioeconomic disadvantage are different problems, and it’s easy to conflate them. Layering an elaborate program onto an already-favorable baseline — of education, health-system access, and time to participate — has a much smaller ceiling to push against than aiming the same idea at a population without stable food access, safe places to exercise, or equivalent support.

That reframes where the bigger opportunity might sit. A far cheaper intervention — reliable access to real food, a free community fitness membership, social infrastructure as a safety net — delivered to a population currently missing all of that could plausibly produce a larger absolute effect than an elaborate protocol layered onto people already doing reasonably well.

The FINGER investigators noticed this population gap. World-Wide FINGERS has expanded to more than 40 countries, including Africa-FINGERS, now piloting structured multidomain coaching across several African nations with genuine community co-design [4]. But what’s being exported is still the FINGER template — coached diet counseling, structured exercise, cognitive training, vascular monitoring — adapted in delivery, not in kind. A different, more urgent question: does closing the gap itself — food security, safe places to exercise, a functioning safety net — produce more benefit, more cheaply, than a coached protocol layered on scarcity? That trial, as far as I can tell, hasn’t been run.

None of that is meant to sound easy. Recruiting and retaining a genuinely underserved population in a multi-year lifestyle trial is its own serious problem — transportation, work schedules, trust in research institutions, and the competing demands of poverty all erode compliance in ways a motivated, retired, academic-medical-center-adjacent volunteer pool doesn’t face. That’s precisely why a standard multi-year RCT of the FINGER or POINTER kind may be the wrong instrument here. Pragmatic and decentralized trial designs — community-embedded cluster randomization through existing food-assistance or primary-care infrastructure, stepped-wedge rollouts, or remote and digital monitoring that cuts down on repeated clinic visits — are better suited to testing a basic-access intervention in a population a conventional trial would struggle to retain.

Choosing the intervention — and the control

FINGER and POINTER both bundled diet, exercise, cognitive training, and vascular risk management together. MIND isolated one variable, testing a hybrid Mediterranean-DASH diet against a mild-calorie-restriction control [3]. But the more revealing design choice, across all three, is what the comparison group actually received — and in every case, it wasn’t nothing. FINGER’s control got “regular health advice” on diet, exercise, and vascular risk. POINTER’s self-guided arm attended six peer group meetings, got general health education across every domain, and received small incentives [2]. MIND’s control diet produced almost exactly as much weight loss as the MIND diet itself — about five kilograms in both arms over three years, with real protective effects on blood pressure, lipids, and insulin sensitivity baked into both arms before the specific food pattern is even considered.

MIND’s design raises a further question, and it’s baked into the score itself, not just the intervention. The MIND diet index sums 15 food groups — 10 to eat more of (leafy greens, other vegetables, nuts, berries, beans, whole grains, fish, poultry, olive oil, wine) and 5 to eat less of (red meat, butter, cheese, pastries and sweets, fried food). It’s an optimization checklist, built to reward a specific idealized eating pattern, not a measure targeted at ultra-processed food intake, added sugar, or fiber as such. Someone can score low on MIND simply by skipping berries, nuts, and olive oil, whether or not their diet is high in ultra-processed food; someone else could hit enough of the positive categories to score reasonably while still eating a lot of packaged, sugar-heavy, low-fiber food the index was never built to catch. The typical U.S. diet MIND’s participants started from derives roughly 55% of calories from ultra-processed foods [5], and both low fiber and high ultra-processed intake are independently linked to worse cognitive trajectories via gut-brain and inflammatory pathways [6,7] — pathways the MIND score, by design, wasn’t built to target directly. Cutting calories, ultra-processed foods, and raising fiber may be the bigger lever than a curated food list of berries, olive oil, and nuts layered on a diet still built around what it should avoid.

That matters for how to read the results. The real question isn’t “does lifestyle change help, compared with nothing.” It’s narrower: does an intensive, expert-coached optimization protocol add much on top of the active ingredient the comparison group already shared — guidance in FINGER, structured support in POINTER, weight loss in MIND? In all three, the honest answer is: a little, but not a lot — a very different message than “breakthrough lifestyle program prevents cognitive decline,” which is closer to how these trials were actually presented.

Reading the effect sizes honestly

This is where the field’s self-presentation runs ahead of the data. FINGER’s primary outcome, a composite neuropsychological z-score, improved 0.20 in the intervention group and 0.16 in control over two years [1] — an absolute difference of about 0.04 standard deviations (see illustration below). Reported as “25% greater improvement than control,” that’s true, but it’s a relative figure built on a very small absolute one, and the same arithmetic produced the widely quoted 83% gain in executive function and 150% gain in processing speed.

A more honest yardstick is Cohen’s d, the standard way to express an effect on a common scale, alongside its plain-language twin, the common language effect size: the odds a random person from the structured group scored higher than a random person from the comparison group. Using FINGER’s own model-based d ≈ 0.13, that’s roughly a 54-in-100 chance — barely better than a coin flip. Given the comparison group wasn’t null but a lighter-touch version of the same idea, that’s about what you’d expect: the marginal value of intensive structuring on top of ordinary guidance, not lifestyle change measured against nothing. (This figure comes from the trial’s own supplementary analyses; check the appendix directly.)

What’s notable is less the size of that number than how it’s been communicated: as a landmark breakthrough, driving funding to replicate the FINGER model in more than 60 countries through World-Wide FINGERS. A modest, real, worth-pursuing signal is not a breakthrough.

U.S. POINTER’s own translation — “equivalent to one to two years of slowed cognitive aging” — deserves similar scrutiny. That figure is a model-based extrapolation against an assumed normative decline rate, not a directly observed outcome — a generous way to make a small number sound tangible.

Set against the wider record, FINGER and POINTER look like the best case, not the norm. MAPT [8] found no significant effect on cognitive decline; preDIVA [9] — the rare trial actually powered to detect a difference in dementia incidence itself — found none either; nor did HATICE [10] (see table below). “Mixed and modest” is a fair summary of this literature.

preDIVA’s null result also points to a structural problem: incident dementia is an extraordinarily hard outcome to power a trial on, requiring thousands of participants and years of follow-up — why most trials fall back on cognitive composites as a proxy, the very proxies producing the small, hard-to-interpret effect sizes above. This is a core reason nutrition and lifestyle trials are structurally harder to design than drug trials: drugs can often show a biomarker signal, while lifestyle change needs years before it plausibly shows up in diagnosable disease.

Effect sizes across the major prevention trials, at a glance:

Trial Comparison Primary effect size Plain-language interpretation FINGER (2015) [1] Coaching vs. general health advice, 2 yrs Composite z: 0.20 vs 0.16 (Δ≈0.04 SD); d ≈ 0.13 ~54-in-100 chance a treated person outscores a control person — barely above a coin flip U.S. POINTER (2025) [2] Structured vs. self-guided coaching, 2 yrs 0.029 SD/yr greater, structured arm Marketed as “1–2 years of slowed cognitive aging” — a model-based extrapolation, not a measured outcome MIND (2023) [3] Diet vs. calorie-restriction control, 3 yrs No significant difference in cognition or MRI Both arms improved and lost ~5 kg; weight loss, not the food pattern, may be the active ingredient MAPT (2017) [8] Coaching + omega-3 vs. placebo, 3 yrs No significant effect on cognitive decline Null on the primary outcome preDIVA (2016) [9] Nurse-led vascular care vs. usual care, 6 yrs Dementia incidence: 7% vs 7%; HR 0.92 Null; the rare trial actually powered on dementia diagnosis, not a proxy HATICE (2019) [10] Internet-coached vascular care vs. control, 18 mo No significant cognitive effect Modest improvement in cardiovascular risk factors only

The two columns look like a verdict — "significant" versus "null" — but the underlying effect sizes are small enough on both sides that the line between them is closer to a statistical threshold than a true difference in what these interventions accomplished. A slightly larger sample, or a slightly different outcome measure, could plausibly have moved several of these trials across the line in either direction.

Choosing outcomes that match the mechanism

Layered on the incidence problem is a measurement problem, and it isn’t just that standard batteries are blunt — they may be tuned to the wrong domain. Look again at FINGER’s own breakdown by cognitive domain: the largest relative gains were in executive function (83%) and processing speed (150%), while memory — the domain most closely tied to classic Alzheimer’s pathology, and the one nearly every AD drug trial is built around — showed the smallest relative gain, at 40%. That pattern fits what a lifestyle intervention plausibly does: support attention, task-switching, and processing speed through cardiovascular and metabolic pathways, rather than directly reversing amyloid-and-tau-driven memory decline.

Standard batteries, though, are built around exactly the domain lifestyle interventions seem to move least. Memory subtests dominate most AD-oriented composites, because memory is what clinicians use to diagnose dementia. Executive function and processing speed are typically assessed with a handful of brief, coarse paper-and-pencil tasks — a few minutes of Trail Making or digit-symbol substitution — far less sensitive to subtle change than a memory test built and validated for detecting AD. If the domains most responsive to diet and exercise are the ones our batteries measure worst, that could make a real effect look smaller and noisier than it is, on top of the floor, ceiling, and practice effects already limiting these tools in healthy populations [11,12].

This is where digital assessment tools have a genuine edge, beyond being reliable or convenient. They can deliver dozens of processing-speed and executive-function trials in a single sitting, repeated over months from a phone, with millisecond-level precision a stopwatch and paper form can’t match — capturing graded change in reaction time and task-switching cost, not one coarse pass-fail score. Repeated at-home testing also averages out day-to-day noise and practice-effect drift, which often swamps small true effects in one-off batteries. A recent scoping review found digital tools improve reliability and ecological validity relative to a single in-clinic assessment [13]; the sharper opportunity is fitting the tool to the domain — fine-grained, repeatable measures of processing speed and executive function as primary outcomes, rather than composite scores weighted toward a memory domain a diet-and-exercise program was never especially likely to move.

What basic science suggests we might be missing

A newly published mouse study adds a mechanistic layer here, not because mouse data settle anything about human trials, but because it shows how cell-type-restricted an exercise response can be. Researchers used single-nucleus RNA sequencing to identify which brain cells respond to exercise in the hippocampal dentate gyrus of an Alzheimer’s mouse model [14]. The response was concentrated in immature neurons, in a neurovascular-associated astrocyte population depleted in Alzheimer’s mice but partly restored by exercise, and in oligodendrocyte progenitor cells, which had the largest share of disease-related genes normalized by exercise of any cell type studied. An earlier study similarly found that long-term voluntary exercise reversed cognitive impairment in an Alzheimer’s mouse model through astrocyte remodeling, without changing amyloid-beta plaque burden [15]. Exercise may not clear the disease’s hallmark pathology so much as build a parallel, cell-specific form of resilience around it.

The clearest thread connecting these findings to the human domain pattern above runs through the neurovasculature unit. Both studies converge on astrocytes, cells that couple neural activity to blood flow and glucose delivery — so an exercise-induced astrocyte response plausibly improves neurovascular efficiency rather than clearing amyloid or tau directly. That fits a well-established pattern in the vascular cognitive impairment literature: circuits depending on blood flow and white-matter integrity, largely underlying processing speed and executive function, are especially sensitive to vascular health, while memory decline is more specifically driven by amyloid and tau damage to hippocampal circuits. If exercise, weight loss, and vascular risk management work substantially through this route, that would explain why FINGER’s and POINTER’s clearest signal was in executive function and processing speed, not memory — a coherent hypothesis, not a demonstrated mechanism, but better-specified than “lifestyle helps the brain.”

Uncertainty and future directions

None of this is settled. The cell-specific effects above come from mice, and the neurovascular story linking them to the human executive-function signal is a plausible synthesis, not a demonstrated chain. Real, useful signals are also getting inflated by relative-risk framing and normative-decline translations, in a field where funding rewards breakthrough narratives over careful ones — not reason to dismiss FINGER or POINTER, but reason to ask, as Africa-FINGERS scales up, whether it’s testing the right thing: an elaborate coached protocol exported wholesale, or a cheaper, gap-filling alternative that hasn’t had its own trial.

Take-Home Messages

• Target the population where the room to improve is largest: FINGER, POINTER, and even diet-screened MIND enrolled cohorts that were educated and well-resourced, whatever their baseline diet score. A poor diet score and genuine socioeconomic disadvantage are not the same thing — the bigger opportunity likely sits in vulnerable populations with limited access to whole foods or exercise infrastructure, and even Africa-FINGERS is still exporting the expensive coached protocol rather than testing whether closing basic gaps works better alone.

• De-emphasize expensive “optimization protocols”: every comparison group already received an active ingredient — health advice, self-guided coaching, or weight loss — and the elaborate add-on produced only a small increment on top. The MIND diet score itself is an optimization checklist, not a measure built around ultra-processed food, added sugar, or fiber; cutting ultra-processed foods and raising fiber may matter more than a curated food list.

• Recruiting underserved populations is genuinely hard, and that’s not a footnote: transportation, work schedules, and trust in research institutions all erode compliance. Pragmatic, decentralized, or community-embedded trial designs may be better suited to testing basic-access interventions than a standard multi-year RCT.

• Matching outcomes to mechanism: interventions working through vascular pathways plausibly act on processing speed and executive function more than on amyloid/tau-driven memory decline — outcomes should be weighted accordingly, not defaulted to memory-heavy composites built for drug trials.

• Digital, phone-based executive-function testing could change the landscape: repeated, low-burden testing averages out day-to-day noise and practice effects, detecting small, domain-specific effects one-off batteries are too noisy to catch.

References

1. Ngandu T, Lehtisalo J, Solomon A, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): a randomised controlled trial. Lancet. 2015;385(9984):2255-2263.

2. Baker LD, Espeland MA, Whitmer RA, et al. Structured vs self-guided multidomain lifestyle interventions for global cognitive function: the U.S. POINTER randomized clinical trial. JAMA. Published online July 28, 2025. doi:10.1001/jama.2025.12923.

3. Barnes LL, Dhana K, Liu X, et al. Trial of the MIND diet for prevention of cognitive decline in older persons. N Engl J Med. 2023;389(7):602-611.

4. Udeh-Momoh CT, Maina R, Anazodo UC, et al. Dementia risk reduction in the African context: multi-national implementation of multimodal strategies to promote healthy brain aging in Africa (the Africa-FINGERS project). Alzheimers Dement. 2024;20(suppl 12). doi:10.1002/alz.14344.

5. National Center for Health Statistics (CDC). Ultra-processed food consumption among adults and youth, United States, August 2021–August 2023. NCHS Data Brief No. 536. Hyattsville, MD: 2025.

6. Zhao Y, Bazan NG. Nutrition and gut-brain axis: opposing effects of dietary fiber and Western-style diets on Alzheimer’s disease. Curr Opin Clin Nutr Metab Care. 2026;29(4):421-430.

7. Gomes Gonçalves N, Vidal Ferreira N, Khandpur N, et al. Association between consumption of ultraprocessed foods and cognitive decline. JAMA Neurol. 2023;80(2):142-150.

8. Andrieu S, Guyonnet S, Coley N, et al. Effect of long-term omega 3 polyunsaturated fatty acid supplementation with or without multidomain intervention on cognitive function in elderly adults with memory complaints (MAPT): a randomised, placebo-controlled trial. Lancet Neurol. 2017;16(5):377-389.

9. Moll van Charante EP, Richard E, Eurelings LS, et al. Effectiveness of a 6-year multidomain vascular care intervention to prevent dementia (preDIVA): a cluster-randomised controlled trial. Lancet. 2016;388(10046):797-805.

10. Richard E, Van den Heuvel E, Moll van Charante EP, et al. Healthy ageing through internet counselling in the elderly (HATICE): a multinational, randomised controlled trial. Lancet Digit Health. 2019;1(8):e424-e434.

11. Cohen S, Cummings J, Knox S, Potashman M, Harrison J. Clinical trial endpoints and their clinical meaningfulness in early stages of Alzheimer’s disease. J Prev Alzheimers Dis. 2022;9(3):507-522.

12. Aschenbrenner AJ, Hassenstab J, Wang G, et al. Avoid or embrace? Practice effects in Alzheimer’s disease prevention trials. Front Aging Neurosci. 2022;14:883131.

13. Berron D, et al. A scoping review of remote and unsupervised digital cognitive assessments in preclinical Alzheimer’s disease. npj Digit Med. 2025;8:266.

14. da Rocha JF, Lance ML, Luo R, et al. Protective exercise responses in the dentate gyrus of Alzheimer’s disease mouse model revealed with single-nucleus RNA-sequencing. Nat Neurosci. 2025;28:1546-1561.

15. Belaya I, Ivanova M, Sorvari A, et al. Astrocyte remodeling in the beneficial effects of long-term voluntary exercise in Alzheimer’s disease. J Neuroinflammation. 2020;17:271.

You are reading because the stakes are personal.

You want to know what to eat, which medications to take or avoid, which supplements might help, whether you should be on anti-amyloid treatments, whether to join a trial, and what you can do now. That urgency is understandable. It also makes people vulnerable to exaggerated claims.

A supplement guru can cite a case report. An influencer can cite a breakthrough observational study. A podcast guest can quote an alarming FDA warning. A headline can promise “the truth,” “what doctors won’t tell you,” or “the end of Alzheimer’s.”

The problem is not curiosity. Patients and families should read, ask questions, debate, challenge claims, and participate in research when possible. That is how the field improves.

The problem is certainty with weak evidence, or worse, certainty without evidence.

Science is cautious because biology is complicated. Breakthroughs are rare. A patient story is not a clinical trial. An association is not causation. A statistically significant result is not always meaningful. A warning label is not proof that a drug causes disease. And one study is rarely the final word.

1. Words Designed to Catch Your Attention, but Should Make You Slow Down

The first warning sign is often the headline.

Be careful when a health claim begins with phrases like:

“The truth about…”

“Breaking news…”

“The cure”

“The end of disease”

“The number one cause…”

“Medication risks nobody warns you about”

“Why doing this changes everything”

“What doctors won’t tell you”

“Doctors have it all wrong”

“Big Pharma has been lying to you”

“The real driver of disease”

“The study they don’t want you to see”

“Clinically proven”

“Reverses Alzheimer’s”

“Targets the root cause”

“Ancient remedy validated by science”

These phrases are designed to create a sense of urgency, distrust, or certainty before the evidence is discussed.

Scientific language usually sounds more modest: “is associated with,” “may reduce,” “suggests,” “requires replication,” “hypothesis-generating,” “not powered to detect,” “exploratory analysis,” or “clinical significance remains unclear.”

That language may sound less exciting, but it is usually more honest.

The late astrophysicist Richard Feynman captured this spirit well: “I think it’s much more interesting to live not knowing than to have answers which might be wrong.”

That is the heart of science. It is better to admit uncertainty than to be certain about something false.

A good scientific claim would not give the impression of certainty without weighing the evidence. It shows you the evidence, the uncertainty, and helps you make an informed decision.

2. Case Reports and Abstracts: Useful Clues, Weak Proof

Some of the most persuasive health claims begin with a story.

A patient was declining. Then they started a protocol, supplement, diet, or program. Their memory improved. Their family noticed. Their test score looked better.

Stories matter. And Social Media platforms are filled with these stories.

Case reports can alert doctors to rare side effects, unusual symptoms, unexpected harms, or new disease mechanisms. Some major discoveries began with careful observation of one patient, one family, or one rare genetic variant.

In Alzheimer’s disease, rare families with inherited mutations helped scientists understand amyloid biology long before amyloid PET scans and blood biomarkers existed, or rare mutations that led to medications.

So the problem is not the case report.

The problem is using a case report to claim that a treatment works.

A case report can say:

“This happened.”

It cannot say:

“This works.”

To know whether something works, we need comparison groups, clear methods, fair outcome measurement, and replication.

This is especially important in cognitive disorders. Memory test scores can fluctuate. People can improve because they have taken the same test before, called a practice effect. Symptoms can improve when sleep, mood, medications, thyroid disease, vitamin B12 deficiency, or sleep apnea are treated. Others may be selected for publication because they improved, while non-responders are left out, and we never hear about the patient who did not improve.

This is also why we need to be fair about popular “Alzheimer’s reversal” programs.

Some advice in these programs is reasonable: exercise, better sleep, a healthier diet, treatment of sleep apnea, attention to vascular risk, hearing care, and checking for reversible causes of cognitive symptoms. These are not fringe ideas. Many dementia clinics already recommend them.

The concern is what gets added around them.

Helpful advice can be bundled with expensive supplements, large panels of blood tests, coaching programs, branded protocols, and a price tag that many families can barely afford. The scientific evidence for those added layers is often much weaker than the evidence for the basic lifestyle and medical care recommendations.

That distinction matters.

Exercise is not the same as an expensive supplement stack.
Sleep apnea treatment is not the same as a branded protocol.
Checking B12 or thyroid function is not the same as ordering unconventional tests on toxins or hormones without clear evidence that acting on them changes dementia outcomes.

A program can feel scientific because it includes lab tests, supplements, diet rules, coaching, and personalized reports. But complexity is not the same as evidence.

A common pattern is to market a protocol with words such as “prevent,” “reverse,” “restore,” or “end,” while the published evidence consists mainly of uncontrolled case reports or case series, or even limited trials. These reports may lack clear methods, inclusion criteria, dosing, duration, blinded outcome assessment, the correct sample size, outcome, complete reporting of non-responders, validated cognitive measures, or transparent financial disclosures.

Those omissions matter.

Without them, readers cannot know whether improvement reflects the intervention, natural fluctuation, practice effects, better sleep, treatment of another medical condition, extra attention, placebo response, or selective reporting.

There is no harm in participating in such programs as long as we are well-informed.

It is prudent to separate reasonable, low-risk health practices from expensive claims that have not yet been proven.

Conference abstracts have similar limitations.

A conference abstract is a summary presented at a scientific meeting. It may describe interesting preliminary findings, but it is not the same as a full peer-reviewed paper. Abstracts often do not provide enough detail to judge the methods, missing data, statistical plan, subgroup analyses, limitations, or conflicts of interest.

Peer review is not perfect. Flawed papers can pass peer review, and good papers can be rejected. Reviewers can miss errors. Journals can favor exciting findings. Scientists have biases. Science is a human process, and humans make mistakes.

But science also has tools for correction: replication, criticism, meta-analysis, reanalysis, corrections, and sometimes retractions.

My personal hero, Carl Sagan, captured this well: “Science is far from a perfect instrument of knowledge. It’s just the best we have.”

That is why we should not treat science as a single paper, a single p-value, or a single headline. Science is the process that allows wrong ideas to be tested, challenged, corrected, or replaced.

3. Observational Studies: Association Is Not Causation

Observational studies are essential.

They can follow large groups of people over time and identify patterns that would be impossible, unethical, or too expensive to test in a randomized trial. We cannot randomize people for decades to poverty, air pollution, poor education, diabetes, hypertension, poor access to healthy food, or social isolation.

But observational studies have limits.

They can show that two things occur together. They cannot automatically prove that one caused the other.

One problem is confounding by indication.

This means the reason someone receives a treatment is also related to the outcome being studied.

A simple example is insulin.

People who use insulin are often sicker than people with early or mild diabetes. They may have had diabetes longer, have higher blood sugar, or have more complications. If insulin users later have more kidney disease, heart disease, or dementia, we cannot conclude that insulin caused those problems.

More often, insulin use is a marker of more advanced diabetes.

The same problem can occur with statins.

People prescribed statins often have high LDL, diabetes, hypertension, obesity, vascular disease, inflammation, or prior cardiovascular events. Those same conditions also increase dementia risk. So if statin users later develop dementia at higher rates, the statin may not be the cause. The statin may identify people who were already at higher vascular and metabolic risk.

Another problem is reverse causation.

Reverse causation means the arrow points in the opposite direction from what we first assume. The early disease process may change the exposure, instead of the exposure causing the disease.

In Alzheimer’s disease, brain and metabolic changes can begin many years before dementia is diagnosed. During that period, people may lose weight, change eating patterns, become less active, start or stop medications, or show changes in cholesterol levels.

So if a study finds that people with lower cholesterol are more likely to develop dementia, one possibility is that low cholesterol contributed to dementia.

But another possibility is the reverse: early disease, frailty, weight loss, or inflammation lowered cholesterol before dementia was diagnosed.

The arrow may point in the opposite direction.

4. The Statin Example: When a Signal Spreads Faster Than the Correction

Statins are a good example of how scientific information can be distorted and how easy misinformation can become viral.

In 2012, the FDA added language to statin labels about reports of memory loss, forgetfulness, and confusion. These reports were largely based on patient complaints after statins were already in use.

Such reports matter. They can alert regulators and clinicians to possible side effects and demand clinical trials. But they are not proof of causation.

This does not mean patient symptoms are imaginary.

Some people do experience side effects from statins. Muscle symptoms, medication intolerance, and individual reactions or cognitive symptoms can be real. A large study may show no evidence of harm at the population level, but that does not mean every individual will tolerate the medication perfectly.

Population evidence answers one question: “Does this drug increase risk on average?”

A patient’s experience raises another question: “Is this drug right for this person?”

Both questions matter.

What population studies and randomized trials do not support is the stronger claim that statins cause Alzheimer’s disease or progressive dementia.

A warning label is not a randomized trial.
A complaint is not proof.
A reversible symptom report is not Alzheimer’s disease.

The warning was real, but it was heavily misused, and that may have misled many patients. Online, it was often presented as proof that statins cause dementia or Alzheimer’s disease.

A later conference abstract added fuel. It reported that among people with early mild cognitive impairment and lower baseline cholesterol, users of lipophilic statins had higher dementia conversion over eight years than non-users.

At first glance, this sounds alarming.

But this was not a randomized trial. It was an observational subgroup analysis presented as an abstract. Participants were not assigned to statins by chance. They were taking statins because clinicians had already judged them to have cardiovascular or lipid risk.

The analysis also focused on people with early mild cognitive impairment, exactly the group where reverse causation is a concern. Early Alzheimer’s biology may already be changing weight, diet, frailty, medical care, medication patterns, and cholesterol levels.

The careful conclusion is not:

“Statins cause dementia.”

The careful conclusion is:

“This subgroup analysis raised a concern that needed replication and better-controlled evidence.”

Later, more careful studies and meta-analyses did not replicate the claim that statins cause dementia. But the viral message had already spread on social media platforms.

This is how misinformation works: a signal spreads faster than the studies that later correct it.

5. Clinical Trials: Better Evidence, Still Not Perfect

Randomized clinical trials are a reasonable way to test whether a treatment works. Randomization makes groups more similar at the start. Blinding reduces expectation effects. A control group shows what might have happened without the treatment.

But even trials need careful reading.

A trial should tell us its primary outcome before it starts. This is the main question. Secondary outcomes are supportive. Exploratory outcomes and subgroup findings are clues, not proof.

This matters when the primary outcome is missed. For example, the ALZ-801/APOLLOE4 trial in APOE4 homozygotes did not meet its primary cognitive endpoint, although secondary and subgroup findings suggested possible signals in milder participants and imaging outcomes. Those signals may guide another trial, but they do not replace the missed primary endpoint.

Subgroups need the same caution. The BROADWAY obicetrapib study reported a promising p-tau217 biomarker signal, especially in APOE4 carriers. But this was a biomarker substudy, not proof that the drug prevents Alzheimer’s disease or slows cognitive decline. The right conclusion is: promising signal, needs a dedicated trial.

So when reading a trial, ask:

Was this the primary outcome?
Was the study large enough?
Was the finding replicated?
Was it a subgroup or an exploratory result?
Did the outcome matter to patients?

A signal is something worth testing.

A finding is something that survives testing.

6. P-Values, Effect Size, and the Word “Breakthrough”

A p-value tells us how surprising a result would be if there were truly no difference between two groups.

A small p-value can suggest that the result is unlikely to be due to chance alone. But a p-value does not tell us whether the effect is large, important, or meaningful to a patient.

That is why we also need the effect size.

An effect size tells us how big the difference is.

A treatment can have a statistically significant effect that is still very small. With enough people in a study, even a tiny difference can produce a low p-value.

One common effect size is Cohen’s d. It expresses the difference between two groups in standard deviation units.

A Cohen’s d of 1.0 is large.
A Cohen’s d of 0.5 is moderate.
A Cohen’s d of 0.2 is small.
A Cohen’s d of around 0.1 is very small.

There is a simple way to think about this.

If two groups are identical, and you randomly pick one person from each group, the chance that the person from the treatment group scores higher is about 50 in 100.

If Cohen’s d is 0.13, that chance rises to about 54 in 100.

That is not nothing.

But it is modest.

And modest effects need modest language.

One example to illustrate effect sizes and p-values is multidomain lifestyle trials such as FINGER and US POINTER, and the recent Anti-amyloid trials.

FINGER was an interesting experiment. It was an intensive multidomain intervention compared with a less intensive control condition that also included health advice. The intervention group did better, but the standardized effect size was small, with a reported Cohen’s d ≈ 0.13.

US POINTER (FINGER experiment in the US) also needs careful framing. It did not compare lifestyle intervention with doing nothing. It compared a structured, intensive lifestyle program with a self-guided lifestyle program. Both groups received lifestyle support. Both groups improved. The structured arm did somewhat better, but like FINGER, the effects were very modest (Cohen’s d ≈ 0.06).

Why interpreting this low effect size matters.

Participants in these trials were often motivated and highly educated. An expensive program that works modestly in a highly educated research population may not automatically translate to communities with limited access to healthy food, safe exercise spaces, preventive medical care, hearing care, sleep treatment, or social support, where cheaper “standard” programs can be good enough.

The same caution applies to anti-amyloid treatments.

Lecanemab and donanemab are scientifically important because they showed that removing amyloid can slow clinical decline in early Alzheimer’s disease. But the effect sizes are modest.

In CLARITY AD, lecanemab slowed decline on the Clinical Dementia Rating–Sum of Boxes, or CDR-SB, by 0.45 points over 18 months compared with placebo. The CDR-SB ranges from 0 to 18, with higher scores meaning worse impairment. In standardized terms, this is roughly a small effect size, around Cohen’s d = 0.18 to 0.20.

In TRAILBLAZER-ALZ 2, donanemab slowed decline by about 3.25 points on the Integrated Alzheimer’s Disease Rating Scale, or iADRS, in the primary analysis population. The iADRS ranges from 0 to 144, with lower scores meaning worse cognition and function. In standardized terms, this is also a small effect size, roughly Cohen’s d = 0.20 to 0.25. On CDR-SB, donanemab’s treatment difference was about 0.7 points, again a modest effect.

These numbers matter.

They show that anti-amyloid drugs are not “nothing.” They changed the trajectory of disease in carefully selected patients with early Alzheimer’s disease and confirmed amyloid pathology.

But they are not cures. They have not shown that they reverse Alzheimer’s disease yet. They modestly slow decline, while also carrying risks such as amyloid-related imaging abnormalities, infusion reactions, monitoring burden, cost, and access barriers.

This is the same lesson as the lifestyle trials.

A result can be real and still modest.

A result can be scientifically important and still require careful language.

Good science does not need exaggeration.

7. A Simple Checklist for Reading Health Claims

Conclusion: Ask for Better Evidence

The goal is not to be dismissive.

The goal is to ask better questions.

A patient's story can raise a question.
An observational study can show a pattern.
A clinical trial can test a treatment.
Replication tells us whether the finding holds up.

If the claim is big — “reverses Alzheimer’s,” “statins cause dementia,” “this supplement prevents disease,” or “this protocol ends cognitive decline” — the evidence needs to be big too.

Patients and families should not be told to stop asking questions.

They should be encouraged to ask sharper ones.

Challenge claims. Ask for the primary outcome. Ask about the effect size. Ask whether the finding was replicated. Ask whether someone is selling something.

Good science does not ask us to believe harder.

It asks us to test better. And embrace uncertainty over false claims.

References

1. Hellmuth J. Can we trust The End of Alzheimer’s? Lancet Neurology. 2020;19(5):389–390.

2. Ioannidis JPA. Why most published research findings are false. PLoS Medicine. 2005;2(8):e124.

3. Jager LR, Leek JT. An estimate of the science-wise false discovery rate and application to the top medical literature. Biostatistics. 2014;15(1):1–12.

4. U.S. Food and Drug Administration. FDA Drug Safety Communication: Important safety label changes to cholesterol-lowering statin drugs. 2012.

5. Swiger KJ, Manalac RJ, Blumenthal RS, Blaha MJ, Martin SS. Statins and cognition: a systematic review and meta-analysis of short- and long-term cognitive effects. Mayo Clinic Proceedings. 2013;88(11):1213–1221.

6. Journal of Nuclear Medicine. Statin use, cholesterol level, and dementia conversion in early mild cognitive impairment. J Nucl Med. 2021;62(Supplement 1):102.

7. Li G, Mayer CL, Morelli D, et al. Effect of simvastatin on CSF Alzheimer disease biomarkers in cognitively normal adults. Neurology. 2017;89:1251–1255.

8. Abushakra S, et al. Clinical efficacy, safety and imaging effects of oral valiltramiprosate in APOE4/4 homozygous individuals with early Alzheimer’s disease: the APOLLOE4 Phase III randomized controlled trial. CNS Drugs. 2025.

9. Davidson MH, Szarek M, Scheltens P, et al. Effect of obicetrapib, a potent cholesteryl ester transfer protein inhibitor, on p-tau217 levels in patients with cardiovascular disease. The Journal of Prevention of Alzheimer’s Disease. 2025. doi:10.1016/j.tjpad.2025.100394.

10. Ngandu T, Lehtisalo J, Solomon A, et al. A 2-year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people: the FINGER randomized controlled trial. Lancet. 2015;385(9984):2255–2263.

11. Baker LD, Snyder HM, et al. U.S. POINTER Study: multidomain lifestyle intervention in older adults at risk for cognitive decline. JAMA. 2025.

12. van Dyck CH, Swanson CJ, Aisen P, et al. Lecanemab in early Alzheimer’s disease. New England Journal of Medicine. 2023;388:9–21.

13. Sims JR, Zimmer JA, Evans CD, et al. Donanemab in early symptomatic Alzheimer disease: the TRAILBLAZER-ALZ 2 randomized clinical trial. JAMA. 2023;330(6):512–527.

14. Feynman RP. The Pleasure of Finding Things Out. Basic Books; 1999.

15. Sagan C. The Demon-Haunted World: Science as a Candle in the Dark. Random House; 1995.

A recent analysis from the BROADWAY trial reported that obicetrapib slowed the rise of plasma p-tau217 over 12 months in people with cardiovascular disease [1].

That finding deserves attention.

P-tau217 is one of the strongest blood biomarkers we have for Alzheimer’s disease biology. It often rises when amyloid and tau pathology are building in the brain. So when a drug changes p-tau217 in a randomized placebo-controlled study, the field should pay attention.

But the interpretation needs to stay proportional to the study design.

BROADWAY was designed as a cardiovascular lipid trial. The Alzheimer’s biomarker analysis was pre-specified, which makes it more meaningful than a purely post-hoc analysis. But it was still an analysis within a cohort built for another primary disease question.

That distinction matters.

A pre-specified biomarker signal can point us toward an important hypothesis. It does not, by itself, prove Alzheimer’s prevention, clinical benefit, or disease modification.

The question now is whether the signal reflects a real effect on Alzheimer’s biology, whether it is strongest in the right biological subgroup, and whether it eventually translates into cognition or function.

Why Obicetrapib Is Biologically Interesting

Obicetrapib blocks cholesterol ester transfer protein (CETP), a protein that helps transfer cholesterol between lipoprotein particles in the blood.

When CETP is inhibited, the lipid system is remodeled. LDL cholesterol falls. ApoB and non-HDL cholesterol fall. Lp(a) can fall. HDL cholesterol rises, and HDL particles may become larger or more functional.

That is the established biology.

This is also why obicetrapib worked in BROADWAY for its main lipid readouts. BROADWAY enrolled people with atherosclerotic cardiovascular disease (ASCVD) and/or heterozygous familial hypercholesterolemia whose LDL cholesterol remained inadequately controlled despite lipid-lowering therapy. In that population, obicetrapib lowered LDL-C, raised HDL-C, and lowered Lp(a) [1].

The Alzheimer’s question is different.

The question is whether this peripheral lipid remodeling can also influence Alzheimer’s-related biology.

There are several plausible pathways. The effect could be vascular, through lower LDL-C, ApoB, Lp(a), and improved vascular risk biology. It could be HDL-mediated, through larger or more functional HDL particles that support cholesterol efflux, antioxidant activity, and anti-inflammatory signaling. It could also involve amyloid handling, including a possible peripheral sink effect where remodeled HDL particles bind or help transport amyloid-related material in blood.

These mechanisms are plausible, but they are not proven.

The key point is that obicetrapib’s clearest biology is in the blood lipid system. The BROADWAY p-tau217 signal may reflect a downstream effect of that remodeling, but we do not yet know whether it is central, peripheral, vascular, HDL-mediated, or mixed.

What BROADWAY Tested

BROADWAY was a large cardiovascular lipid trial.

The parent trial evaluated obicetrapib in adults with established ASCVD, and/or heterozygous familial hypercholesterolemia. Participants were recruited from cardiology clinics and lipid specialty centers. They were already receiving maximally tolerated lipid-lowering therapy, but their LDL cholesterol remained inadequately controlled [1].

So BROADWAY did not enroll a general aging population. It enrolled people with residual lipid and cardiovascular risk.

The trial randomized 2,530 participants in a 2:1 ratio to obicetrapib 10 mg daily or placebo for 12 months. The parent trial’s main lipid readout was LDL-C lowering. At day 84, LDL-C decreased by about 30% with obicetrapib and increased slightly with placebo, producing a placebo-adjusted LDL-C difference of about 33%. Obicetrapib also increased HDL-C by more than 130% and lowered Lp(a) by about one-third [1].

The BROADWAY study materials summarize the key inclusion criteria as ASCVD or HeFH, LDL-C of at least 55 mg/dL with additional risk factors or LDL-C of at least 100 mg/dL, and maximally tolerated lipid-lowering therapy [1]. The parent trial endpoints included LDL-C at 12 weeks as the primary endpoint, with ApoB, Lp(a), non-HDL-C, and safety as key readouts [1].

That context is important.

The Alzheimer’s biomarker substudy did not arise from a general Alzheimer’s prevention cohort. It arose from a lipid-risk population, which may be exactly where a lipid-modifying drug is most likely to show a signal.

What the Biomarker Substudy Found

The Alzheimer’s biomarker analysis used stored blood samples from participants who had known APOE status and p-tau217 measured at baseline and after 12 months. The analysis included 1,535 of the 2,530 randomized BROADWAY participants, or 61% of the randomized population [1].

In the overall biomarker group, p-tau217 rose less in people who received obicetrapib than in people who received placebo. The signal appeared stronger in APOE4 carriers and strongest in APOE4/E4 participants.

Among APOE4/E4 participants, p-tau217 decreased with obicetrapib and increased with placebo, producing a placebo-adjusted difference of about 20%. Other biomarkers also moved in a favorable direction in APOE4/E4 participants, including GFAP and NfL [1].

This is biologically interesting. It supports the idea that lipid modulation may affect Alzheimer’s-related biomarkers, especially in people with APOE4.

But the subgroup details are important.

Of the 1,535 participants in the biomarker analysis, 1,045 were APOE3/E3, 338 were APOE3/E4, 20 were APOE2/E4, and only 29 were APOE4/E4 [1].

That means the most striking APOE4/E4 finding came from a small subgroup.

Small subgroups can reveal real biology, but they can also produce unstable estimates. A large effect in 29 people is worth following, but it requires replication before it can carry the weight of any clinical claim.

The cardiovascular enrichment also matters. In the BROADWAY biomarker subset, most APOE4 carriers had established ASCVD because the parent trial required ASCVD and/or HeFH. From the subgroup table, approximately 87% of APOE4 carriers in the biomarker substudy had ASCVD. About 81% had hypertension, and about 38% had diabetes. Among APOE4/E4 participants specifically, about 86% had ASCVD, about 83% had hypertension, and about 45% had diabetes [1].

That is not a typical APOE4 prevention cohort.

BROADWAY’s APOE4 carriers were genetically at risk, but they were also heavily enriched for vascular and metabolic risk. That combination may be central to why the signal appeared.

How to Read the BROADWAY Analysis Fairly

The BROADWAY biomarker analysis has real strengths. It came from a randomized, double-blind, placebo-controlled parent trial. The Alzheimer’s biomarker analysis was pre-specified. The overall biomarker sample was substantial. The biological rationale is plausible. And the main p-tau217 finding is worth following.

A fair reading also requires caution.

First, this was a biomarker substudy of a cardiovascular lipid trial, not a dedicated Alzheimer’s prevention trial. It did not test whether obicetrapib slows memory decline, prevents mild cognitive impairment, or prevents dementia.

Second, the substudy examined many outcomes: p-tau217 overall, APOE subgroups, APOE4/E4 participants, age subgroups, p-tau217/Aβ42:40, GFAP, NfL, p-tau181, lipid correlations, and obicetrapib blood-level correlations. These analyses are useful, but the paper states that p values were not adjusted for multiplicity [1]. That means the pre-specified main biomarker result deserves the most weight, while subgroup and secondary biomarker findings are best treated as hypothesis-generating.

Third, a subgroup pattern is not the same as a formal interaction analysis. If the treatment effect looks larger in APOE4 carriers, that does not automatically prove that APOE genotype modifies treatment response. That requires a treatment-by-APOE interaction analysis. Without that, the careful statement is that the signal appeared stronger in APOE4 carriers, not that APOE4 has been proven to modify response.

Finally, baseline p-tau217 may matter. APOE4/E4 participants had higher p-tau217 at baseline, and the paper reported that participants with higher baseline p-tau217 appeared to show greater treatment effects [1]. That raises an important question: was the stronger APOE4/E4 signal due to APOE4 biology, higher baseline p-tau217, lipid risk, ASCVD, or some combination of all these factors?

That is the key precision-prevention question.

The best target population may not be APOE4 alone. It may be APOE4 plus elevated p-tau217 plus lipid dysfunction. Or it may be elevated Alzheimer’s biomarkers plus measurable lipid dysfunction, even without APOE4.

The Statin Lesson: LDL May Define the Responder

Statins provide a useful lesson for obicetrapib.

In a 12-month randomized, placebo-controlled trial, Li and colleagues tested simvastatin in cognitively normal adults. Participants were 45 to 64 years old, statin-naive, cognitively normal, and had normal or mildly elevated cholesterol. Forty-nine people were randomized, and 46 completed the study per protocol: 25 in the simvastatin group and 21 in the placebo group. Simvastatin was titrated to 40 mg per day. The primary outcomes were CSF Aβ42, total tau, and p-tau181 measured at baseline and after 12 months [2].

Overall, simvastatin did not significantly change CSF Aβ42, total tau, or p-tau181 compared with placebo. It also did not produce significant differences in neuropsychological test changes.

But the most informative result may have been hidden in the biology of the participants. Baseline LDL modified the treatment effect on CSF p-tau181. People with higher baseline LDL appeared to have greater reductions in CSF p-tau181 on simvastatin. The authors concluded that simvastatin-related reductions in CSF p-tau181 may be modulated by LDL cholesterol and merit further study in people with hypercholesterolemia [2].

That is directly relevant to obicetrapib.

A lipid-lowering drug may show its clearest brain biomarker effect in people whose lipid biology is abnormal. If a trial includes many people without elevated LDL, ApoB, Lp(a), or other lipid risk, the average effect may look smaller than it really is in the subgroup most likely to respond.

The lesson is not that statins prevent Alzheimer’s disease.

The lesson is that the responder population matters.

For obicetrapib, the key question may not be whether it works in all people with APOE4 or all people with elevated p-tau217. The more precise question is whether it works best in people with Alzheimer’s biology plus measurable lipid dysfunction.

The EVOKE Lesson: Biomarkers Need the Right Biology and the Right Outcome

EVOKE and EVOKE+ provide a second lesson.

These were large phase 3 trials of oral semaglutide in early Alzheimer’s disease, enrolling a combined 3,808 adults with early-stage Alzheimer’s symptoms and following participants for about two years [4].

The trials were important because GLP-1 receptor agonists have strong biological rationale. They improve metabolic health, insulin resistance, inflammation, vascular risk, and weight — pathways that may matter for brain aging.

The biomarker readout was encouraging. Semaglutide reduced several CSF biomarkers compared with placebo, including p-tau217 and other Alzheimer’s-related markers.

But the clinical readout was negative: semaglutide did not significantly slow disease progression compared with placebo [4].

This matters for two reasons.

First, it shows that moving Alzheimer’s biomarkers does not guarantee cognitive benefit.

Second, EVOKE was an Alzheimer’s disease trial, but it was not primarily designed around the metabolic phenotype where GLP-1 receptor agonists are most powerful. GLP-1 drugs often work best in people with obesity, insulin resistance, diabetes, cardiometabolic disease, or related inflammatory-metabolic risk. If those features are not central to trial enrichment, the average effect may dilute the benefit in the subgroup most likely to respond.

That is another lesson for obicetrapib.

Obicetrapib is a lipid-modifying drug. Its best test may require more than Alzheimer’s biomarkers or APOE genotype. It may require identifying the people whose Alzheimer’s risk is tied to lipid or vascular dysfunction.

In other words, the question is not only whether the drug can move p-tau217.

The question is whether the trial has selected the population most likely to benefit from the way the drug works.

The Mechanism Is Still Uncertain

There are several possible explanations for the BROADWAY p-tau217 signal.

It could be a central effect, meaning obicetrapib changes processes inside the brain. But we do not yet know whether obicetrapib enters the human brain in meaningful amounts.

It could be a vascular effect. BROADWAY participants had high cardiovascular risk, and obicetrapib improved lipid markers linked to vascular disease.

It could be an HDL-mediated peripheral effect. CETP inhibition may create larger or more functional HDL particles that improve cholesterol efflux, antioxidant defense, and inflammatory signaling.

It could also involve a peripheral sink effect, where remodeled HDL particles bind or transport amyloid-related proteins in blood, changing plasma biomarker dynamics without proving that tau pathology inside the brain has changed.

These possibilities are not mutually exclusive.

The next studies need to connect plasma p-tau217 to CSF biomarkers, amyloid PET, tau PET, vascular markers, HDL function, drug exposure, and cognition.

The key question is not only whether obicetrapib lowers p-tau217.

It is how it lowers p-tau217 — and whether that mechanism leads to clinical benefit.

APOE4, Lipids, and Trial Design

APOE4 makes obicetrapib interesting because APOE is a lipid transport gene.

In the brain, APOE helps move cholesterol and phospholipids between astrocytes, neurons, microglia, and other cells. These lipids are needed for synapses, membrane repair, myelin, and normal cellular signaling. APOE4 appears less efficient than APOE3 in several aspects of lipid handling and repair.

So it is plausible that some APOE4-related brain risk is mediated through lipid biology.

But APOE4 is not the same as a lipid phenotype.

Population studies show that APOE4 carriers, on average, have higher total cholesterol and LDL cholesterol than APOE3 carriers [6]. But averages do not define individuals. The percentage of APOE4 carriers with elevated LDL, low HDL, high ApoB, high Lp(a), metabolic syndrome, or ASCVD depends on the population being studied.

That matters for obicetrapib.

A trial enriched for APOE4, p-tau217, or GFAP may identify people at Alzheimer’s risk, but it may not identify the people most likely to respond to a lipid-modifying drug.

For obicetrapib, the most informative population may be people with both Alzheimer’s biology and lipid biology: APOE4 plus elevated p-tau217, high LDL-C or ApoB, elevated Lp(a), low HDL function, ASCVD, metabolic syndrome, or combinations of these factors.

The trial design should match how the drug works.

SPINOZA is coming next

The name SPINOZA also carries a useful meaning. Baruch Spinoza was a 17th-century Dutch philosopher whose work helped shape Western ideas about reason, nature, freedom, and the unity of mind and body. That connection is fitting here: the trial asks whether changing biology outside the brain — lipid metabolism in the blood and vasculature — might influence Alzheimer’s biology inside or around the brain.

SPINOZA is a phase 2b randomized, double-blind, placebo-controlled trial testing obicetrapib in people with preclinical Alzheimer’s disease. Unlike BROADWAY, it is designed around Alzheimer’s prevention rather than cardiovascular lipid lowering [5].

The primary endpoint is change in plasma p-tau217 at week 52. Secondary endpoints include Aβ42/40, p-tau217/Aβ42, p-tau217/Aβ42/40, p-tau181, and brain-derived tau. Cognition is exploratory and measured with the Preclinical Alzheimer Cognitive Composite 5, or PACC5 [5].

SPINOZA plans to enroll up to 400 participants, including at least 100 APOE4/E4 participants, plus 125–150 APOE3/E4 and 125–150 APOE3/E3 participants [5].

That is an important design. It prospectively tests the signal in a preclinical Alzheimer’s population and intentionally includes a larger APOE4/E4 group.

The key point is not to criticize SPINOZA before it starts. The key point is that its most informative analyses may depend on careful phenotyping.

Since obicetrapib is a lipid-modifying drug, the trial could be especially informative if it reports and analyzes LDL-C, ApoB, Lp(a), HDL-C, triglycerides, metabolic syndrome, diabetes, ASCVD, statin use, PCSK9 inhibitor use, baseline p-tau217, GFAP, and drug exposure.

Not all APOE4 carriers have high LDL, low HDL, high ApoB, high Lp(a), metabolic syndrome, or ASCVD. If many participants have preclinical Alzheimer’s biology but relatively normal lipid biology, the average treatment effect could be diluted.

That does not make the trial weak. It makes subgroup and interaction analyses essential.

For a lipid drug, the most informative trial design considers both Alzheimer’s biology and lipid biology.

Conclusion: From Signal to Reproducible Science

BROADWAY produced a promising p-tau217 signal. The signal appeared strongest in APOE4 carriers, especially APOE4/E4 participants. The finding is biologically plausible and deserves prospective testing.

But it should not be overstated.

BROADWAY does not prove that obicetrapib prevents Alzheimer’s disease. It does not prove that obicetrapib slows memory decline. It does not prove that APOE4 carriers, as a broad group, benefit more. And it does not tell us whether the p-tau217 effect is central, peripheral, or both.

What BROADWAY does is point to a serious hypothesis: lipid biology may be a modifiable pathway in Alzheimer’s prevention, especially in people with APOE4, vascular risk, lipid dysfunction, and elevated Alzheimer’s biomarkers.

That is an important hypothesis.

Now it needs reproducible testing.

The next studies could select participants based on biology, test APOE interactions directly, test lipid interactions directly, measure biomarkers carefully, include cognitive outcomes, follow people long enough, and avoid turning subgroup findings into clinical claims before they are reproduced.

The main lesson is that Alzheimer’s prevention trials need to match the drug to the biology it is designed to modify.

For obicetrapib, that biology is lipid metabolism.

The right conclusion is not:

“Obicetrapib prevents Alzheimer’s disease.”

The right conclusion is:

“Obicetrapib produced a promising lipid-biology signal that now needs rigorous prospective testing in the right biological population.”

That is how reproducible prevention science can move forward.

Take-Home Messages

• Trial context matters. BROADWAY was a cardiovascular lipid trial, not a dedicated Alzheimer’s prevention trial.

• Mechanism matters. Obicetrapib’s clearest biology is peripheral lipid remodeling.

• Subgroups need discipline. The APOE4/E4 signal is promising, but the subgroup was small and needs interaction testing and replication.

• Biomarkers are not cognition. EVOKE and EVOKE+ showed that p-tau217 and other CSF biomarkers can improve without clinical benefit.

• APOE4 is not a lipid phenotype. The most informative population may be people with Alzheimer’s biology plus measurable lipid or vascular dysfunction.

• The mechanism remains open. The p-tau217 effect could be central, vascular, peripheral, HDL-mediated, or mixed.

• Precision prevention means matching the drug to the biology and the responder population. Obicetrapib acts on lipid metabolism, so the most informative trials may be those that identify people whose Alzheimer’s risk is tied to lipid or vascular dysfunction, not only APOE genotype or p-tau status.

Relevant Disclosure:

I sit on the advisory board of New Amsterdam Pharma, which is developing Obicetrapib

References

1. Davidson MH, Szarek M, Scheltens P, et al. Effect of obicetrapib, a potent cholesteryl ester transfer protein inhibitor, on p-tau217 levels in patients with cardiovascular disease. Journal of Prevention of Alzheimer’s Disease. 2025. doi:10.1016/j.tjpad.2025.100394.

2. Li G, Mayer CL, Morelli D, et al. Effect of simvastatin on CSF Alzheimer disease biomarkers in cognitively normal adults. Neurology. 2017;89:1251–1255.

3. Corcoran E, Kettlety M, Mogul U, Azah JN, Cork SC. The effects of GLP-1 receptor agonists on Alzheimer’s pathophysiology: a systematic review. Molecular and Cellular Neuroscience. 2026;137:104091.

4. Novo Nordisk. Headline results from the EVOKE and EVOKE+ phase 3 trials of oral semaglutide in early Alzheimer’s disease. Reported 2025.

5. SPINOZA Study. A Phase 2b, randomized, double-blind, placebo-controlled, parallel group study to assess the efficacy and safety of obicetrapib in participants with preclinical Alzheimer’s disease. Protocol Synopsis Version 1.0, 01 Apr 2026.

6. Bennet AM, Di Angelantonio E, Ye Z, et al. Association of apolipoprotein E genotypes with lipid levels and coronary risk. JAMA. 2007;298(11):1300–1311.

7. Alberti KGMM, Eckel RH, Grundy SM, et al. Harmonizing the metabolic syndrome: a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity. Circulation. 2009;120(16):1640–1645.

Before discussing our recent trial, I want to thank the hundreds of research participants who made PreventE4 possible. This trial asked a lot of people: screening visits, blood draws, cognitive testing, MRI scans, study capsules, follow-up visits, and, for many, lumbar punctures twice. Prevention trials depend on people who are willing to contribute their time, often for the benefit of future generations. We are deeply grateful.

Should You Take DHA Supplements?

Many readers will come to this question first: Should I take a DHA or fish oil supplement for brain health?

PreventE4 does not give a simple yes or no answer.

It was a two-year prevention trial in cognitively healthy older adults at risk for dementia. That matters. Two years may be too short to detect whether a nutrient changes the long course of Alzheimer’s disease, especially before memory symptoms begin.

Whether someone chooses to take an omega-3 supplement is still a personal decision. We hope that the information gained from this trial will help you make that decision.

But the goal of PreventE4 was not only to answer whether people at risk of dementia should take DHA.

The deeper goal was to gain a better understanding of DHA biology and Alzheimer’s risk.

Why We Did PreventE4

DHA is an omega-3 fat that the brain needs.

It helps build brain cell membranes. It supports synapses, the connections between brain cells. It also helps regulate inflammation.

For years, studies have suggested that people with higher omega-3 levels tend to have better brain health.

From 1994 to 2008, my friend and mentor, Helena Chui, led the Aging Brain Study at USC. I joined that team later. In the Aging Brain Study, we found that cognitively healthy older adults with lower blood DHA had more amyloid on brain scans, smaller memory-related brain regions, and worse memory performance [1].

That finding was important, but it also showed the limits of observational studies.

When we see low DHA in the blood, it can mean at least two things.

It may mean a person is eating too little DHA, usually from fatty fish or other omega-3 sources.

Or it may mean the body or brain is breaking DHA down faster — what scientists call catabolism.

Both can also be true at the same time.

That distinction matters because a blood DHA level is not just a measure of intake. It may also be a clue about how the body and brain are handling DHA.

To answer this question we needed to do a DHA supplementation trial.

Why APOE4 Matters

APOE4 is the strongest common genetic risk factor for late-onset Alzheimer’s disease.

APOE is involved in moving lipids, or fats, around the brain. Since DHA is a brain lipid, we wondered whether APOE4 changes how DHA gets into the brain or how the brain uses it.

Earlier work suggested that APOE4 and Alzheimer’s disease stage may affect DHA delivery to the brain [2,3].

We conducted PET imaging with radiolabeled DHA to find more clues. Cognitively healthy APOE4 carriers showed higher DHA incorporation into gray matter, especially in the entorhinal cortex, a brain region affected early in Alzheimer’s disease [4].

One interpretation is that the APOE4 brain may be pulling in more DHA because it is using it faster, losing it faster, or trying to compensate for stress.

A later high-dose DHA pilot trial measured DHA in cerebrospinal fluid, or CSF, the fluid that surrounds the brain. DHA supplementation increased CSF DHA, but the increase was modest, and APOE4 carriers showed signals of lower omega-3 delivery than non-carriers [5].

Together, these studies led to PreventE4.

The question was simple: if we give enough DHA, can we get it into the central nervous system in people at risk, especially APOE4 carriers, before dementia begins?

Does the Form of DHA Matter?

Another common question is whether the form of DHA matters.

Some people argue that triglyceride-based DHA is the wrong form for brain delivery, and that phospholipid DHA would be superior. This is an important scientific question. Different DHA forms may behave differently.

But it is not correct to say that triglyceride DHA cannot enter phospholipid pools.

In our earlier pilot work, participants took triglyceride-based algal DHA. After supplementation, DHA increased not only in triglycerides, but also in phosphatidylcholine and cholesteryl ester lipid pools in both plasma and CSF [6]. These lipid pools appear to exchange with one another.

So the form of DHA may matter, but triglyceride DHA is not biologically isolated from phospholipid DHA. And they both reach the brain.

The Omega-3 Index

One practical way to measure omega-3 status is the omega-3 index.

This is the percentage of EPA plus DHA in red blood cell membranes. It is useful because red blood cells give a longer-term picture of omega-3 status than a single diet questionnaire.

Some brain health clinics now use the omega-3 index as part of a broader prevention workup [8], with a low or suboptimal index defined as < 6 and an optimal index>10. But these cut-offs are not validated in intervention trials.

The omega-3 index is not an Alzheimer’s test. A low number does not mean someone will develop dementia. And raising the number with supplements has not been proven to prevent cognitive decline.

It is best understood as one piece of a larger picture: diet, exercise, APOE genotype, vascular risk, inflammation, and brain biomarkers.

The Bigger Question: How Does the Brain Use DHA?

PreventE4 was designed to answer the delivery question and its implications on AD risk: once DHA reaches the brain, do we see signs of decreased AD risk?

This matters especially for APOE4.

In Alzheimer’s disease brain tissue, we see signs that polyunsaturated fatty acids are being broken down and remodeled in an inflammatory environment. This includes changes in omega-3 and omega-6 balance, lower DHA-related resolving mediators, and activation of inflammatory lipid pathways [9].

One enzyme that may be involved is cPLA2. This enzyme cuts fatty acids out of cell membranes. When it is overactive, it can release arachidonic acid and drive inflammatory lipid signaling.

In APOE4 and Alzheimer’s disease, we found that cPLA2 appears to be more active [9,14].

Put simply, the APOE4 brain may not only need more DHA delivered to it. It may also need help using DHA in the right way.

This is an important shift in thinking.

The question is not only: “How much omega-3 do you eat?”

It is also: “How does your brain metabolize omega-3s?”

What PreventE4 Was Designed to Test

I am deeply grateful to Lon Schneider for helping me design this trial. We had intense discussions on who we should recruit, what the outcomes should be, and how we interpret these findings. Wendy Mack was instrumental in planning and estimating sample sizes and analysis plans.

PreventE4 was a double-blind, placebo-controlled prevention trial in cognitively healthy older adults between ages 55 and 80 [10].

Participants had to have at least one dementia-related risk factor. These included obesity, high blood pressure, high cholesterol, low physical activity, or fewer than 12 years of education. They also had to have preserved daily function and no clinical dementia.

The trial intentionally looked for people with limited omega-3 intake. Participants were excluded if they were taking omega-3 supplements or consuming more than 200 mg per day of DHA [10].

In plain language, we were trying to find people who did not already eat much fish, had dementia risk factors, but did not yet have dementia.

That sounds straightforward.

It was not.

What It Took to Run the Trial

PreventE4 was conceived in 2016. Funding came in 2017 from NIH, two years after the application was submitted. This was followed by funding from ADDF. Recruitment began in May 2018.

The plan sounded simple: recruit cognitively healthy older adults with dementia risk factors, low omega-3 intake, and enough APOE4 carriers to test whether genotype changed DHA delivery.

In practice, it was enormously difficult.

We had to screen thousands of people to find the right participants. They had to be older adults who did not like to eat much fish, had dementia risk factors, were cognitively healthy, were willing to be followed for two years, and, in many cases, were willing to undergo a lumbar puncture.

This was especially challenging for APOE4 homozygotes, who are much less common than non-carriers or people with only one APOE4 copy.

The trial required constant outreach, prescreening, saliva collection, genotyping, phone calls, follow-up visits, and reassurance. Staff members worked tirelessly to call participants, answer questions, mail saliva kits, receive tubes, organize samples, and prepare batches for genotyping.

I remember the mechanics of it vividly.

Once a week, the genotyping workflow had to keep moving. I remember visiting the lab and finding packages dropped off by FedEx, sometimes many of them, waiting to be processed. On Fridays, I would pick up those packages and help deliver them to the genotyping team.

It was not glamorous work. It was not the part of science that people usually see. But without those tubes, labels, calls, packages, and staff members, there would be no PreventE4.

Then COVID changed everything.

Recruitment slowed. In-person visits became harder. Some participants dropped out. Safety rules changed how we could see people, collect samples, and keep the study moving. In 2021, the sample size had to be increased to account for the unexpected dropouts.

What began as a carefully powered prevention trial became a test of persistence.

Recruitment eventually ended in May 2024.

The Intervention

Participants were randomized to either 2 grams per day of DHA or an identical placebo for two years [10].

The DHA was algal-derived, which allowed a high, standardized dose.

Both the DHA and placebo groups also received a high-dose vitamin B complex. This was intentional. B-vitamin status and homocysteine metabolism can affect brain aging and may interact with omega-3 biology. By giving vitamin B supplementation to both groups, the trial reduced the chance that low B-vitamin status would confound the DHA results.

The primary outcome was measured at six months in the lumbar puncture subset. The question was whether DHA supplementation changed the ratio of DHA to arachidonic acid, or DHA/AA, in CSF [10].

CSF is not the brain itself. But it is one of the closest fluids we can safely measure in living people.

Secondary outcomes were measured over two years and focused on the brain. These included resting-state functional MRI, diffusion tensor imaging, and volumetric MRI [10]. In simpler terms, the trial asked whether DHA affected brain connectivity, white matter structure, hippocampal volume, and other imaging markers.

Exploratory outcomes included cognition, measured with the Repeatable Battery for the Assessment of Neuropsychological Status, or RBANS, along with blood biomarkers, inflammatory lipids, oxylipins, amyloid-related markers, phosphorylated tau, and microbiome sampling [10].

What We Found

The omega-3 index helped confirm that we recruited the right population and that the intervention did what it was supposed to do.

Testing showed that omega-3 levels in red blood cells climbed dramatically, from 4.9% to 11%. A 6% increase (shown in figure below)

DHA levels in CSF, the fluid that surrounds the brain, rose by an average of 17% after six months [13].

This confirmed that DHA reached its intended target.

The same increases were seen in people with an APOE4 gene, the group thought most likely to benefit from supplementation [13].

That finding matters.

At the level of the blood, people responded to DHA. The supplement was taken, absorbed, and incorporated into red blood cell membranes. At the level of CSF, DHA also increased.

So the simplest explanation is not that the dose failed to reach the body or the central nervous system.

But over two years, DHA supplementation did not improve cognition or slow hippocampal atrophy.

As you can see, there is a complete overlap between the treatment and placebo arms on cognition. No difference!

APOE4 carriers did show signs of greater vulnerability. They started slightly higher in the begining and learned less on the RBANS cognitive battery toward the end (Figure below). Something similar happened with hippocampal volumes. That finding is consistent with what we know about APOE4’s effect on cognition and hippocampal volume during preclinical (predementia) phases.

But being in the DHA treatment arm did not change that pattern.

This point is important.

Prevention trials in cognitively healthy people are hard because ceiling effects can make it difficult to detect change over only two years. But RBANS is a sensitive cognitive tool, and PreventE4 was a well-designed randomized trial. So while ceiling effects may have limited our ability to detect a small signal, we also have to be clear: we did not see a treatment signal on cognition or hippocampal volume.

Raising the omega-3 index to >10 did not affect AD biomarkers in 2 years. This is a real limitation to relying on the omega-3 index as an AD diagnostic.

That is the central result of PreventE4.

The intervention worked biologically. It changed omega-3 levels in blood and CSF. But it did not produce a measurable cognitive or hippocampal-volume benefit over two years in cognitively healthy older adults.

Why Prevention Trials Are So Hard

The first thing to remember is that PreventE4 was a prevention trial conducted in people without dementia.

That is both the strength and the challenge.

When people are cognitively normal at the start of a study, there may be very little room to improve on standard memory tests over only two years. This is called a ceiling effect. And participants do better on repeated testing. This is known as practice or learning effects. If a person is already doing well, a test may not be sensitive enough to show whether the brain is being protected.

Alzheimer’s disease also does not always progress in a straight line.

For years, changes may be slow and difficult to detect. Then, as someone approaches mild cognitive impairment or dementia, decline can accelerate.

In other words, there may be two slopes: a slow early slope and a much steeper later slope (Figure above).

A two-year study in cognitively healthy people may happen during the slow slope, before standard cognitive tests or hippocampal volume show large treatment effects. But waiting until the steep slope begins may be too late, because neurodegeneration may be harder to reverse once neurons and synapses are already being lost.

That is the central dilemma of prevention research.

The best time to intervene may be before we can easily measure clinical decline.

Still, this limitation does not erase the finding. PreventE4 used sensitive cognitive testing, MRI measures, biomarkers, and a randomized placebo-controlled design, and APOE4 carriers had a signal that they learned less and had worse hippocampal volume changes. The study showed clear biological target engagement, but it did not show that DHA changed the cognitive or hippocampal-volume trajectory over two years.

Why We Tested One High Dose

One design decision is worth explaining.

We chose to test a high dose of DHA rather than several different doses.

That was intentional.

Since one of the central questions was whether APOE4 modifies DHA delivery and response, we chose a design that gave us the best chance of detecting that interaction.

With this dose, the biological signal was convincing. DHA increased in blood, and DHA increased in CSF, but we did not see an interaction with APOE4.

The effect of APOE4 on DHA brain transport is stronger in patients with mild cognitive impairment or dementia.

So the problem was not simply an inadequate dose or failed delivery at this clinical stage.

The harder question is what CSF DHA actually tells us.

CSF is not the same as a neuron. It is not a direct biopsy of a brain cell membrane. We cannot say with certainty that a rise in CSF DHA means the same rise occurred inside neurons or synapses.

But CSF is one of the best windows we have into the living human brain. It surrounds the brain and spinal cord, and changes in CSF often reflect changes in central nervous system biology. So while CSF DHA is not a perfect measure of neuronal DHA, it is a reasonable marker of brain delivery.

What This Means for Omega-3 Supplements

PreventE4 gives us a clearer way to think about omega-3s and brain health.

The message is not that omega-3s are unimportant. DHA is essential for the brain. It is part of neuronal membranes, synapses, and inflammatory signaling.

The message is also not that every person should or should not take a supplement. That decision is personal.

The more important lesson is that omega-3 supplementation alone is not a blunt solution for Alzheimer’s prevention.

A capsule cannot fully replace the biological context in which omega-3s seem to be most helpful.

In Mediterranean-style dietary patterns, omega-3s come with a broader lifestyle: fatty fish, vegetables, legumes, nuts, olive oil, physical activity, social connection, and lower cardiometabolic risk.

Whole foods bring more than one nutrient. Salmon contains DHA and EPA, but also protein, vitamin D, vitamin B12, selenium, potassium, and other nutrients. Walnuts and seeds provide plant-based omega-3s along with fiber, minerals, and polyphenols.

That context may matter as much as the omega-3 number itself.

It is plausible that a healthy lifestyle does more than raise omega-3 levels. It may also create the biology needed to use DHA effectively.

A person who eats fatty fish, exercises, sleeps well, has good vascular health, and has lower chronic inflammation may not only have higher omega-3 levels. They may also have a brain environment that transports, retains, remodels, and uses DHA better.

That is different from taking a capsule in the setting of poor sleep, high stress, sedentary behavior, obesity, insulin resistance, and an ultraprocessed diet.

In that setting, DHA may reach the blood and even the central nervous system, but the broader biology may not be ready to use it effectively.

This is the key shift.

Omega-3s are not magic. They are part of a biological system. And a quick fix with supplements may not work.

Where We Go Next

PreventE4 did not end the omega-3 story. It sharpened the next question.

The problem may not simply be getting DHA into the body or even into CSF during preclinical AD. The problem may be whether the brain can retain, remodel, and use omega-3s in the right way.

This is why we think the next generation of research needs to focus on DHA metabolism.

How does inflammation change brain DHA use? What is the role of gut dysbiosis?

And can we change the brain environment so omega-3s work better?

This is where lifestyle becomes important.

A healthy lifestyle may not only raise omega-3 levels. It may also create the biology needed to use DHA effectively. A person who eats fatty fish, exercises, sleeps well, has good vascular health, and has lower chronic inflammation may have a brain environment that transports, retains, remodels, and uses DHA better.

That may be more meaningful than any single omega-3 index number, supplement label, or blood level.

This is also where our work on brain-penetrant cPLA2 inhibitors comes in.

cPLA2 is an enzyme that cuts polyunsaturated fatty acids out of cell membranes. When it is overactive, it can release both arachidonic acid and DHA and drive unresolved inflammation. In APOE4 and Alzheimer’s disease, cPLA2 appears to be more active. That may push the brain toward a state where polyunsaturated fatty acids are being broken down, remodeled, or consumed faster than they can be used for healthy membrane repair and anti-inflammatory signaling [9,14].

In simple terms, giving DHA may not be enough if the brain is stuck in a lipid-inflammatory state that keeps burning through these fats.

Our future direction is to target that biology directly.

We are developing brain-penetrant cPLA2 inhibitors designed to reduce harmful inflammatory lipid signaling and help restore a healthier omega-6 to omega-3 balance in the brain. The goal is not to replace omega-3 nutrition. The goal is to change the brain environment so omega-3s can be used more effectively [14].

That is still early-stage work. It is not yet a human Alzheimer’s treatment.

That is where nutrition, lipid biology, neuroinflammation, lifestyle, and drug discovery now come together.

Take-Home Messages

• PreventE4 does not give a simple yes or no answer about whether an individual should take DHA supplements.

• The trial showed that high-dose DHA reached its biological targets: omega-3 levels in red blood cells rose from 4.9% to 11%, and CSF DHA rose by about 17%.

• The same biological response was seen in APOE4 carriers.

• DHA did not improve cognition or slow hippocampal atrophy over two years.

• APOE4 carriers learned less on RBANS and had greater hippocampal atrophy, but DHA treatment did not change that pattern.

• That does not mean omega-3 biology is unimportant. It means delivery alone may not be enough.

• The next question is DHA metabolism: how the brain transports, retains, remodels, and uses omega-3s.

• A holistic lifestyle may help create the biology needed to use DHA effectively.

• Future prevention may require combining nutrition, lifestyle, biomarkers, and therapies that target inflammatory lipid metabolism, including cPLA2.

Publication

Yassine HN, Ghasem Pour S, Juarez M, Arellanes IC, Ali N, Dikeman D, Sanchez A, Park J, Kerman B, Duro MV, Asante I, Louie S, Kono N, D’Orazio L, Chui H, Mack WJ, Harrington MG, Braskie MN, Schneider LS.
CNS Target Engagement of High-dose DHA Supplementation in Older Adults at Risk for Dementia: A Randomised, Double-blind, Placebo-controlled Trial.

Acknowledgments

PreventE4 was a team effort. Hussein N. Yassine designed and supervised the trial, obtained funding, and wrote the manuscript. Sara Ghasem Pour helped with data analysis and manuscript drafting. Nada Ali assisted with manuscript drafting. Marlene Juarez helped with cognitive testing. Isabella C. Arellanes and Dante Dikeman coordinated study visits. Ashley Sanchez assisted with blood sample collection and storage. Bilal Kerman, Marlon V. Duro, and Isaac Asante assisted with biomarker measurements. Stan Louie supervised biomarker management. Jackson Park assisted with study coordination. Naoko Kono led data management. Lina D’Orazio supervised cognitive testing. Helena Chui helped with study design. Wendy J. Mack supervised study design and all data analysis. Michael G. Harrington performed lumbar punctures. Meredith N. Braskie oversaw the imaging biomarkers. Lon S. Schneider supervised study design, grant applications, and manuscript writing. Naoko Kono and Wendy J. Mack accessed and verified the underlying data. All authors read and approved the final version of the manuscript.

This study was funded in part by the National Institutes of Health/National Institute on Aging, including RF1AG076124, R01AG055770, R01AG067063, R01AG054434, R21AG056518, and P30AG066530 to Hussein N. Yassine; the Alzheimer’s Drug Discovery Foundation, GC-201711–2014197 to Hussein N. Yassine and Lon S Schneider; and donations from the Vranos and Tiny Foundations and Ms. Lynne Nauss to Hussein N. Yassine. Wendy J. Mack, Meredith N. Braskie, and Lon S. Schneider also received funding from NIH R01AG054434 and the Alzheimer’s Drug Discovery Foundation. We also thank the members of the Data and Safety Monitoring Board for their oversight and guidance.

References

1. Yassine HN, Feng Q, Azizkhanian I, et al. Association of serum docosahexaenoic acid with cerebral amyloidosis. JAMA Neurology. 2016.

2. Yassine HN, Rawat V, Mack WJ, et al. The effect of APOE genotype on the delivery of DHA to cerebrospinal fluid in Alzheimer’s disease. Alzheimer’s Research & Therapy. 2016;8:25.

3. Yassine HN, Braskie MN, Mack WJ, et al. Association of docosahexaenoic acid supplementation with Alzheimer disease stage in apolipoprotein E ε4 carriers: a review. JAMA Neurology. 2017;74(3):339–347.

4. Yassine HN, Croteau E, Rawat V, et al. DHA brain uptake and APOE4 status: a PET study with [1-¹¹C]-DHA. Alzheimer’s Research & Therapy. 2017;9:23.

5. Arellanes IC, Choe N, Solomon V, et al. Brain delivery of supplemental docosahexaenoic acid (DHA): a randomized placebo-controlled clinical trial. EBioMedicine. 2020;59:102883.

6. Bantugan MA, Xian H, Solomon V, et al. Associations of ApoE4 status and DHA supplementation on plasma and CSF lipid profiles and entorhinal cortex thickness. Journal of Lipid Research. 2023;64(6):100354. doi:10.1016/j.jlr.2023.100354.

7. Yassine HN, Samieri C, Livingston G, et al. Nutrition state of science and dementia prevention: recommendations of the Nutrition for Dementia Prevention Working Group. The Lancet Healthy Longevity. 2022;3:e501–e512.

8. Yassine HN. The omega-3 index in Alzheimer’s disease: ready for prime time? American Journal of Clinical Nutrition. 2022.

9. Ebright B, Assante I, Poblete RA, et al. Eicosanoid lipidome activation in post-mortem brain tissues of individuals with APOE4 and Alzheimer’s dementia. Alzheimer’s Research & Therapy. 2022;14:152.

10. Yassine HN, Arellanes IC, Mazmanian A, et al. Baseline findings of PreventE4: a double-blind placebo controlled clinical trial testing high dose DHA in APOE4 carriers before the onset of dementia. Journal of Prevention of Alzheimer’s Disease. 2023.

11. Yassine HN, Carrasco AS, Badie DS. Designing newer omega-3 supplementation trials for cognitive outcomes: a systematic review guided analysis. Journal of Alzheimer’s Disease. 2024.

12. Ebright B, Duro MV, Chen K, Louie S, Yassine HN. Effects of APOE4 on omega-3 brain metabolism across the lifespan. Trends in Endocrinology & Metabolism. 2024.

13. Yassine HN, Ghasem Pour S, Juarez M, et al. CNS target engagement of high-dose DHA supplementation in older adults at risk for dementia: a randomized, double-blind, placebo-controlled trial. EBioMedicine. 2026.

14. Sadybekov AV, Duro MV, Wang S, et al. Development of potent, selective cPLA2 inhibitors for targeting neuroinflammation in Alzheimer’s disease and other neurodegenerative disorders. npj Drug Discovery. 2026.

She carries a genetic risk for Alzheimer’s disease and wanted to know whether she should take rapamycin to reduce her risk.

This post is dedicated to her.

My answer today is: no.

Not because rapamycin is uninteresting. It is one of the most important drugs in aging biology. It extends lifespan in mice, affects a major growth-and-repair pathway, and may improve some aspects of immune function in older adults.

But the human evidence does not show that rapamycin works in people the way it works in mice.

We do not yet have convincing evidence that rapamycin slows human aging, improves Alzheimer’s biomarkers, prevents cognitive decline, or reduces dementia risk. The strongest human data are much narrower: vaccine response, infection-related outcomes, and tissue-specific markers such as skin aging.

That is not enough to justify taking rapamycin today for Alzheimer’s prevention. But it gets more complicated. How do we know it’s working for aging or AD?

What Is Rapamycin?

Rapamycin, also called sirolimus, is a drug originally discovered from bacteria found on Easter Island, also known as Rapa Nui.

It blocks part of a pathway called mTOR, short for mammalian target of rapamycin.

Think of mTOR as one of the cell’s main growth switches. When food and energy are plentiful, mTOR tells cells to grow, build proteins, divide, and store nutrients. When mTOR activity goes down, cells may shift more toward repair, recycling, and maintenance.

This is why rapamycin became so interesting to aging researchers.

Aging is not only about time passing. It is also about how cells handle damage, inflammation, protein buildup, and repair. By lowering mTOR activity, rapamycin may push cells away from constant growth and toward cellular maintenance.

But this does not mean that “slowing metabolism” automatically slows aging.

That is too simple.

A better way to think about rapamycin is that it changes the balance between growth mode and repair mode. That may help in some tissues, at some ages, and in some disease states. But growth signals are also needed for muscle repair, wound healing, immune defense, and recovery after exercise.

The real question is not whether mTOR is good or bad.

The real question is: how much mTOR activity, in which tissue, at what age, and at what time?

Why Rapamycin Became Famous

The strongest evidence for rapamycin comes from mice.

The landmark study was published in 2009 through the National Institute on Aging’s Interventions Testing Program, or ITP.

This was not a small mouse experiment. It was a large, multi-site study using genetically diverse mice. More than 1,900 mice were included. Rapamycin treatment began late in life, at about 20 months of age.

The result was striking: rapamycin extended lifespan in both male and female mice.

Figure 1: from Harrison et al. 2009 — t rapamycin extended lifespan in genetically heterogeneous mice when started late in life

One way the investigators reported this was by examining the age at which 90% of the mice had died. That age increased from about 1,078 to 1,179 days in males, and from about 1,094 to 1,245 days in females.

This mattered because the drug worked even when started late in life. That suggested rapamycin might affect the biology of aging itself, not just growth or development.

Later mouse studies supported the same general idea. Rapamycin became one of the most reproducible lifespan-extending drugs in mice. But those studies also showed that the effect depends on dose, sex, timing, and treatment schedule.

That is impressive.

But mouse lifespan extension is not the same as human disease prevention.

Rapamycin and Alzheimer’s Mouse Models

Rapamycin has also been tested in Alzheimer’s mouse models.

In some models, including 3xTg-AD mice, rapamycin or mTOR inhibition has been reported to reduce amyloid and tau-related pathology and improve memory-like performance.

The mechanism makes sense.

mTOR inhibition can increase autophagy, the cell’s recycling system. Autophagy helps cells clear damaged proteins and worn-out cell parts. Since Alzheimer’s disease involves protein buildup and cellular stress, improving cleanup is an attractive idea.

But Alzheimer’s mouse models are not the same as human Alzheimer’s disease.

They are simplified versions of the disease. Many are engineered to develop amyloid or tau changes quickly. Human Alzheimer’s develops over decades and is shaped by aging, blood vessels, sleep, infections, metabolism, immune history, and genetics such as APOE4.

So the mouse studies are useful for understanding possible mechanisms.

They are not proof that rapamycin prevents Alzheimer’s disease in humans.

Why Mice Are Not Small Humans

The mouse data justify human trials.

They do not justify routine human use.

Mice and humans age differently. Mice are short-lived and fast-growing. Humans grow slowly and live for decades. A drug that slows a growth pathway in a mouse may not have the same effect in a long-lived human.

Laboratory mice also live in controlled environments. Their food, movement, temperature, and exposure to germs are carefully managed.

Humans live in the real world.

We experience infections, stress, poor sleep, medications, injuries, exercise, obesity, diabetes, dental disease, and decades of immune memory.

This matters especially for the immune system.

By older age, each person has a unique immune biography. One person may have had shingles, repeated respiratory infections, autoimmune disease, insulin resistance, or chronic inflammation. Another person may not.

So two 70-year-olds may have very different immune systems.

If rapamycin helps one kind of immune aging, it may not help another. It may even hurt some people if the dose, timing, or baseline biology is wrong.

That is why biomarkers matter. And for immune aging, we are learning that they need to be tissue or organ-specific.

In mice, we can give rapamycin and measure lifespan. In humans, we need to know what biology we are trying to improve.

Is it high mTOR activity? Poor vaccine response? Chronic inflammation? Weak autophagy? Senescent cells? Metabolic dysfunction? Brain inflammation?

Right now, we do not have a simple clinical test that answers this.

How Can Rapamycin Improve Immunity?

Rapamycin can seem confusing.

At high or continuous doses, it can suppress the immune system. That is why rapamycin and related drugs have been used in transplant medicine.

But low-dose or intermittent mTOR inhibition may act differently, especially in older adults.

Older immune cells may be chronically stimulated, inflamed, and inefficient. They may be “on” all the time, but not very good at responding to a new threat.

A short course of mTOR inhibition may reduce some of that background noise. It may help immune cells clean up damaged parts and respond more effectively to a challenge, such as a vaccine.

So rapamycin may not “boost” immunity like pressing the gas pedal.

It may help some older immune cells respond more cleanly by taking them out of a stressed, inefficient, or exhausted state. I have dedicated a previous post to why immune exhaustion is a hot topic in AD.

That is the idea behind the vaccine-response studies, and why it may be relevant for the brain. But we still cannot link them.

What Do Human Studies Show?

Human studies of rapamycin and rapamycin-like drugs do exist. They are interesting, but limited.

Mannick 2014: Vaccine Response

The best-known positive human study was published by Mannick and colleagues in 2014.

This randomized trial tested RAD001, also called everolimus, a rapamycin-like drug, in older adults before influenza vaccination. Participants received short-term treatment before vaccination. The main outcome was antibody response to the flu vaccine.

The result was encouraging: short-term mTOR inhibition improved vaccine response in older adults.

This was important because rapamycin is often thought of as an immunosuppressing drug. The study suggested that dose and schedule matter. Low-dose or intermittent mTOR inhibition may improve some aspects of immune function in older people.

This is a positive study because the outcome is well defined: antibody titers. And it has clear implications. If our goal is to improve the immune response to a vaccine in older adults, rapamycin works. We need to stop and not extrapolate beyond what the study showed.

Mannick 2018: Infections

A later Mannick study, published in 2018, extended this idea, but the design was more complicated.

This was a Phase 2a randomized, placebo-controlled trial in 264 older adults. Participants received low-dose mTOR-inhibitor therapy for 6 weeks.

This study did not test rapamycin itself. It tested rapamycin-pathway drugs, including everolimus and dactolisib, alone or in combination.

The strongest signal came from a low-dose combination treatment. This group had fewer reported infections over the following year. The study also found increased activity of antiviral gene programs and improved flu-vaccine response.

Together, the 2014 and 2018 studies suggest that mTOR-pathway drugs can affect immune aging in humans in relation to an immune response to a vaccine and infections. That is the strongest positive human signal so far.

Topical Rapamycin in Skin

Another human study tested topical rapamycin for skin aging.

In this small randomized study, adults over 40 applied rapamycin to one hand and a placebo to the other. The study reported reductions in p16INK4A, a marker linked to cellular senescence, and increases in collagen VII, a protein important for skin structure.

This suggests rapamycin can affect aging-related markers in human tissue. But it was a local skin study, not a systemic aging or brain-aging study. Another example where rapamycin wins in well-defined outcomes with clear readouts.

Rapamycin and Exercise

However, a small exercise study raises an important caution. One effect on skin or a vaccine response does not mean better health.

In this 13-week randomized trial, 40 sedentary older adults received either once-weekly low-dose rapamycin or a placebo while doing a light home exercise program.

The expectation was that rapamycin might improve exercise benefits by improving aging biology. The signal went the other way.

The placebo group improved more than the rapamycin group on some physical-function measures. The clearest difference was in the 30-second sit-to-stand test, a simple measure of lower-body strength and function. Grip strength and self-rated well-being also reportedly favored the placebo. Side effects such as soreness and fatigue were more common in the rapamycin group, and one serious infection was reported.

This study was small and short. It does not prove that rapamycin always blunts exercise benefits. But it matters because exercise is one of the strongest proven interventions for healthy aging and dementia prevention.

Exercise works by creating a controlled stress. Muscle is challenged, then rebuilds stronger. That rebuilding requires mTOR. If rapamycin is active during the recovery window, it may interfere with the muscle’s ability to adapt. We saw something similar with metformin, albeit via a different mechanism.

So, To Rapamycin or Not To Rapamycin? Paraphrasing Shakespeare

In humans, the clearest positive signals are the immune response and selected tissue markers. Rapamycin-like drugs have improved vaccine response in older adults. A later study suggested fewer infections in selected treatment groups. Topical rapamycin changed some aging-related markers in skin.

These are real biological signals.

But they are not the same as showing that rapamycin slows human aging, protects the brain, or prevents Alzheimer’s disease.

That distinction is the main point.

For Alzheimer’s prevention, we would want evidence that rapamycin improves meaningful brain-related outcomes: amyloid, tau, neurodegeneration, brain inflammation, cognition, or dementia risk.

We do not have that evidence.

Instead, the field has mostly shown that rapamycin can change selected immune or tissue readouts under specific conditions. That is interesting biology. It is not yet a prevention therapy.

Where the Field Seems to Be Moving

The most active human work no longer seems to be “rapamycin for everyone to slow aging.”

It seems to be moving toward more specific questions:

Can rapamycin improve immune aging?

Can it slow ovarian aging?

Can it affect periodontal disease?

Can it change skin aging markers?

Can it be timed or dosed in a way that preserves benefit while avoiding harm?

That shift is important. It suggests that the field recognizes the main problem: aging is not one pathway, and rapamycin is not one simple anti-aging switch.

A useful rapamycin trial needs a clear biological target and a clear readout.

For Alzheimer’s prevention, we do not have that.

The Real Missing Piece: Immune-Aging Biomarkers

The biggest lesson from rapamycin is not only about rapamycin.

It is about measurement.

For cardiovascular disease, we can measure blood pressure, ApoB, LDL cholesterol, HbA1C, and other risk factors. We can treat those risk factors and track whether the treatment worked.

For immune aging, we do not have an equivalent.

We can measure antibody response after a vaccine. That is useful, but it is only one test of one part of the immune system. It does not tell us whether rapamycin improved brain aging, neuronal autophagy, senescent-cell burden, or Alzheimer’s risk.

This is the central challenge.

To use rapamycin intelligently, we would need biomarkers that tell us who has the kind of aging biology rapamycin can improve, what dose is enough, whether the drug hit the intended pathway, and whether the benefit is worth the risk.

That is hard because the human immune system is not one number. It is shaped by age, sex, infections, vaccines, metabolism, obesity, sleep, exercise, medications, genetics, chronic inflammation, and decades of immune memory.

So the future of rapamycin may depend less on enthusiasm and more on biomarkers.

Not “Who wants to try rapamycin?”

But: Who has the biology that rapamycin is likely to improve, and how will we know it worked? And in what organ? Topical for skin aging? Adjuvant to a vaccine?

Rapamycin and Alzheimer’s Prevention

Rapamycin has a plausible connection to Alzheimer’s biology. mTOR is linked to autophagy, metabolism, inflammation, immune aging, and synaptic function. These all matter for brain aging.

But plausibility is not proof.

For someone with a genetic risk for Alzheimer’s disease, the key question is not whether rapamycin is biologically interesting. It is. The question is whether the existing human studies show a prevention effect.

They do not.

Today, I would not recommend rapamycin for Alzheimer’s prevention outside a clinical trial.

The Risk Side

Rapamycin is not a supplement. It is a real drug.

It can affect immunity, cholesterol and triglycerides, blood sugar, wound healing, mouth ulcers, infection risk, and drug interactions. The risk depends on dose, schedule, age, baseline health, and other medications.

Some people argue that once-weekly low-dose rapamycin may avoid many problems seen with daily immune-suppressing doses.

That may be true.

But “may be safer” is not the same as “proven safe for long-term prevention in healthy people.”

Prevention raises the bar. A prevention drug must be very safe, because many people taking it may never develop the disease they are trying to avoid.

Areas of Uncertainty

Several questions remain unanswered.

What is the right dose?

Is weekly dosing better than daily dosing?

Should rapamycin be used only in older adults?

Does APOE genotype matter?

Does baseline inflammation matter?

Does metabolic health matter?

Does rapamycin help people who already exercise, sleep well, and have good metabolic health?

Could it blunt exercise adaptation in some settings?

Can it improve Alzheimer’s biomarkers?

Most importantly: what specific outcome? That determines whether it is working.

Until we can answer these questions, use for aging or Alzheimer’s prevention is not recommended.

What I Would Tell My Friend

I would tell my friend that rapamycin is scientifically important, but not ready for Alzheimer’s prevention.

For now, I would focus on interventions with stronger human evidence: exercise, blood pressure control, sleep, metabolic health, hearing, social connection, vascular risk reduction, and treatment of insulin resistance or diabetes when present.

The next step is better and targeted trials with better immune-aging biomarkers tied to outcomes more concrete than human aging.

Take-Home Messages

• Rapamycin is one of the most compelling anti-aging drugs in mouse studies.

• Mouse lifespan extension does not prove human benefit.

• Alzheimer’s mouse-model data are useful for understanding mechanisms, but they do not prove Alzheimer’s prevention in humans.

• Human trials exist, but the strongest positive evidence is improved vaccine response and fewer infections in selected settings.

• Low-dose rapamycin may improve immune response by reducing age-related immune dysfunction, not by simply “boosting” the immune system.

• Rapamycin may interfere with exercise adaptation if mTOR is suppressed during the muscle-rebuilding window.

• The human studies so far do not show that rapamycin works in people the way it works in mice.

• We do not yet have convincing evidence that rapamycin improves Alzheimer’s biomarkers such as amyloid, tau, neurodegeneration, or cognition.

• The biggest missing piece is a practical immune-aging biomarker that tells us who might benefit and whether the drug worked.

• Rapamycin is a real drug with real risks, especially if used long-term.

• For aging or Alzheimer’s prevention, the current evidence does not justify rapamycin use today.

• The right conclusion is not “rapamycin does not work.” The better conclusion is: what is a better indication for rapamycin? And how do we design smarter trials?

Key References

1. Harrison DE, Strong R, Sharp ZD, et al. Rapamycin fed late in life extends lifespan in genetically heterogeneous mice. Nature. 2009.

2. Miller RA, Harrison DE, Astle CM, et al. Rapamycin-mediated lifespan increase in mice is dose and sex dependent and metabolically distinct from dietary restriction. Aging Cell. 2014.

3. Caccamo A, Majumder S, Richardson A, Strong R, Oddo S. Molecular interplay between mTOR, amyloid-beta, and tau: effects on cognitive impairments. Journal of Biological Chemistry. 2010.

4. Mannick JB, Del Giudice G, Lattanzi M, et al. mTOR inhibition improves immune function in the elderly. Science Translational Medicine. 2014.

5. Mannick JB, Morris M, Hockey HP, et al. TORC1 inhibition enhances immune function and reduces infections in the elderly. Science Translational Medicine. 2018.

6. Chung CL, Lawrence I, Hoffman M, et al. Topical rapamycin reduces markers of senescence and aging in human skin: an exploratory, prospective, randomized trial. GeroScience. 2019.

7. Kaeberlein M, Galvan V. Rapamycin and Alzheimer’s disease: time for a clinical trial? Science Translational Medicine. 2019.

If the Alzheimer’s brain is inflamed, should we reduce inflammation? First, we need to answer a more basic question:

How do we know which patient actually has harmful brain inflammation?

That is the main problem this post is about.

The field is moving toward immune-based treatments before we have reliable tools to identify the right immune state in the right patient. The recent XPro1595 trial is a useful case study. It is not just a trial of one drug. It is a test of whether Alzheimer’s research is ready for precision neuroinflammation trials.

The Main Problem: We Lack a Good Brain Inflammation Marker

In Alzheimer’s disease, we now have increasingly useful biomarkers for amyloid and tau. They are not perfect, but they tell us something fairly specific about Alzheimer’s pathology.

Inflammation is harder.

We do have good markers of general inflammation. CRP, for example, is a useful blood marker. If CRP is high, it tells us that the body may be in an inflammatory state.

But CRP does not tell us where the inflammation is coming from.

A high CRP could reflect arthritis, infection, obesity, heart disease, gum disease, poor sleep, or many other conditions outside the brain. It may matter for brain health, but it does not prove that the Alzheimer’s brain itself is inflamed.

The same problem applies to ESR and HbA1C. They may capture body-wide inflammation, metabolic stress, or vascular risk. But they do not measure brain inflammation directly.

APOE4 is more closely tied to Alzheimer’s biology and may shape immune vulnerability in the brain. But APOE4 is a genetic risk marker, not a real-time inflammation test.

This creates a major problem for trials.

If we cannot measure brain inflammation well, then we may not know who should receive an immune-targeted treatment. We also may not know whether the treatment worked biologically, even if the clinical results move in the right direction.

In drug development, that matters.

A proper Phase 2 trial should do more than look for a clinical signal. It should help define the population, the dose, the target, and the biological readout. It should show target engagement: evidence that the drug hit the pathway it was designed to hit.

There Is Inflammation in the Alzheimer’s Brain. But Is It Good or Bad?

There is strong evidence that the brain’s immune system is active in Alzheimer’s disease.

In brain tissue, immune cells cluster around amyloid plaques. Genetic studies point toward immune pathways, including TREM2, CD33, INPP5D, PLCG2, and APOE.

APOE4, the strongest common genetic risk factor for late-onset Alzheimer’s disease, is especially important here. It is not only about cholesterol. In the brain, APOE4 may affect lipid handling, amyloid clearance, blood-brain barrier function, and immune tone.

Single-cell studies also show that microglia can shift into disease-associated microglia, or DAM. DAM are often found near amyloid plaques. Early on, this may be protective: surrounding plaques, clearing debris, and containing injury.

But a protective response can become costly if it is pushed for too long.

In an APOE4 brain, where amyloid handling and immune signaling may already be stressed, DAM activation may be harder to sustain. Over time, a cleanup response may become less efficient, more inflammatory, or exhausted.

So the question is not whether inflammation exists in Alzheimer’s disease.

It does.

The better question is whether that inflammation is protective, harmful, or failing.

The Stage-Specific Model

One way to think about Alzheimer’s inflammation is by stage.

Early in the disease, inflammation may be mostly reactive and compensatory. The brain sees amyloid, damaged synapses, or cellular stress, and immune cells respond. In this setting, more immune activity may sometimes help.

This was the logic behind TREM2 agonists: enhance microglial cleanup and improve the brain’s response to amyloid. But recent TREM2 agonist trials have been disappointing, reminding us that boosting immune activity is not automatically helpful.

Later in the disease, the immune response may change. It may become chronic, senescent, or exhausted. The cells are still activated, but they may no longer be solving the problem. Instead of clearing damage, they may keep sending alarm signals and contribute to synapse loss.

This creates a different treatment goal.

Early disease may require supporting protective immune responses.

Later or inflammation-heavy disease may require inhibiting or re-aligning harmful immune signals.

This is why “anti-inflammatory treatment” is too blunt a phrase. The better goal may be stage-specific immune rebalancing.

The MINDFuL trial and the rationale for Soluble TNF

TNF, or tumor necrosis factor, is an immune signaling molecule. It is not automatically bad. The body and brain use TNF signaling for immune defense, repair, and communication between cells.

But TNF comes in different forms.

Soluble TNF can move through tissue and amplify inflammatory signaling.

Transmembrane TNF stays attached to cells and may help support normal immune balance, repair, and protective glial function.

This distinction is the rationale for XPro1595. The drug was designed to block soluble TNF while preserving transmembrane TNF. Mechanistically, XP is a dominant-negative TNF biologic. It binds to soluble TNF and forms inactive complexes, preventing soluble TNF from properly activating its receptors.

That is a reasonable biological idea. But it also raises a hard question:

Do we know which Alzheimer’s patients actually have excess soluble TNF signaling in the brain?

Right now, that answer is not clear.

XPro1595 is not really a DAM drug. It is not simply an immune exhaustion drug either. It is a soluble-TNF pathway drug. It assumes that, in some Alzheimer’s patients, inflammatory signaling has become maladaptive and that blocking soluble TNF may restore a healthier immune balance.

That assumption needs biomarkers.

What Justified Testing XPro1595?

The rationale for testing XPro1595 had three parts.

First, the biology of TNF suggested that selective soluble TNF blockade might be safer and more precise than broad TNF suppression.

Second, preclinical studies suggested that soluble TNF blockade could reduce glial activation, improve debris clearance, and restore synaptic plasticity.

Third, an earlier Phase 1b Alzheimer’s study reportedly showed dose-dependent reductions in cerebrospinal fluid biomarkers of neuroinflammation and neurodegeneration.

That last point is important because cerebrospinal fluid is closer to the brain than blood. But there is a caveat: those Phase 1b data were described as a manuscript in preparation. That means readers cannot fully judge the sample size, the biomarkers, the exposure-response relationship, or whether soluble TNF itself was measured.

So the rationale for MINDFuL was reasonable.

But it was not airtight.

What MINDFuL Tested

The MINDFuL trial was a Phase 2, randomized, double-blind, placebo-controlled study in early Alzheimer’s disease.

Participants had mild cognitive impairment or mild Alzheimer’s dementia. Treatment lasted 24 weeks. XPro1595 was given once weekly by subcutaneous injection.

The primary endpoint was EMACC, a cognitive composite designed to detect change in early Alzheimer’s disease.

That endpoint makes sense for Phase 2 signal detection. But it also has limits. EMACC is not the same as showing clear benefit on a more familiar clinical endpoint such as CDR-SB, which connects cognition and daily function.

The overall trial did not meet the primary endpoint-Figure below. Nillsh..

The study then focused on a subgroup called ADi, meaning Alzheimer’s disease with inflammation. These participants were amyloid-positive and had at least two inflammation-related markers.

The Core Critique

The core critique is not that XPro1595 had no rationale. It did.

The core critique is that the trial moved into Phase 2 without a sufficiently clear way to define the inflammatory population or prove target engagement in the brain.

The ADi subgroup was defined using CRP, ESR, HbA1C, and APOE4.

These markers may enrich for risk. They may identify people with systemic inflammation, metabolic stress, APOE4 biology, or faster expected decline.

But they do not prove harmful brain inflammation.

They also do not prove excess soluble TNF signaling in the central nervous system.

That distinction matters.

If the drug targets soluble TNF, then the ideal Phase 2 trial would show several linked steps:

The selected patients had evidence of relevant brain inflammation.

The drug reached the central nervous system.

The drug engaged soluble TNF signaling.

That engagement changed downstream biomarkers.

Those biomarker changes tracked with clinical benefit.

MINDFuL showed possible downstream biological activity, especially through p-tau217 and GFAP signals in the ADi subgroup. But these changes are not convincing for target engagement, even in the dose compliant ADi, an effect size of 0.26-0.22 is modest at best (figure below).

What do we do with these results?

Better Markers of Brain Inflammation

If this field is going to move forward, we need better biomarkers.

GFAP is one useful blood marker, but as seen above, not exactly responding to inhibition of soluble TNF. It reflects astrocyte activation and is closer to brain biology than CRP or ESR. But GFAP does not tell us which inflammatory pathway is active. It does not prove soluble TNF signaling. And worse, participants were not selected or stratified by baseline GFAP.

PET imaging may help. Some tracers, such as TSPO PET, aim to measure microglial activation. These are more brain-specific than CRP, but they still have limitations. TSPO is not exclusive to one cell type, tracer performance varies, and genetic differences can affect binding.

Experimental PET approaches may also help measure brain lipid biology.

Arachidonic acid, or AA, is involved in inflammatory lipid signaling. DHA, an omega-3 fatty acid, is important for brain membranes and inflammation resolution. AA and DHA PET are not routine clinical tools, but they point to an important idea: unresolved inflammation may involve lipid signaling, membrane turnover, and failed resolution, not just cytokines.

The field needs biomarkers that can distinguish:

body-wide inflammation

astrocyte activation

microglial activation

lipid inflammatory signaling

soluble TNF pathway activity

protective versus harmful immune states

Without this, we are partly guessing.

The Uncertainty

This is where the field needs humility.

Inflammation in Alzheimer’s may be protective early and harmful later.

Peripheral inflammation is not the same as brain inflammation.

APOE4 may increase inflammatory vulnerability, but it is not a real-time inflammation measure.

GFAP, PET imaging, and lipid PET approaches may improve the picture, but none is yet a perfect clinical tool.

Most importantly, we do not yet have a routine biomarker showing that a patient has excess soluble TNF signaling in the brain.

So XPro1595 remains scientifically interesting. But MINDFuL does not prove that we know how to identify the right patients or measure treatment efficacy with enough precision.

The Bigger Lesson

The future of Alzheimer’s inflammation treatment may not depend only on better drugs.

It may depend on better measurement.

The right patient.

The right disease stage.

The right immune pathway.

The right biomarker.

The right timing.

XPro1595 may or may not become an approved Alzheimer’s treatment. But it already teaches an important lesson:

A precision immune therapy requires precision immune biomarkers.

Without them, we risk running trials in vaguely defined populations and interpreting subgroup signals without knowing whether the drug reached the brain, hit the target, or treated the biology we intended to treat.

Take-Home Messages

• The main barrier to targeting inflammation in Alzheimer’s disease is not only treatment. It is measurement.

• We have useful markers of general inflammation, such as CRP, but we lack reliable markers of brain inflammation.

• A blood marker can tell us that the body is inflamed without proving that the brain is inflamed.

• Inflammation in Alzheimer’s may be protective early, harmful later, or ineffective when the immune response becomes exhausted.

• APOE4 may shape brain immune vulnerability, but it is not a real-time inflammation test.

• A precision immune therapy requires precision immune biomarkers.

• A strong Phase 2 trial should show more than a clinical signal. It should show brain penetration, target engagement, biomarker change, and clinical benefit.

• Without better biomarkers, Alzheimer’s neuroinflammation trials risk enrolling vaguely defined populations and producing results that are hard to interpret.

• We will likely need multiple blood-based biomarkers to detect protective or harmful inflammation markers, and senescent markers may offer clues into the latter.

Key References

1. Jaeger J, Staats KA, Barnum S, et al. XPro1595 in early Alzheimer’s disease with inflammation: results from the phase 2 MINDFuL trial. npj Dementia. 2026.

2. Heneka MT, Carson MJ, El Khoury J, et al. Neuroinflammation in Alzheimer’s disease. Lancet Neurology. 2015.

3. Hansen DV, Hanson JE, Sheng M. Microglia in Alzheimer’s disease. Journal of Cell Biology. 2018.

4. Keren-Shaul H, Spinrad A, Weiner A, et al. A unique microglia type associated with restricting development of Alzheimer’s disease. Cell. 2017.

5. Bellenguez C, Küçükali F, Jansen IE, et al. New insights into the genetic etiology of Alzheimer’s disease and related dementias. Nature Genetics. 2022.

6. Vitek MP, Brown CM, Colton CA. APOE genotype-specific differences in the innate immune response. Neurobiology of Aging. 2009.

7. Ising C, Venegas C, Zhang S, et al. NLRP3 inflammasome activation drives tau pathology. Nature. 2019.

8. Reviews on TSPO PET, GFAP, and brain lipid metabolism imaging may be useful additions for readers who want a deeper dive into neuroinflammation biomarkers.

A husband and wife came to see me in clinic.

She was 63 years old. She had two copies of APOE4, the strongest common genetic risk factor for late-onset Alzheimer’s disease. Over the previous three years, her function had declined significantly. She had been diagnosed with Alzheimer’s dementia.

Her husband had seen information about ALZ-801, an investigational oral Alzheimer’s drug being developed for people with two copies of APOE4. The message was powerful: his wife had the risk gene copy, she had Alzheimer’s, and this drug seemed designed for someone like her.

He asked me a direct question: How can we get this drug to her as soon as possible?

That question is why I am writing this.

Alzheimer’s disease creates urgency. APOE4 creates a sense of precision. How can we help her?

This post is about two related stories: taurine, a naturally occurring molecule now being studied for ApoE4 biology, and ALZ-801, an investigational oral drug built around homotaurine/tramiprosate chemistry.

Both are scientifically interesting. Neither should be treated as proven.

Why ApoE4 matters

ApoE is a protein involved in fat transport and brain health. Most people carry one of three common forms: ApoE2, ApoE3, or ApoE4. ApoE4 is a major AD risk factor

One idea is that ApoE4 may be more likely than ApoE3 to misfold or clump together. That is where taurine enters the story.

A recent paper reported that taurine can reduce ApoE4 aggregation in laboratory systems, change ApoE4 protein behavior, and shift some gene patterns in ApoE4 brain organoids toward patterns seen with ApoE3.

That is interesting. But it does not mean taurine treats Alzheimer’s disease.

Why a promising lab result is not the same as a treatment

A lab result can be exciting because it gives scientists a lead. It says: this pathway might matter. But a lead is not a medicine.

To become a medicine, a molecule has to answer many questions.

Does it work in a clear dose-response pattern? Scientists ask whether more drug produces more effect, and whether the concentration needed is realistic. For many drugs, this is measured with an IC50, the amount needed to block a process by 50 percent.

But protein clumping is messy biology. It depends on time, temperature, protein concentration, salt, assay conditions, and whether the molecule binds one specific target or weakly touches many surfaces.

That matters for taurine. In the ApoE4 paper, taurine appears to interact weakly and in multiple places. That can still matter biologically, but it is not the same as a strong, clean, highly selective drug effect.

Does it reach the brain? A molecule can work in a test tube and fail in the body. The brain is protected by the blood-brain barrier, which limits what gets from the blood into brain tissue. For an Alzheimer’s drug, reaching the brain is essential.

Does the lab dose match what people get from food or supplements? This is where translation often breaks down. In the taurine paper, the aggregation assay used 20 millimolar taurine to suppress ApoE4 aggregation, and 30 millimolar taurine to test whether pre-formed aggregates could be reduced. Those are test-tube concentrations, not dietary exposures.

A 20 millimolar taurine solution contains about 2,500 milligrams of taurine per liter. One liter of a 20 millimolar taurine solution contains roughly the taurine found in about 3.7 to 31 kilograms of beef; a 30 millimolar solution corresponds to about 5.5 to 47 kilograms. This is an imperfect comparison because a test tube is not a human body, but that is the point: eating steak does not recreate a millimolar concentration to break the apoE aggregation shown in the test tube. At least after a beefy meal!

The point is not that dietary taurine is irrelevant. The point is that foods and supplements must not be treated as if they automatically reproduce the concentrations, timing, and exposure used in laboratory experiments.

Does it have off-target effects? A molecule that weakly interacts with many surfaces may affect many pathways. Sometimes that is useful. Sometimes it creates unwanted effects. Either way, it has to be measured.

PK means pharmacokinetics: what the body does to the drug. Is it absorbed? How long does it last? Does it enter the brain? How is it cleared?

PD means pharmacodynamics: what the drug does to the body. Does it hit the intended biology? Does it move a meaningful biomarker? Does that translate into better memory or function?

A molecule can look promising and still fail because the PK is poor, the PD is weak, or the effect does not matter clinically.

This does not make the biology unimportant. It means the biology must be tested carefully before it becomes a claim about treatment.

ALZ-801: applying the drug-development ladder

ALZ-801 is different from taurine because it is not simply a supplement story. It is a drug-development story.

ALZ-801, also called valiltramiprosate, is a prodrug of tramiprosate. Tramiprosate is also known as homotaurine or 3-APS. Homotaurine is chemically related to taurine, but ALZ-801 is not taurine. It was designed to deliver tramiprosate more predictably in the body.

Tramiprosate was originally developed as an Alzheimer’s drug called Alzhemed. The idea was to interfere with amyloid-beta oligomers, small soluble amyloid clusters that may damage synapses before large plaques form.

The broad tramiprosate Phase 3 program did not show convincing benefit in the overall Alzheimer’s population. Later, homotaurine was sold over the counter in Canada as a memory supplement called Vivimind.

So how did this older molecule return as ALZ-801? The public story has four pieces: lab studies suggesting anti-amyloid aggregation activity; subgroup analyses suggesting a possible APOE4/4 signal; a prodrug designed for better absorption and tolerability; and a Phase 3 trial in people with early Alzheimer’s disease and two copies of APOE4.

That is a reasonable path to a Phase 3 trial. It is not a reason to assume the drug works before the randomized trial answers the question.

What ALZ-801 still had to prove

Is there a clean IC50? Not in the simple way people may imagine. ALZ-801 is not like a classic drug that blocks one enzyme at one active site. Its proposed effect is to reduce amyloid oligomer formation, which is a complex clumping process. The public development story is built around aggregation assays, stoichiometry, PK modeling, and projected brain exposure, not a single easy-to-interpret human IC50.

How was the dose chosen? Public papers describe a PK bridging strategy. The Phase 3 dose, 265 mg twice daily, was selected to produce tramiprosate exposure similar to 150 mg twice daily tramiprosate, the dose later emphasized from older APOE4/4 subgroup analyses. In other words, the dose was not chosen from a clean prospective human brain target-engagement trial; it was bridged from the older tramiprosate program and from projected exposure needed to inhibit amyloid oligomer formation.

Does it cross the blood-brain barrier? The public evidence is supportive but still indirect. The development literature describes tramiprosate and 3-SPA as brain-penetrant and refers to CSF/metabolite data, animal brain-penetration work, and projected human brain exposure. I did not find a public FDA review package that independently lays out a definitive human BBB and target-engagement analysis for ALZ-801.

What about off-target effects? ALZ-801 is usually discussed as an amyloid-oligomer drug, but homotaurine-like molecules may have other biological effects, including possible effects on GABA receptors. That does not mean the drug is unsafe or ineffective. It means the mechanism should not be oversimplified.

What is the PD readout? Cognition is not the PD readout. Cognition is the clinical outcome. A cleaner PD readout would show that the drug changed Alzheimer’s biology in humans, such as a prespecified change in p-tau217, p-tau181, amyloid measures, GFAP, NfL, or another biomarker connected to disease activity. A convincing story would line up exposure, brain penetration, target engagement, biomarker movement, and clinical benefit.

The data on ALZ-801 are still incomplete.

Timing matters: preventing amyloid clumps is not the same as removing them

Amyloid biology unfolds over years. Amyloid-beta can misfold, form small oligomers, seed further aggregation, and eventually contribute to plaques. By the time someone has memory symptoms, especially someone with two copies of APOE4, amyloid plaques may already be widespread. The amyloid plaques appear decades before one has a diagnosis of mild cognitive impairment.

That matters for ALZ-801. If a drug mainly prevents new amyloid oligomers from forming, when is the best time to use it? Before plaques form? Around the time amyloid is seeding? After plaques are already present?

Does ALZ-801 dissolve existing plaques? Does it reduce amyloid PET signal? Does it prevent plaques from shedding toxic oligomers? Does it slow downstream tau spread? Or does it need to be used earlier than the stage when most patients are diagnosed?

These questions remain open. This does not mean ALZ-801 cannot work. It means the timing of treatment may be critical and that there are many unaswered questions still.

The problem with confidence before placebo-controlled evidence

The path to ALZ-801 included lab studies, improved PK, brain-penetration arguments, APOE4 subgroup signals, and open-label Phase 2 biomarker studies. That may be enough to justify a Phase 3 trial. It is not enough to assume success.

Open-label Phase 2 studies can help with safety, dosing, tolerability, and biomarker hypotheses. But they cannot prove clinical efficacy in Alzheimer’s disease. Without a placebo group, it is hard to know whether apparent cognitive stabilization reflects the drug, patient selection, natural history, missing data, expectations, or analysis choices.

The company’s 2023 press release claimed 31% reduction in plasma p-tau181 at 24 months, 28% preservation of hippocampal volume compared with matched ADNI external controls, and improvement on cognitive tests at 6 months with sustained stabilization above baseline for 24 months. It also stated that treated patients had a 24% improvement on RAVLT Total Score compared with matched ADNI subjects

But there was no placebo arm. Matching to ADNI is not the same as having a placebo arm. FDA standards for drug approval still require evidence that can distinguish a real drug effect from placebo effect, spontaneous change, or biased observation.

But it can become a loophole in the public narrative. When open-label data are described as clinical improvement or treatment benefit, readers may hear proof where the study design only supports a hypothesis. That is especially risky in Alzheimer’s disease, where patients and families are actively searching for options.

What did APOLLOE4 report?

APOLLOE4 tested valiltramiprosate in people with early Alzheimer’s disease who had two copies of APOE4. It was randomized, double-blind, and placebo-controlled. That was the right test.

The overall clinical result was negative. The trial did not meet its primary cognitive endpoint in the overall study population. That is the anchor result.

Fig: In the overall APOLLOE4 Phase 3 population, ALZ-801 did not show a statistically significant clinical benefit over placebo on the primary cognitive endpoint, ADAS-Cog13, or on key functional measures. The ALZ-801 and placebo curves separate modestly on some outcomes, but the trial did not meet its primary endpoint, so these results do not establish clinical efficacy.

The trial did report slowing of hippocampal atrophy on MRI and nominally significant clinical findings in a prespecified mild cognitive impairment subgroup. These findings are worth studying. They may justify a new trial focused on earlier-stage patients.

But they do not erase the main result: the overall Phase 3 clinical endpoint was not met. Secondary endpoints and subgroup findings can fail when tested directly in a new trial. The responsible next step is to test the MCI hypothesis prospectively, not to promote the drug publicly as if an MCI indication has already been proven.

What about biomarkers? The peer-reviewed Phase 3 paper reports that plasma biomarkers were collected, including Aβ42, Aβ40, p-tau181, p-tau217, GFAP, and NfL. But it also says that serial assessments of core Alzheimer’s plasma biomarkers were still in progress and would be reported separately.

Alzheon later announced p-tau217 results, including decreases beginning at 26 weeks, larger effects in the MCI subgroup, and correlations with cognition, function, and hippocampal volume. Those results should be taken seriously, but they are not the same as a complete, independently interpreted, peer-reviewed biomarker package with all methods, subgroup definitions, missing-data handling, and statistical corrections clearly shown.

Being conservative is not being dismissive

There is an urgent need for safe, accessible treatments for APOE4 patients. The approved anti-amyloid antibodies are difficult treatments. They require infusions, MRI monitoring, infrastructure, and careful safety discussions. APOE4 carriers, especially APOE4 homozygotes, face particular safety concerns.

An oral drug with good safety, brain penetration, and real disease-modifying benefit would be a major advance. So the goal is not to dismiss ALZ-801. The goal is to leave room for it to succeed without treating success as already proven.

If future trials in earlier APOE4/4 patients show consistent clinical benefit, supported by clear biomarker movement and acceptable safety, that would matter.

But until then, the fairest summary is this: ALZ-801 has a rational mechanism, improved PK compared with older tramiprosate, a plausible but not fully independently verified brain-exposure argument, and company-reported biomarker signals. It also missed its primary clinical endpoint in the overall APOLLOE4 Phase 3 population. The MCI signal is a reason to study the drug further, not a reason to market it to the public as proven for MCI.

What if a patient wants to take the risk?

This is the hardest question in clinic. Some APOE4 carriers and families may say: if the drug appears safe and the disease is devastating, why not try it?

That instinct is understandable. But risk is not only side effects. There is also the risk of false hope, cost, distraction from proven care, delay in planning, and loss of trust if the promise does not hold. A drug with few obvious side effects can still fail to help, and an unproven treatment can still change decisions.

The right answer is not to block hope. It is to keep hope inside a system that can tell whether the drug truly works. For now, that means carefully designed trials, transparent communication, and avoiding public claims that go beyond the evidence.

The danger of direct-to-consumer promises

The husband in my clinic was responding to a message that seemed to match his wife’s biology. That is why direct-to-consumer communication about investigational Alzheimer’s drugs must be handled carefully.

Alzheimer’s disease is devastating. APOE4 status can make a drug feel personally targeted. Families may hear developed for APOE4/4 and ask why they cannot access it now. A company may communicate through emails, websites, newsletters, webinars, advocacy groups, trial-recruitment materials, or advertisements. Some of that communication may be legal. Some of it may be framed as education.

But legal is not the same as scientifically responsible.

When a company contacts consumers with language that implies benefit before benefit has been proven, or highlights a subgroup such as MCI after the overall Phase 3 trial missed its primary endpoint, it risks turning scientific uncertainty into marketing. The audience includes patients, spouses, adult children, and caregivers making decisions under pressure.

False or overstated promises can shape decisions. Families may pursue access, interpret subgroup findings as proof, delay planning, or organize expectations around a treatment whose benefit is still uncertain.

They also damage trust. If patients are led to believe that a drug is likely to help, and then Phase 3 results disappoint, many will remember that science raised their hopes and then changed the message.

Science can survive negative trials. Negative trials are part of progress. What science cannot afford is a cycle in which early biology becomes marketing, subgroup signals become promises, and families are asked to believe before the evidence is ready.

Take-home messages

• ApoE4 biology matters, and therapies tailored to APOE4 patients are urgently needed.

• Taurine is interesting, but not proven. The taurine-ApoE4 paper is a mechanistic study, not evidence that taurine prevents or treats Alzheimer’s disease.

• Food molecules are not automatically therapies. Red meat contains taurine, wine contains resveratrol, and green tea contains EGCG, but eating or drinking these foods does not reproduce the high, controlled concentrations used in many preclinical aggregation experiments.

• ALZ-801 is a serious drug-development story, but still unproven. Its dose selection and brain-exposure argument are based on public PK and bridging data, not a clean public IC50-to-human-brain target-engagement chain.

• APOLLOE4 did not meet its primary cognitive endpoint overall. MCI subgroup and secondary findings are worth studying, but they should not be promoted to the public as a proven MCI indication without a direct confirmatory trial.

• Conservative interpretation is not dismissal. An oral, safe, effective therapy for APOE4 patients would be a major advance. The evidence should be strong enough that patients can trust the claim.

References

1. Legrand A, Cerna KA, Marques SM, et al. Taurine inhibits apolipoprotein E4 aggregation. Biomedicine & Pharmacotherapy. 2026.

2. Hey JA, et al. Clinical Pharmacokinetics and Safety of ALZ-801, a Novel Prodrug of Tramiprosate in Development for the Treatment of Alzheimer’s Disease. Clinical Pharmacokinetics. 2018.

3. Hey JA, Kocis P, Hort J, et al. Discovery and Identification of an Endogenous Metabolite of Tramiprosate and Its Prodrug ALZ-801 that Inhibits Beta Amyloid Oligomer Formation in the Human Brain. CNS Drugs. 2018.

4. Abushakra S, et al. Clinical Pharmacokinetics of Oral ALZ-801/Valiltramiprosate and its Active Agents in APOE4/4 Subjects with Early Alzheimer’s Disease. Clinical Pharmacokinetics. 2025.

5. Abushakra S, et al. Clinical Efficacy, Safety and Imaging Effects of Oral Valiltramiprosate in APOE4/4 Homozygous Individuals with Early Alzheimer’s Disease: The APOLLOE4 Phase III Randomized Controlled Trial. CNS Drugs. 2025.

6. Alzforum. Alzhemed / tramiprosate therapeutic profile.

7. Alzforum. ALZ-801 therapeutic profile.

8. FDA / eCFR. 21 CFR 314.126: Adequate and well-controlled studies.

9. FDA. Fast Track designation information and FDA Global Substance Registration System note that UNII/substance records do not imply regulatory review or approval.

10. FDA. Unproven Alzheimer’s Disease Products.

11. Cleveland Clinic Journal of Medicine. Table: Taurine content of meats, seafood, and dairy products.

12. Hey JA, et al. Effects of Oral ALZ-801/Valiltramiprosate on Plasma Biomarkers, Brain Hippocampal Volume, and Cognition: Results of 2-Year Single-Arm, Open-Label, Phase 2 Trial in APOE4 Carriers with Early Alzheimer’s Disease. Drugs. 2024.

13. Alzheon. Topline Results from Pivotal APOLLOE4 Phase 3 Trial of Oral Valiltramiprosate/ALZ-801. 2025.

It is also a good moment to think about rest, recovery, and the quiet biological work the brain does when the day finally ends.

Consider a 56-year-old man -- let’s call him David.

David works in the media industry, where deadlines, breaking news, public pressure, and long screen-heavy days can make work unusually stressful.

He recently learned that he carries two copies of APOE4, the strongest common genetic risk factor for late-onset Alzheimer’s disease.

He occasionally loses a word mid-sentence.

He eats well, exercises daily, and keeps a healthy weight.

But one pattern stands out: he wakes around 4 a.m., lies there exhausted, and cannot fall back asleep.

Most evenings, he watches television late.

He feels anxious at night.

A few times a week, he has caffeine in the afternoon or wine with dinner.

This post is written for him -- and for anyone wondering whether sleep is just rest, or something more active and protective.

I am grateful to my colleague and friend Dr. Miranda Lim -- neuroscientist, sleep researcher, and expert in sleep and neurodegeneration -- for her generous feedback on this piece.

1. The Brain’s Nightly Cleaning Crew

Sleep is active brain maintenance.

It is not simply the absence of being awake.

During sleep, fluid moves through the spaces around brain cells and helps wash away metabolic waste.

This system is called the glymphatic system -- glial because it depends on support cells in the brain, and lymphatic because it behaves like a drainage system.

Among the waste products it helps clear are amyloid-beta and tau, two proteins central to Alzheimer’s disease.

In a landmark mouse study, the brain cleared waste far more efficiently during sleep than during wakefulness [1].

Other studies have shown that amyloid and tau follow sleep-wake rhythms, rising during wakefulness and falling during sleep [2,3].

One bad night does not cause Alzheimer’s.

But repeated sleep disruption, year after year, may remove one layer of protection from a brain that is already aging.

The relationship runs in both directions.

Poor sleep may worsen amyloid and tau biology.

Early amyloid and tau changes may also damage the brain circuits that regulate sleep, making sleep more fragmented over time [4].

2. APOE4 and Sleep: Why David’s Night Matters

Why might sleep matter more in APOE4 carriers?

One way to think about it is this: APOE4 may make amyloid handling less efficient and the brain’s immune response more reactive.

If the nightly clean-up system is also weakened by insomnia, sleep apnea, alcohol, irregular timing, or anxiety, the same amount of sleep disruption may carry a higher cost.

Human studies suggest that sleep and APOE4 interact.

In older adults, better sleep consolidation appeared to weaken the relationship between APOE4 and Alzheimer ’s-related pathology [5].

Newer work also suggests that sleep disturbance may be more strongly linked to amyloid burden in APOE4 carriers than in non-carriers [6].

Animal work supports the same idea: APOE4 and sleep disruption can act together to accelerate amyloid and tau-related changes [7].

The practical point is leverage.

David cannot change his genes.

But he can change the nightly environment his genes operate in.

For an APOE4 carrier, treating insomnia, sleep apnea, irregular sleep timing, alcohol-related sleep fragmentation, and nighttime anxiety is not cosmetic wellness.

It is risk management.

3. Not All Sleep Is Equal: Understanding the Cycles

Adults usually need seven to nine hours of sleep.

But quality matters as much as quantity.

Sleep moves through stages in cycles of roughly 90 to 120 minutes.

Most people cycle through these stages three to five times per night.

A typical night is not flat.

It looks like a staircase: deeper sleep early in the night, then longer stretches of dream-rich REM sleep toward morning.

Figure: Sleep cycles across a typical night. Deep sleep is more common early in the night. REM sleep lengthens toward morning. Adapted from Shah et al., Portable Monitoring, 2013.

The light green zone at the beginning is the hypnagogic state. This is the drifting period when thoughts loosen, images appear, and the brain crosses from wakefulness into sleep. Its mirror image is the hypnopompic state, the groggy transition that happens while waking. Both are normal borderlands between consciousness and sleep.

NREM Stage 1, or N1, is the lightest stage of sleep — the brief doorway between wakefulness and sleep, when thoughts loosen, awareness fades, and the hypnagogic state begins.

NREM Stage 2 is the workhorse of the night. It often makes up about half of total sleep and is marked by sleep spindles, brief bursts of brain activity involved in memory processing.

NREM Stage 3 is deep sleep, also called slow-wave sleep (SWS). This stage is linked to physical restoration, immune signaling, memory stabilization, and possibly glymphatic clearance. The purple segments in the figure refer to lucid dreamless sleep. This is a rare state, reported especially among experienced meditators, in which a person may have clear awareness during deep sleep without the usual story, images, or dream content.

Deep sleep declines with age. A long-term study in JAMA Neurology found that lower slow-wave sleep was associated with higher dementia risk, although the study was observational and cannot prove cause and effect [8].

REM sleep is the stage of vivid dreams. The brain is highly active, but the body is temporarily paralyzed, so we do not act out dreams. Some people experience lucid dreaming during REM: they realize they are dreaming while the dream is still happening, and sometimes can influence the dream. REM helps process emotion, fear, anxiety, and complex memories. Alcohol, sleep apnea, and some medications can fragment or suppress REM.

The dashed arrows mark brief returns toward wakefulness, often called WASO -- wake after sleep onset. Tiny awakenings happen to everyone. The problem is not waking for a moment. The problem is waking and staying awake for 20 minutes or more.

The goal is not to force one stage at the expense of another. The brain needs the full natural sequence: light sleep, Stage 2, deep sleep, REM, and brief awakenings that quickly return to sleep.

David’s 4 a.m. awakening may begin as a normal sleep-cycle boundary. It also occurs in the part of the night when REM sleep is usually longer and more frequent. If anxiety fragments this REM-rich period, the night may feel as if it ends too early.

4. Start With a Sleep Audit

Before buying devices or supplements, start with a simple sleep diary.

Track for at least two to three weeks, including weekends.

Each morning, write down bedtime, wake time, estimated total sleep, number of awakenings, and how refreshed you feel.

If you wake during the night, estimate how long it takes to fall back asleep.

Each evening, note caffeine timing, alcohol, exercise, screen exposure, and anxiety level.

Anxiety deserves its own line in the diary.

It is one of the most common and underrecognized sleep disruptors.

For David, the diary might reveal a pattern: caffeine after 2 p.m., television until 11 p.m., wine with dinner, and then a 4 a.m. awakening.

None of these is dramatic alone.

Together, they can fragment sleep.

The goal is not perfection.

The goal is pattern recognition.

5. Six Habits With the Best Evidence

Keep a steady schedule.

Wake up at the same time every day, including weekends.

A stable wake time is one of the strongest signals to the body clock.

Protect the last hour.

Dim lights after dinner.

Avoid emotionally charged television, work email, and bright screens close to bed.

Blue-wavelength light suppresses melatonin, the hormone that tells the brain darkness has arrived.

Keep the bedroom cool.

A room around 65 to 68 degrees F, or 18 to 20 degrees C, helps the body drop its core temperature, which supports sleep onset and deep sleep.

Cut caffeine early.

Caffeine can last five to seven hours, and often longer in older adults.

What was harmless at 35 may be disruptive at 56.

Do not use alcohol as a sleep aid.

Alcohol may help you fall asleep faster, but it fragments the second half of the night and suppresses deep sleep and REM.

Move your body daily.

Regular exercise improves sleep pressure, mood, insulin sensitivity, and cardiovascular health.

Avoid vigorous exercise close to bedtime if it leaves you wired.

6. Treat Anxiety and Insomnia Together: Why CBT-I Matters

Many people think insomnia is a nighttime problem.

Often, it is also a daytime anxiety loop.

The brain starts to learn that the bed is a place for effort, monitoring, frustration, and worry.

That learned association is called conditioned arousal.

CBT-I -- Cognitive Behavioral Therapy for Insomnia -- is designed to break that loop.

It is not simply sleep hygiene.

CBT-I combines stimulus control, sleep scheduling, cognitive therapy, relaxation skills, and careful adjustment of time in bed.

It helps the brain relearn a simple rule: bed equals sleep, not struggle.

The American College of Physicians recommends CBT-I as the initial treatment for chronic insomnia in adults [9].

AASM guidance also supports behavioral-psychological treatment and, when medication is needed, pairing it with CBT-I rather than using medication alone [10].

For David, CBT-I is especially relevant because anxiety may be shortening or fragmenting the REM-rich final third of his night.

The goal is not to convince him to stop worrying by willpower.

The goal is to change the sleep system so his body no longer practices worry at 4 a.m.

CBT-I can be done with a trained therapist, through some sleep clinics, or through validated digital CBT-I programs when access is limited.

It can feel uncomfortable at first because it asks people to change habits that feel protective.

But unlike sedatives, it teaches a durable skill.

7. When to Refer to a Sleep Specialist

Not every sleep problem needs a sleep lab.

But some do.

Refer for evaluation if there is loud snoring, witnessed pauses in breathing, gasping, choking, morning headaches, dry mouth, high blood pressure, or daytime sleepiness despite enough time in bed.

These are red flags for obstructive sleep apnea.

Sleep apnea repeatedly drops oxygen and forces the brain to wake just enough to breathe.

That can destroy deep sleep and REM without the person remembering the awakenings.

Referral is also appropriate when insomnia lasts more than three months, causes daytime impairment, or does not improve with basic behavioral steps.

Other reasons include restless legs, violent dream enactment, narcolepsy symptoms, shift-work disorder, complex medication questions, or persistent early-morning awakening with depression or severe anxiety.

For older adults and APOE4 carriers, the threshold can be lower.

A treatable sleep disorder is not something to watch passively.

8. Sleep Medications: Useful, but Not the Same as Natural Sleep

There are times when sleep medication is reasonable.

Jet lag, a hospital stay, acute grief, or a short period of severe insomnia may justify short-term help.

The problem is assuming that a pill that causes unconsciousness is the same as healthy sleep.

Many common sleep aids -- benzodiazepines, Z-drugs such as zolpidem, and over-the-counter antihistamines such as diphenhydramine -- sedate the brain.

They may reduce the memory of waking up.

But they can also alter sleep architecture, suppress REM or deep sleep, increase falls, and create dependence or rebound insomnia.

This matters even more in older adults.

Antihistamine sleep aids have anticholinergic effects, which are generally a poor fit for brain health when used chronically.

Newer orexin receptor antagonists, such as suvorexant and lemborexant, work differently.

They reduce the brain’s wakefulness signal rather than broadly sedating the brain.

They may preserve sleep architecture more faithfully, but they still require medical supervision and their long-term effects are still uncertain.

Melatonin belongs in a separate category.

It is not a sedative.

It is a darkness signal that can help shift the body clock for jet lag or delayed sleep timing.

Low doses, often 0.5 to 1 mg, are usually enough for circadian purposes.

More is not necessarily better.

Do not take melatonin before driving.

9. Rings, Watches, and New Sleep Technologies

Sleep devices can be helpful when they are used for trends, not truth.

A ring or watch can show bedtime regularity, total sleep opportunity, heart rate patterns, movement, and whether weekends are drifting later.

For David, a device might show that his sleep is shorter after wine or caffeine, or that his weekend schedule shifts by two hours.

But consumer devices are not sleep labs.

They usually estimate sleep from movement, heart rate, temperature, and algorithms.

They do not directly measure brain waves the way polysomnography does.

Recent validation studies suggest many devices are fairly good at estimating sleep versus wake, but less reliable at distinguishing deep sleep from REM or light sleep [11].

This is why a low deep sleep score should not create panic.

And a beautiful sleep score should not override symptoms like snoring, gasping, or daytime sleepiness.

The best use is pattern recognition.

Use the device to ask: what happens when I keep a steady wake time, stop alcohol, treat anxiety, or get morning light?

The worst use is sleep perfectionism.

Some people become more anxious because they chase a score, a problem sometimes called orthosomnia.

If tracking makes sleep feel like a test, stop tracking for a while and return to a diary.

On the research frontier, more advanced tools are emerging.

Wearable EEG headbands can measure brain activity more directly and may eventually guide gentle stimulation during naturally occurring deep sleep transitions [12].

Other experimental tools are trying to measure glymphatic function in humans during sleep [13,14].

They are not yet a substitute for the basics: regular timing, darkness at night, morning light, exercise, apnea treatment, and anxiety care.

10. What We Still Do Not Know

Sleep and Alzheimer’s influence each other in both directions.

Poor sleep may increase amyloid and tau.

But early Alzheimer’s changes can also damage the brain circuits that regulate sleep.

That makes cause and effect hard to untangle.

We do not yet know the exact amount of deep sleep needed for brain clearance.

We do not know whether every APOE4 carrier benefits equally from sleep intervention.

We do not know whether fixing sleep after symptoms begin can slow disease progression.

So the honest answer is this: sleep is not a cure for Alzheimer’s.

But it is one of the most biologically plausible, low-cost, and broadly beneficial targets we have.

Take-Home Messages

· Sleep is active brain maintenance. The glymphatic system helps clear amyloid-beta, tau, and other waste during sleep.

· APOE4 raises the stakes, not destiny. Genes are not fate, but APOE4 may make good sleep more important.

· All sleep stages matter. Deep sleep may support clearance, REM supports emotional and memory processing, and the full cycle matters more than any single stage.

· Brief awakenings are normal. Lying awake for 20 minutes or more is the pattern to address.

· Treat anxiety as part of treating sleep. CBT-I directly targets the 4 a.m. worry loop.

· Refer when red flags appear. Snoring, gasping, witnessed pauses, daytime sleepiness, restless legs, unusual nighttime behaviors, or chronic insomnia deserve evaluation.

· Use devices wisely. Rings and watches can reveal patterns, but they are not diagnostic tests and should not become another source of anxiety.

· Be careful with sedatives. Unconsciousness is not the same as natural sleep.

For David, the prescription is not fear.

It is a plan: protect the night, treat the disruptors, and give the brain the conditions it needs to do its quiet work.

References

1. Xie L, Kang H, Xu Q, et al. Sleep drives metabolite clearance from the adult brain. Science. 2013;342(6156):373-377.

2. Kang JE, Lim MM, Bateman RJ, et al. Amyloid-beta dynamics are regulated by orexin and the sleep-wake cycle. Science. 2009;326(5955):1005-1007.

3. Holth JK, Fritschi SK, Wang C, et al. The sleep-wake cycle regulates brain interstitial fluid tau in mice and CSF tau in humans. Science. 2019;363(6429):880-884.

4. Vanderheyden WM, Lim MM, Musiek ES, Gerstner JR. Alzheimer’s disease and sleep-wake disturbances: amyloid, astrocytes, and animal models. Journal of Neuroscience. 2018;38(12):2901-2910.

5. Lim ASP, Yu L, Kowgier M, Schneider JA, Buchman AS, Bennett DA. Sleep modifies the relation of APOE to risk of Alzheimer disease and neurofibrillary tangle pathology. JAMA Neurology. 2013;70(12):1544-1551.

6. Liang Q, et al. Moderating effect of APOE epsilon4 on the association of sleep disturbance with regional amyloid-beta burden in cognitively normal older adults. Frontiers in Aging Neuroscience. 2025.

7. Wang C, Nambiar A, Strickland MR, et al. APOE-epsilon4 synergizes with sleep disruption to accelerate amyloid-beta deposition and plaque-associated tau seeding and spreading. Journal of Clinical Investigation. 2025.

8. Himali JJ, Baril A-A, Cavuoto MG, et al. Association between slow-wave sleep loss and incident dementia. JAMA Neurology. 2023;80(11):1326-1333.

9. Qaseem A, Kansagara D, Forciea MA, Cooke M, Denberg TD. Management of chronic insomnia disorder in adults: a clinical practice guideline from the American College of Physicians. Annals of Internal Medicine. 2016;165(2):125-133.

10. Buysse DJ, et al. Combination treatment for chronic insomnia disorder in adults: an American Academy of Sleep Medicine clinical practice guideline. Journal of Clinical Sleep Medicine. 2026.

11. Robbins R, Weaver MD, Sullivan JP, et al. Accuracy of three commercial wearable devices for sleep tracking in healthy adults. Sensors. 2024;24(20):6532.

12. Hathaway E, Morgan K, Carson M, et al. Transcranial electrical stimulation targeting limbic cortex increases the duration of human deep sleep. Sleep Medicine. 2021;81:350-357.

13. Dagum P, Giovangrandi L, Levendovszky SR, et al. A wireless device for continuous measurement of brain parenchymal resistance tracks glymphatic function in humans. Nature Biomedical Engineering. 2025;9:1656-1676.

14. Dagum P, Elbert DL, Giovangrandi L, et al. The glymphatic system clears amyloid beta and tau from brain to plasma in humans. Nature Communications. 2026;17(1):715.

The arrival of anti-amyloid therapies has revived the field of Alzheimer’s disease, but whether they truly qualify as disease-modifying remains an open and unresolved question.

This post is the summary of a discussion with my friend and mentor, Lon Schneider, whose work on Alzheimer’s trials has shaped my interpretation of AD clinical trial readouts.

To get a better understanding of the background of this post, I suggest you take a look at two past posts:

1. Where does amyloid fit in the pathogenesis of AD?

2. How do we interpret blood amyloid biomarkers?

In 2023, the U.S. Food and Drug Administration approved a new drug for early Alzheimer’s disease called lecanemab (sold as Leqembi). A second drug, donanemab (Kisunla), followed in 2024. Both are antibodies — proteins made in a lab and given by IV — designed to clear amyloid, the sticky protein that builds up in Alzheimer’s brains. Some regulators and drugmakers describe them as “disease-modifying.”

Their arrival is the first significant progress in Alzheimer’s drug development in decades. After many years of failed trials, these are the first treatments to clear amyloid from patients’ brains and to produce statistically significant benefits on composite, cognitive, and daily-living measures in large Phase 3 studies, and on core AD biomarkers. Those results justified FDA approval and are not in dispute.

Figure 1. Changes in plasma Alzheimer’s disease biomarkers with donanemab treatment.
Least-squares mean (±SE) change from baseline over 76 weeks in (A) plasma p-tau217, (B) GFAP, (C) neurofilament light chain (NfL), and (D) Aβ42/40 ratio in patients receiving donanemab versus placebo. Donanemab treatment is associated with sustained reductions in p-tau217 and GFAP, stabilization of NfL compared with placebo, and an increase in the Aβ42/40 ratio, consistent with amyloid clearance. Numbers at risk are shown below each panel.

The “disease-modifying” label, though, sets a higher bar than the trial evidence has yet met. By the European Medicines Agency’s standard, a disease-modifying drug has to do more than help while you’re taking it — it has to actually change the course of the illness. No anti-amyloid antibody has been shown to do that in a trial designed to test it. Most data so far support that these drugs may modestly reduce symptoms while patients receive them.

After 18 months, patients on lecanemab or donanemab decline a little less than placebo patients on tests of memory and daily function. The difference is real and statistically reliable — but statistical reliability isn’t the same as a noticeable change, and a measurable difference on a test isn’t the same as a meaningful change in the disease.

What “disease-modifying” actually means

A 2017 paper by my colleague Dr. Jeffrey Cummings offered a framework for evaluating disease-modifying therapies (DMT) in Alzheimer’s. It uses the “slope of decline” to separate drugs that truly slow the disease from those that only ease symptoms.

Figure 2 above (adapted from Cummings, J. (2017). JPAD) illustrates one framework for how a disease-modifying treatment (DMT) can change the trajectory of Alzheimer’s.

The Solid Line: The typical path of the disease, with memory and thinking declining steadily over time. The arrow marks DMT initiation — when a treatment that targets the biology of the disease begins.

The Dashed Line: The goal of treatment is to slow the rate of decline so a person keeps their independence and cognitive health longer.

The European Medicines Agency — Europe’s drug regulator — has since set out a clear working standard. A symptomatic drug helps you while you take it; stop it, and the benefit fades. A disease-modifying drug actually changes the disease, so that people who took it remain better off even after stopping. The disease has been put on a different track.

To prove this, regulators look for specific trial designs — typically ones in which some patients start on a placebo and are later switched to the active drug. If the late-starters never catch up, the early treatment must have changed something durable. If they do catch up, the drug was probably just helping with symptoms while it was being given.

This “slope separation” picture above is a useful starting point, but on its own it doesn’t meet the EMA’s bar. The agency expects a real slowing of clinical decline alongside a matching change in biological markers like amyloid or tau. It also asks that the slowing be meaningful in patients’ daily lives, not just on a graph — and that it reflect a true shift in the disease, not an ongoing symptomatic effect. That is why regulators favor “delayed-start” or “withdrawal” designs over simple slope comparisons.

An example of a delayed start is shown in Figure 3 below:

• Early Starters: Patients who start the active treatment on day one.

• Late Starters: Patients who start on a placebo and switch to active treatment after a set period.

• The Outcome: If late starters never catch up, the drug likely slows the disease itself rather than just masking symptoms.

In the TRAILBLAZER-ALZ 2 donanemab trial, the delayed-start and early-start groups appear to converge over time, Figure 4 (Modified from Mintun MA. Donanemab TRAILBLAZER-ALZ in Early Symptomatic Alzheimer’s Disease: Efficacy and Safety from the Long-Term Extension. Presented at: 2023 International Conference on Alzheimer’s and Parkinson’s Diseases (AD/PD); March 28–April 1, 2023; Gothenburg, Sweden. Eli Lilly and Company, Indianapolis, IN. Available from Eli Lilly medical materials.

“Time Saved” Does Not Mean the Disease Is Changed

Another framing of the benefit is “time saved” (see figure 5 below). It describes a delay in reaching a given level of symptoms — not a fundamental change in the disease itself. “Time saved” only reflects true disease modification if the benefit persists after treatment stops, or if the trajectory is permanently changed. Current studies have shown neither. So while the phrase sounds meaningful, it may simply represent a temporary slowing of symptoms.

Adapted from Dickson SP, et al. Time saved as a demonstration of clinical meaningfulness and illustrated using the donanemab TRAILBLAZER-ALZ study findings. Journal of Prevention of Alzheimer’s Disease, 2023

What the drugs appear to do

Two things show up consistently in the trials.

First, the drugs clear amyloid from the brain. PET scans confirm that plaques drop, often back to normal levels, and other markers of brain damage — including the protein tau — also move in the expected direction.

Second, after 18 months, patients on the drug score slightly better than patients on placebo on standard tests of memory, thinking, and daily activities. The relative slowing of decline is roughly 25 to 30 percent. Supporters frame this as about seven to eight extra months in the milder stages of disease, with simulation models projecting up to two and a half years if the slowing keeps compounding. “Time saved” is one framing that patients and families may likely hear from a doctor.

Two cautions about that framing.

First, the absolute differences on the tests are small. On the cognitive test most often used (ADAS-Cog, 0–70), the average difference between groups was less than one point. On a common dementia severity scale (CDR-SB, 0–18), it was about a third of a point. Most clinicians say a patient or family wouldn’t notice anything smaller than 2–4 points on the cognitive test or 1–2 points on the dementia scale — the trial differences fall below those thresholds. A 2026 Cochrane review pooling 17 trials and more than 20,000 patients called the effect on memory and thinking “trivial.”

Second, the “time saved” calculation only really makes sense if the slowing is true disease modification. If a patient takes lecanemab for two years, comes out a bit ahead on cognitive testing, and then stops — does the gap hold, close, or grow? The answer determines whether “time saved” is a real bank balance or a temporary read on the meter. We don’t yet have one.

Why the disease-modifying claim outruns the data

Two further problems push the other way.

First, the trials weren’t set up to test it. After CLARITY AD ended, all participants were offered the active drug in an open-label extension. The manufacturer has used that follow-up to argue that the benefit accumulates over three years, and the placebo group never catches up. A 2026 commentary coming out soon in Alzheimer’s & Dementia points out that the statistical analysis needed to support this claim was never done; the argument relies on visual separation in figures rather than formal modeling.

Second, where the analysis has been done, it points the other way. In the 36-month donanemab data, the placebo-then-treated group catches up to the early-treated group — the pattern of a symptomatic drug. And in the DIAN-TU trial, in people with rare genetic forms of Alzheimer’s caused unambiguously by amyloid, treatment cleared amyloid but produced no clinical benefit over nearly five years.

A useful comparison: cholinesterase inhibitors

For more than two decades, an older class of Alzheimer’s drugs — cholinesterase inhibitors taken as daily pills (donepezil, rivastigmine, galantamine) — has been used to treat the disease. These have always been called “symptomatic”: they help while they’re taken but don’t change the underlying disease.

A recent analysis across more than 40 trials compared the older pills with the newer antibodies. On the same tests of memory and daily function, the two classes produce similar effects — on some measures, the older pills produce slightly larger differences.

A 2026 study in The Lancet Regional Health – Europe found that patients who stayed on cholinesterase inhibitors for four years were nearly two points ahead on a standard cognitive test (the MMSE) compared with patients who stopped. By the same reasoning used to argue lecanemab is disease-modifying, that sustained benefit would also count — yet no one calls cholinesterase inhibitors disease-modifying.

This isn’t to say the two classes are interchangeable. The biology differs. Cholinesterase inhibitors boost acetylcholine, a brain chemical that supports memory and thinking; they ease symptoms but don’t touch amyloid, tau, or the underlying pathology. The anti-amyloid antibodies clear amyloid from the brain and lower tau biomarkers in blood and spinal fluid. That biological effect is unprecedented in Alzheimer’s care, and it is the foundation for regulatory approval. The unsettled question — the one ongoing trials are designed to answer — is whether that biological change will translate into a durable, disease-modifying clinical benefit. On the scales that matter for patients and families today, the two classes still look similar in size of effect, even though the underlying biology is different.

What could resolve this

Two ongoing trials are worth watching.

The first is TRAILBLAZER-ALZ 3, testing donanemab in people who don’t yet have Alzheimer’s symptoms but whose brain scans show elevated amyloid. The endpoint is whether participants reach a clinical diagnosis over roughly three and a half years. Because the trial measures whether disease itself emerges, a positive result would be hard to explain as merely symptomatic — and the disease-modifying claim would become much stronger. A negative result would be a serious blow to the amyloid hypothesis as a basis for prevention.

AHEAD 3-45 is a similar prevention trial of lecanemab. It is important, but on its own, it cannot prove disease modification — even a statistically significant cognitive benefit could be explained as a temporary effect.

Trials combining anti-amyloid drugs with treatments targeting tau or brain inflammation are also underway. Larger and more durable effects from those combinations would strengthen the case for amyloid as part of a disease-modifying strategy.

Take-home messages

• FDA approval rests on Phase 3 trials showing statistically significant benefits on composite, cognitive, and daily-living measures at 18 months. After decades of failed Alzheimer’s trials, that is real progress.

• By Europe’s regulatory standard, however, no anti-amyloid antibody has yet been formally shown to be disease-modifying — that is, to change the underlying course of the illness. The 18-month evidence is consistent with a symptomatic effect; ongoing trials may shift that picture.

• The differences favoring treatment are real but small in absolute terms — below what many clinicians consider noticeable on standard scales.

• Advocates frame the benefit as “time saved” — months of additional independence. That framing only fully holds if the slowing is durable, which hasn’t yet been demonstrated.

• The antibodies differ biologically from older cholinesterase inhibitors: they clear amyloid and lower tau, while the older drugs don’t touch this pathology. Whether that biological difference translates into a durable disease-modifying clinical benefit is the central open question.

• Costs and risks are real: about 1 in 8 treated patients develops brain swelling or microbleeds, and treatment requires IV infusions, repeat MRIs, genetic testing, and around $26,000 a year.

• Even if current evidence does not yet meet strict definitions of disease modification, anti-amyloid antibodies represent progress: they show that targeting a biological feature associated with Alzheimer’s disease can translate into consistent, measurable slowing of clinical decline in large trials.

• The TRAILBLAZER-ALZ 3 prevention trial of donanemab could potentially settle the disease-modifying question over the next few years.

• Decisions about starting, continuing, or stopping treatment are personal, best made with a clinician who can weigh the evidence, costs, and individual circumstances. Nothing here, on its own, is a reason to stop a treatment that is working for someone.

Sources

• European Medicines Agency. Guideline on the clinical investigation of medicines for the treatment of Alzheimer’s disease (CPMP/EWP/553/95 Rev. 2). 2018.

• Cummings, J. (2017). “Defining Disease Modifying Therapy for Alzheimer’s Disease.” The Journal of Prevention of Alzheimer’s Disease, 4(2), 109–115. doi.org.

• Aisen P, Bateman RJ, Crowther D, et al. The case for regulatory approval of amyloid-lowering immunotherapies in Alzheimer’s disease based on clearcut biomarker evidence. Alzheimer’s & Dementia 2025;21:e14342.

• Cummings J. Anti-amyloid monoclonal antibodies are transformative treatments that redefine Alzheimer’s disease therapeutics. Drugs 2023;83:569–576.

• Nonino F, Minozzi S, Sambati L, et al. Amyloid-beta-targeting monoclonal antibodies for people with mild cognitive impairment or mild dementia due to Alzheimer’s disease. Cochrane Database of Systematic Reviews 2026, Issue 4: CD016297.

• Schneider LS, Kennedy RE, Cutter G. Caution in interpreting disease-modification claims with lecanemab: selective reporting and causal inference. Alzheimer’s & Dementia 2026 (Letter to the Editor).

• The Lancet Neurology. Stopping Alzheimer’s disease before symptoms start [Editorial]. Lancet Neurology 2026;25:213.

• Giacobini E, Schneider LS. Cholinesterase inhibitors and amyloid-targeting antibody treatments show similar clinical effect. Presented at ADPD, Copenhagen, March 2026.

• Lecerf S, Guinebretiere O, Bentegeac R, et al. The Lancet Regional Health – Europe 2026;62.

• Dickson SP, et al. Time saved as a demonstration of clinical meaningfulness and illustrated using the donanemab TRAILBLAZER-ALZ study findings. Journal of Prevention of Alzheimer’s Disease, JPAD 2023.

Today is Mother's Day — and for many of us, the fear of watching someone we love lose their memory is very real. Wishing all mothers a beautiful day.

A new study published in the Journal of Prevention of Alzheimer’s Disease made headlines with a finding that seems to defy common sense: people who took omega-3 supplements in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort showed faster cognitive decline than those who didn’t. If true, should millions of people stop taking their fish oil capsules?

Not so fast. While the findings deserve attention, understanding what this study can and cannot tell us is essential before drawing any conclusions.

What the Study Found

The researchers analyzed data from 819 older adults in the ADNI database, comparing 273 omega-3 users — primarily fish oil — to 546 matched non-users over a median of five years. Using statistical techniques designed to create comparable groups (propensity score matching), they found that omega-3 users showed faster declines on three standard cognitive tests: the MMSE, ADAS-Cog13, and CDR-SB.

Importantly, this accelerated decline appeared linked not to the classic hallmarks of Alzheimer’s disease — amyloid plaques or tau tangles — but to reduced glucose metabolism in key brain regions, a marker of synaptic dysfunction or damaged mitochondria where omega-3s may get oxidized. The authors proposed that commercially available fish oil, which is particularly vulnerable to oxidation, might be generating harmful byproducts that damage synaptic function. It’s a provocative hypothesis. But the study design makes it very difficult to draw that conclusion.

This figure is striking visually, but it illustrates the core limitation — an observational design cannot tell us whether omega-3s caused the faster metabolic decline, or whether people already heading toward faster decline were the ones who reached for supplements.

The Problem with Observational Studies

The ADNI analysis is an observational study — it watches what people do in real life and looks for patterns. This is a valuable starting point, but it comes with a fundamental limitation: people who take supplements are systematically different from those who don’t, and no amount of statistical adjustment can perfectly account for all those differences.

The authors used propensity score matching, a technique that pairs omega-3 users with non-users who look similar on measured characteristics like age, sex, APOE genetic status, and diagnosis. But “similar on paper” is not the same as “truly comparable.” People who reach for a fish oil bottle may do so because a physician flagged concerns about their cognition, because they are managing cardiovascular or inflammatory conditions, or because they’ve noticed early memory changes themselves. These hidden motivations — what researchers call unmeasured confounders — can create a spurious association between a supplement and a bad outcome, even when the supplement itself is entirely innocent.

There is also a well-known phenomenon in observational research called “healthy user bias,” where supplement takers tend to be healthier overall than non-takers. In this case, however, the study found the opposite — omega-3 users appeared to do worse. The authors acknowledge this paradox themselves, noting that it “presents a puzzle” since healthy user bias would typically favor detecting a protective effect. This unusual direction of effect makes unmeasured confounding an especially important concern.

Propensity Scoring Does Not Eliminate Reverse Causation

This brings us to one of the most critical points in interpreting this kind of research: propensity score matching reduces baseline imbalances on measured variables, but it cannot solve the problem of reverse causation. That is, people may begin taking omega-3 supplements because they are already experiencing subtle cognitive concerns — concerns that haven’t yet shown up on formal tests but are real enough to prompt self-medication.

The authors tried to address this by comparing cognitive trajectories during the pre-supplementation period, finding no significant differences. That’s a thoughtful step. But it is an incomplete test. Pre-clinical Alzheimer’s pathology — the kind that silently accumulates for years before symptoms emerge — can generate subtle functional concerns that motivate lifestyle changes long before they register on cognitive scales. No retrospective analysis of pre-treatment trajectories can fully rule out this possibility. The authors themselves noted that their analysis “may miss subtle trajectories that prompt individuals to start supplements.” That is not a minor caveat — it is the central interpretive challenge.

Putting It in Context: What the Rest of the Evidence Says

Perhaps the most important corrective to this study is the broader evidence base. In the UK Biobank — one of the largest prospective cohorts ever assembled, following hundreds of thousands of participants — omega-3 supplement use was associated with lower rates of dementia, not higher. If omega-3 supplements were genuinely accelerating neurodegeneration, we would expect to see that signal consistently across different populations and methodologies. We don’t.

Large randomized controlled trials, which are far less vulnerable to confounding, have consistently found something different and more mundane: omega-3 supplements produce no significant benefit in people with established Alzheimer’s disease or mild cognitive impairment, but they also cause no meaningful harm. Multiple meta-analyses have reached similar conclusions. “Ineffective” and “harmful” are very different verdicts, and the RCT evidence supports the former, not the latter.

The ADNI findings may also reflect a quirk of the supplement itself. Studies have found that a large proportion of commercially available fish oil products are already oxidized before they are consumed — meaning the capsule delivers not clean DHA and EPA, but a mixture that includes pro-inflammatory breakdown products. If the ADNI signal reflects anything real, it may tell us more about supplement quality than about omega-3 biology.

Why Omega-3 Supplements May Have Only a Limited Role

Even setting aside these methodological concerns, there are strong reasons to think that omega-3 supplements taken in isolation may simply not be potent enough to meaningfully alter Alzheimer’s trajectories in either direction. The biology is more complicated than the supplement industry suggests.

DHA — the primary omega-3 fatty acid in the brain — has genuine neuroprotective functions: maintaining membrane fluidity, supporting synaptic signaling, resolving inflammation, and facilitating amyloid clearance. But how effectively a supplement translates into actual brain DHA levels depends heavily on individual factors: APOE genetic status, age, existing brain pathology, baseline omega-3 levels, and the health of the gut microbiome. In people with already-established Alzheimer’s pathology — where glucose metabolism is disrupted, oxidative stress is high, and the blood-brain barrier is compromised — the situation becomes even more complex. Under these conditions, additional DHA may face a hostile metabolic environment, and any benefit is easily overwhelmed.

This is not a reason to conclude that omega-3s are harmful. It is a reason to conclude that supplements, delivered as isolated capsules outside the context of overall dietary patterns, are likely to have a modest — and possibly negligible — effect on cognitive aging.

There is also a little-known biological wrinkle: high-dose DHA supplementation can paradoxically suppress the body’s own DHA production. Research published by the Bazinet group demonstrates that dietary DHA inhibits a liver enzyme called ELOVL2, which elongates EPA along the DHA biosynthesis pathway. In plain terms, when you take a DHA supplement, you send a signal to your liver to stop making its own DHA from plant-based precursors found in foods like flaxseed and walnuts. The body is tuned for a steady dietary supply — not a daily bolus from a capsule — and flooding it with preformed DHA suppresses the more nuanced, on-demand synthesis the liver would otherwise perform. Long-term supplementation may therefore not simply add DHA on top of what the body makes; it may partly replace endogenous production with an external dependency.

Compounding this is the problem of oxidation. DHA and EPA are polyunsaturated fats — chemically fragile and prone to oxidative damage. When the metabolic environment is already compromised, as in an aging brain facing elevated oxidative stress, disrupted glucose metabolism, and mitochondrial dysfunction, supplemental omega-3s may not arrive at their destination intact. Rather than being incorporated into neuronal membranes to support synaptic function, they can be oxidized into pro-inflammatory byproducts. This is why the context in which omega-3s are consumed may matter as much as the omega-3s themselves. A healthy dietary and lifestyle pattern — regular physical activity, a diet rich in fiber and polyphenols, restorative sleep, metabolic health — creates the cellular conditions under which DHA can actually be used well. The same fatty acids delivered into a dysfunctional metabolic environment may have a very different fate.

Uncertainty and Future Directions

There is much we genuinely do not know. Whether omega-3 supplementation could prevent Alzheimer’s if started early enough — potentially decades before symptoms — in people who are genetically vulnerable and nutritionally deficient remains an open question. The PreventE4 trial — testing high-dose DHA in cognitively normal APOE4 carriers well before the onset of cognitive decline — is expected to publish its results within the next two months. If high-dose DHA shows meaningful benefit in this genetically at-risk, presymptomatic population, it would reframe the entire debate: not whether omega-3s work, but when and for whom the intervention needs to start. We also don’t understand well how the gut microbiome shapes the response to omega-3 supplementation, or whether certain genetic profiles (such as APOE4 carriers with low baseline omega-3 levels) might benefit more than others.

Supplement quality is also an underappreciated problem. If fish oil products are frequently oxidized by the time of consumption, then what we are testing in many observational studies is not omega-3s per se, but a degraded product. Rigorous studies with verified, high-quality formulations are needed.

Finally, the broader question remains unresolved: do omega-3s work better as part of a complex dietary pattern than as a standalone supplement? Emerging evidence consistently suggests yes. Whole omega-3 dietary patterns — both of which feature regular fatty fish consumption alongside fiber, polyphenols, and other anti-inflammatory nutrients — show more consistent associations with cognitive resilience than any single nutrient taken in isolation. The whole appears to be greater than the sum of its parts.

Take-Home Messages

The ADNI finding is thought-provoking but should not cause alarm. Observational studies cannot establish causation, and the pattern seen here — where omega-3 users appear sicker than non-users — is more consistent with reverse causation than with genuine neurotoxicity. The UK Biobank and randomized trial evidence do not support the conclusion that fish oil harms the brain.

Omega-3 supplements are unlikely to have a major effect in either direction for most people. The balance of evidence suggests that they don’t meaningfully slow Alzheimer’s disease in people who already have cognitive impairment or established pathology, but they also don’t accelerate it. The story is mostly one of modest-to-neutral impact.

A diet naturally rich in omega-3s is almost certainly more beneficial than any supplement. Eating fatty fish regularly, or omega-3s from plants, alongside a diet high in vegetables, fiber, and plant-based foods, delivers omega-3s in an unoxidized form, within a matrix of other nutrients, in a gut environment that supports beneficial metabolism. That package — not a capsule — is what the observational evidence consistently associates with better brain aging.

Don’t throw away your fish oil based on a single observational study. But do invest more in your overall diet. No supplement replaces the complexity of real food, and this study — whatever its ultimate interpretation — is a useful reminder of that.

References

1. Liao et al. “The association between omega-3 supplementation and cognitive decline in older adults.” Journal of Prevention of Alzheimer’s Disease, 2026.

2. Metherel AH et al. “Dietary docosahexaenoic acid (DHA) downregulates liver DHA synthesis by inhibiting eicosapentaenoic acid elongation.” Journal of Lipid Research, 65(6):100548, 2024.

3. Huang Y et al. "Associations of fish oil supplementation with incident dementia: Evidence from the UK Biobank cohort study." Frontiers in Neuroscience, 2022.

4. Kerman BE, Self WK, Yassine HN. “Can the gut microbiome inform the effects of omega-3 supplementation?” Current Opinion in Clinical Nutrition and Metabolic Care, 2023.

5. Ebright B, Duro MV, Chen J, Louie G, Yassine HN. “APOE4 effects on DHA brain metabolism across the Alzheimer’s disease spectrum.” Trends in Endocrinology & Metabolism, 2024.

Your brain is not a passive organ. Tucked between neurons, constantly on patrol, is a population of cells called microglia — the brain’s resident immune cells. Think of them as a hybrid between a security guard and a janitor. When something goes wrong — a dead cell, a misfolded protein, a threatening invader — microglia are the first responders. They sense the problem, move toward it, surround it, and break it down.

In Alzheimer’s disease, one of the earliest events is the accumulation of a sticky protein called amyloid, which clumps into plaques outside neurons. Microglia are supposed to recognize those plaques and clear them. For decades, most Alzheimer’s research focused almost exclusively on amyloid itself, as if the cleanup crew didn’t exist. That changed in 2013.

A Signal Hidden in the Genome

That year, two independent research teams published back-to-back papers in the New England Journal of Medicine, each arriving at the same unexpected finding. They had been scanning the entire genetic code of tens of thousands of people — comparing those with Alzheimer’s to those without — searching for places where the DNA spelling differed. This kind of large-scale comparison, called a genome-wide association study or GWAS, is like comparing two manuscripts letter by letter across three billion characters, looking for the discrepancies that explain why one reader ends up with dementia and another doesn’t.

Both teams kept landing on the same gene: TREM2. A specific variant — called R47H — increased the risk of developing Alzheimer’s by two to four times. TREM2 encodes a receptor that sits on the surface of microglia, acting as the sensor that tells them when to swing into action. When TREM2 detects amyloid, cellular debris, or the molecular fingerprints of a dying cell, it instructs the microglia to cluster around the threat and clear it. TREM2, in short, is the alarm system of the brain’s cleanup crew.

A Tempting Hypothesis — and Its Early Warning Signs

The R47H variant blunts that alarm. It is a loss-of-function mutation: the sensor becomes less sensitive. People carrying it have microglia that respond sluggishly to amyloid plaques. Their brains show less microglial clustering around plaques, more diffuse and toxic forms of amyloid, and worse damage to nerve cell branches. Less TREM2 activity. More disease.

The therapeutic idea writes itself: if a broken alarm means the janitors don’t show up, what if you make the alarm louder? Could boosting TREM2 activity help the brain clear amyloid and slow the disease?

The first warning signs were easy to miss. One study found that chronically activating TREM2 could actually worsen the spread of tau — the second major toxic protein in Alzheimer’s, which propagates between neurons and drives irreversible neurodegeneration. Another comprehensive analysis tested TREM2-activating antibodies across multiple mouse models and found neutral or even detrimental results. But the genetic rationale was compelling enough that a clinical-grade drug was built, and a major trial was launched.

Building AL002: The Drug and Its Preclinical Story

What the drug does

AL002, developed by Alector, is a monoclonal antibody — an engineered protein designed to bind to TREM2 and activate it. When it binds, the receptor gets pulled inside the cell and degraded, triggering a burst of downstream signaling. A fragment of TREM2 that normally floats in spinal fluid — called soluble TREM2, or sTREM2 — drops sharply after treatment. Counterintuitively, that drop means the drug is working: the receptor was activated, did its job, and was cleared.

Why the animal studies were misleading

In mice engineered to carry both human TREM2 and five familial Alzheimer’s mutations — the 5xFAD model — a related antibody caused microglia to multiply around plaques, reduced the most toxic forms of amyloid, and decreased nerve cell damage. The mice looked better.

But several important gaps were quietly baked into this evidence. The molecule tested in mice was not AL002 itself — it differed in the region of the antibody that interacts with the broader immune system. The actual human drug was never tested in animals. The 5xFAD mice also accumulate amyloid at extraordinary speed driven by five stacked mutations borrowed from rare inherited forms of disease — an engineered catastrophe that bears little resemblance to the decades-long, genetically complex disease unfolding in a 70-year-old patient. And the mice were treated early, before significant pathology had accumulated, while patients in the trial already had confirmed amyloid and measurable cognitive impairment. Testing a fire suppression system in a controlled drill, then expecting it to work the same way in a building already fully engulfed, is not a reliable methodology. Mouse microglia also differ from human microglia at the level of gene expression and activation states — a gap that further undermines the predictive value of these models.

A Phase 1 trial in healthy volunteers confirmed AL002 was safe and produced the expected spinal fluid signals. But healthy volunteers in their thirties tell you about safety, not whether a drug will slow neurodegeneration in elderly patients with established disease.

What the Trial Found: Active Drug, Passive Disease

The bottom line

Target engagement occurred — CSF sTREM2 fell and CSF osteopontin rose — but amyloid PET showed zero reduction, and nearly one in three treated patients developed amyloid-related imaging abnormalities (ARIA).

INVOKE-2 enrolled 381 people with early Alzheimer’s confirmed by amyloid biomarkers at 69 sites worldwide. Participants received AL002 at one of three doses or placebo by intravenous infusion every four weeks for up to two years.

The drug hit its pharmacological targets. sTREM2 fell in spinal fluid across all treated groups. Osteopontin — a protein linked to TREM2-driven microglial activation — rose. AL002 was reaching the brain and engaging TREM2. The biology was happening.

Figure 1: AL1002 engaged its target as shown with the drop in sTREM2, and raised Osteopontin, inflammatory markers. From Mummery et al, 2026

It just wasn’t helping. After 96 weeks, cognitive and functional decline was identical between treated and placebo groups across every measure tested — memory, orientation, daily function, composite scales. Blood and spinal fluid biomarkers of amyloid, tau, and neurodegeneration showed no difference. The disease progressed as if the drug were not there.

Fig 2: Clinical progression did not change after treatment, despite target engagement. From Mummery et al, 2026

What AL002 Actually Did to Microglia

Activated but not eating

The pharmacodynamic data tells a story, and it is not the story researchers hoped to tell. AL002 activated microglia — they moved toward plaques, enclosed nearby neurons, and protected them from the most acutely toxic fragments of amyloid. But critically, the drug did not activate the intracellular machinery that microglia need to actually engulf and digest an amyloid plaque. The downstream signaling cascade required — involving a pathway called PI3K and the physical restructuring of the cell’s internal skeleton — was not sufficiently engaged. The microglia arrived at the scene. They did not clean it up.

The result was what researchers now recognize as a Toxic Disease-Associated Microglia state — microglia that are biologically loud (secreting cytokines and inflammatory signals) but mechanically lazy (not phagocytosing the amyloid they were sent to clear). Sterile inflammation without productive action.

Osteopontin: a marker we misread

This reinterpretation fundamentally changes how we read one of the trial’s key biomarkers. CSF osteopontin (encoded by the SPP1 gene) was treated as a positive pharmacodynamic signal — evidence that TREM2 signaling had been activated and that microglia were responding. And in a narrow sense, it was. But osteopontin rising without a corresponding fall in amyloid is not a success marker. It is a failure marker.

High osteopontin drives microglia to activate the complement cascade — a set of immune proteins that tag structures for removal. The problem is that complement tags are not specific to amyloid. Microglia hypersensitized by complement activation mistake the molecular markers on healthy synapses — the connections between neurons — for debris, and prune them. The cleanup crew, activated but directionless, starts dismantling functional connections instead of toxic plaques. That pruning of healthy synapses is itself a driver of cognitive decline.

The lesson is subtle but important: moving toward a plaque is not the same as eating it. And eating the wrong thing is worse than not eating at all.

How the Drug Damaged Blood Vessels

The drug’s first encounter is not with microglia

AL002 is given by intravenous infusion — it enters the bloodstream. Before it ever reaches a microglial cell sitting next to an amyloid plaque deep in brain tissue, it must pass through the brain’s vasculature. Lining the outside of those blood vessels is a separate population of immune cells called perivascular macrophages (PVMs). PVMs also express high levels of TREM2, and AL002 binds them just as readily as it binds microglia.

When AL002 cross-links TREM2 receptors on a PVM, it triggers a pathway called Syk signaling — a molecular alarm that, in a PVM, is interpreted as a pathogen trying to breach the blood-brain barrier. The cell does not know it is responding to a drug. It responds as if an infection is trying to force its way into the brain. It begins preparing to demolish the vessel wall so that reinforcements — peripheral immune cells from the blood — can flood in.

In patients who also carry amyloid deposits within their blood vessel walls, a condition called cerebral amyloid angiopathy (CAA), the alarm is compounded. The PVM is simultaneously sensing the drug (interpreted as an invader) and vascular amyloid (interpreted as structural damage). The result is maximum alert: a large release of an enzyme called MMP-9 — molecular scissors that cut the structural collagen holding blood vessel walls together.

When the vessel wall fails

As MMP-9 dissolves the vessel wall, the blood-brain barrier fails. Fluid leaks from the bloodstream into surrounding brain tissue, causing swelling — this is ARIA-E (edema). Red blood cells also escape, releasing heme and toxic iron into brain tissue — this is ARIA-H (microhemorrhage). In APOE4 homozygotes, whose vessel walls are already structurally weakened and who carry the most CAA, the same mechanism produced the most severe outcomes: high rates of both ARIA types, and in some cases seizures, as heme disrupted the electrical signaling of neurons nearby.

This mechanism also explains why testing in standard APOE3 mouse models gave a false sense of safety. APOE3 blood vessels are structurally robust — the same MMP-9 release that ruptures weakened APOE4 vessels simply does not cause visible damage in a mouse with healthier vasculature. The risk was not detectable in the model used.

What This Means for the Next Generation of Microglial Drugs

New benchmarks — and a higher bar

AL002’s failure demands a reckoning with what we actually ask of drugs before they enter trials. Activation alone is no longer an acceptable endpoint. The field now needs to require three things that AL002 could not demonstrate: that a drug suppresses sterile inflammatory signaling rather than amplifying it; that it drives genuine amyloid phagocytosis — actual engulfment and lysosomal digestion — not just microglial recruitment; and that it is vascular-safe in APOE4-bearing models with pre-weakened vessel walls, not just in APOE3 mice whose stronger vessels will hide the damage.

A practical early-screening signal is emerging from this: before advancing any microglial drug, examine its transcriptomic fingerprint. If the SPP1 gene — which encodes osteopontin — dominates the signature without a parallel rise in amyloid-digestion and lysosomal genes, the drug is likely pushing microglia toward inflammatory noise rather than productive cleanup. That is the AL002 pattern. Catching it in the dish or the mouse, before a Phase 2 trial, could save years and patients.

Strategy 1: Refining TREM2 itself

Not everyone has abandoned TREM2 as a target — but the next wave of TREM2 drugs is being designed with AL002’s failures explicitly in mind. VHB937, currently in Phase 2 trials, takes a different approach: rather than activating and degrading the TREM2 receptor (as AL002 did), it stabilizes TREM2 at the cell surface, sustaining signaling without the receptor disappearing. Early data suggest it reduces pro-inflammatory biomarkers, which is the opposite of what AL002 accomplished. Whether stabilizing TREM2 drives more productive amyloid phagocytosis — rather than the inflammatory state AL002 induced — is the question the trial will answer.

A separate approach, VG-3927, recently acquired by Sanofi after positive Phase 1 data, takes a striking detour from the antibody format entirely. It is an oral small molecule TREM2 agonist. An oral drug distributes through the body differently from an intravenous antibody — potentially reaching microglia at lower, more sustained levels rather than flooding the system with a large pulse every four weeks. Whether that changes the vascular risk profile, and whether it drives phagocytosis or inflammation, are open questions for its upcoming Phase 2.

Strategy 2: Release the brakes instead of flooring the gas

Rather than forcing TREM2 to signal louder, a more physiological strategy is to remove the proteins that naturally suppress microglial phagocytosis. Three genetic targets identified in GWAS point this way.

SHIP1 (encoded by INPP5D) acts as a molecular brake downstream of TREM2 — it damps the PI3K signaling cascade that drives microglial engulfment of amyloid. Inhibiting SHIP1 removes that brake, enhancing amyloid uptake and lysosomal capacity in primary microglia without requiring TREM2 agonism. The catch, as described above, is that SHIP1 inhibition would also disinhibit perivascular macrophages, risking the same MMP-9-driven vascular damage seen with AL002. One proposed solution is combining SHIP1 inhibition with cPLA2 blockade — which would physically prevent PVMs from producing the enzyme that damages blood vessel walls, leaving microglia empowered to eat plaques while the vascular risk is contained.

PILRA is a second inhibitory receptor on microglia whose protective variant (G78R) is associated with reduced Alzheimer’s risk, particularly in APOE4 carriers. Recent work published in Science Translational Medicine showed that blocking PILRA with a high-affinity antibody rescued the metabolic and phagocytic deficits that APOE4 specifically causes in microglia — reducing amyloid pathology and restoring synaptic integrity in mouse models transplanted with human microglia. This makes PILRA especially interesting as a target in APOE4 carriers, potentially addressing a genotype that drove the most severe complications in INVOKE-2.

CD33 (Siglec-3) is a third inhibitory receptor whose loss-of-function variants are protective in Alzheimer’s GWAS. CD33 suppresses TREM2 signaling and blocks microglial phagocytosis partly through interaction with SHIP1. A CD33-blocking antibody (AL003) was discontinued in 2022, but the target remains biologically valid — and given what we now know about what AL002 lacked, a drug that disinhibits phagocytosis through CD33 blockade rather than activating inflammation through TREM2 agonism may deserve fresh attention.

Strategy 3: Dampen the inflammatory fire

A third approach asks a different question: rather than trying to redirect microglial activity toward phagocytosis, what if we simply reduce the inflammatory damage those microglia are causing? Two drugs in Phase 2 trials test this idea directly. XPro1595 is a selective TNFα inhibitor designed to block type 1 TNF receptors — the ones that drive neuroinflammation — while leaving type 2 receptors intact, which support myelin maintenance and protective immune functions. The goal is to cool the inflammatory microglial state without broadly suppressing the immune system. Canakinumab, an antibody against IL-1β, targets a specific inflammatory cytokine that activated microglia produce in abundance. Neither drug is designed to drive amyloid clearance directly — they aim to reduce the collateral damage that inflamed microglia cause while other mechanisms handle the plaques.

Strategy 4: Fix the digestive machinery

Even microglia that successfully engulf amyloid may fail to destroy it if the cellular machinery for digestion is impaired. This is an underappreciated dimension of the problem, and it points toward a fourth class of targets: the lysosomal system inside the cell that is supposed to break down ingested material. Progranulin, encoded by GRN, is a protein that supports lysosomal health in microglia. Loss-of-function mutations in GRN cause frontotemporal dementia, and progranulin deficiency impairs the microglial ability to clear amyloid even when phagocytosis is initiated. More broadly, activating TFEB — a master transcription factor that drives production of lysosomal enzymes — could theoretically turn phagocytically recruited microglia into effective digesters, completing the process that AL002 started but could not finish.

These are not fully formed clinical programs yet, but they represent the conceptual shift that INVOKE-2 has accelerated: from asking ‘did we activate microglia?’ to asking ‘did the microglia actually eat the right thing, and digest it?’

Take-Home Messages

Finding a risk gene is not a treatment roadmap. TREM2 is genetically validated and biologically important. But knowing that a gene matters is very different from knowing when, how, and in which patients to intervene on it.

Mouse models have limits that matter. The 5xFAD model produces aggressive, artificial disease in pre-symptomatic mice using a molecule that wasn’t AL002. This model is useful for certain readouts, but not sufficient. Preclinical AD testing remains a formidable task.

Engaging a target is not the same as helping a patient. AL002 hit TREM2, activated microglia, and reached the brain. None of that translated into clinical benefit. Amyloid clearance may prove to be a useful efficacy biomarker.

Microglia that move are not microglia that eat. AL002 drove recruitment without phagocytosis. The next generation of drugs must demonstrate actual amyloid digestion — not just inflammatory activation — before advancing to trials.

Rising osteopontin was not a success signal — it was a failure signal. High CSF osteopontin reflects inflammatory, synapse-pruning microglia, not productive amyloid clearance. Inflammation is a tough target and a complex readout.

The vascular damage had a specific mechanism. AL002 activated TREM2 on blood vessel-lining macrophages, not just neurons. Those cells released MMP-9, cut structural collagen, and damaged the blood-brain barrier — an effect invisible in standard mouse models but devastating in APOE4 patients.

Future microglial drugs need a vascular safety test. ARIA-like damage only appears in APOE4 models with pre-weakened vessels. Testing in APOE3 mice provides false reassurance. That benchmark must be built into preclinical programs.

The next drugs must prove they actually clean up, not just show up. The field is now pursuing four distinct strategies and fixing the cellular digestive machinery (lysosomal and progranulin approaches) is in the pipeline. We have reasons to be hopeful for a breakthrough.

Sources

Mummery CJ et al. The TREM2 agonistic antibody AL002 in early Alzheimer’s disease: a phase 2 randomized trial. Nature Medicine (2026). https://doi.org/10.1038/s41591-026-04273-1

Guerreiro R et al. TREM2 variants in Alzheimer’s disease. N Engl J Med 368, 117–127 (2013).

Jonsson T et al. Variant of TREM2 associated with the risk of Alzheimer’s disease. N Engl J Med 368, 107–116 (2013).

Jain N et al. Chronic TREM2 activation exacerbates Aβ-associated tau seeding and spreading. J Exp Med 220, e20220654 (2023).

Etxeberria A et al. Neutral or detrimental effects of TREM2 agonist antibodies in preclinical models of Alzheimer’s disease and multiple sclerosis. J Neurosci 44, e2347232024 (2024).

Wang S et al. Anti-human TREM2 induces microglia proliferation and reduces pathology in an Alzheimer’s disease model. J Exp Med 217, e20200785 (2020).

Zhong L et al. Soluble TREM2 ameliorates pathological phenotypes by modulating microglial functions in an Alzheimer’s disease model. Nat Commun 10, 1365 (2019).

Edwin TH et al. A high cerebrospinal fluid soluble TREM2 level is associated with slow clinical progression of Alzheimer’s disease. Alzheimers Dement (Amst) 12, e12128 (2020).

Cummings J et al. Alzheimer’s disease drug development pipeline: 2025. Alzheimers Dement (N Y) (2025). https://pmc.ncbi.nlm.nih.gov/articles/PMC12131090/

Loss of PILRA promotes microglial immunometabolism to reduce amyloid pathology in cell and mouse models of Alzheimer’s disease. Science Translational Medicine (2025). https://doi.org/10.1126/scitranslmed.adw7428

VHB937 Phase 2 trial in Alzheimer’s disease — NeurologyLive (2025). https://www.neurologylive.com/view/new-phase-2-trial-test-trem2-stabilizing-agent-vhb937-alzheimer-disease

Vigil Neuroscience VG-3927 Phase 1 data (2025). https://www.globenewswire.com/news-release/2025/1/23/3014135/0/en/Vigil-Neuroscience-Reports-Positive-Data-from-its-Phase-1-Clinical-Trial-Evaluating-VG-3927-for-the-Potential-Treatment-of-Alzheimer-s-Disease.html

Sanofi acquires Vigil Neuroscience (2025). https://www.sanofi.com/en/media-room/press-releases/2025/2025-05-21-23-15-31-3086232

Optimization of SHIP1 inhibitors for the treatment of Alzheimer’s disease. PMC (2025). https://pmc.ncbi.nlm.nih.gov/articles/PMC11713407/

But a large new study published in the American Journal of Preventive Medicine adds an interesting wrinkle to that story. It’s not just how long you sit, it suggests. It might be what you’re doing while you’re sitting.

The Study

Researchers followed over 20,000 Swedish adults for nearly 20 years, tracking who developed dementia along the way. What made this study different from previous research is that it distinguished between two types of sedentary behavior:

Mentally passive: TV watching, listening to music, sitting in a bath — activities where your brain is largely in low gear.

Mentally active: Office work, attending meetings, knitting or sewing — activities that require sustained attention, planning, or problem-solving, even if your body isn’t moving.

The question they asked was simple: does it matter which kind of sitting you do?

What They Found

The answer appears to be yes — at least when it comes to the mentally active kind.

People who spent more time in mentally active sedentary behaviors had a measurably lower risk of developing dementia (Table 2). Each additional hour per day of this type of sitting was associated with roughly a 4% reduction in risk, since HR=0.96 means a 4% lower hazard per 60 min/day increment. And when researchers modeled what would happen if people swapped one hour of passive sitting for one hour of mentally active sitting, they found about a 7% reduction in dementia risk with HR=0.93 (95% CI: 0.87–0.99).

That’s a modest effect, but it was statistically meaningful and held up even after accounting for age, education, smoking, diet, and other factors.

The protective effect was notably stronger among older participants (ages 50–64), which the researchers speculate may be because mentally active behaviors build up what’s called “cognitive reserve” — essentially a buffer of mental resilience that only becomes apparent later in life.

Think of cognitive reserve as a savings account for your brain. Decades of research show that people who accumulate more of it — through education, demanding work, intellectually stimulating leisure, and social engagement — can sustain significantly more physical brain damage before showing clinical symptoms of dementia. Their brains have, in a sense, built alternative routes. A 2024 meta-analysis found that higher cognitive reserve across the life course was consistently associated with reduced dementia risk, and that its benefits compound over time — meaning that what you do in your 40s and 50s may matter as much as, or more than, what you do in your 70s. The catch is that cognitive reserve isn’t something you can bank quickly. It’s built slowly, through years of habits, not months of effort.

What This Doesn’t Mean

Before you swap your gym membership for a puzzle subscription, a few important caveats.

This is observational research. The study can show an association, but it can’t prove that mentally active sitting causes lower dementia risk. It’s possible — and this is a real concern — that people who are already cognitively sharper simply tend to choose more engaging activities. In other words, the direction of causation could run the other way.

The exercise findings are puzzling. Strangely, the study found no significant protective effect from physical activity — even vigorous exercise. That contradicts a mountain of previous research showing PA is one of the strongest dementia protectors we know of. The authors suggest this may be a quirk of how they captured dementia cases (through specialist registers, which miss milder cases). But it’s a flag worth noting: if the model can’t detect a well-known effect, it invites some caution about what it does detect.

The measurement was basic. Sedentary behavior was assessed once, in 1997, using a questionnaire that lumped together very different activities. Knitting and a high-stakes work presentation both counted as “mentally active.” Listening to a podcast and zoning out in front of reality TV both counted as “mentally passive.” The real world is messier than these categories suggest.

19 years is a long time. People’s habits change. The study couldn’t account for how sedentary behavior evolved over two decades of follow-up.

The Elephant in the Room: Your Phone

There’s one glaring gap in this study that researchers and outside experts have been quick to point out: it was designed in 1997, when smartphones didn’t exist, social media hadn’t been invented, and short-form video was science fiction. The passive sedentary behaviors measured back then — TV, music, a long bath — look almost quaint by today’s standards.

Scientists studying dementia and cognitive decline have raised concerns that the kind of passive consumption most of us now do for hours each day — endless scrolling, short-form video, algorithmic feeds designed to hold attention without requiring it — may be doing something more insidious than old-fashioned TV watching. The worry isn’t just that it’s passive. It’s that it may actively train your brain away from the kind of sustained, focused attention that mentally active behaviors depend on.

The idea is that when you spend large amounts of time in a state of passive, low-effort reception — jumping from clip to clip without really concentrating — you may be degrading the very neural pathways you need for concentration, learning, and memory consolidation. The next time you sit down to do something genuinely demanding, your brain has gotten a little worse at it.

Whether this translates into measurable long-term dementia risk is still an open question. The research on “brain rot” and attention spans is early and contested. But the underlying biological logic — that the brain adapts to what it repeatedly does — is well-established. And the sheer volume of passive screen time that now fills daily life dwarfs anything studied in this or most other research.

It’s worth sitting with that for a moment. The study’s data ends in 2016. The TikTok era hadn’t even begun.

Not All Brain Activity Is the Same

Not all “mentally active” activities are equal, and the brain training industry has learned this the hard way. The FTC fined the makers of Lumosity for falsely claiming their games could stave off cognitive decline — because the evidence simply didn’t support it. The core problem is transfer: getting better at a brain-training app mostly makes you better at that app. The skills don’t generalize. What does appear to generalize are activities with genuine complexity, novelty, and depth — learning a new language, picking up a musical instrument, taking up knitting or woodworking. These demand the kind of sustained, layered attention that exercises multiple cognitive systems at once: memory, sequencing, spatial reasoning, problem-solving, and often social engagement too. A crossword is better than scrolling, but learning to play chess or speak conversational Spanish is probably better than a crossword.

The Takeaway

Here’s what the evidence, taken together, suggests:

• Type of sitting matters. Mentally active sedentary behavior — work that demands focus, creative hobbies, learning — is associated with lower dementia risk. Passive sitting is not.

• Cognitive reserve is built over decades. The habits you build in midlife compound quietly. There is no shortcut.

• Brain training apps don’t cut it. The benefit of games like Lumosity doesn’t transfer to real-world cognitive function. Genuine novelty and complexity are what count.

• Your phone may be working against you. Passive scrolling and short-form video may erode the very capacity for focused attention that protects the brain. This is speculative but biologically plausible — and the scale of modern screen time is unprecedented.

• This is one study with real limits. It can’t prove causation, it missed the smartphone era entirely, and it couldn’t even detect the well-established link between exercise and dementia. Hold the findings with appropriate humility.

• The simplest advice still holds. Sit less. Move more. And when you do sit, make it count.

References

1. Werneck et al., “Mentally Active Versus Passive Sedentary Behavior and Risk of Dementia: 19-Year Cohort Study,” American Journal of Preventive Medicine (2026). https://doi.org/10.1016/j.amepre.2026.108317

2. Livingston G et al., “Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission,” The Lancet (2024). https://doi.org/10.1016/S0140-6736(24)01296-0

3. Fracolli LA et al., “Cognitive reserve over the life course and risk of dementia: a systematic review and meta-analysis,” Frontiers in Aging Neuroscience (2024). https://doi.org/10.3389/fnagi.2024.1358992

4. “How Cognitive Reserve Could Protect from Dementia? An Analysis of Everyday Activities and Social Behaviors During Lifespan,” Brain Sciences (2025). https://www.mdpi.com/2076-3425/15/6/652

5. “Brain-training games remain unproven, but research shows what sorts of activities do benefit cognitive functioning,” The Conversation. https://theconversation.com/brain-training-games-remain-unproven-240499

6. “U.S. Cracking Down on ‘Brain Training’ Games,” Scientific American. https://www.scientificamerican.com/article/u-s-cracking-down-on-brain-training-games/

7. Xu C et al., “Associations between recreational screen time and brain health in middle-aged and older adults,” J Am Med Dir Assoc (2024). https://doi.org/10.1016/j.jamda.2024.03.010

8. Fehring D et al., “Changes in prefrontal hemodynamics and mood states during screen use,” Scientific Reports (2025). https://doi.org/10.1038/s41598-025-09360-w

This vascular dimension of Alzheimer’s has long been underappreciated and undertreated. We do not yet have a drug that directly targets the vascular inflammation driving it. Ambreen Kanwal and Bilal Kerman, leading this work at USC, set out to ask whether a specific enzyme — called cPLA2 — might be that target. Their perspective article, just published in Alzheimer’s & Dementia, argues that cPLA2 sits at the intersection of amyloid burden, APOE4-driven lipid dysregulation, and blood-brain barrier breakdown, and that it may be a key upstream driver of vascular injury in Alzheimer’s disease. This post walks through the biology behind that hypothesis, the evidence supporting it, and what would be needed to test it in patients.

Cerebral Amyloid Angiopathy (CAA): Amyloid in the Wrong Place

Most people understand that Alzheimer’s disease involves amyloid plaques building up between brain cells. What is less appreciated is that amyloid also accumulates in the walls of the brain’s blood vessels — the arteries, arterioles, and capillaries that supply the brain with oxygen and nutrients. This condition is called cerebral amyloid angiopathy, or CAA, and it is far more common than most people realize.

Neuropathology studies show that CAA is present in the brain tissue of nearly all patients with Alzheimer’s disease, especially in older patients and in those who carry the APOE4 genetic variant. Even in cognitively normal elderly adults, CAA can be found in 20 to 40 percent of brains examined at autopsy. It is, in this sense, a near-universal companion of aging — but in Alzheimer’s, it is particularly prominent and particularly consequential.

CAA is not one uniform condition. Researchers distinguish between two main subtypes based on which vessels are affected. Type 1 CAA involves amyloid deposits in the basement membranes of the tiny capillaries — the smallest blood vessels deep in the brain tissue. Type 2 CAA, which is more common in older, cognitively normal individuals, involves amyloid depositing in the walls of larger leptomeningeal and cortical arteries. These two subtypes have different relationships to Alzheimer’s pathology: Type 1, capillary CAA, is more specifically linked to AD and to the APOE4 genotype, and it is associated with more active perivascular inflammation.

During life, most CAA is silent. Patients with CAA often have no symptoms, and the condition is discovered incidentally — either on a brain MRI or, posthumously, at autopsy. But when CAA becomes symptomatic, the consequences can be serious: transient episodes of confusion or focal neurological symptoms, recurrent headaches, seizures, and a pattern of cognitive decline. The symptomatic inflammatory form, called CAA-related inflammation (CAA-ri), is characterized by perivascular immune activation — the body’s immune cells attacking the amyloid-laden vessels — and this process disrupts the blood-brain barrier (BBB), the tightly regulated interface that normally keeps blood-borne molecules out of the brain.

Diagnosing CAA in living patients relies almost entirely on brain MRI. The updated Boston Criteria version 2.0 provides a framework for classifying probable CAA based on imaging findings: lobar microbleeds (tiny hemorrhages appearing as dark spots on susceptibility-weighted MRI sequences), cortical superficial siderosis (iron deposits along the brain surface from prior small hemorrhages), and severely enlarged perivascular spaces in specific brain regions. No blood test or spinal fluid marker can currently diagnose CAA with certainty in a living person, and definitive diagnosis requires brain tissue — which is why advances in imaging biomarkers are so important for this field.

ARIA: When Treatment Triggers the Problem

Enter ARIA. When anti-amyloid antibody therapies clear amyloid from plaques and from vessel walls, the mechanical and inflammatory stress on already-fragile CAA-affected vessels can cause visible abnormalities on MRI. Clinicians divide ARIA into two types based on what appears on imaging. ARIA-E refers to edema or effusion — swelling around blood vessels visible as bright signal on FLAIR MRI sequences. ARIA-H refers to hemorrhagic changes — microbleeds or superficial siderosis appearing as dark lesions on susceptibility-weighted imaging.

The majority of ARIA events detected on monitoring MRI scans are asymptomatic. The patient feels nothing, and the abnormalities resolve on their own, often within weeks to a few months. But a meaningful minority of patients develop symptomatic ARIA, which can include headache, confusion, dizziness, visual disturbances, focal neurological symptoms resembling a stroke, and in severe cases, seizures or hospitalization. Symptomatic ARIA requires drug hold and sometimes permanent discontinuation, which eliminates the therapeutic benefit the patient was just beginning to receive.

The rates of ARIA with the current generation of anti-amyloid antibodies are substantial. In the pivotal clinical trial of lecanemab, approximately 21 percent of treated patients developed ARIA-E and 36 percent developed ARIA-H at some point during treatment. With donanemab, similarly high rates were observed. These numbers are already high across the general trial population — but they are dramatically higher in specific subgroups.

The strongest risk factor for ARIA, by a considerable margin, is the APOE4 genotype. APOE4 is the most common genetic risk factor for late-onset Alzheimer’s disease, and carrying one copy raises the risk of developing Alzheimer’s roughly threefold. Carrying two copies — being a homozygote, designated APOE4/4 — raises the risk by eight to twelve times. In the anti-amyloid trials, APOE4/4 homozygotes face ARIA rates of 33 to 67 percent, depending on the drug and the dose. In the APOLLOE4 trial specifically designed to study this population, 32 percent of APOE4/4 participants with early Alzheimer’s already had at least one lobar microbleed at baseline — before receiving any treatment — reflecting how extensive the underlying vascular disease already is. After treatment, the vulnerability compounds dramatically.

Other risk factors for ARIA include higher antibody doses, the presence of pre-existing CAA on baseline MRI, and treatment earlier in the disease course. But APOE4 status dominates all of them. This genetic specificity strongly suggests there is a biology underlying ARIA vulnerability that is tied to APOE4’s effects on the vasculature — and understanding that biology might reveal how to mitigate the risk.

Why APOE4 Makes Blood Vessels So Vulnerable

The APOE4 protein — the product of the APOE4 gene — has several well-established functions in the brain, most notably in transporting cholesterol and facilitating the clearance of amyloid-beta from the brain and from blood vessel walls. APOE4 performs these functions less efficiently than the more common APOE3 variant, which is part of why APOE4 carriers accumulate more amyloid. But APOE4’s effects on blood vessels go beyond amyloid clearance.

APOE4 promotes BBB dysfunction through several mechanisms. It activates a cellular pathway involving matrix metalloproteinase-9 (MMP9), an enzyme that degrades the proteins holding the BBB’s tight junctions together. It is associated with the loss of pericytes — the specialized cells that wrap around blood vessel walls and are essential for BBB maintenance. APOE4 also shifts the endothelium toward a chronic low-grade inflammatory state, even before significant amyloid deposits have formed. The result, in APOE4 carriers, is a blood-brain barrier that is pre-existing compromised — more permeable, more inflamed, and more structurally fragile — before any drug or disease complication adds further stress.

This vascular vulnerability is not just theoretical. In human brain tissue, APOE4 carriers with high CAA burden show measurably reduced astrocyte end-foot coverage of blood vessels, impaired tight junctions, increased MMP9 activity, and lower PDGFRβ levels (a marker of pericyte health) in their cerebrospinal fluid. When anti-amyloid antibodies begin to strip amyloid from these already-vulnerable vessel walls, the resulting inflammatory response occurs on top of a compromised foundation.

The cPLA2 Hypothesis: A Proposed Molecular Culprit

This is where the new perspective paper enters with a specific mechanistic proposal. We hypothesize that an enzyme called cytosolic phospholipase A2 — cPLA2, encoded by the gene PLA2G4A — sits at the intersection of these vascular problems and may be a central driver of the vascular inflammation underlying CAA-ri and ARIA susceptibility in APOE4 carriers.

It is important to say upfront: this is a scientific hypothesis. The evidence marshaled in this perspective is compelling enough to justify serious investigation, but it does not yet prove that cPLA2 causes ARIA or that inhibiting it will prevent ARIA in humans. We argue that the existing evidence is strong enough to propose cPLA2 as a priority target and to outline how the hypothesis should be tested.

The figure above shows cPLA2 protein (red) covering the proximal areas to abeta (green) in blood vessels (white). Microglia and macrophages (brain immune cells) are lining up along the blood vessel, together with the enzyme MMP9, which breaks down the blood-brain barrier junctions. Credit to Bilal Kerman for creating these high-resolution confocal images.

So what is cPLA2, and why does it matter? cPLA2 is an enzyme that sits inside cells — including brain endothelial cells, astrocytes, microglia, and pericytes — and cleaves fat molecules from cell membranes. Specifically, it liberates arachidonic acid (AA) from membrane phospholipids when it is activated by calcium and inflammatory signals. AA is the starting material for a large family of pro-inflammatory signaling molecules called oxylipins, which include prostaglandins, leukotrienes, and hydroxyeicosatetraenoic acids (HETEs). These molecules recruit immune cells, activate MMP9 (which degrades the BBB), and amplify the inflammatory cascade.

But cPLA2’s effects are not purely through what it releases. There is also the matter of what it removes. The same membranes that contain arachidonic acid also contain protective lipids — including plasmalogens (a special type of fat with a vinyl-ether bond that confers antioxidant properties and structural integrity) and docosahexaenoic acid (DHA), the omega-3 fatty acid essential for neuronal membrane health. cPLA2, when overactive, depletes these protective lipids while simultaneously generating pro-inflammatory ones. It is, in effect, a molecular switch that simultaneously amplifies the fire and consumes the fireproofing.

Elevations in the downstream products of cPLA2 — particularly 12-HETE and 15-HETE, measurable in blood and cerebrospinal fluid — have been found in Alzheimer’s patients and correlate with microglial activation and cognitive decline. In APOE4 carriers and in APOE4 mouse models, cPLA2 activity and cPLA2 phosphorylation (the activated form of the enzyme) are measurably higher compared to non-APOE4 controls. These are associations, observed in tissue samples and animal models — they do not prove cPLA2 is the driver of disease, but they establish that the enzyme is more active in precisely the context where vascular vulnerability is greatest.

The perspective paper further points to data from human post-mortem brain tissue: in brains with high CAA burden (confirmed AD patients with CAA scores of 2 or 3 out of 3), activated cPLA2 clusters visibly in and around the blood vessel walls at the sites of amyloid deposition — seen in high-resolution three-dimensional microscopy images included in the paper. Additionally, the level of cPLA2 activity in the vessel walls correlates with evidence of BBB leakage measured by extravascular fibrinogen staining. APOE4 carriers with definite CAA show significantly higher perivascular cPLA2 activity than those without CAA, independent of amyloid and tau pathology.

Animal model experiments add mechanistic plausibility. Genetic knockout of PLA2G4A in mouse models of Alzheimer’s disease reduces neuroinflammation, improves learning and memory, and decreases premature death. Amyloid-oligomer-induced neurotoxicity in neurons can be partly blocked by cPLA2 inhibition in culture. In the E4FAD mouse model — a mouse carrying human APOE4 and familial AD mutations — pharmacological inhibition of cPLA2 reduced amyloid accumulation in vessel walls and reduced hemorrhagic load as measured by MRI. These experiments show that the pathway is tractable, but mice are imperfect models of human Alzheimer’s disease, and such findings do not guarantee that the same results will hold in human clinical trials.

The paper also situates cPLA2 within the complement cascade — the innate immune system’s rapid-response arm. In CAA, amyloid deposits in vessel walls activate complement through the classical pathway, and complement components including MAC (membrane attack complex) cause direct cell lysis. Components C5b-9 and C6, downstream of this cascade, are associated with subcortical hemorrhage and cortical superficial siderosis. Activation of complement by amyloid and APOE may further activate cPLA2, potentially creating a self-reinforcing cycle of vascular injury. Again, this remains mechanistically proposed — the convergence is biologically coherent but not yet causally proven in humans.

Why Previous Anti-Inflammatory Approaches Failed

One of the most instructive threads in the perspective paper is its analysis of why previous anti-inflammatory treatments have not worked in Alzheimer’s disease — and what that failure tells us about where to look next.

The most extensively studied approach has been COX inhibitors: drugs like aspirin, naproxen, and celecoxib, which block the cyclooxygenase enzymes that convert arachidonic acid into prostaglandins. The reasoning seemed sound: if prostaglandins cause inflammation, blocking their production should be protective. The Alzheimer’s Disease Anti-Inflammatory Prevention Trial (ADAPT) randomized over 2,500 cognitively normal older adults to naproxen, celecoxib, or placebo and followed them for two years — until the trial was halted early because both active treatment arms trended toward higher AD rates and showed worse performance on global cognitive scores compared to placebo.

The failure of COX inhibitors is actually informative. These drugs work downstream of cPLA2 (As shown in Figure above), blocking only one branch of the arachidonic acid cascade. But when you block COX enzymes, the freed arachidonic acid is still being produced by cPLA2, and it can be redirected into the lipoxygenase (LOX) pathway, generating leukotrienes and HETEs — other pro-inflammatory oxylipins that COX inhibitors do not touch. You may also inhibit the COX-1 pathway, which produces some beneficial prostaglandins, and you leave intact the cytochrome P450 pathway, which generates yet another class of oxylipins. The result is a pharmacologically incomplete intervention that can cause compensatory activation of the very pathways it is trying to suppress. Furthermore, COX inhibitors have well-documented cardiovascular risks with long-term use, complicating their use in elderly patients.

The perspective paper argues that this history supports moving further upstream — to cPLA2 itself, which sits at the branch point before all these pathways diverge. Inhibiting cPLA2 would simultaneously suppress the lipoxygenase pathway, the COX pathway, and the production of lysophosphatidylcholine (LPC), while also — if the enzyme is sufficiently inhibited — preserving more of the protective membrane lipids that cPLA2 otherwise depletes. The logic is appealing. Whether it holds in human trials remains to be seen.

What Would It Take to Test This Hypothesis?

The perspective paper devotes considerable space to biomarker development and patient stratification — the infrastructure needed to actually test whether cPLA2 inhibition does anything useful in humans.

The proposed biomarker strategy has several layers. In the blood and cerebrospinal fluid, oxylipin profiling can measure the direct products of cPLA2 activity: PGE2, LTB4, 12-HETE, and 15-HETE. Elevated levels of these molecules in APOE4 carriers and their correlation with CAA severity and ARIA susceptibility would strengthen the hypothesis. Plasma oxylipin profiling offers a practical non-invasive readout; CSF sampling provides more direct access to what is happening near brain tissue.

Plasmalogen levels — the protective lipids depleted by overactive cPLA2 — offer a complementary readout. Plasmalogen deficiency is already recognized as a metabolic signature of Alzheimer’s disease progression. If cPLA2 inhibition restores membrane plasmalogens, this would provide mechanistic confirmation of target engagement in the right direction.

Imaging offers the most direct window into the brain. The paper describes PET tracers labeled with fluorine-18 attached to arachidonic acid and DHA. In preliminary studies in APOE4 knock-in mice, these tracers show elevated arachidonic acid uptake in regions consistent with heightened cPLA2 activity, particularly in cortical and perivascular regions. If these tracers can be validated in humans, they would allow clinicians to identify which patients have the most active cPLA2-mediated lipid dysregulation in their brain vessels, and to monitor directly whether a treatment is having an effect on that activity.

A clinical trial testing this hypothesis would ideally enrich for APOE4/4 homozygotes with imaging-confirmed CAA — precisely the patients at highest ARIA risk — and would measure oxylipin and plasmalogen biomarkers as primary pharmacodynamic endpoints before moving to clinical outcomes like ARIA incidence. The question at the center of such a trial would not be “does this drug slow cognitive decline?” (too hard, too long, too expensive for an early-phase trial) but rather “does inhibiting cPLA2 reduce the biochemical and imaging evidence of vascular inflammation in these patients?” That is a testable hypothesis with defined biomarkers and a realistic timeline.

Uncertainty and Future Directions

The cPLA2 hypothesis is scientifically well-constructed, biologically coherent, and supported by a convergence of associative and experimental evidence. It is not yet proven. The gap between a compelling mechanistic framework and a validated human therapeutic is wide, and the history of Alzheimer’s drug development is filled with hypotheses that survived every preclinical test and then failed in people.

Several specific uncertainties deserve emphasis. First, all the human tissue evidence is associative: elevated cPLA2 activity near amyloid deposits in vessel walls could reflect cPLA2 as a cause of vascular damage, or it could reflect cPLA2 activity as a consequence of damage caused by something else — or both, in a feedback loop. The causal arrow is not yet established.

Second, complete inhibition of cPLA2 could be problematic. cPLA2 has normal physiological roles in immune function, wound healing, and lipid metabolism. Loss-of-function mutations in humans cause a platelet dysfunction disorder and intestinal problems. The expectation is that partial inhibition — enough to dampen pathological vascular inflammation without eliminating normal immune signaling — will be both safe and effective, but this therapeutic window has not been established in humans.

Third, the oxylipin and plasmalogen biomarkers, while promising, have not yet been prospectively validated as predictors of ARIA risk in individual patients. The correlations observed in patient cohorts are encouraging, but whether a low-HETE or high-plasmalogen signature in a given patient’s blood or CSF reliably predicts that they will or won’t develop ARIA remains to be tested.

Looking ahead, the paper identifies three areas where progress is needed in parallel. Brain-penetrant cPLA2 inhibitors are the prerequisite — drugs that can reach the perivascular space in adequate concentrations, are selective for cPLA2 over related enzymes, and have a tolerable side effect profile with long-term use. (A companion paper in npj Drug Discovery describes early-stage compounds along these lines, which we covered in a previous post.) Biomarkers for target engagement must be validated — particularly plasma oxylipin profiling and the [18F]-labeled PET tracers — so that clinical trials can confirm drug activity before committing to large, expensive outcome studies. And the most immediate opportunity may lie in clinical integration with the existing anti-amyloid immunotherapy trials: APOE4/4 patients starting lecanemab or donanemab represent a precisely defined high-risk group where an ARIA-prevention trial could be conducted with a meaningful primary endpoint within 12 to 18 months.

The paper also raises an intriguing possibility about timing. Recent scholarship suggests there may be two distinct stages of inflammation in Alzheimer’s disease — an early, pre-clinical stage that is still amenable to anti-inflammatory intervention, and a later stage that has progressed beyond the reach of simple suppression. If that model is correct, earlier intervention targeting cPLA2 in presymptomatic APOE4 carriers with elevated amyloid burden might be far more effective than intervening after established CAA-ri.

Take-Home Messages

CAA — amyloid in the brain’s blood vessels — is nearly universal in Alzheimer’s disease, especially in APOE4 carriers. It is mostly silent, but it creates a vulnerable vascular environment that predisposes to ARIA when anti-amyloid therapies are given. APOE4/4 homozygotes face the highest burden of both CAA and ARIA, with rates of brain hemorrhage-related imaging abnormalities between 33% and 67% on current therapies.

Understanding why APOE4 carriers are so much more vulnerable requires understanding the specific molecular biology that makes their vessel walls fragile and inflamed. The new perspective paper proposes that the enzyme cPLA2 is a key upstream driver of that process, converting amyloid deposits in vessel walls into a chemical cascade of inflammatory lipid mediators while simultaneously depleting the protective fats that keep membranes resilient.

This is a hypothesis, not yet a proven mechanism. The evidence supporting it — from post-mortem human brain tissue, from mouse models, and from cell biology experiments — is substantial and convergent. But demonstrating causality in humans will require the development of validated biomarkers and a well-designed clinical trial in the right patient population.

The failure of COX inhibitors and aspirin in Alzheimer’s trials is reframed by this hypothesis not as evidence that inflammation doesn’t matter, but as evidence that targeting inflammation too far downstream is pharmacologically incomplete. Going upstream — to cPLA2 itself — offers a more comprehensive suppression of the pro-inflammatory oxylipin cascade, combined with the potential benefit of preserving protective membrane lipids.

The ultimate test of this idea will come from humans. We are working on candidate drugs that penetrate the brain in sufficient concentrations and inhibit cPLA2 with the potency and selectivity needed for a rigorous trial.

About the Paper

Kanwal A, Kerman BE, Wang S, Camey K, Li B, Flores-Aguilar L, Ali N, McIntire LB, Shu CA, Louie SG, Head E, Arvanitakis Z, Yassine HN. “A perspective: PLA2G4A as drug target for vascular inflammation in Alzheimer’s disease.” Alzheimer’s & Dementia. 2026;22:e71320. https://doi.org/10.1002/alz.71320

Ambreen Kanwal and Bilal E. Kerman are co-first authors.

Acknowledgments

This work was supported by the National Institute on Aging (RF1AG076124, R01AG055770, R01AG067063, R01AG054434, R21AG056518, P30AG066530, R01AG082362, P30AG10161, P30AG72975, and R01AG15819), the Alzheimer’s Drug Discovery Foundation (ADDF; GC-201711-2014197), donations from the Vranos and Tiny Foundations, and Ms. Lynne Nauss. Additional support was provided by the National Institutes of Health (P50AG05142, R01AG074549, R01AG078800, R01AG072794, and RF1AG059621).

What made her remarkable was not just how long she lived but how. Despite her extraordinary age, she never developed Alzheimer’s disease, never had cancer, and showed no cardiovascular disease. A team of researchers in Spain performed what may be the most comprehensive biological study ever conducted on a single human being — analyzing her genome, transcriptome, metabolome, proteome, gut microbiome, and epigenome, comparing results against multiple matched cohorts. They published their findings in October 2025 in Cell Reports Medicine.¹ What emerged was not a simple formula for longevity. It was a portrait of a life in which many biological systems remained, against considerable odds, in functional equilibrium.

Aging and disease can be decoupled

The data revealed an immediate paradox. This woman showed clear molecular signatures of extreme old age: her telomeres — the protective caps at chromosome ends that shorten with each cell division — were the shortest ever recorded in a healthy person. Her blood showed clonal hematopoiesis, in which a mutated blood stem cell expands over decades and is associated with elevated risk of hematological malignancy and cardiovascular disease. Her immune cells included an expanded population of age-associated B cells, which accumulate with aging and are linked to pro-inflammatory and autoimmune activity.

And yet she had none of the diseases these markers are associated with. The paper’s most important conceptual contribution is demonstrating that molecular aging and age-associated disease are not synonymous. They can be decoupled. The hallmarks of aging — telomere attrition, clonal mutations, immune cell remodeling — can accumulate while the body maintains functional health, provided that something is holding the downstream pathological consequences at bay.

Across every layer of her biology, that something kept appearing.

Low inflammation as the common signal

Measured across blood proteins, lipid particles, metabolites, and gene expression, one feature of her biology was consistently preserved: remarkably low systemic inflammation.

The most direct evidence came from two composite markers, GlycA and GlycB, measured by proton nuclear magnetic resonance spectroscopy. These signals reflect the circulating concentration of acute-phase glycoproteins — including haptoglobin, α1-antitrypsin, and transferrin — that the liver releases in response to inflammatory stimuli. In short-term immune responses, this is adaptive. When chronically elevated even at low levels, it reflects what researchers call inflammaging: the persistent, low-grade, unresolved inflammation that accumulates with age and drives deterioration across organ systems.

In large prospective cohort studies — including analyses of over 250,000 individuals from the UK Biobank — elevated GlycA independently predicts cardiovascular mortality, all-cause mortality, and dementia, beyond traditional risk markers.² Her GlycA and GlycB were both exceptionally low at age 116. Her acute-phase inflammatory response was minimal.

This is not a trivial finding. It is the biological output of an entire life. And when examined alongside how she actually lived, it begins to make sense.

Diet: pattern, not protocol

She followed a Mediterranean dietary pattern and consumed three yogurts daily for at least the final two decades of her life.

The researchers found her gut microbiome was dominated by Bifidobacterium, particularly the family Bifidobacteriaceae — a pattern more typical of younger individuals. Most older adults show progressive Bifidobacterium decline with age, replaced by more pro-inflammatory species. In her case, this was reversed, and the abundance of Bifidobacterium tracked closely with her low-inflammation metabolomic profile. Bifidobacterium produces short-chain fatty acids and conjugated linoleic acid, both of which dampen inflammatory signaling in the gut mucosa and systemically. The researchers drew a direct link between her microbial composition and her low GlycA and GlycB values.

The yogurt strains she consumed — Streptococcus thermophilus and Lactobacillus delbrueckii subsp. bulgaricus — are known to support Bifidobacterium growth. But the yogurt did not function in isolation. It was embedded in a broader Mediterranean pattern: olive oil, vegetables, legumes, fish, minimal ultra-processed food. The anti-inflammatory lipid profile she displayed — very low VLDL and triglycerides, high HDL, favorable lipoprotein particle size distribution — is consistent with decades of this dietary pattern, not with any single food or nutrient.

Movement: habitual, not structured

She did not follow a structured exercise program. She walked, gardened, played piano, and cared for her dogs — continuous, low-intensity, purposeful movement embedded in daily life.

The distinction between habitual movement and structured exercise is biologically meaningful. Evidence from large cohort studies and accelerometry data suggests that reducing prolonged sedentary time and increasing low-intensity habitual movement confer substantial cardiovascular and metabolic benefits, including reductions in circulating TNF-α and IL-6, improved insulin sensitivity, and myokine-mediated anti-inflammatory signaling — effects that do not require high exercise intensity and that accumulate with consistency over years rather than sessions.³

Her movement was also cognitively rich. Piano performance engages motor control, auditory processing, working memory, attention, and emotional regulation simultaneously. Gardening requires ongoing problem-solving, seasonal planning, and fine motor adaptation. This overlap between physical and cognitive engagement is unlikely to be incidental.

Sleep: glymphatic clearance and immune maintenance

The paper notes that she maintained good sleep habits throughout her life. The biological significance of this extends well beyond rest.

During slow-wave sleep, the glymphatic system — a network of perivascular channels that becomes dramatically more active during sleep — clears metabolic waste from the brain, including amyloid-beta and tau, the proteins that aggregate into Alzheimer’s pathology. Even a single night of sleep deprivation measurably elevates amyloid-beta in cerebrospinal fluid. Chronic sleep disruption accelerates epigenetic aging, raises inflammatory markers including GlycA, impairs immune consolidation, and substantially increases dementia risk. The 2024 Lancet Commission on dementia prevention identifies sleep-disordered conditions among its 14 modifiable risk factors, which together account for approximately 45% of dementia cases worldwide.⁴

Sleep is also when the immune system consolidates memory and regulates inflammatory tone. As we discussed in our previous post, VZV reactivation — a driver of chronic inflammaging — is itself promoted by sleep disruption, creating a bidirectional loop between poor sleep and elevated systemic inflammation.

Chronic stress and psychological resilience

Her personal history included significant adversity, including the death of a son. The authors note she maintained strong physical and mental health throughout, remaining socially engaged and purposeful.

Chronic psychological stress activates the hypothalamic-pituitary-adrenal axis, sustaining elevated cortisol that progressively dysregulates immune function. Prolonged cortisol elevation activates NF-κB — the central transcription factor driving inflammatory gene expression — in circulating immune cells, accelerates telomere attrition, promotes hippocampal atrophy via reduced neurogenesis, and advances epigenetic aging. In longitudinal studies, high perceived stress in midlife independently predicts dementia decades later.⁵ Psychological resilience, operationally defined as the capacity to maintain regulatory function under adversity, is associated with lower cortisol reactivity, preserved immune regulation, and slower biological aging by multiple epigenetic clock measures.

Her resilience — likely supported by the same social network, sense of purpose, and daily structure that characterized her life more broadly — appears to have attenuated this pathway. These are not independent factors. Chronic stress disrupts sleep. Poor sleep elevates cortisol and inflammatory markers. Both impair glymphatic clearance and immune resolution. The feedback between stress, sleep, and inflammation is one of the most important amplification loops in biological aging.

Social connection and its biological effects

She remained embedded in a social network — family, friends, and caregivers — throughout extreme old age, including contact with animals.

Social isolation activates the same hypothalamic-pituitary-adrenal stress axis as physical threat, raising circulating inflammatory cytokines, accelerating telomere shortening, and predicting faster epigenetic aging. A major meta-analysis found loneliness associated with a 26% increase in all-cause mortality.⁶ The Lancet Commission identifies social isolation as a modifiable dementia risk factor. Mechanistically, perceived social connection downregulates NF-κB activity in immune cells, reduces cortisol, and promotes oxytocin signaling, which has direct anti-inflammatory effects. The biology does not appear to respond primarily to physical proximity but to the subjective experience of meaningful, reciprocal connection.

Cognitive engagement and reserve

She read books and played piano into her final years, and tended a garden that required continuous adaptation and problem-solving.

Formal education — measured by years of schooling — is identified by the Lancet Commission as the single largest modifiable dementia risk factor. The operative mechanism, however, is cognitive reserve: the brain’s accumulated capacity to sustain function in the presence of neuropathological damage. A brain with more synaptic redundancy and practiced cognitive flexibility can absorb more amyloid deposition, vascular injury, and neuronal loss before clinical symptoms appear. Post-mortem studies consistently show that individuals with high cognitive reserve may carry full Alzheimer’s pathology without clinical diagnosis.

Cognitive reserve is not fixed by formal schooling. It accrues across the lifespan through any genuinely demanding mental engagement. Piano performance is among the most neurologically complex activities available, requiring concurrent engagement of motor, auditory, mnemonic, attentional, and emotional systems. Her sustained engagement with cognitively demanding activities well past conventional retirement age likely continued building and maintaining the neural redundancy that may have protected her from symptomatic neurodegeneration.

A lifetime of infections, resolved

Her immune profile bore the signature of long experience. T cells were dominated by effector and memory subtypes. Immunoglobulin G levels — particularly IGHG2 and IGHG4 — were elevated, consistent with efficient and mature humoral memory. This immune phenotype reflects an organism that has encountered many pathogens, resolved the encounters, and consolidated the memory.

This is distinct from immune exhaustion, which we described in our previous post on microglial biology and Alzheimer’s disease: chronic, unresolved immune stimulation — from VZV reactivation, persistent amyloid burden, or other sources — drives immune cells toward a dysfunctional, pro-inflammatory, senescent state. Her immune system appeared experienced rather than exhausted: it had fought and cleared, repeatedly, across a century. This is consistent with the evolutionary framing introduced in our previous post — that APOE4 confers protection in high-pathogen environments by enhancing immune responsiveness, and becomes a liability in low-pathogen modern environments where the same responsiveness generates unresolved chronic inflammation.

She did not carry APOE4. Her APOE genotype was at the protective end of the spectrum, which almost certainly mattered for her lipid metabolism and neurological resilience.

Genetics helped, but gene-environment interaction was the mechanism

The researchers identified seven rare homozygous variants in her genome absent from European control populations, and protective configurations at APOE, FOXO3A, and genes involved in immune regulation, mitochondrial oxidative phosphorylation, cardiovascular function, and DNA repair.

Her genome was unusually favorable. But the paper is explicit: no single genetic variant explains her longevity. The identified variants span disparate biological processes — immune surveillance, energy metabolism, cardiovascular resilience, neuroprotection — and none appears individually sufficient. All appear to have operated in concert with each other and with her environment. MAP4K3, a longevity-regulating gene whose variant she carried, regulates lifespan in C. elegans through pathways that are sensitive to diet, stress, and metabolic state. The gene does not act independently of the environment in which it is expressed.

Gene-environment interaction is a core principle of quantitative genetics. The same variant that is protective in one environment may be neutral or deleterious in another. Her protective genome was expressed through — and likely required — the environmental context of her life: diet, physical activity, sleep, stress management, social engagement, and a century of immune experience.

The epigenetic clocks: an integrated readout

When the researchers estimated her biological age using six different epigenetic clocks — algorithms that infer biological age from DNA methylation patterns — every clock placed her substantially below her chronological age of 116. The discrepancy ranged from approximately 10 to 27 years across methods and tissues. Using the rDNAm clock, her biological age deceleration was 23.17 years.

Epigenetic clocks are accelerated by chronic inflammation, sleep disruption, psychological stress, social isolation, sedentary behavior, poor diet, and chronic unresolved infection — the same factors addressed throughout this post. Her 23-year deceleration represents the integrated biological output of a life in which these inputs were consistently attenuated. It is not attributable to any single variable. It reflects the cumulative effect of many factors operating in concert over a very long time.

What we do not know

Before drawing conclusions, it is important to be direct about the limits of this evidence.

This is a study of a single individual. No causal inferences can be drawn from an N of 1. We cannot determine whether her lifestyle choices produced her low inflammatory state, or whether her genetic endowment independently determined that state and also happened to promote the behaviors we observed — an instance of reverse causation that cannot be resolved in a cross-sectional study of one person.

We do not know the relative contribution of each factor discussed here. The multiomics data reveal associations between her biological profile and various features of her life and genome; they cannot quantify how much each factor mattered. The gut microbiome findings, for instance, are from a single time point late in life. We do not know whether her Bifidobacterium dominance was lifelong, whether it was a cause or consequence of her metabolic health, or whether it would have the same implications in a genetically different individual.

Epigenetic clocks, while powerful population-level tools, have been trained primarily on datasets that do not include many individuals past age 100. Their accuracy and interpretation at extreme ages remains uncertain. The variability and predictive power of these clocks are uncertain.

It is also worth noting that supercentenarians are not a homogeneous group. Other individuals who have lived to comparable ages have smoked, consumed alcohol heavily, and been largely sedentary. Extreme longevity likely has multiple biological routes, and the profile described here may represent one path among several. Survivorship bias is a genuine concern: we study those who reached extreme age, but we cannot easily study the many individuals who adopted similar lifestyles and did not.

Finally, most of the lifestyle factors described — diet, exercise, sleep, social connection — have been studied in relatively short-term randomized trials or observational studies spanning years to decades. Whether the effect sizes from those studies compound as expected over a full human lifespan is not established.

What the paper offers is not a proof of mechanism or recommendations for certain tests. It is a richly detailed portrait of one person whose biology was measurably younger than her age, in which low systemic inflammation emerges as the most consistent cross-cutting feature. That observation is worth taking seriously, even in the absence of causal certainty.

The sum is greater than the parts

What is most striking about this biological portrait is its refusal to yield a single active ingredient.

Not the yogurt. Not the Mediterranean diet. Not the walking. Not the piano. Not the protective genome. Not the Bifidobacterium. Not the good sleep. Not the family connections. No single factor, isolated, accounts for 117 years of health. The biology supports a systems view: these factors are mutually reinforcing, operating through shared mechanisms — principally, the regulation of chronic inflammation — and their combined effect is greater than any individual contribution.

A good high fiber fermented diet with an active lifestyle supports a healthy microbiome, which reduces systemic inflammation, which preserves epigenetic stability, which maintains mitochondrial function, which supports immune competence, which allows infections to resolve rather than persist, which protects sleep quality, which lowers cortisol, which further attenuates inflammation. These are not parallel pathways. They are a network. Single interventions improve one node in the network; a lifetime of consistent, modest, mutually reinforcing inputs keeps the network itself functional.

She did not optimize. She lived — with consistency, engagement, connection, and purpose — and the biology reflects it.

Illustration made using Claude

Take-home messages

There is no single intervention that confers longevity or health span. The most comprehensive biological study of an extreme long-liver found a confluence of genetic, dietary, microbial, immunological, and behavioral factors. The evidence does not support a protocol-based view of aging.

Inflammation is the final common pathway. Poor sleep, chronic stress, social isolation, physical inactivity, unresolved infection, and poor diet each generate chronic low-grade inflammation through overlapping mechanisms. Reducing the sources of that inflammation — rather than pharmacologically suppressing the immune response — is the central challenge of healthy aging.

Acute, resolving inflammation is not the problem — chronic, unresolved inflammation is. The immune system requires activation. Acute infections that clear, physical effort that triggers repair, and transient physiological stress that drives adaptation are all beneficial. The harmful signal is the one that never resolves.

Diet operates as a pattern over decades, not as an ingredient. The gut microbiome responds to the cumulative input of dietary diversity, fermented foods, fiber, and reduced ultra-processed intake. The evidence does not support supplementing individual components as a substitute for the overall pattern.

Habitual movement — independent of structured exercise — has measurable anti-inflammatory effects. Reducing prolonged sedentary time and sustaining low-intensity daily movement consistently improves metabolic and inflammatory biomarkers. Duration and consistency appear to matter more than intensity for most of the outcomes relevant to aging.

Sleep quality and duration affect amyloid clearance, immune regulation, and epigenetic aging. These are not independent consequences of aging — they are modifiable inputs that feed back into the aging process itself.

Chronic stress activates inflammatory signaling and accelerates biological aging by measurable epigenetic and telomeric markers. Psychological resilience and a sense of sustained purpose appear to attenuate this pathway.

Social connection downregulates NF-κB-mediated inflammatory signaling and is independently associated with dementia risk, cardiovascular outcomes, and mortality. The effect appears to be mediated by perceived meaningful connection rather than physical proximity alone.

Cognitive reserve — built by sustained engagement with mentally demanding activities — provides measurable protection against the clinical expression of neurodegeneration, even in the presence of pathological burden. It is not fixed by formal education and accumulates throughout the lifespan.

Genetics shapes vulnerability and resilience but does not determine outcomes independently of environment. Gene-environment interactions, particularly for variants like APOE4, mean that lifestyle factors may matter more — not less — for individuals with elevated genetic risk. The environment is where the greatest modifiable leverage lies.

The goal is the compression of morbidity — maintaining biological function close to the end of life — not the maximization of lifespan per se. She spent her final months managing bronchiectasis and osteoarthritis but reached that point without Alzheimer’s disease, cancer, or cardiovascular disease. That trajectory, more than the final number, is what the evidence points toward.

This post draws on: Santos-Pujol et al., “The multiomics blueprint of the individual with the most extreme lifespan,” Cell Reports Medicine 6, 102368 (2025); the Lancet Commission on Dementia Prevention, Intervention, and Care (2024 update); and concepts from our previous post on immune exhaustion and Alzheimer’s disease, including Kim & Crimmins (J Gerontol, 2026) and Millet et al. (Immunity, 2024). The post was constructed using Claude, Sonnet 4.6.

Footnotes

1. Santos-Pujol E et al. The multiomics blueprint of the individual with the most extreme lifespan. Cell Rep Med. 2025;6:102368. ↩

2. Wang Z et al. Plasma metabolomic profiles associated with mortality and longevity in a prospective analysis of 13,512 individuals. Nat Commun. 2023;14:5744.

3. Church S et al. Glycoprotein acetyls as a novel inflammatory biomarker of early cardiovascular risk. J Am Heart Assoc. 2022;11:e024380. ↩

4. Ungvari Z et al. The multifaceted benefits of walking for healthy aging. Geroscience. 2023;45:3211–3239. ↩

5. Livingston G et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. Lancet. 2024. ↩

6. Johansson L et al. Midlife psychological stress and risk of dementia: a 35-year longitudinal population study. Brain. 2010;133(8):2217–2224. ↩

7. Holt-Lunstad J et al. Loneliness and social isolation as risk factors for mortality. Perspect Psychol Sci. 2015;10(2):227–237. ↩