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. ↩

Over the past ten years, a new picture has emerged from genetics, from single-cell biology, and from some surprising places — the Bolivian Amazon, a Norwegian natural experiment on shingles, a ward of bladder cancer patients getting TB vaccines. The evidence is pointing toward a unified idea: that Alzheimer’s disease is, at least in part, a disease of immune exhaustion. The brain’s immune cells — microglia — become overwhelmed, worn out, and eventually unable to do the maintenance work that keeps Alzheimer’s pathology from taking hold. And some of our most powerful genetic risk factors, including the notorious APOE4 gene, may be accelerating exactly that process.

Understanding this reframes a lot of things: what our most recent drugs are actually doing, why certain vaccines seem to protect against dementia, and most importantly, what we should be developing next.

The Alzheimer’s Genome Is a Microglia Blueprint

To understand why the immune system matters so much in Alzheimer’s, start with genetics. Over the past fifteen years, genome-wide association studies (GWAS) — massive hunts through the DNA of hundreds of thousands of people for variants linked to disease — have uncovered more than 40 genetic locations associated with late-onset Alzheimer’s disease risk.

When researchers look at what those genes actually do, a striking pattern emerges. A large fraction of them are not expressed primarily in neurons. They are expressed in microglia — the resident immune cells of the brain. Genes like TREM2, CD33, BIN1, INPP5D, ABI3, CR1, and SPI1 all have primary functions in microglial biology: regulating how these cells respond to danger, how they engulf and degrade debris, how they signal to one another and to neurons.

The GWAS picture, in other words, tells us that the genetic architecture of Alzheimer’s is, to a remarkable degree, an immune architecture. Whatever is going wrong in the disease, the immune system is centrally implicated — not as a bystander, but as a principal actor.

TREM2 (Triggering Receptor Expressed on Myeloid Cells 2) deserves special mention. Rare variants in TREM2, particularly the R47H variant, confer a risk of late-onset Alzheimer’s roughly comparable to carrying one copy of APOE4. TREM2 is a receptor that sits on the surface of microglia and helps them sense and phagocytose — essentially, eat — cellular debris including amyloid plaques. When TREM2 is impaired, microglia lose their ability to effectively wrap around and dismantle plaques. They also struggle to survive in the inflammatory milieu of the Alzheimer’s brain. The result is a brain that is literally less capable of cleaning itself.

Microglia: The Brain’s Tireless Caretakers

To appreciate what goes wrong, it helps to appreciate what goes right — at least when things are working well.

Microglia make up roughly 10–15% of all brain cells. They are the central nervous system’s resident immune cells, derived from progenitor cells in the yolk sac during fetal development and maintained largely independently of the circulating bloodstream thereafter. They are, in the most literal sense, the brain’s dedicated immune workforce — present throughout life, patrolling constantly, responding to disturbances.

In a healthy brain, microglia perform ongoing surveillance. They extend fine processes that probe the local environment, sensing cellular damage, pathogen-associated signals, or the accumulation of abnormal proteins. When they detect a problem, they can migrate to the site, engulf and degrade the problematic material, and coordinate a local immune response to resolve the threat. They also play important roles in normal brain homeostasis — pruning synapses during development, supporting neuronal survival, and maintaining the extracellular environment.

In the context of Alzheimer’s disease, microglia are responsible for surveilling and clearing amyloid-beta plaques and tau tangles — the canonical hallmarks of the disease. When amyloid accumulates, microglia cluster around plaques in what is called a disease-associated microglia (DAM) state, upregulating genes involved in phagocytosis and inflammatory signaling. This response is protective, at least initially. It slows plaque expansion and contains the damage.

The problem is that this activated state has a cost. Sustained microglial activation is metabolically demanding and pro-inflammatory. If the stimulus — the accumulating amyloid, the chronic low-grade infection, the cellular stress — never resolves, the microglia cannot return to a resting homeostatic state. They are trapped in a cycle of activation without resolution.

When Immune Cells Run Out of Steam: Terminally Inflammatory Microglia

What does immune exhaustion look like in the brain? A landmark 2024 paper in Immunity by Millet, Ledo, and Tavazoie at Rockefeller University provides one of the most detailed answers yet (Millet et al., Immunity, 2024). Using single-cell RNA sequencing across an entire atlas of brain immune cells in Alzheimer’s model mice bearing different human APOE alleles, the researchers identified a distinct population of microglia they termed Terminally Inflammatory Microglia, or TIMs.

TIMs are characterized by simultaneous expression of inflammatory signaling markers and cellular stress markers. They are not simply “activated” microglia in the conventional sense — they appear to be something more advanced and more problematic: a population that has exhausted its adaptive capacity. Using trajectory analysis, the authors showed that TIMs represent a terminal state in microglial biology, arising from homeostatic microglia through acutely and chronically inflammatory intermediate states. Critically, once microglia reach the TIM state, it appears to be largely irreversible.

The frequency of TIMs was markedly elevated with age and with APOE4 genotype. In very old AD model mice, TIMs made up 45% of all microglia in APOE3 mice — but a striking 69% in APOE4 mice. This population was also detectable in human Alzheimer’s brains, with higher frequency in APOE4 carriers and at more advanced stages of disease (higher Braak score). Perhaps most damning: TIMs showed severely impaired capacity to phagocytose amyloid-beta. In direct ex vivo phagocytosis assays, TIMs were dramatically underrepresented among cells that successfully cleared amyloid. And in APOE4 mice, even TIMs that retained some phagocytic capacity were less effective than their counterparts in other genotypes.

The picture that emerges is stark. APOE4 does not merely accelerate amyloid production or impair its clearance through lipid metabolism. It also drives microglia faster toward a state of immune exhaustion — a state from which they cannot effectively do the work that might protect the brain.

Illustration created using Claude.

APOE4: A Risk Gene That Made Evolutionary Sense

The APOE4 story has long been puzzling from an evolutionary perspective. If the allele is so harmful — raising lifetime Alzheimer’s risk by 3-4 fold, reducing longevity — why does it exist in 20–25% of most human populations? Why hasn’t natural selection eliminated it?

A remarkable 2017 study by Trumble, Stieglitz, Blackwell, and colleagues at Arizona State University and UC Santa Barbara helps answer this question (Trumble et al., FASEB Journal, 2017). They studied 372 members of the Tsimane people — Amazonian forager-horticulturalists in Bolivia who live without access to sanitation, clean water, or modern medicine. More than two-thirds of Tsimane adults carry active intestinal helminth (parasite) infections at any given time.

The finding was startling. In industrial populations, APOE4 reliably predicts cognitive decline. But among Tsimane adults with high parasite burden, the opposite was true: APOE4 carriers showed better cognitive performance than non-carriers, particularly on measures of fluid cognition. The higher the parasitic load (measured by eosinophil counts, a biomarker of infection), the more pronounced the cognitive advantage of carrying APOE4. In parallel, APOE4 carriers in this population had significantly lower eosinophil counts overall — suggesting the allele was also helping to actually clear parasitic infections.

The interpretation is that APOE4 evolved in an environment of heavy pathogen burden. Its effects on immune function — including its effects on cholesterol metabolism, lipid transport, and immune signaling — were adaptive when the immune system was constantly challenged by real parasitic and infectious threats. These are conditions that characterized the vast majority of human evolutionary history. What has changed is the environment. In modern industrialized populations with low pathogen burden, the immune-stimulating properties of APOE4 no longer have a useful target. Instead of keeping a useful immune response calibrated against real threats, APOE4 may push microglia toward a state of chronic, unresolved inflammation — the very condition that drives immune exhaustion.

This is the mismatch hypothesis applied to the immune system: APOE4 was built for a world of worms and bacteria, and in a world of processed food and antibiotics, it may be driving the brain’s immune system toward burnout.

Microglial Senescence: When Tiredness Becomes Permanent

Immune exhaustion and cellular senescence are related but distinct concepts, and both are increasingly implicated in Alzheimer’s disease.

Cellular senescence is a state in which cells permanently exit the cell cycle, lose normal function, and begin secreting a cocktail of pro-inflammatory molecules — a phenomenon known as the senescence-associated secretory phenotype (SASP). Senescent microglia have been identified in aging human and mouse brains, and their numbers increase dramatically with age and in the presence of AD pathology. They are characterized by dystrophic morphology (fragmented, bulbous processes rather than the fine ramified extensions of healthy microglia), impaired phagocytic capacity, and elevated expression of stress-response genes.

There is a vicious cycle at work here. Chronic microglial activation in response to accumulating amyloid and tau generates persistent inflammatory signals. These signals promote cellular stress and, eventually, senescence. Senescent microglia then cannot clear pathology — allowing it to accumulate further — and actively contribute to neuroinflammation through SASP. The TIM population described by Millet et al. appears to represent something analogous: a late-stage, functionally compromised microglial state that can no longer effectively contribute to amyloid clearance, and that may be enriched for senescent markers.

The key point is that this is not simply “inflammation” in the cartoon sense of the immune system being turned on. It is chronic, unresolved, dysregulated inflammation — a system that has lost the capacity to mount and then resolve an effective immune response. It is inflammation as dysfunction, not inflammation as defense.

The Shingles Vaccine: An Accidental Experiment in Immune Rescue

One of the most intriguing recent threads in Alzheimer’s prevention comes not from neuroscience labs, but from large population studies of vaccination.

In 2025, a landmark natural experiment published in Nature Medicine exploited a Welsh policy quirk: the live-attenuated shingles vaccine (Zostavax) was rolled out in age-cohort order, meaning that people born just before a certain date were eligible and people born just after were not. This created a near-randomized comparison. The result: vaccine eligibility was associated with a roughly 20% reduction in dementia incidence over the next several years — a striking effect size for a vaccine not designed to protect against dementia.

This finding was complemented by a 2026 study by Kim and Crimmins in the Journals of Gerontology, using data from the Health and Retirement Study (Kim & Crimmins, J Gerontol, 2026). Among nearly 4,000 older Americans, those who had received the shingles vaccine showed significantly more favorable biological aging profiles, including lower systemic inflammation, slower epigenetic aging (as measured by DNA methylation clocks), and lower transcriptomic aging scores. Crucially, the inflammation reduction was robust even after adjusting for socioeconomic factors, health behaviors, and multimorbidity. The effect was most pronounced in the years immediately following vaccination, consistent with the hypothesis that suppressing varicella-zoster virus (VZV) reactivation reduces chronic immune stimulation.

The mechanism, the researchers propose, is one of immune recalibration. The herpes zoster virus lies dormant in nerve tissue after childhood chickenpox infection, and it periodically reactivates — sometimes producing shingles, often subclinically. Each reactivation triggers an immune response, elevating inflammatory cytokines like IL-6, TNF-alpha, and CRP. Over decades, this chronic, low-level immune activation contributes to inflammaging — the background hum of systemic inflammation that accelerates biological aging and may progressively exhaust immune capacity.

By suppressing VZV reactivation, shingles vaccination may reduce this chronic immune burden. Microglia, relieved of one source of sustained peripheral immune activation, may be better positioned to maintain homeostasis and respond effectively to amyloid and tau. In essence, the vaccine may be protecting the brain by giving the immune system one fewer thing to fight.

Importantly, Kim and Crimmins also note that influenza and pneumococcal vaccines showed some similar — if weaker — benefits, particularly in the cardiovascular domain. This raises the question of whether reducing the burden of latent and recurrent infections more broadly might constitute a strategy for preserving immune capacity in aging. Whether BCG vaccination (the tuberculosis vaccine, which has broad non-specific immune-modulating effects and has been associated with roughly a 20–45% reduction in Alzheimer’s risk in some retrospective studies), herpes simplex vaccines (not yet approved), or earlier pneumococcal vaccination might offer parallel protections is not yet known, but the hypothesis is increasingly testable.

What Anti-Amyloid Antibodies Are Really Doing

The approval of aducanumab in 2021 and lecanemab in 2023 brought the first disease-modifying treatments for Alzheimer’s disease to the clinic. These are monoclonal antibodies designed to bind amyloid-beta and accelerate its removal from the brain. They work — lecanemab reduced the rate of clinical decline by about 27% in phase 3 trials. But they also come with a significant side effect: Amyloid-Related Imaging Abnormalities, or ARIA.

ARIA manifests as brain edema (ARIA-E) or microhemorrhages (ARIA-H) on MRI. In clinical trials, ARIA-E occurred in about 12–35% of patients depending on the antibody and APOE4 genotype, with APOE4 carriers at substantially higher risk. It is usually asymptomatic but can cause headache, confusion, and in rare cases, serious neurological events.

The mechanism of ARIA illuminates something important about what these antibodies are actually doing. Anti-amyloid antibodies opsonize amyloid aggregates — they tag them for destruction. Microglia express Fc receptors that recognize the antibody-tagged amyloid and engulf it. But amyloid also accumulates extensively in cerebral blood vessels (cerebral amyloid angiopathy), and when the antibody binds this vascular amyloid, it triggers an immune response at the vessel wall — complement activation, microglial Fc receptor-mediated phagocytosis, and the resulting inflammation increases vascular permeability, producing the edema and microbleeds of ARIA.

The Millet et al. study sheds light on what happens to microglia during aducanumab treatment. Acute treatment drove microglia toward disease-associated (DAM) states and effector-hi TIMs — an expanded immune response. Crucially, aducanumab and APOE4 genotype were both associated with a higher number and strength of predicted cell-cell interactions in the microglial compartment, suggesting that the drug is doing more than simply tagging amyloid for clearance. It is engaging and reshaping the brain’s immune landscape, including expanding adaptive immune cell interactions (particularly CD8 T cells) in an APOE4-dependent manner.

This is both promising and cautionary. It suggests that anti-amyloid antibodies may be exerting some of their benefits — and their risks — by broadly engaging microglial biology beyond amyloid clearance. Whether they can help restore a more homeostatic microglial state, or whether they risk further exhausting already-compromised TIM populations, remains an important open question. In APOE4 carriers, where TIMs are more abundant and more profoundly impaired, these dynamics may be particularly consequential.

The Future: Restoring Microglial Biology

If the central problem in Alzheimer’s — at least through the lens of APOE4 and many GWAS hits — is the progressive exhaustion and senescence of microglia, what can we do about it?

Several approaches are in active development:

TREM2 agonists. TREM2 is a critical survival and activation signal for microglia, and its expression is reduced in aging and in advanced disease. Agonistic antibodies that activate TREM2 signaling — including AL002 (tested in the INVOKE-2 phase 2 trial) and VHB937 (currently in phase 2) — aim to restore microglial function by boosting this pathway. The INVOKE-2 results were disappointing on primary endpoints, raising questions about timing: TREM2 agonism may need to be applied earlier in disease progression, before the TIM state becomes dominant.

Senolytics. Drugs that selectively eliminate senescent cells are a rapidly growing field. In preclinical models, clearing senescent microglia reduced neuroinflammation and improved cognitive outcomes. Clinical trials in Alzheimer’s are in early stages, but the rationale is compelling: if a proportion of microglia have become permanently dysfunctional, removing them might allow the remaining healthy population — or newly recruited monocyte-derived cells — to function more effectively.

CSF1R modulators. CSF1R (Colony Stimulating Factor 1 Receptor) governs microglial survival and proliferation. Transient pharmacological depletion of microglia followed by withdrawal of the drug allows repopulation from progenitor cells — essentially resetting the microglial compartment. This approach has shown promise in preclinical settings, though getting the timing and context right for human application is an active area of investigation.

Cell therapy. More ambitiously, some researchers have explored transplanting healthy microglia or microglial precursors to replace exhausted resident populations. A recent study demonstrated that systemic hematopoietic cell transplantation can restore microglial TREM2 function in mouse models.

Metabolic support. TIMs, as Millet et al. showed, are characterized by profound metabolic derangement — depleted glycolysis, pentose phosphate pathway activity, and TCA cycle function. Whether interventions that restore microglial metabolic fitness (through mTOR modulation, NAD+ precursors, or other approaches) could help prevent or reverse the TIM state is an exciting but still largely unexplored question.

cPLA2 inhibition. One particularly compelling metabolic-inflammatory target is calcium-dependent phospholipase A2 (cPLA2), an enzyme that sits at the intersection of lipid metabolism and immune activation. When microglia are chronically stimulated, cPLA2 is activated — either by oxidative stress, MAPK signaling, or elevated intracellular calcium — and begins liberating arachidonic acid (AA) from membrane phospholipids. AA is then metabolized by cyclooxygenases and lipoxygenases into a cascade of pro-inflammatory eicosanoids: prostaglandins, leukotrienes, and thromboxanes. In the short term, this is part of a normal immune response. But when cPLA2 stays chronically active, the sustained eicosanoid output generates reactive oxygen species, promotes lipid accumulation in lysosomes, activates NF-κB, and drives microglia toward the very senescent phenotype described above — a vicious cycle of inflammation and cellular damage that is difficult to escape. Critically, APOE4 specifically activates cPLA2 through MAPK p38 signaling, a finding confirmed in both animal models and postmortem human brain tissue (Wang et al., 2022; Ebright et al., 2022, reviewed in Hugo et al., Antioxid. Redox Signal., 2024). In AD mouse models, cPLA2 protein levels are elevated in astrocytes surrounding amyloid plaques, and cPLA2 knockout mice show rescued cognition on maze tasks. This makes cPLA2 inhibition an attractive strategy — not simply to reduce inflammation downstream, but to interrupt the lipid-driven senescence program at its source, potentially before irreversible TIM states are reached. The key practical challenge is developing inhibitors that are potent, selective for cPLA2 over related phospholipase isoforms, and capable of crossing the blood-brain barrier. Several lead compounds exist (including ASB14780 and the structurally diverse series emerging from virtual screening platforms), and the goal is not to abolish cPLA2 activity entirely — which would impair normal immune signaling — but to bring its chronic overactivation back into a physiological range. APOE4 carriers, who show measurably higher brain cPLA2 activity, may represent the population most likely to benefit; [18F]fluoro-AA PET imaging, which tracks arachidonic acid kinetics as a surrogate of brain cPLA2 activity, is being developed as a companion biomarker to identify and monitor this at-risk group.

Neither Good nor Bad: Reframing Inflammation

Throughout this discussion, you may have noticed that we keep running into a paradox: inflammation is implicated in Alzheimer’s disease, but so is immune exhaustion. How can both be true?

The answer is that the popular framing of inflammation as simply “good” or “bad” is not adequate to describe what is happening in the aging brain.

In the context of Alzheimer’s disease, we need good inflammation: the acute, targeted, resolving immune response that allows microglia to detect, engulf, and degrade amyloid and tau, then stand down. This is the physiological function the system was designed for. We also need the capacity to resolve inflammation — to return to homeostasis after a threat has been addressed.

What we do not want — and what appears to be accumulating in aging APOE4 brains — is chronic, unresolved inflammation: the persistent low-grade immune activation that exhausts microglial capacity, drives cells into senescent and terminally inflammatory states, and ultimately impairs the very clearance functions that were supposed to protect against pathology.

This is why the shingles vaccine finding is so conceptually important. The vaccine is not directly anti-inflammatory — in fact, it triggers a robust immune response. But by resolving a source of chronic latent viral pressure, it may restore the immune system’s capacity to function effectively: acute responses that resolve, rather than chronic smoldering that never does.

The same logic applies to APOE4. In an environment of high pathogen burden (the Tsimane Bolivians with their helminth infections), the allele’s immune-enhancing properties are channeled into useful, targeted immune work that keeps cognitive function intact. In a modern, low-pathogen environment, the same properties may tip the immune system toward chronic activation without resolution — driving microglia toward burnout.

And it is the same logic that should guide how we think about the new anti-amyloid antibodies. They are not simply “reducing inflammation” — they are restructuring the brain’s immune response, engaging microglia, and reshaping cell-cell interactions. Whether they do so in a way that ultimately restores or further compromises microglial resilience will likely depend on when in the disease course they are given, in whom, and at what dose.

Where This Leaves Us

We are at an early but genuinely exciting moment in Alzheimer’s research. The convergence of GWAS genetics, single-cell biology, evolutionary medicine, and epidemiology is drawing us toward a coherent picture of what goes wrong in the most common genetic form of Alzheimer’s disease: the progressive failure of the brain’s immune workforce.

The discovery of TIMs — exhausted-like microglia that accumulate with age and APOE4 burden and cannot effectively clear amyloid — provides a compelling cellular mechanism for much of what we have observed clinically. The finding that APOE4 functioned differently — indeed, protectively — in environments with high parasite burden puts this in evolutionary context. The association between shingles vaccination and slower biological aging suggests that the immune system’s capacity to protect the brain may be more modifiable than we thought.

The implications for therapy are profound. Rather than asking only “how do we clear amyloid?” we should also be asking: “how do we keep the immune cells that clear amyloid from burning out?” Those are different questions, and they may call for different — and complementary — answers.

A brain in which the right inflammation happens at the right time, resolves completely, and leaves microglia refreshed and ready for the next challenge: that is what we are trying to preserve. Not a brain with no inflammation. Not a brain with relentless, unresolved inflammation. But a brain whose immune system can do its job — and then rest.

This post draws on four key papers:

• Millet, Ledo & Tavazoie. “An Exhausted-Like Microglial Population Accumulates in Aged and APOE4 Genotype Alzheimer’s Brains.” Immunity 57(1): 153–170, 2024.

• Trumble et al. “Apolipoprotein E4 is associated with improved cognitive function in Amazonian forager-horticulturalists with a high parasite burden.” FASEB Journal 31: 1508–1515, 2017.

• Kim & Crimmins. “Association between shingles vaccination and slower biological aging: evidence from a US population-based cohort study.” Journals of Gerontology, Series A, 81(3): glag008, 2026.

• Hugo, Asante, Sadybekov, Katritch & Yassine. “Development of Calcium-Dependent Phospholipase A2 Inhibitors to Target Cellular Senescence and Oxidative Stress in Neurodegenerative Diseases.” Antioxid. Redox Signal., 2024. DOI: 10.1089/ars.2024.0794

Additional references:

• Late-onset AD genetics implicates microglial pathways — Molecular Neurodegeneration, 2017.

• Functional characterization of AD genetic variants in microglia — Nature Genetics, 2023.

• Senescent microglial accumulation in AD — Neuron, 2024.

• Varicella-zoster reactivation and dementia risk — Nature Medicine, 2025.

• Vaccines and dementia — BCG and beyond — J Alzheimers Dis, 2024.

• TREM2 and sTREM2: mechanisms to therapies — Molecular Neurodegeneration, 2025.

• Aducanumab induces sustained microglial alterations — Journal of Experimental Medicine, 2024.

Claude Sonnet 4.6 (Anthropic) was used to help put this article together and create the illustration. AI is not used to generate content.

The honest truth is that academic life had swallowed me whole at the time. Grant deadlines, manuscript revisions, back-to-back committee meetings, study section, a clinical schedule that doesn’t pause for anyone. I saw her message, thought I need to think carefully about this one. So belatedly, here is my take:

What the Study Did and Found

“Meat Consumption and Cognitive Health by APOE Genotype” by Norgren, Carballo-Casla, and colleagues at Karolinska Institutet is a prospective cohort study published March 19, 2026 in JAMA Network Open, drawing from the Swedish National Study on Aging and Care, Kungsholmen (SNAC-K). Of 5,111 randomly selected adults aged 60 and older, 2,157 met inclusion criteria — no dementia at baseline, with available diet, cognition, and APOE data. Mean age was 71.2 years; 62% were women; 569 participants (26.4%) carried APOE ε3/ε4 or ε4/ε4 genotypes (APOE34/44). Follow-up extended up to 15 years, during which 296 participants developed dementia.

Dietary exposure was assessed using a validated 98-item food frequency questionnaire at baseline and at 3- and 6-year follow-ups, with total meat consumption expressed as grams per total kilocalories. Secondary exposures included the ratio of processed to total meat and the log ratio of unprocessed red meat to poultry. Covariates were selected using causal inference principles and included age, sex, education, physical activity, smoking, alcohol, total energy intake, morbidity, baseline cognition, and — methodologically important — the Alternative Healthy Eating Index (AHEI) score excluding meat items, ensuring the association was not simply a marker of overall dietary quality.

Cognitive trajectories (composite z-score across episodic memory, semantic memory, verbal fluency, and perceptual speed over 15 years, available in 1,680 participants) were analyzed by linear regression with quintile-based exposure. Dementia incidence was analyzed using Fine and Gray subdistribution hazard ratios (sHRs), treating non-dementia death as a competing risk. To address reverse causation, the authors used a triangulation approach — combining longitudinal between-participant, within-participant fixed-effects, and cross-sectional baseline analyses — and confirmed that dietary changes did not differ between those who did and did not develop dementia.

Main findings: in APOE34/44 carriers, the top versus bottom quintile of meat intake was associated with better cognitive trajectories (β = 0.32; 95% CI, 0.07–0.56; P = .01), most pronounced for episodic memory (β = 0.52; 95% CI, 0.12–0.92). No association was found in non-carriers (β = −0.11; P = .20). The APOE interaction on cognitive trajectory: P = .004. For dementia, the sHR in APOE34/44 carriers was 0.45 (95% CI, 0.21–0.95; P = .04) versus 0.95 in non-carriers — but the APOE-by-meat interaction on dementia did not reach statistical significance (P = .10). Processed meat increased dementia risk regardless of genotype (sHR, 1.14; 95% CI, 1.01–1.29, Figure below). Post-hoc analyses suggested a consistent APOE interaction for all-cause mortality (HR, 0.85; P = .04) and greater vitamin B12 absorption from meat in APOE34/44 carriers.

Does P = .10 for the Dementia Interaction Undermine the Whole Paper?

This is the right question to ask, and it deserves a direct answer.

The P = .10 for the APOE-by-meat interaction applies specifically to the dementia diagnosis endpoint. With only 569 APOE34/44 participants and 296 total dementia cases across all genotypes, the study was underpowered to reliably detect a genotype interaction for incident dementia — this is a harder, rarer endpoint than a continuously measured cognitive score. The directionality of the dementia finding is fully consistent with the cognitive trajectory result; it simply does not clear the significance threshold for that specific secondary outcome.

The paper’s primary pre-specified outcome was cognitive trajectory, analyzed in 1,680 participants with 15 years of longitudinal data, where the interaction is P = .004 and survives multiple sensitivity analyses including participants with higher HbA1c, lower AHEI scores, and higher vascular risk scores. The triangulation approach yielded consistent results across three analytical methods. The dementia result is best understood as directionally supportive but statistically inconclusive — not as evidence against the cognitive finding.

That said, the P = .10 is a legitimate reason for interpretive caution. The paper supports the hypothesis that APOE genotype modifies the association between meat consumption and cognitive aging trajectories in a Swedish elderly cohort. That is what it shows — not that red meat prevents Alzheimer’s disease.

Why This Paper Is Still Worth Taking Seriously

Beyond the pre-specified hypothesis and triangulation design, several features add credibility. The AHEI adjustment (excluding meat) means the cognitive signal is not merely reflecting that high meat consumers also eat more vegetables and fish. The effect was robust among women, participants under 72, those with higher HbA1c, and those with higher cardiovascular risk scores — subgroups where confounding by general healthiness is less likely. Post-hoc findings for mortality and B12 absorption point in the same direction. And the distinction between processed and unprocessed meat — with processed meat conferring risk regardless of genotype — is internally coherent and consistent with the broader literature.

The limitations are real: self-reported diet, a single urban Swedish cohort with limited generalizability across ancestries, a modest APOE34/44 subgroup, and the inherent inability of observational data to establish causality despite the methodological care. The study is best received as a well-grounded hypothesis-generating contribution that makes a compelling case for APOE-stratified randomized trials.

The Biology: Why APOE4 Carriers May Respond Differently to Dietary Fat

The observational finding has a plausible biological basis, though the mechanism points somewhere more specific than “APOE4 carriers need more fat.”

In a review I wrote with Caleb “Tuck” Finch in Frontiers in Aging Neuroscience, we examined what apoE4’s cell biology implies for brain energy metabolism. APOE4 impairs the brain’s ability to use glucose efficiently — APOE4 astrocytes have roughly half the glucose uptake capacity of APOE3 astrocytes, and neuroimaging confirms regional brain glucose hypometabolism in APOE4 carriers detectable decades before symptoms. In response, the APOE4 brain shifts toward dietary fat as its preferred energy source. This may be one reason why older APOE4 carriers who lose significant weight show accelerated cognitive decline — adipose fat stores appear to serve as an important fuel reservoir for a brain that can no longer rely as heavily on glucose.

This energy-preference hypothesis has direct clinical support. Hanson and colleagues (2015) showed in a blinded feeding study that an acute high-fat meal improved cognitive performance and plasma Alzheimer’s disease biomarkers in APOE4 carriers with mild cognitive impairment — while non-carriers showed the opposite pattern. Fat, in a glucose-deprived APOE4 brain, appears to be a better fuel.

This provides a biologically coherent explanation for the SNAC-K finding: if the APOE4 brain runs better on fat, a dietary pattern that includes more fat — including from unprocessed meat — may support cognitive function in ways that do not apply to non-carriers.

The important qualification: the APOE4 brain’s preference for fat does not translate into a recommendation to eat more red meat or to adopt a ketogenic diet. The fat that matters most for the APOE4 brain is high-quality polyunsaturated fat, particularly the omega-3 fatty acid DHA. Using PET imaging, we showed that brain DHA uptake is approximately 20% greater in younger cognitively normal APOE4 carriers versus non-carriers — a signal of heightened biological demand. That demand is best met through fish and seafood. Red meat, particularly unprocessed, may contribute to overall fat and protein intake in ways that modestly support this, but it is not the primary source, and processed red meat — high in sodium, nitrates, and additives — increases dementia risk across all genotypes and should be minimized regardless of APOE status.

Dietary Patterns, Not Individual Foods

The most important and most frequently overlooked principle in nutrition and dementia research is that the relevant unit of analysis is a dietary pattern, not a single food. In our Lancet Healthy Longevity paper, we reviewed the full evidence base — observational studies, clinical trials, and mechanistic data — and found that no individual nutrient or food has demonstrated consistent cognitive benefit in adequately powered randomized trials when evaluated in isolation. The Mediterranean and MIND dietary patterns show the most consistent associations with cognitive health, not because of any single component but because of their structural coherence: high vegetables, whole grains, legumes, fish, olive oil, moderate unprocessed animal protein, and minimal processed food and added sugar. The SNAC-K finding that processed meat increases dementia risk regardless of genotype fits squarely within this framework. For APOE4 carriers, the most evidence-supported approach is a pattern that provides adequate omega-3 fatty acids, preserves insulin sensitivity, and minimizes ultra-processed foods — not one defined by a target intake of any single protein source.

There is, however, a complication in applying Mediterranean diet evidence to the SNAC-K cohort that is worth naming. Sweden is not a Mediterranean country, and long-term adherence to a Mediterranean dietary pattern in Scandinavian populations is low. In a cohort where few participants habitually consume olive oil, legumes, and fish in the combinations that define Mediterranean eating, the comparison group for "high meat consumers" is not people eating a Mediterranean diet — it is people eating less meat within a Northern European dietary context, which may look quite different nutritionally. This makes it difficult to use Mediterranean diet evidence as the interpretive frame for what the SNAC-K results mean in practice. It also underscores a broader point that our field has been slow to act on: dietary recommendations need to account for cultural and regional food environments. A precision nutrition approach for APOE4 carriers cannot simply export one population's dietary pattern to another; it must identify the underlying nutritional principles — adequate high-quality fat, omega-3 sufficiency, low processed food burden — and allow those to be achieved within the foods and traditions that people actually eat.

Take-Home Messages

• The primary finding — a statistically significant APOE interaction for cognitive trajectories (P = .004) — is the paper’s substantive claim. The dementia interaction did not reach conventional significance (P = .10), which reflects limited power for that secondary endpoint rather than evidence against the cognitive finding.

• The study is observational. Self-reported diet, a single Swedish cohort, and modest APOE34/44 subgroup sizes constrain interpretation. The triangulation design partially addresses reverse causation, but residual confounding cannot be excluded.

• Processed and unprocessed red meat are not the same exposure. The cognitive benefit was most pronounced for unprocessed meat; processed meat increased dementia risk regardless of genotype. This distinction is often lost in coverage of this study.

• There is biologically plausible support for APOE4-specific dietary fat responses. The APOE4 brain has impaired glucose metabolism and compensates by preferring fat as an energy substrate. Hanson and colleagues showed that an acute high-fat meal improved cognition and AD biomarkers in APOE4 carriers with MCI but worsened them in non-carriers — exactly the kind of genotype-specific response the SNAC-K data reflect at a population level.

• This does not mean APOE4 carriers should eat more red meat or adopt a ketogenic diet. The fats that matter most for the APOE4 brain are heart-healthy good fats, obtained from fish and seafood, but also from unprocessed red meat or plant-based foods. Processed red meat increases dementia risk regardless of genotype and should be avoided.

• Dietary patterns are more important than individual foods. No single food drives or prevents Alzheimer’s disease. One example of a healthy pattern is the Mediterranean-style pattern — rich in fish, vegetables, whole grains, and olive oil, with modest unprocessed animal protein — remains the best-supported framework for cognitive health, including in APOE4 carriers.

• “Mediterranean pattern” evidence does not map cleanly onto a Swedish cohort. Few SNAC-K participants habitually follow a Mediterranean dietary pattern, so the high-meat comparison group is not eating Mediterranean — it is eating less meat within a Northern European food context. This limits cross-study interpretation and highlights the need for personalized, culturally adapted dietary guidance that identifies the underlying nutritional principles — adequate high-quality fat including red meat, omega-3 sufficiency, low processed food burden — and allows people to meet them through the foods and traditions they actually live with.

• The right response to this paper is to design better trials, not to change dietary recommendations. Genotype-stratified randomized nutrition trials are urgently needed. This study makes a compelling case for them.

References

1. Norgren J, Carballo-Casla A, Grande G, et al. Meat Consumption and Cognitive Health by APOE Genotype. JAMA Network Open. 2026;9(3):e266489. doi:10.1001/jamanetworkopen.2026.6489

2. Yassine HN, Finch CE. APOE Alleles and Diet in Brain Aging and Alzheimer’s Disease. Front Aging Neurosci. 2020;12:150. doi:10.3389/fnagi.2020.00150

3. Yassine HN, et al. Nutrition State of Science and Dementia Prevention: Recommendations of the Nutrition for Dementia Prevention Working Group. Lancet Healthy Longevity. 2022;3(8):e501–e512.

4. Hanson AJ, Bayer JL, Baker LD, et al. Differential effects of meal challenges on cognition, metabolism and biomarkers for apolipoprotein E varepsilon4 carriers and adults with mild cognitive impairment. J Alzheimers Dis. 2015;48(1):205–218. doi:10.3233/jad-150218

5. 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 Neurol. 2017;74(3):339–347. doi:10.1001/jamaneurol.2016.4899

6. Yassine HN, Croteau E, Rawat V, et al. DHA Brain Uptake and APOE4 Status: A PET Study With [1-11C]-DHA. Alzheimers Res Ther. 2017;9(1):23. doi:10.1186/s13195-017-0250-1

To understand why the field of Alzheimer’s research has spent three decades pursuing a single protein, you need to appreciate how consistent the original clues were. In 1992, John Hardy and Gerald Higgins formally proposed what became known as the amyloid cascade hypothesis—the idea that the abnormal buildup of a protein fragment called amyloid-beta (Aβ) in the brain is the earliest and most critical trigger of Alzheimer’s disease (AD). This was not speculation. It was a framework built on converging lines of genetic evidence that, taken together, pointed toward amyloid as the initiating event.

The first and perhaps most powerful piece of evidence came from people with Down syndrome. Individuals with trisomy 21 carry three copies of chromosome 21, which also happens to be where the gene encoding the amyloid precursor protein (APP) resides. Because of this gene-dosage effect, people with Down syndrome produce more Aβ throughout their lives—and virtually all of them develop the hallmark plaques and tangles of Alzheimer’s pathology by their late thirties and forties. Those who die young from other causes in their early-to-mid teens already show diffuse amyloid deposits in the brain, with tangles and neurodegeneration following later. This natural experiment—triple the dose of APP, and you observe early, progressive Alzheimer’s pathology—provided a foundational biological anchor for the hypothesis.

The second line of evidence came from rare inherited forms of early-onset Alzheimer’s disease (familial AD, or FAD). Scientists discovered that mutations in the APP gene itself, or in proteins called presenilin-1 and presenilin-2 (the enzyme that cuts APP to generate Aβ), all caused aggressive Alzheimer’s in people in their forties and fifties. These mutations share a unifying biochemical consequence: they shift the balance of Aβ production toward longer, stickier, more aggregation-prone forms, particularly Aβ42. Strikingly, a rare protective mutation in the APP gene (A673T) was later identified in Icelandic families: carriers produced slightly less Aβ throughout their lives and were substantially protected from both Alzheimer’s dementia and age-related cognitive decline, with some showing very few amyloid plaques even at age 100. As Selkoe and Hardy wrote at the 25-year anniversary of the hypothesis, “definitive proof... could only come from clinical trials that selectively target Aβ and produce slowing and ultimately arrest of cognitive decline.” The hypothesis also gained important genetic credibility when the discovery of ApoE4 as a major risk factor for late-onset AD was found to act, at least in part, by impairing the brain’s ability to clear Aβ effectively. Taken together, these observations across Down syndrome, inherited mutations, protective variants, and population genetics built a coherent biological case: Aβ dyshomeostasis—too much production, too little clearance, or both—sits at the beginning of the Alzheimer’s disease cascade, preceding tangles, neurodegeneration, and cognitive decline by years or even decades.

Validation at Last: How Recent Clinical Trials Supported the Amyloid Cascade Hypothesis

The road from biological hypothesis to therapeutic validation was long and littered with failure. For roughly two decades, nearly every major pharmaceutical effort targeting amyloid came up empty. Anti-amyloid vaccines were halted due to dangerous brain inflammation. Drugs designed to block the enzymes that produce Aβ (beta- and gamma-secretase inhibitors) either failed to show benefit or actually worsened outcomes. Dozens of monoclonal antibodies—proteins designed to bind and clear Aβ from the brain—missed their primary endpoints in large Phase 3 trials, including failures for solanezumab, bapineuzumab, crenezumab, gantenerumab, and others. These repeated setbacks raised serious questions about whether targeting amyloid was a viable therapeutic strategy at all.

But the failures contained instructive lessons. They indicated that treating patients who already had moderate-to-severe dementia was likely too late, and that not all anti-amyloid antibodies were equivalent—those that powerfully cleared fibrillar plaque, the dense aggregated form most associated with downstream pathology, appeared more likely to produce measurable biological effects. Armed with these refinements, a new generation of trials targeting earlier disease stages with more potent plaque-clearing antibodies produced the long-awaited result. The CLARITY AD trial of lecanemab (2022) enrolled 1,795 patients with early symptomatic Alzheimer’s and demonstrated a statistically significant 27% slowing of clinical decline at 18 months, as measured by the Clinical Dementia Rating Sum of Boxes (CDR-SB). The TRAILBLAZER-ALZ 2 trial of donanemab (2023) enrolled 1,736 early-stage patients and showed a 35% slowing of decline. Both drugs not only cleared amyloid plaques robustly—they also produced reductions in phosphorylated tau (p-tau) biomarkers in blood and cerebrospinal fluid. This biomarker finding was among the most scientifically significant: if reducing Aβ also lowers downstream p-tau—a protein that becomes pathologically modified and spreads through the brain as a consequence of amyloid accumulation—it provides biological support for amyloid’s position upstream in the disease cascade, rather than as a parallel or unrelated process.

Not the Whole Story: Why Amyloid Doesn’t Fully Explain Alzheimer’s Dementia

Validating the amyloid cascade hypothesis is not the same as claiming it fully explains the clinical syndrome of Alzheimer’s dementia across all affected people. There are several important reasons to interpret the clinical trial results with measured expectations.

Even in the genetic forms of Alzheimer’s that originally anchored the hypothesis, the disease mechanism appears more complex than Aβ overproduction alone. In Down syndrome, the extra copy of APP on chromosome 21 was long assumed to cause AD primarily by increasing Aβ42 production. But work from Ralph Nixon’s laboratory has shown that the β-secretase-cleaved carboxyterminal fragment of APP (APP-βCTF, or C99)—generated from the same excess APP—directly elevates lysosomal pH by approximately 0.6 pH units in cells from individuals with Down syndrome, inactivating cathepsin D and other lysosomal hydrolases that depend on acidic conditions to function. This lysosomal dysfunction is detectable perinatally, preceding amyloid plaque formation by years, and is reversed by normalizing lysosomal pH or by reducing APP expression with siRNA or BACE1 inhibition. Similarly, in familial AD caused by PSEN1 mutations, the presenilin-1 protein has a role in lysosomal acidification entirely separate from its function in the γ-secretase complex that cleaves APP: PSEN1 acts as a chaperone for the v-ATPase V0a1 subunit and is required for proper lysosomal acidification. Disease-causing PSEN1 mutations impair this function, elevating lysosomal pH and creating a cascade of autophagic substrate accumulation and progressive neurodegeneration that Nixon and Rubinsztein characterize as a primary driver of AD pathogenesis, not merely a consequence of Aβ . These findings introduce a meaningful qualification: even in the most penetrant genetic forms of Alzheimer’s, the path from gene mutation to neurodegeneration may run through lysosomal failure in parallel with—or even upstream of—Aβ accumulation. The distinction between “amyloid-driven” and “endolysosomal-driven” pathology may be less clean than the original formulation assumed.

More than 99% of Alzheimer’s cases are nonetheless late-onset and sporadic, arising from the complex interaction of genetics, aging, lifestyle, vascular health, and environmental factors. In this much larger population, amyloid accumulation appears to be a contributing factor but not the sole sufficient cause of dementia, and other pathological processes operate in parallel.

The first consideration is the magnitude of clinical benefit from anti-amyloid therapies. The CDR-SB is a scale ranging from 0 (no impairment) to 18 (severe dementia), where clinicians rate how well a person manages six life domains: memory, orientation, judgment, community activities, home life, and personal care. In early Alzheimer’s, scores typically fall between 2 and 8. The minimum difference on this scale that most clinicians consider meaningful to a patient’s daily life is estimated at 1 to 2 points. Lecanemab slowed decline by approximately 0.5 points over 18 months relative to placebo; donanemab by approximately 0.7 points over roughly 20 months. In plain terms: treated patients declined somewhat more slowly, but both groups continued to decline, and the average difference between treated and untreated patients was below what most clinicians or caregivers would detect day to day. For any individual patient, the chance of experiencing a perceptible benefit is modest, and it must be weighed against the risk of ARIA—amyloid-related imaging abnormalities including brain swelling and microbleeds—that affected 13–37% of treated patients across the trials, with a subset requiring discontinuation or hospitalization. This does not diminish the scientific significance of a positive trial, which after decades of failure is genuinely important. It does, however, raise a practical and conceptual question: if amyloid is the central driver of Alzheimer’s dementia, why does substantially clearing it from the brain produce relatively small effects on the rate of cognitive decline? The most likely answer is that by the time patients are enrolled in these trials—even at the “early” stage—substantial neurodegeneration, tau pathology, and inflammation have already accumulated, and removing amyloid at that point is analogous to extinguishing a fire after the structure has burned. The disease has become self-sustaining through mechanisms that are no longer amyloid-dependent.

This complexity becomes further apparent when we examine racial and ethnic diversity in Alzheimer’s presentation. A 2026 study by Wheeler and colleagues examined 1,181 cognitively unimpaired and 383 mildly cognitively impaired participants from the Health and Aging Brain Study–Health Disparities (HABS-HD), encompassing non-Hispanic White (NHW), Hispanic, and Black older adults. Using tau PET imaging to map tau pathology in the medial temporal lobe (MTL)—a brain region critical for memory—and amyloid PET to characterize amyloid burden, the study revealed important group differences. After controlling for age, sex, and education, Black and Hispanic participants showed significantly higher MTL tau levels than NHW participants, even after accounting for choroid plexus off-target signal (a technical artifact that can artificially inflate tau measurements in certain brain regions). More notable still was the nature of the amyloid-tau-cognition relationship across groups: in NHW and Hispanic participants, amyloid positivity significantly amplified the relationship between tau and memory impairment—amyloid-positive individuals with high tau showed substantially lower cognitive scores. But in Black participants, this interaction between amyloid positivity and the tau-cognition relationship was attenuated and did not reach statistical significance (p = 0.067). Additionally, APOE4 significantly predicted higher MTL tau in NHW participants but showed no significant association with tau in Black or Hispanic participants. These findings suggest that in Black participants, factors other than amyloid and tau may be more dominant contributors to cognitive impairment—possibly cerebrovascular disease, greater burden of cardiometabolic risk factors like hypertension and diabetes, chronic stress exposure (which animal models show can promote tau hyperphosphorylation), or mixed brain pathologies. Prior neuropathological studies have shown that mixed pathology is more common in Black decedents with dementia. The amyloid cascade hypothesis, developed primarily from data in predominantly NHW populations, requires expansion to adequately account for the etiological diversity of Alzheimer’s dementia across different populations.

What Genetics Is Telling Us: Endolysosomal Dysfunction and Lipid Metabolism as Underlying Risk Factors

If amyloid does not fully account for Alzheimer’s dementia—particularly in late-onset, sporadic cases—where should we look for a more complete explanation? An important set of answers has been emerging from genome-wide association studies (GWAS), large-scale genetic analyses that identify common genetic variants associated with increased Alzheimer’s risk across tens of thousands of people. When researchers catalogued the genes most consistently implicated in late-onset AD risk, they found not a collection of genes involved primarily in amyloid production or clearance, but rather two distinct biological themes: lipid transport and metabolism, and endolysosomal trafficking—the cellular machinery responsible for sorting, recycling, and degrading molecular cargo inside cells.

The landscape of late-onset GWAS genes reads like a chapter on cellular waste management and membrane biology. APOE itself—encoding the brain’s primary lipid carrier—represents by far the strongest genetic signal. SORL1 regulates the trafficking of amyloid precursor protein through endolysosomal compartments. BIN1 facilitates membrane trafficking and endocytosis. PICALM coordinates clathrin-mediated endocytosis relevant to Aβ removal. TREM2 is a receptor expressed by microglia—the brain’s immune cells—that governs their capacity to phagocytose cellular debris, including Aβ and damaged lipid-laden membranes. ABCA7, a lipid transporter closely related to ABCA1, carries loss-of-function mutations that substantially increase AD risk, likely by impairing cellular lipid efflux. CLU (clusterin) participates in lipid metabolism and Aβ clearance. GRN (granulin) is required for lysosomal function and inflammatory regulation. CTSF is a lysosomal protease involved in intracellular protein degradation. Taken together, these genes point consistently toward a shared biological pathway: the endolysosomal system—the cell’s internal sorting and disposal apparatus—is disturbed in Alzheimer’s disease in ways that may operate in parallel to, or upstream of, amyloid accumulation. The enlargement of early endosomes is among the earliest cytopathological features of AD, observable years before plaques appear, and is exacerbated by APOE4. This pattern suggests a complementary framework: that failure of endolysosomal recycling (Figure below), particularly in the astrocytes and microglia responsible for lipid transport and waste clearance, represents a biologically distinct pathway to Alzheimer’s pathology—one that the amyloid hypothesis was not designed to capture. Here, we can conclude that amyloid contributes to the endolysosomal recycling failure but is one of many drivers, and vice versa, endolysosomal failure accelerates amyloid accumulation.

Figure from Scott and Petsko, Science Translational Medicine, 2020

APOE4: A Gene With Many Faces, Neuroinflammation Foremost Among Them

Of all the genetic factors in late-onset Alzheimer’s disease, none has a larger population-level impact than APOE4. Carrying one copy of the APOE4 allele roughly triples the lifetime risk of Alzheimer’s; carrying two copies increases risk more than tenfold. The initial explanation for this risk centered on amyloid: APOE4 impairs the brain’s ability to clear Aβ, contributing to earlier and more abundant plaque accumulation. That relationship is well-supported. But APOE4’s contribution to Alzheimer’s risk involves multiple interconnected biological mechanisms—organized around three distinct but overlapping “hits” on brain health.

The first hit is disrupted lipid handling. ApoE is the brain’s primary lipid transport protein, produced mainly by astrocytes and microglia to deliver cholesterol and other lipids essential for synaptic maintenance and membrane repair. The APOE4 isoform has an altered protein conformation that reduces its lipid-loading efficiency, particularly through a transporter called ABCA1. Research from our lab demonstrated that in APOE4 astrocytes, ABCA1 becomes abnormally retained in lysosomes rather than recycling to the cell surface, impairing the formation of small, lipidated HDL particles that are associated with protection against amyloid accumulation. This lysosomal sequestration is driven in part by co-aggregation of ApoE4 and ABCA1 proteins in an environment enriched with oxysterols (oxidized cholesterol metabolites), which are elevated in APOE4 brain tissue. The resulting cholesterol dyshomeostasis sets off a compensatory over-expression of ABCA1—which then activates mTORC1 in lysosomes, triggering cellular senescence in astrocytes and microglia. Senescent brain cells surrounding amyloid plaques have been observed in human AD brain tissue, staining positive for lipofuscin and GPNMB (classical senescence markers), suggesting that APOE4-driven lipid dysregulation accelerates cellular aging in the very cells whose function is to protect neurons.

The second hit is neurovascular inflammation. APOE4-expressing microglia are oriented toward a chronically activated, pro-inflammatory state. In the brain, APOE4 amplifies activation of the classical complement cascade—a molecular immune recognition system—which, when chronically overactivated, accelerates excessive synaptic pruning by microglia and promotes neurodegeneration. APOE4 also upregulates matrix metalloproteinase-9 (MMP9), an enzyme that degrades the tight junctions of the blood-brain barrier, allowing inflammatory molecules to enter the brain parenchyma and fibrinogen to accumulate near neurons. These neurovascular effects appear to operate at least partly independently of amyloid pathology and may help explain why Alzheimer’s dementia in APOE4 carriers has been more difficult to treat with amyloid-targeted therapies alone.

The third hit is neuronal dysfunction, including impaired insulin receptor trafficking (APOE4 sequesters insulin receptors in early endosomes), reduced apoE receptor signaling important for synaptic plasticity, and mitochondrial vulnerability. Together, these three hits illustrate why clearing amyloid from an APOE4 carrier’s brain may not be sufficient to reverse or halt dementia: the biological disruption involves lipid pathways, immune activation, and vascular integrity in ways that extend well beyond what the amyloid hypothesis alone was designed to address.

What We Still Don’t Know—and Where the Science Is Heading

Three decades after the amyloid hypothesis was formalized, the field of Alzheimer’s research is confronting a more complex biological reality than the original framework anticipated. Evidence now supports that amyloid is an important upstream contributor to Alzheimer’s pathology—at least in the populations and disease stages studied to date. But several major uncertainties limit our ability to translate this understanding into broadly effective therapies.

A central open question concerns treating people before symptoms develop. The trajectory of Alzheimer’s pathology is now understood to unfold over 15 to 20 years before cognitive symptoms appear, with amyloid accumulating silently in the brain throughout midlife and late life. This raises a logical but practically challenging question: would removing amyloid at this pre-symptomatic stage—before tau pathology spreads and neurons begin to die—produce substantially larger clinical benefits? The A4 (Anti-Amyloid Treatment in Asymptomatic Alzheimer’s) trial tested this directly, enrolling cognitively normal older adults with elevated amyloid on PET imaging and treating them with solanezumab for 4.5 years. The result was negative—no significant slowing of cognitive decline—though solanezumab was a relatively weak amyloid-clearing agent. Ongoing trials are now testing more potent antibodies, including lecanemab, in cognitively normal individuals with elevated amyloid (the AHEAD 3-45 and A45 trials). The scientific rationale is sound, but the practical challenges are formidable. Because the base rate of progression from amyloid positivity to dementia is low in any given multi-year window, extremely large trials over many years are required to detect a meaningful difference—and most participants who receive treatment (and its risks) may not have developed dementia within the trial period regardless. Identifying who carries silent amyloid accumulation at scale requires either amyloid PET imaging, which is expensive and not universally accessible, or emerging blood-based biomarkers for Aβ and phospho-tau that are still being validated for broad clinical use. The ARIA risks from potent anti-amyloid antibodies in a population that currently has no symptoms—and no guarantee they would develop dementia—raises difficult questions about acceptable risk-benefit balance. Prevention is the intellectually compelling endpoint for the amyloid hypothesis, but achieving it equitably and safely remains one of the field’s most pressing unresolved challenges.

Beyond prevention, significant uncertainty remains about which forms of Aβ and tau are most toxic, and through what precise molecular mechanisms. We do not yet adequately understand how amyloid pathology interacts with cerebrovascular disease, neuroinflammation, and metabolic disruption in the complex mix of causes that characterizes most late-onset dementia. We do not know why amyloid appears to drive tau pathology with different efficiency across racial and ethnic populations, or whether ancestral differences in APOE4 haplotype structure (documented across African and Latin American populations) modify the downstream consequences of amyloid accumulation. We do not yet know whether therapies targeting the endolysosomal system, APOE4 lipidation, neuroinflammation, or cholesterol metabolism will demonstrate clinical efficacy—though first-generation therapies directly targeting APOE4 biology, including APOE4 structure correctors that convert APOE4 toward a less pathogenic conformation and lipidation-enhancing agents designed to restore ABCA1-mediated cholesterol transport, are now in early clinical testing. And we face a persistent practical dilemma: the therapies currently approved for disease modification are expensive, require intravenous infusion, carry meaningful safety risks, and provide modest average benefit that may not be detectable in individual patients.

Take-Home Messages

• The amyloid cascade hypothesis has now been clinically validated. Anti-amyloid clinical trials demonstrated that clearing Aβ from the brain slows cognitive decline and simultaneously lowers downstream phospho-tau biomarkers—supporting the position of amyloid as an upstream driver in the disease cascade.

• The clinical benefits of current anti-amyloid therapies are real but modest. Lecanemab and donanemab slowed decline by approximately 0.5–0.7 points on an 18-point scale over 18–20 months—below the threshold many clinicians consider perceptible to patients and caregivers. Both groups continued to decline. This likely reflects that by the time patients enroll in trials, the disease has progressed beyond amyloid dependence.

• Average trial results do not tell the whole story: some individuals may respond meaningfully. Population averages can obscure the fact that a subset of patients may derive substantially greater benefit from anti-amyloid therapy—particularly those treated earliest, with lower tau burden. Identifying who is most likely to respond is an active and critical area of research. These approvals represent the first disease-modifying treatments in Alzheimer's history and reflect decades of hard-won scientific progress that should not be minimized.

• Amyloid may be necessary but not sufficient for dementia. In late-onset, sporadic Alzheimer’s—which represents over 99% of cases—amyloid accumulation interacts with vascular disease, mixed pathology, neuroinflammation, and other factors that the amyloid hypothesis alone does not fully account for.

• Alzheimer’s dementia is not the same disease in all populations. Black participants showed a distinct relationship between amyloid, tau, and cognition compared to non-Hispanic White participants, suggesting that other biological or sociostructural factors may be primary contributors to cognitive decline in this group. The vast majority of foundational Alzheimer’s research has been conducted in predominantly White cohorts.

• Late-onset GWAS genes tell a story that extends beyond amyloid. Risk variants in APOE, BIN1, SORL1, PICALM, TREM2, and ABCA7 collectively implicate endolysosomal dysfunction and lipid metabolism as major disease pathways—suggesting that cellular waste clearance and lipid transport are as central to AD risk as amyloid dynamics, and open the path for new drug targets.

• APOE4 drives Alzheimer’s risk through at least three parallel mechanisms: disrupted lipid handling and HDL formation, neurovascular inflammation and blood-brain barrier disruption, and direct neuronal dysfunction. Anti-amyloid therapies may be less effective in APOE4 carriers partly because they do not address these additional pathways.

• Neuroinflammation is a central and potentially modifiable driver of disease, particularly in APOE4 carriers. Chronically activated microglia, primed by lipid dysregulation, may propagate tau pathology and neurodegeneration independently of ongoing amyloid accumulation.

• Prevention is the most scientifically compelling application of the amyloid hypothesis, but also the most difficult to achieve. Treating cognitively normal individuals with elevated amyloid requires very large, long trials; effective and accessible screening tools; and a careful reassessment of acceptable risk-benefit balance in a population that currently has no symptoms.

• The next generation of Alzheimer’s therapies will likely need to be combinatorial—addressing amyloid and tau alongside lipid metabolism, endolysosomal function, neuroinflammation, and cerebrovascular health. The validation of the amyloid hypothesis is a scientific milestone, but the path to broadly effective dementia prevention will require a substantially broader biological framework.

A 63-year-old woman came to see me with questions about hormone replacement therapy (HRT) and brain health. She recently found out that she has two copies of APOE4.

She began using a transdermal estrogen patch with micronized progesterone at age 56, around the time she entered menopause. Like many women at that stage, she was experiencing hot flashes, night sweats, and disrupted sleep. Her gynecologist prescribed hormone therapy and her symptoms improved significantly.

She has now been on therapy for seven years without complications.

Recently, however, she began noticing changes in memory, occasionally forgetting names or pausing to remember why she walked into a room. Nothing dramatic, but enough to make her concerned.

She came to the clinic with three straightforward questions:

• Is hormone therapy safe for the brain?

• Can it prevent dementia?

• How long should I stay on it?

These questions are becoming increasingly common as more women remain on hormone therapy longer and awareness of Alzheimer’s disease grows.

But answering them leads to a careful discussion on benefit versus risk.

A conversation with Dr. Howard Hodis

Shortly after seeing this patient, I discussed the case with Dr. Howard Hodis, Director of the Atherosclerosis Research Unit at the University of Southern California and one of the leading investigators studying the cardiovascular biology of menopausal hormone therapy.

Dr. Hodis has spent decades studying how estrogen interacts with vascular aging, particularly how the timing of hormone therapy relative to menopause may determine whether it helps or harms the cardiovascular system.

One of the key points from our conversation was that hormone therapy is often discussed as if it were a single treatment with a single risk profile.

Biologically, it is not.

The effects of estrogen depend on several factors:

• when therapy begins

• the health of the vascular system

• the hormone formulation

• the biology of the individual patient

And when the discussion shifts from the cardiovascular system to the brain, the uncertainty increases even further.

Why Hormone Therapy Exists

Hormone therapy was originally developed to treat symptoms of menopause.

As ovarian estrogen production declines, many women develop:

• hot flashes

• night sweats

• sleep disruption

• mood changes and depression

• vaginal dryness

• accelerated bone loss

Among available treatments, estrogen therapy remains the most effective therapy for vasomotor symptoms.

Hormone therapy also plays an important role in bone health. After menopause, bone loss accelerates rapidly. Roughly one in two women will experience an osteoporotic fracture during their lifetime. Hip fractures in particular can be devastating—about one-third of women die within a year, and many never fully recover functional independence.

Randomized trials show estrogen therapy reduces fractures by 30–50% in postmenopausal women.

For symptom relief and bone protection, the benefits of hormone therapy are well established.

The controversy arises when hormone therapy is discussed as a preventive therapy for chronic diseases, particularly cardiovascular disease and dementia.

Who Should Not Take Hormone Therapy

Although hormone therapy is highly effective for menopausal symptoms, it is not appropriate for everyone.

Most clinical guidelines recommend avoiding systemic estrogen therapy in women with certain conditions where risks may outweigh benefits.

These include:

• Estrogen-dependent cancers, particularly breast cancer

• A history of venous thromboembolism (blood clots) such as deep vein thrombosis or pulmonary embolism

• Stroke or significant cardiovascular disease

• Active liver disease

• Unexplained vaginal bleeding

The concern in these situations is that estrogen can influence coagulation pathways, hormone-sensitive tissues, and vascular biology, potentially increasing the risk of recurrent disease.

For women with a history of breast cancer, systemic hormone therapy is generally avoided because of concerns that estrogen could stimulate hormone-sensitive tumor cells.

Similarly, women with a history of blood clots may have increased thrombotic risk with oral estrogen therapy, although some studies suggest that transdermal estrogen may carry lower clotting risk.

Importantly, these recommendations apply primarily to systemic hormone therapy. In some cases, local vaginal estrogen preparations may still be considered because systemic absorption is minimal.

As with many areas of menopause care, decisions often require individualized risk assessment and shared decision-making between patient and clinician.

The Cardiovascular Context

One key fact about women’s health is often overlooked:

Cardiovascular disease is the leading cause of death in women.

Before menopause, women experience significantly lower rates of coronary heart disease than men. After menopause, that protection declines.

This observation led researchers decades ago to propose the estrogen cardioprotective hypothesis.

Early observational studies suggested women using hormone therapy had 30–50% lower rates of coronary heart disease.

To test this hypothesis more rigorously, researchers launched one of the largest clinical trials ever conducted.

The Women’s Health Initiative

The Women’s Health Initiative (WHI) randomized more than 27,000 postmenopausal women to hormone therapy or placebo.

Initial results suggested increased risks of:

• stroke

• blood clots

• breast cancer (with combined therapy)

These findings dramatically changed clinical practice.

However, an important detail emerged later: most participants were older, with an average age of 63, and had begun therapy more than a decade after menopause.

This raised a critical question:

Does the timing of hormone therapy matter?

The Timing Hypothesis

The timing hypothesis proposes that estrogen’s effects depend on the stage of vascular—and possibly other organ—health when therapy begins.

Early after menopause, when arteries are relatively healthy, estrogen appears to support vascular function through:

• improved endothelial function

• reduced LDL oxidation

• improved vasodilation

• reduced inflammation

But once atherosclerosis is established, estrogen may behave differently.

Experimental work suggests estrogen can increase matrix metalloproteinases such as MMP-9, enzymes involved in plaque instability.

In other words, estrogen may help prevent early vascular disease, but may not help—and could theoretically worsen—advanced disease.

Evidence for Timing

The ELITE trial (Early vs Late Intervention Trial with Estradiol) directly tested this hypothesis.

Women were randomized to estradiol or placebo and stratified according to time since menopause.

Women who began estradiol within six years of menopause showed slower progression of carotid atherosclerosis.

Women who began therapy more than ten years after menopause showed no benefit.

Figure. Effect of estradiol therapy on progression of carotid intima–media thickness (CIMT) according to time since menopause (ELITE Trial).
Mean carotid intima–media thickness (CIMT) as a measure of atherosclerosis measured over six years in women randomized to oral estradiol or placebo, stratified by time since menopause. In women in early postmenopause (<6 years since menopause), estradiol significantly slowed the progression of CIMT compared with placebo. In women in late postmenopause (≥10 years since menopause), estradiol did not reduce CIMT progression compared with placebo. These findings support the timing hypothesis, suggesting that the vascular effects of estrogen depend on when therapy is initiated relative to menopause. Data adapted from the Early vs Late Intervention Trial with Estradiol (ELITE)

Cardiovascular Outcomes

Meta-analyses of randomized trials have examined clinical outcomes.

Among women who started hormone therapy:

• before age 60, or

• within 10 years of menopause

studies show approximately:

• 30–40% reduction in coronary heart disease events

• ~30% reduction in all-cause mortality

A Danish randomized trial similarly found reduced cardiovascular events in recently postmenopausal women receiving hormone therapy for up to 16 years.

Hormone Therapy and the Brain

While much research has focused on cardiovascular disease, patients are often most concerned about brain health.

Estrogen influences several biological processes relevant to cognition:

• synaptic plasticity

• mitochondrial function

• cerebral blood flow

• inflammation

• neuronal metabolism

Laboratory studies suggest estrogen may support neuronal resilience and synaptic function.

This biological rationale led many researchers to believe hormone therapy might help prevent Alzheimer’s disease.

However, clinical trials have not confirmed that expectation.

Hormone Therapy and Dementia Prevention

The most important trial examining this question was the Women’s Health Initiative Memory Study (WHIMS).

WHIMS enrolled more than 7,000 women aged 65 and older to test whether hormone therapy could prevent dementia.

The results were unexpected.

Women receiving combined estrogen-progestin therapy had approximately twice the risk of dementia compared with placebo when therapy was started after age 65.

Importantly, this finding does not take into account the role of timing.

Later studies suggest that hormone therapy begun earlier in life does not increase dementia risk, but it also has not convincingly reduced it.

At present, the conclusion is straightforward:

There is no strong randomized evidence that hormone therapy prevents dementia.

APOE4 and Brain Vulnerability

The picture becomes more complex when genetics are considered.

The APOE4 allele is the strongest common genetic risk factor for Alzheimer’s disease.

APOE4 carriers often show:

• earlier amyloid deposition

• greater vascular dysfunction

• increased blood–brain barrier fragility

Importantly, not all APOE4 carriers are the same.

A cognitively normal APOE4 carrier may still have significant neural reserve.

But once someone develops mild cognitive impairment or dementia, the brain becomes far less resilient.

Estrogen, MMP-9, and Neurovascular Biology

One pathway linking estrogen to both cardiovascular and neurological disease involves matrix metalloproteinase-9 (MMP-9).

MMP-9 regulates:

• extracellular matrix remodeling

• vascular stability

• blood–brain barrier permeability

Elevated MMP-9 activity has been associated with:

• plaque rupture in cardiovascular disease

• blood–brain barrier breakdown

• neuroinflammation

Estrogen signaling can influence MMP-9 activity, suggesting that estrogen’s effects may differ depending on the stage of vascular or neurodegenerative disease.

The Problem of Uncertainty

Despite decades of research, hormone therapy remains one of the most debated topics in medicine.

Several factors contribute to this uncertainty.

Different studies examine different populations. Many randomized trials studied women who began therapy long after menopause, while clinical practice typically begins therapy earlier.

Hormone formulations also differ. Earlier trials studied oral conjugated estrogens and synthetic progestins, whereas modern regimens often use transdermal estradiol and micronized progesterone.

Most hormone trials were designed to study cardiovascular disease and fractures, not dementia.

And despite the major role of APOE genetics in Alzheimer’s disease, very few studies have stratified hormone therapy outcomes by genotype.

Current evidence on hormone therapy and dementia has important gaps, particularly for APOE4 carriers. Most randomized trials were not designed to study genetic subgroups and rarely stratified participants by APOE genotype. In addition, many trials initiated hormone therapy years after menopause, limiting their ability to test the timing hypothesis in the brain.

As a result, we still lack trials that start hormone therapy near menopause, follow women long-term, and specifically evaluate cognitive outcomes in APOE4 carriers. Studies using modern hormone formulations and neuroimaging or biomarker endpoints are also limited. These gaps mean that the effects of hormone therapy on dementia risk, especially in genetically vulnerable populations, remain uncertain.

Taken together, these limitations mean that hormone therapy decisions still require clinical judgment rather than simple rules.

Returning to the Patient

After discussing the science, the question comes back to the woman who started this conversation.

Like many medical interventions, the value of hormone therapy depends on several factors:

• timing

• biology

• individual risk factors

• patient goals

For my patient, the answer is not straightforward.

Hormone therapy likely helped her menopausal symptoms, and, given that she started treatment near menopause and has tolerated it well, it may still be reasonable to continue.

But when it comes to protecting the brain from dementia, science has not yet provided a definitive answer.

New memory concerns should prompt a formal cognitive assessment and a thoughtful discussion about whether continuing hormone therapy remains appropriate.

Take-Home Messages

Hormone therapy remains the most effective treatment for menopausal symptoms.

• Timing matters. Starting hormone therapy near the menopausal transition appears biologically different from starting it decades later. Much of the risk seen in earlier trials occurred in women who initiated therapy long after menopause, highlighting the importance of timing.

• Cardiovascular effects depend on vascular health. Estrogen may support vascular function in relatively healthy arteries, but its effects may differ once atherosclerosis is established, potentially altering plaque biology.

• Hormone therapy is primarily a treatment for menopausal symptoms. Its strongest evidence supports relief of vasomotor symptoms and prevention of bone loss. Current evidence does not support using HRT as a general longevity or dementia-prevention therapy.

• Cognitive symptoms warrant reassessment. When mild cognitive impairment or dementia emerges, there is less brain resilience, which may alter the risk–benefit balance, and continued therapy should be reassessed with a clinician.

• Genetics may modify risk. Factors such as the APOE4 genotype, along with vascular health and baseline cognitive status, may influence how hormone therapy affects the brain, though this area remains incompletely studied.

• Ultimately, hormone therapy should not be viewed as either a miracle therapy or a dangerous drug.

Until the science becomes clearer, the most important step is not a universal rule, but careful conversations between patients and their physicians, weighing symptoms, risks, and individual goals.

What brought him in was a lab value that did not fit his lifestyle: LDL cholesterol 175 mg/dL. He had read conflicting messages, including claims that statins might worsen memory. He wanted to understand whether he should do something about his LDL, and how to think about that decision in the context of APOE4/4 and dementia risk.

So the question became:

If an APOE4/4 carrier has elevated LDL but no dementia, does lowering LDL with a statin meaningfully help brain health, and is there evidence that statins harm cognition?

This post is not medical advice. It is a way to read the evidence with the patient’s real-world decision in mind.

1) APOE4, high cholesterol, and ischemic heart disease: the “why” behind the lab

APOE is not just an “Alzheimer’s gene.” It is also a lipid transport gene. One practical consequence is that APOE4, especially APOE4/4, shifts lipid handling in ways that can raise LDL, even in people who eat well and exercise consistently.

Mechanistically, APOE sits on lipoprotein particles and helps them interact with hepatic receptors involved in clearance. In a randomized human study using a surface plasmon resonance approach plus hepatocyte uptake assays led by Anne Marie Mihihane and her team, APOE4 carriers showed altered lipoprotein–LDL receptor binding and downstream handling consistent with a propensity toward higher LDL under certain dietary fat conditions. This is one concrete example of how genotype can constrain physiology even when behavior is optimal.

The next step in the logic is vascular risk. Large population data support that APOE genotypes associate not only with dementia risk but also with ischemic vascular disease. For example, in UK Biobank (about 391,992 white British participants), APOE genotypes were associated with ischemic heart disease and relevant vascular risk factors, with effects that vary by genotype, age, and sex.

That matters for brain health because vascular injury, clinical stroke and “silent” small vessel disease, can lower cognitive reserve and amplify neurodegenerative trajectories. If someone is APOE4/4 and carrying a high LDL burden over time, it is reasonable to ask whether LDL lowering could reduce downstream vascular contributions to cognitive risk, even if it does not directly target Alzheimer pathology.

This is the first frame shift for the patient: his LDL cholesterol level may be partially genetic, and the brain relevance may be vascular as much as, or more than, direct Alzheimer prevention.

2) Genetic epidemiology signals: direction-of-effect, not proof

When randomized prevention trials are difficult, researchers often turn to genetic epidemiology to test whether a risk factor plausibly sits upstream of disease.

One example used ADNI data (N = 1,534 across cognitively normal, early MCI, late MCI, and AD) and performed a Mendelian randomization (MR) analysis using APOE as a genetic instrument to examine whether higher total cholesterol might contribute to risk along the clinical spectrum. They reported that higher total cholesterol was associated with higher odds of MCI and AD diagnoses in APOE4-positive versus APOE3 carriers in their MR framework.

How should we treat this?

• It is not a statin trial.

• MR is best read as a direction-of-effect argument: lipid biology is plausibly upstream of risk, particularly in APOE4.

• APOE is biologically pleiotropic, so MR assumptions can be strained. The takeaway is plausibility, not certainty.

Next comes the observational pharmaco-epidemiology signal (association, not proof): In the Chicago Health and Aging Project (CHAP; N = 4,807), statin initiation was associated with a lower hazard of incident clinical Alzheimer’s disease overall, and the association appeared stronger in APOE ε4 carriers (HR about 0.60 in ε4 carriers vs about 0.96 in non-carriers; interaction reported).

The figure above shows the covariate-adjusted cumulative incidence of AD Dementia, by statin initiation and the APOE e4 Allele

This type of result motivates clinicians and patients while also raising methodological caution, because healthy-user bias, confounding by indication, and time-varying treatment patterns are hard to fully eliminate.

So the honest interpretation is: there are credible signals suggesting subgroup benefit (notably in APOE4), but they are not definitive.

3) Why so many clinical trials are null, and why that may not settle prevention

If you only looked at randomized trials in people with established Alzheimer’s disease, you might walk away thinking the case is closed: statins do not help.

Two large symptomatic AD trials illustrate the point:

• LEADe randomized 640 patients with mild-to-moderate AD to atorvastatin 80 mg/day vs placebo for 72 weeks and found no significant benefit on cognition (ADAS-Cog) or global function (ADCS-CGIC).

• A separate randomized trial in 406 patients with mild-to-moderate AD using simvastatin vs placebo similarly reported no meaningful differences in cognitive or functional decline over 18 months.

These are important results, but they answer a treatment question (does this help after diagnosis) more than a prevention question (does decades of lower LDL reduce the probability of reaching symptomatic disease).

Why prevention is hard to prove in standard randomized clinical trials:

• Dementia unfolds over years to decades, so trials need long follow-up and large samples.

• People may cross over: they start and stop statins, and the cognitive risk benefit may change if the change diets, change blood pressure meds.

• Cognitive endpoints can be relatively insensitive over short windows, especially in healthy or mildly impaired individuals.

• If a therapy primarily reduces vascular events, the cognitive signal may be indirect and diluted, especially if neurodegeneration is already advanced.

So the null symptomatic trials are informative, but they do not fully answer what our patient cares about: risk modification earlier in life. One important finding from these trials is that even in patients with AD, statins do not worsen memory (more on that below). Newer trials for brain outcomes using statins should consider more brain biomarkers than cognition during the predementia phases, and especially those that reflect the effects of statins on the brain.

4) Debunking the idea that statins “worsen memory”

Patients commonly report cognitive symptoms while taking statins, but the key scientific question is whether statins cause clinically meaningful cognitive decline in controlled data.

A useful anchor here is that large randomized trials that included cognitive testing have generally not shown a consistent cognitive harm signal. For example, in PROSPER (5,804 participants aged 70 to 82 randomized to pravastatin vs placebo), detailed cognitive outcomes were assessed repeatedly and did not support a clinically meaningful adverse effect on cognition attributable to pravastatin.

A practical way to say this in clinic is: the best available evidence does not support the claim that statins generally worsen memory. If a specific patient experiences cognitive symptoms temporally associated with a medication, that should be evaluated carefully, but it should not be assumed to be the expected or typical effect.

5) Lipophilic and hydrophilic statins: why the distinction persists

Statins differ in chemical properties that influence distribution. Lipophilic statins are often discussed as having greater potential CNS penetration, while hydrophilic statins are generally more hepatoselective and do not enter the brain.

Does this translate into meaningfully different dementia outcomes? The evidence is not definitive, but large observational datasets have explored the question.

One widely cited example is a retrospective Medicare cohort of 694,672 beneficiaries that compared dementia risk across combinations of statins and antihypertensive classes. The authors reported that combinations involving pravastatin or rosuvastatin (hydrophilic statins) with renin-angiotensin system acting antihypertensives, particularly ARBs, were associated with lower dementia risk compared with some other combinations.

This is not causal proof, but it supports a reasonable scientific posture: drug class and combinations are plausible modifiers, and the question deserves prospective testing.

6) Uncertainties that remain, especially once Alzheimer’s disease is diagnosed

Two ideas should be kept separate:

• Risk modification: lowering LDL and reducing vascular events, which plausibly supports brain resilience.

• Disease modification: changing core Alzheimer’s pathology trajectories (amyloid, tau, neurodegeneration).

The large symptomatic AD trials above were largely null on cognition, which argues against presenting statins as established disease-modifying Alzheimer therapies once dementia is present.

For an APOE4/4 carrier without dementia, the strongest rationale remains vascular risk reduction. Whether LDL-lowering meaningfully changes dementia incidence is still uncertain, in part because it is difficult to run long, well-powered prevention trials for a peripheral target. The implication of this analysis makes us question APOE4 carriers with low LDL levels. Would they benefit from statin treatment if they had a lower cardiovascular risk profile?

Back to the case: risk-benefit framing

For this 65-year-old APOE4/4 man with LDL 175 mg/dL and no dementia, the most evidence-based framing is vascular.

• APOE4/4 is associated with higher ischemic heart disease risk in large population data.

• Elevated LDL is a modifiable vascular risk factor, and vascular events are relevant to brain health.

• The best controlled evidence does not support a general claim that statins worsen cognition.

So, should he take a statin?

What can be said responsibly, based on the evidence reviewed, is that an LDL of 175 mg/dL at age 65 commonly triggers a clinician-patient discussion about lipid-lowering for vascular risk reduction. APOE4/4 and family history make the prevention context more salient, but they do not turn statins into a proven dementia-prevention drug.

The decision belongs in a discussion with his clinician that incorporates his overall cardiovascular risk profile, family history, preferences, and tolerance, while keeping expectations calibrated: the clearest benefit is vascular, the long-term dementia prevention-related benefit is plausible but not proven.

Take-home messages

• APOE4/4 can raise LDL through lipid handling biology, even when lifestyle is excellent.

• APOE genotypes, including ε4ε4, are associated with ischemic heart disease risk in large population data.

• Genetic and observational studies provide plausible signals that lipid biology and statin exposure may matter more in APOE4, but they are not definitive prevention proof.

• Large symptomatic Alzheimer’s trials of statins have been largely null, which does not settle prevention questions.

• The best controlled evidence does not support the claim that statins generally worsen memory.

• Whether to start a statin should be decided in discussion with a clinician, weighing vascular risk, LDL level, patient values, and tolerance, without assuming statins are a proven dementia-prevention therapy.

• The benefit of statin initiation in those with low cardiovascular risk or low LDL levels for AD is uncertain.

Repeatedly, one enzyme emerged as central to that biology: calcium-dependent phospholipase A2 (cPLA2).

The remaining question was both anatomical and clinical:

Is cPLA2 activated at human synapses—the structures most closely linked to cognition—and is that activation associated with impairment?

Our recent study was designed to address that question directly. Dr. Qiulan Ma led this study at USC over the past 3 years.

What Is a Synaptosome?

Synapses are the specialized junctions where neurons communicate. Synaptic density and integrity are among the strongest pathological correlates of cognitive function in Alzheimer’s disease.

Because living human synapses cannot be directly studied, we isolate synaptosomes from postmortem brain tissue. Synaptosomes are subcellular particles enriched in intact presynaptic and postsynaptic components. They preserve synaptic membranes, receptors, scaffold proteins, and signaling enzymes.

Figure 1. Synaptosome markers and corresponding human synaptic ultrastructure.
Left: Simplified schematic of a human synaptosome illustrating preserved presynaptic and postsynaptic components. The presynaptic membrane (upper thick dark line) contains SVs (synaptic vesicles) and synaptic proteins including synapsin, as well as mitochondria that support neurotransmission. The synaptic cleft is the extracellular space between the pre- and postsynaptic membranes. The postsynaptic membrane (lower thick dark line) contains key excitatory synaptic markers including NMDARs (N-methyl-D-aspartate receptors), AMPARs (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors), PSD-95 (postsynaptic density protein 95), and CaMKIIα (calcium/calmodulin-dependent protein kinase II alpha), which are central to synaptic signaling and plasticity.

Right: Representative transmission electron microscopy images of isolated human synapses from brains with low and high Alzheimer’s disease neuropathologic change (ADNC). Labeled structures include Mito (mitochondria), SVs (synaptic vesicles), PSD (postsynaptic density), presynaptic membrane, postsynaptic membrane, and the synaptic cleft. Scale bar = 0.5 μm.Although not complete neurons, they retain the biochemical machinery of synapses and allow direct measurement of synaptic protein levels and lipid mediators.

Using synaptosomes from participants in a longitudinal aging cohort with detailed cognitive assessments prior to death (Religous Order Study), we examined cPLA2 levels, downstream lipid signaling, and their relationship to cognition.

Selective Synaptic Vulnerability

Synaptosome yield was significantly reduced in individuals with mild cognitive impairment (MCI) and Alzheimer’s dementia compared with cognitively normal controls. In contrast, glial- and myelin-enriched fractions were not similarly reduced.

Fig. 2 Reduction of synaptosome yield in MCI and AD cases.

This finding supports selective synaptic vulnerability during disease progression and confirms that synaptic loss is detectable even at the MCI stage.

Elevated cPLA2 at Human Synapses

We measured two isoforms: cPLA2α and cPLA2β.

Both were elevated in synaptosomes from Alzheimer’s dementia cases. Notably, cPLA2β was already elevated in MCI.

Fig. 3 Elevation of cPLA2 in synaptosomes of AD.

This suggests that synaptic cPLA2 activation is not restricted to advanced disease and may emerge during earlier clinical phases.

Association With Cognitive Performance

Higher synaptosomal cPLA2β levels were associated with worse global cognition and episodic memory after adjusting for age, sex, and education. These associations were particularly strong in male participants.

Fig. 4 Correlation between synaptosomal cPLA2 levels and cognitive dysfunction. a. p-cPLA2a (all). b. p-cPLA2a (male participants). c. cPLA2b (all). d. cPLA2b (male participants). Residuals were calculated from linear regression models in which cognitive function was regressed on age at last visit, sex and education. For sex-based analyses, residuals were calculated from models regressing cognitive function on age at last visit and education. *P < 0.05, **P < 0.01. ****P < 0.0001. ROS Cohort.

These findings are correlational and do not establish causation. However, they demonstrate that synaptic cPLA2 levels closely align with clinical impairment in humans.

Biochemical Evidence of Pathway Activation

cPLA2 hydrolyzes membrane phospholipids to release arachidonic acid (AA), which is then converted into eicosanoids—prostaglandins that regulate inflammatory signaling.

In Alzheimer’s synaptosomes, we observed:

• Increased arachidonic acid

• Increased AA-derived eicosanoids

• Strong correlations between cPLA2 levels and AA metabolites

Other lipid pathways, including cholesterol metabolism, were not similarly altered. This supports the specificity of the cPLA2–arachidonic acid pathway.

Fig 5. Alterations of AA and AA metabolites but not cholesterol and oxysterols in AD synaptosomes. a, b. Heatmap generated from lipidomic analysis of PUFAs and PUFAs’ metabolites in synaptosomes comparing AD vs. NCI by Wilcox rank sum test. The color represents the Wilcox test estimate, where redness indicated the increased lipid species in AD, and vice versa. Overall, relative to NCI, AD showed high arachidonic acid (AA) and AA metabolites, including increased TXB2 and PGF2a, which were highly corelated with cPLA2a (c). AA, PGD2, PGE2, and adrenic acid were corelated with both cPLA2a (c) and cPLA2b (d). Cholesterol and oxysterols, including 7-dehydrocholesterol (7-HC), 25-hydroxycholesterol (25-HC), and 7a, 25-hydroxychlesterol (7a, 25-HC), were not changed (e). f. A diagram of cPLA2 activation involved AA metabolic pathway. c-d Graphs show Pearson correlation between two variables, as described on the axis, *P < 0.05, **P < 0.01, ****P < 0.0001. ROS cohort.

These findings indicate that synaptic cPLA2 is not only elevated but enzymatically active.

Excitotoxicity and Postsynaptic Localization

Excitotoxicity refers to neuronal injury resulting from sustained activation of excitatory glutamate receptors and excessive calcium influx. Elevated intracellular calcium activates downstream enzymes that destabilize synaptic structure.

cPLA2 is calcium-dependent.

In human iPSC-derived neurons, amyloid-β oligomers increased phosphorylation of cPLA2 and induced its redistribution to dendritic membranes. Activated cPLA2 colocalized with PSD-95 and CaMKIIα—proteins central to postsynaptic organization and synaptic plasticity.

Inhibition of cPLA2 suppressed amyloid-induced synaptic alterations in these models.

These observations support a mechanistic framework in which pathological calcium signaling activates cPLA2 at postsynaptic sites, leading to localized inflammatory lipid release and synaptic destabilization.

Interpretation and Uncertainty

Several limitations are essential to acknowledge:

Correlation does not establish causation.
Elevated synaptic cPLA2 may contribute to dysfunction, but it may also represent a downstream response to amyloid, tau, or other pathological processes.

Cross-sectional human data cannot determine temporal sequence.
We cannot establish whether synaptic cPLA2 activation precedes cognitive decline or follows it.

Experimental models simplify disease biology.
Cell-based systems allow mechanistic interrogation but do not replicate the full complexity of aging human brain networks.

Physiological roles of cPLA2 must be preserved.
cPLA2 participates in normal membrane remodeling and signaling. Excessive inhibition could disrupt essential processes.

These uncertainties require that translational claims remain cautious.

Why These Findings Support Testing a Disease-Modifying Hypothesis

A disease-modifying therapy alters the biological trajectory of a disease rather than temporarily improving symptoms.

The present study demonstrates that:

• cPLA2 activation occurs directly at human synapses

• Synaptic cPLA2 levels are associated with cognitive impairment

• The arachidonic acid inflammatory pathway is upregulated at those sites

• Amyloid-induced cPLA2 activation and synaptic alterations can be pharmacologically suppressed in human neurons

In our prior inhibitor study, selective brain-penetrant cPLA2 inhibition reduced amyloid-induced tau hyperphosphorylation and preserved synaptic markers in human neuronal systems. Those findings established that this pathway is modifiable and that its modulation affects downstream tau and synaptic biology.

Taken together, the data support a specific mechanistic proposition:

Pathological activation of cPLA2 may represent a convergence point linking lipid dysregulation, excitotoxic signaling, tau phosphorylation, and synaptic loss.

If correct, then selective modulation of cPLA2 activity—particularly during early clinical stages—could alter the rate of synaptic deterioration. That possibility meets the conceptual definition of a disease-modifying strategy.

This hypothesis has not yet been tested in clinical trials. It requires rigorous evaluation of safety, timing, and long-term impact.

However, the convergence of human synaptic data and pharmacologic modulation provides a biologically coherent framework for such testing.

Conclusion

This study establishes that cPLA2 activation occurs at human synapses in Alzheimer’s disease and is associated with cognitive impairment.

It connects earlier observations of lipid imbalance to a defined synaptic mechanism and links that mechanism to tau and excitatory signaling pathways known to drive neurodegeneration.

These findings do not establish therapeutic efficacy. They provide mechanistic support for evaluating whether selective, brain-penetrant modulation of cPLA2 can slow synaptic decline and modify disease progression.

The next step is careful translational testing.

Reference:

Ma, QL., Ebright, B., Li, B. et al. Evidence for cPLA2 activation in Alzheimer’s disease synaptic pathology. acta neuropathol commun (2026). https://doi.org/10.1186/s40478-025-02214-6

Funding:

This work was support by grants from by the National Institute on Aging. R01AG076124 (HNY, ZA), P30AG066530 to HNY; P30AG066530 subaward (QLM); R21AG089611(QLM) P30AG066530 to Neuropathology Core, R01AG070255 (AL), ROS is supported by P30AG10161, P30AG72975, and R01AG15819, and U01AG094622 to HNY and SL.

ROS resources can be requested at

https://www.radc.rush.edu

and www.synpase.org.

He was worried for understandable reasons. Both of his parents developed dementia in their 80s. He had also undergone neuropsychological testing because he noticed cognitive changes. The testing showed objective mild cognitive impairment (MCI). Genetic testing showed one copy of APOE ε4.

After that evaluation, he started a ketogenic protocol that was presented to him as a way to prevent dementia. He followed it consistently for a full year.

Over that year, he lost ~25 pounds, and his BMI is now ~19.

His question was straightforward:

“Given my genetics and MCI, is staying in ketosis helping my brain—or am I pushing my body into a state that could backfire?”

This is the central tension: ketosis is often discussed as a brain-fuel strategy, but in real life it frequently comes bundled with weight loss, and weight loss can be biologically meaningful in older adults—especially those already symptomatic.

What ketosis is actually trying to accomplish

The ketogenic argument for brain health rests on a plausible idea: the brain can use ketones, and in Alzheimer’s disease brain glucose metabolism is impaired. Raising ketones might provide alternative substrate support.

The clinical question is not whether ketones can be used. It is whether raising ketones translates into meaningful, durable benefit in the population in front of us—older, symptomatic, and APOE ε4-positive—and whether the cost of maintaining ketosis (often weight loss and physiologic stress) is acceptable.

A second constraint matters in APOE ε4 and later-stage disease: there is evidence and mechanistic concern that ketone delivery and/or utilization in the brain may be impaired due to blood–brain barrier changes, transporter limitations, and mitochondrial dysfunction.

What human intervention studies suggest (focusing on larger samples)

Ketone-raising interventions in Alzheimer’s dementia: inconsistent efficacy, genotype signal

A randomized trial in 152 patients with mild-to-moderate Alzheimer’s disease tested a ketone-raising compound (AC-1202). The cognitive signal was reported primarily in APOE ε4–negative participants, with less evidence of benefit in ε4 carriers despite ketone elevation.

The figure above shows the mean change from baseline in ADAS-Cog score (which measures the severity of memory, language, and thinking problems in people with dementia, with a higher score indicating worse performance) over 90 days in participants randomized to AC-1202 (red) or placebo (blue). Assessments were performed at baseline (Day 0), Day 45, and Day 90. Panel A: all genotypes combined. Panel B: APOE ε4 non-carriers (APOE4−). Panel C: APOE ε4 carriers (APOE4+). Values represent mean ± error bars (as reported in the original study). The y-axis is oriented so that movement upward (more negative change) reflects clinical improvement (lower ADAS-Cog score), and downward reflects worsening. Asterisks indicate statistically significant differences between groups at the specified time point (typically p < 0.05).

A later multi-site trial using a related formulation (AC-1204) in 413 patients did not show overall cognitive benefit, with practical issues such as attrition and formulation/bioavailability discussed as limiting factors.

A careful, clinically honest read of this literature is: ketone-raising strategies can produce symptomatic signals in some settings, but the evidence is not consistent enough to treat ketosis as a broadly effective therapy for established dementia—and APOE genotype likely influences response.

Ketone-based intervention in MCI: a more plausible window, still not definitive

In MCI, ketone interventions may have a more plausible biologic window. A 6-month randomized trial in 82 participants with MCI using a ketogenic medium-chain triglyceride intervention reported cognitive benefits alongside evidence of increased ketone availability.

This matters for the case because the patient has MCI, not dementia. But it does not settle the question of long-term outcomes, and it does not answer whether sustained nutritional ketosis is required (or safe) in an older, lean APOE ε4 carrier.

The part that changes the risk–benefit analysis in this case: catabolic vulnerability

This case is not “ketosis in the abstract.” It is ketosis plus substantial weight loss in a 70-year-old with MCI.

In older adults, significant weight loss can reflect or contribute to:

• lower physiologic reserve,

• loss of lean mass and strength,

• vulnerability to illness and stressors,

• and (in some cases) a prodromal neurodegenerative trajectory.

Several large human datasets link late-life weight loss with higher dementia risk, and APOE ε4 can modify those relationships. For example, a long-term cohort of 1,462 women reported that APOE ε4 carriers with greater late-life weight loss had higher dementia risk.

Interventional data add nuance. In Look AHEAD (5,145 adults with type 2 diabetes and overweight/obesity), subgroup analyses suggest that cognitive effects of intensive lifestyle weight loss are not uniform across groups, and have been interpreted in the literature as consistent with the idea that weight loss may deprive some APOE ε4 brains of a relevant energy buffer.

None of this proves that weight loss causes dementia. But it supports a practical clinical posture:

In an older APOE ε4 carrier who is already symptomatic, substantial weight loss—especially to low BMI—should not be assumed to be beneficial for brain health.

Mechanisms: why APOE ε4 may increase sensitivity to catabolism

The mechanistic concern is not “thin is bad.” It is that APOE ε4 may be associated with differences in metabolic handling and adipose biology that reduce buffering capacity during stress. One mechanism includes the blunted PPARγ-related responses and adipose remodeling differences in APOE ε4 contexts.

Add one more layer: if later-stage disease limits brain ketone uptake/oxidation (BBB/transport/mitochondria),

then a patient may incur the systemic costs of restriction and weight loss while receiving less of the intended energetic benefit.

That is the catabolic vulnerability problem in one sentence.

Chronic ketosis can be physiologically stressful (especially in lean older adults)

Ketosis is not inherently harmful, but chronic nutritional ketosis is an active physiologic state with predictable stress points:

• early diuresis (water/sodium shifts),

• electrolyte needs,

• and potential for muscle loss if protein and total energy intake are not adequate.

Early weight loss contributions include glycogen-linked water loss and ketone-associated natriuresis/diuresis, with clinical concerns such as dehydration and muscle loss risk with ketogenic dieting—along with mitigation strategies.

There is also evidence framed as a biphasic oxidative stress response (initial rise in oxidative stress markers with later adaptive signaling), consistent with a hormetic model.

Clinically, the relevance is straightforward: a physiologic stressor can be adaptive in one person (younger, metabolically impaired, high reserve) and costly in another (older, low BMI, already cognitively vulnerable).

Long-term feasibility and the LDL question in APOE ε4

Even if ketosis were metabolically helpful short term, long-term lifestyle ketosis raises two practical issues: adherence and cardiometabolic tradeoffs.

A 12-month randomized trial (n=118) comparing very-low-carbohydrate versus low-fat diets illustrates lipid tradeoffs over longer duration.

Across longer-duration comparisons, LDL-C can rise on very-low-carbohydrate ketogenic patterns in aggregate (with large meta-analytic datasets reporting LDL increases).

For APOE ε4 carriers, LDL is not a side detail. APOE ε4 is associated with higher LDL levels with diet-response considerations relevant to saturated fat intake and lipid handling.

So, even if a patient experiences some symptomatic benefit, a long-term pattern that worsens LDL may introduce competing vascular risk—highly relevant to brain aging.

Symptomatic benefit vs disease modification

This distinction should be explicit.

Most ketosis/ketone intervention studies are designed to test symptomatic outcomes (cognitive scales, function) over months.

A disease-modifying claim would require evidence that an intervention changes core pathobiology trajectories. At present, it is fair to say we do not have strong evidence that nutritional ketosis reduces phosphorylated tau or amyloid levels in humans or clearly modifies Alzheimer’s disease course over long horizons.

That does not rule out benefit. It simply sets the appropriate scientific boundary for “prevention” claims.

Who may benefit from ketosis—and who should be cautious?

This is where the discussion should land: matching the tool to the patient.

More plausible benefit profile

Ketosis (or other structured carbohydrate restriction) is most defensible when the primary target is a metabolic disorder:

• obesity,

• insulin resistance,

• type 2 diabetes,

• hypertriglyceridemia/metabolic syndrome.

In that setting, short-term metabolic improvements are consistently reported, while longer-term differences often attenuate and require careful attention to adherence and lipid response.

If a patient is younger, metabolically at risk, and not yet demented, the risk–benefit balance can be favorable—especially if weight loss is therapeutic rather than destabilizing.

Higher caution profile

For an older adult with established cognitive impairment and low BMI or ongoing weight loss, particularly with APOE ε4:

• the downside of catabolism is higher,

• the uncertainty around brain ketone utilization is greater,

• and long-term lipid tradeoffs matter more.

Take-home messages

• Ketosis can be helpful for some metabolically at-risk people—but it is not yet a proven disease-modifying tool for Alzheimer’s pathology.

• In older adults with MCI, low BMI should be treated as a clinical risk signal, and the cost-benefit from ketosis should be discussed with a clinician.

• APOE ε4 may shift the cost–benefit balance because of metabolic buffering differences and because ketone utilization in later-stage disease may be constrained.

• Chronic ketosis can be physiologically stressful (fluid/electrolyte shifts, potential muscle loss), which matters more in lean older adults.

• Long-term ketosis requires metabolic vigilance—especially monitoring muscle mass, metabolic stress signals and LDL cholesterol levels.