|

Evidence Watch, September 2026: The APOE4 and Alzheimer’s Papers That Matter, Graded

Evidence Watch Physician-curated · All audiences

Evidence Watch, September 2026
The APOE4 and Alzheimer’s papers that matter, graded

Each edition of Evidence Watch selects the most important new peer-reviewed APOE ε4 and Alzheimer’s research, grades the strength of each study, and explains what changed, why it matters, and what is worth reading in full. This edition covers papers indexed in PubMed from mid-2025 through September 25, 2026. Systematic reviews, meta-analyses, and randomized trials are listed first, followed by large human studies, then mechanistic work.

How to read the grades. High randomized trial or meta-analysis of randomized trials. Moderate large prospective human cohorts, pooled participant-level analyses, meta-analyses of observational studies, or large genetic studies. Hypothesis-generating human-tissue, stem-cell, or animal work, or subgroup findings that need confirmation. Red flags mark important caveats. The full method is on the Evidence Watch page.

What changed this month

  1. ARIA risk tracks APOE ε4 dose, in trials and in routine care. Real-world lecanemab data reproduce the trial pattern: about 1.5 times the ARIA risk with one copy and about 3.5 times with two.
  2. Whether anti-amyloid antibodies work less well in carriers is unresolved. Two 2025 syntheses reach different conclusions.
  3. The only Phase 3 trial run exclusively in ε4/ε4 homozygotes (APOLLOE4) missed its primary endpoint. Signals in the MCI subgroup and in blood biomarkers are hypothesis-generating.
  4. High-dose DHA reached the central nervous system but did not change cognition or brain volume over 24 months in the PreventE4 randomized trial.
  5. The same plasma p-tau217 level signals more near-term risk in ε4 carriers (pooled analysis in Lancet Neurology).
  6. Vascular biology is moving to the center of APOE4 research, with two new studies converging on fibronectin and blood-vessel scarring.

Tier 1: Meta-analyses and randomized trials

The strongest designs available. Read the caveats: a strong design can still answer a narrow question.

Anti-amyloid therapy · 1 of 13

Anti-amyloid antibodies by patient subgroup: carriers face more ARIA, and possibly less benefit

Shim GH, Lau ECY, Huynh ALH, Lu CY, Tan ECK. Ageing Research Reviews. 2025;114:102981. doi:10.1016/j.arr.2025.102981

High · safety meta-analysisEfficacy by genotype: descriptive only

What it found. This systematic review covered 16 reports from six randomized trials of lecanemab and donanemab (5,633 participants). Compared with non-carriers, APOE ε4 carriers had 2.19 times the risk of ARIA-E, meaning brain swelling (95% CI 1.91–2.50), and 3.45 times the risk of ARIA-H, meaning microbleeds (95% CI 1.35–8.72). The largest slowing of decline was reported in non-carriers and in White participants. Lecanemab improved quality of life in heterozygotes and in people aged 65–74.

Why it matters. The safety estimates are pooled and robust. The efficacy comparisons, however, summarize trial subgroups rather than formally pooling them, so they suggest but do not prove smaller benefit in carriers.

Read it if you are weighing lecanemab or donanemab and want the subgroup data in one place.

Anti-amyloid therapy · 2 of 13

Network meta-analysis: antibodies vs. cholinesterase inhibitors, with no efficacy difference by carrier status

Hsu CW, Hsu TW, Kao YC, et al. Journal of Prevention of Alzheimer’s Disease. 2025;12(8):100195. doi:10.1016/j.tjpad.2025.100195

High · with caveatsIncludes aducanumab · indirect comparison

What it found. This Bayesian network meta-analysis pooled seven antibody trials (8,010 participants) and nine cholinesterase-inhibitor trials (4,993 participants). Against placebo, antibodies slowed decline on the CDR-SB by 0.41 points (95% credible interval −0.61 to −0.22) and on the ADAS-Cog by 1.35 points. Neither reached the minimally important difference (1 and 2 points respectively). Efficacy did not differ by APOE4 carrier status, but homozygotes had 5.53 times the odds of ARIA-E compared with non-carriers (95% credible interval 2.48–13.07).

Why it matters. This result contradicts the Shim review on efficacy. Taken together, the two papers show that whether carriers benefit less is still an open question. Both agree on higher ARIA risk.

Read it if you want a realistic sense of effect size against the minimal difference patients can notice.

Anti-amyloid therapy · 3 of 13

Lecanemab safety in trials and routine care: a clear gene-dose effect for ARIA

Qi L, Zheng F, Tu M, et al. Journal of Prevention of Alzheimer’s Disease. 2026;13(3):100473. doi:10.1016/j.tjpad.2025.100473

Moderate–High · RCT + real-world pooling

What it found. This PROSPERO-registered meta-analysis combined two randomized trials with five real-world studies (1,576 patients). Pooled ARIA incidence was 19% (95% CI 16–23%), and symptomatic ARIA occurred in 3%. Compared with non-carriers, the relative risk of ARIA was 1.45 for heterozygotes and 3.54 for homozygotes. Infusion reactions occurred in 26% of patients. In real-world studies, 5% stopped treatment because of ARIA.

Why it matters. ARIA risk in routine clinical practice matches what the trials predicted. This supports routine APOE genotyping before treatment and genotype-specific consent.

Read it if you are a carrier discussing lecanemab. Pair it with our case discussions of ARIA-E after donanemab and candidacy with prior microhemorrhages.

Anti-amyloid therapy · 4 of 13

Donanemab meta-analysis: modest benefit, with ARIA-E concentrated in homozygotes

Pereira da Silva AM, Falcão L, Virgilio F, et al. Journal of Alzheimer’s Disease. 2025;107(2):477–493. doi:10.1177/13872877251361044

High · three trials only

What it found. Across three randomized trials (2,054 participants), donanemab modestly slowed decline (CDR-SB mean difference −0.36; iADRS 3.19). The risk of ARIA-E was about twelve times that of placebo (RR 12.39), and highest in ε4 homozygotes.

Why it matters. It reinforces the genotype gradient in ARIA-E for donanemab. See our comparison Lecanemab vs. Donanemab and our appraisal Do Leqembi and Kisunla Meaningfully Help?

APOE4-specific therapy · 5 of 13

APOLLOE4: oral valiltramiprosate (ALZ-801) in ε4/ε4 homozygotes, plus a 2026 biomarker follow-up

Primary results: Abushakra S, Power A, Watson D, et al. Drugs. 2025;85(11):1455–1472. doi:10.1007/s40265-025-02250-5

Biomarker analysis: Hey JA, Abushakra S, Power A, et al. Drugs. 2026;86(10):1705–1721. doi:10.1007/s40265-026-02381-3

High · negative primary resultMCI subgroup: hypothesis-generatingSponsor-authored

What it found. In this 78-week Phase 3 trial, 325 ε4/ε4 homozygotes with early Alzheimer’s disease received valiltramiprosate or placebo. The drug did not meet its primary endpoint: ADAS-Cog13 decline was slowed by 11% (p = 0.607). It did slow hippocampal atrophy (18%, p = 0.017). In the prespecified MCI subgroup (125 participants), cognitive and functional results were nominally favorable (ADAS-Cog13 nominal p = 0.041; CDR-SB p = 0.053), but these p-values are not corrected for multiple comparisons. There was no increase in ARIA-E or microhemorrhages. Nausea, vomiting, and decreased appetite occurred at more than twice the placebo rate. The 2026 follow-up reports that plasma p-tau217 fell 36% on drug and rose 17% on placebo in the MCI subgroup (observed cases). Neurofilament light was also lower on drug at week 78 in the MCI subgroup, and the biomarker changes correlated modestly with clinical and MRI outcomes.

Why it matters. This is the first Phase 3 trial designed entirely around a genotype, and its safety profile is a real advantage for homozygotes. However, a biomarker response in a subgroup is not proof of clinical benefit. A confirmatory trial restricted to the MCI stage is needed. Both papers are authored largely by the sponsor.

Read it if you are a homozygote tracking APOE4-specific therapies. For context, see I Have Two Copies of APOE4 and our Clinical Trial Evaluation Guide.

Nutrition · 6 of 13

PreventE4: high-dose DHA reached the central nervous system but did not change cognition or brain volume

Yassine HN, Pour SG, Juarez M, et al. eBioMedicine. 2026;129:106316. doi:10.1016/j.ebiom.2026.106316 (NCT03613844)

High · for brain delivery38% dropout · cognition not the primary outcome

What it found. This 24-month Phase IIa trial randomized 365 adults aged 55–80 without dementia, all with low dietary DHA and at least one dementia risk factor (47% were ε4 carriers), to 2 g/day of DHA or placebo. At 6 months DHA raised the DHA-to-arachidonic-acid ratio in spinal fluid, and the rise did not depend on ε4 status (interaction p = 0.71). Over 24 months there were no differences in brain volumes or cognitive performance. Dropout was 38%, mostly because of COVID-19, and there were no treatment-related serious adverse events.

Why it matters. This was the first randomized test of high-dose DHA started before dementia in a population enriched for carriers. It was designed and powered to show brain delivery, not cognitive benefit. Its null secondary results therefore do not rule out a benefit, but they do not support one either. It also found that the rise in spinal-fluid DHA was similar in carriers and non-carriers. The observational associations between fish intake and slower decline in carriers remain, but causality is unproven.

Read it if you take or recommend omega-3 supplements. Our DHA guidance is now combined in one free guide: Supplements and APOE ε4: DHA, Vitamins, and What the Evidence Really Shows.

Risk and staging · 7 of 13

APOE ε4 in subjective cognitive decline: a true intermediate stage

Alonge P, Baiamonte L, Gerardi G, et al. Journal of Neurology. 2026;273(9). doi:10.1007/s00415-026-14077-5

Moderate · observational meta-analysis

What it found. Across 49 studies, the share of people carrying at least one ε4 allele was 22.0% among cognitively normal people, 28.3% in subjective cognitive decline (SCD), and 41.0% in MCI. The odds of carrying ε4 were higher in SCD than in controls (OR 1.28) and higher in MCI than in SCD (OR 1.48). These results held when the analysis was restricted to the stricter “SCD-plus” definitions.

Why it matters. For carriers who notice subtle memory changes that normal tests do not detect, this supports taking those concerns seriously while keeping perspective: most people with SCD are not carriers.

Tier 2: High-quality human studies

Large cohorts and genetic studies. These show associations, not proof of cause, but they often change how we interpret tests and risk.

Editor’s pick · Blood biomarkers · 8 of 13

The same p-tau217 level carries more near-term risk in APOE ε4 carriers

Xu Y, Gunasekaran TI, Gu Y, et al. Lancet Neurology. Published online September 2026. doi:10.1016/S1474-4422(26)00313-3

Moderate–High · pooled prospective cohorts

What it found. This participant-level analysis pooled seven cohorts from the US, Canada, and the Dominican Republic (8,582 participants). The population was diverse: 16.1% Black, and 36.7% of Hispanic or other ethnic origin. Among 4,569 people who were cognitively unimpaired at baseline, each 1-standard-deviation rise in p-tau217 predicted incident impairment more strongly in ε4 carriers (HR 1.76, 95% CI 1.36–2.26) than in non-carriers (HR 1.26, 1.12–1.42). It corresponded to a 24% shorter time to impairment in carriers, compared with 13% in non-carriers. Differences in impairment-free survival appeared 3–4 years after the blood draw.

Why it matters. A p-tau217 result should be interpreted in light of genotype. This is the most practical paper in this edition for carriers who are having blood biomarker testing.

Read it if you have had, or are considering, a p-tau217 test. We wrote a plain-language explainer: Your p-tau217 Blood Test and APOE4.

Blood biomarkers · 9 of 13

Predicting dementia in subjective cognitive decline: p-tau217, memory tests, and APOE4

Rivera Sánchez M, Mastenbroek SE, Janelidze S, et al. Neurology. 2026;106(11):e214983. doi:10.1212/WNL.0000000000214983

Moderate · BioFINDER cohorts

What it found. The study followed 469 people with SCD from the Swedish BioFINDER cohorts, of whom 84 progressed to dementia over about 4 years. Plasma p-tau217 was the strongest single predictor of Alzheimer’s dementia (C-index 0.86). Combining p-tau217 with cognitive scores and APOE4 status reached a C-index of 0.91. Adding MRI improved prediction of Alzheimer’s dementia only marginally, though it helped predict all-cause dementia.

Why it matters. A simple, clinically feasible combination of tests stratified risk well in people with SCD, and the authors see a role in memory clinics. Results from research cohorts may not generalize to the wider population.

Genetics · 10 of 13

APOE-stratified genome-wide study reveals hidden risk and protective genes

Thomassen JQ, Leonard H, Ulms B, et al. Nature Genetics. Published online September 2026. doi:10.1038/s41588-026-02713-9

Moderate · genetic discovery

What it found. Running genome-wide analyses separately by ε4 and ε2 carrier status revealed nine new Alzheimer’s risk signals (including HP1BP3, PTPRC, NPAS3, DDHD1, and GFRA1) that were invisible in combined analyses. DDHD1 appeared protective specifically in ε4 carriers.

Why it matters. This is a step toward explaining why some carriers never develop dementia. These are candidate genes and still need functional confirmation.

Proteomics · 11 of 13

Blood-protein “aging clocks” and APOE: carriers show older-appearing astrocytes

Ding DY, Bot VA, Chen KL, et al. Nature Medicine. 2026;32(6):2060–2072. doi:10.1038/s41591-026-04446-y

Related: Lu L, Pichet Binette A, Hristovska I, et al. Nature Aging. 2026;6(5):1138–1157. doi:10.1038/s43587-026-01123-0

Moderate · large cohorts, novel methods

What they found. Ding and colleagues estimated the biological age of more than 40 cell types from more than 7,000 plasma proteins in 60,542 people. APOE4 was linked to older-appearing astrocytes and younger-appearing macrophages. In people with two ε4 copies, extreme astrocyte aging tripled the risk of incident Alzheimer’s disease, while youthful astrocyte profiles lowered it. In a separate multicohort analysis, Lu and colleagues found that APOE-related protein changes appear before amyloid builds up and that ε2 and ε4 act through largely distinct molecular programs.

Why it matters. These findings help explain why genotype alone does not determine outcome. The methods are new and not ready for clinical use. Background: APOE ε2, ε3, and ε4.

Nutrition · 12 of 13

Mediterranean diet, the blood metabolome, and dementia risk in APOE4 homozygotes

Liu Y, Gu X, Li Y, et al. Nature Medicine. 2025;31(11):3790–3800. doi:10.1038/s41591-025-03891-5

Moderate · prospective cohorts

What it found. In 5,705 participants from prospective cohorts, the associations of 57 metabolites with dementia risk varied by genotype. Cholesteryl esters and sphingomyelins were most strongly tied to dementia risk in homozygotes, and Mediterranean-diet adherence shifted dementia-related metabolites more in homozygotes than in others.

Why it matters. This is observational support for genotype-informed nutrition, which remains a hypothesis rather than a proven intervention. See The APOE4 Carrier and the Nutrition Dilemma and heart health and genotype.

Tier 3: Mechanistic science

Important for where the field is heading. None of these findings changes clinical care today.

Blood vessels · 13 of 13

APOE4, fibronectin, and blood-vessel scarring: two studies converge

Schuldt BR, Haworth-Staines D, Perez-Arevalo A, et al. Cell. Published online September 2026. doi:10.1016/j.cell.2026.08.058

Bhattarai P, Yilmaz E, Cakir EÖ, et al. Nature Aging. 2026;6(9):1901–1927. doi:10.1038/s43587-026-01204-0

Context: Laing KK, Fialova N, Wardlaw J, Montagne A. Alzheimer’s Research & Therapy. 2026;18(1). doi:10.1186/s13195-026-02018-3

Hypothesis-generating

What they found. The Mount Sinai team (Schuldt, Cell) built a single-cell atlas of human brain blood vessels. In APOE4 carriers, pericytes, the support cells that wrap small vessels, were depleted and appeared to transform into scar-forming myofibroblast-like cells that secrete fibronectin, which promoted amyloid buildup in vessel walls. The process was driven by TGF-β signaling. Blocking TGF-β in laboratory models restored pericytes and reduced vascular scarring and amyloid. The Columbia team (Bhattarai, Nature Aging) found that astrocyte-derived fibronectin drove APOE4-related blood–brain barrier leakage, and that reducing it restored barrier function in experimental models. A systematic review of APOE4 mouse studies (Laing) found consistently reduced cerebral blood flow in APOE4 mice, although the studies had frequent methodological weaknesses.

Why it matters. Two teams working with different cell types arriving at the same molecule makes fibronectin a credible candidate drug target. The work also links APOE4 to amyloid in blood-vessel walls (cerebral amyloid angiopathy), which is closely tied to ARIA risk. There are no human intervention data yet.

Worth reading first

  1. Xu et al., Lancet Neurology — directly actionable for carriers having blood biomarker testing.
  2. Shim et al. and Hsu et al., read together — the honest state of the anti-amyloid efficacy question.
  3. Qi et al. — current ARIA numbers by genotype for lecanemab.
  4. Yassine et al. (PreventE4) — a rigorous result that tempers enthusiasm for supplements.
  5. Abushakra et al. (APOLLOE4) — a case study in reading subgroup claims critically.

Method. Articles were identified by structured PubMed searches for APOE ε4 and Alzheimer’s disease, prioritizing systematic reviews, meta-analyses, and randomized trials. Every citation, figure, and DOI was checked against the PubMed record. Preprints were excluded. Figures are reported as published; “nominal” p-values are not corrected for multiple comparisons.

This article is for educational purposes only and does not constitute medical advice. Treatment decisions should be made with a qualified clinician who knows your history, imaging, and genotype. © 2026 APOE4 Insights.

Just learned you carry APOE4?

Get the free physician’s guide Your First 30 Days After an APOE4 Result, plus one evidence update a month.

No spam. Unsubscribe anytime.

Just learned you carry APOE4?

Get the free physician’s guide Your First 30 Days After an APOE4 Result, plus one evidence update a month.

No spam. Unsubscribe anytime.

Similar Posts