The same p-tau217 number, a blood fragment of phosphorylated tau that tracks Alzheimer biology but is not the tangle itself, does not mean the same thing if you carry APOE4. In two independent imaging groups, people with two ε4 copies crossed into early tau-PET abnormality (PET is positron emission tomography, a tracer scan of a biological process; tau-PET maps the fibrillar tangle form inside neurons) at plasma levels 42% to 76% lower than non-carriers, depending on the lab test. [1]
That is a threshold finding, not a new blood test you should order this week. It is a clean human answer to a mechanism question mouse work has raised for years: does APOE4 make more phosphorylated tau, or does it make the tau you already have spread faster? Anna Steward, Nicolai Franzmeier, and colleagues say it is the second.
What this page answers
- Where on the amyloid-to-tau path APOE4 seems to act
- Why a “normal” p-tau217 cut-off may be too high for carriers
- What a clinician conversation looks like, and what it does not
Two ideas, one that held
Alzheimer’s biology is often sketched as a relay. Amyloid builds up first. Soluble phosphorylated tau (p-tau) then rises. Tangle-form tau then spreads along connected brain networks, and that spread tracks thinking much more tightly than amyloid does. APOE4 carriers tend to accumulate tau at lower amyloid-PET levels, which speeds the whole timeline. The open question was which baton APOE4 grabs.
The team analyzed two APOE-genotyped samples with amyloid-PET, plasma p-tau217, and follow-up tau-PET. One was 201 people from the Alzheimer’s Disease Neuroimaging Initiative, covering the clinical spectrum. The other was 200 from A4-LEARN, a group still thinking normally. They tested two ideas. First, that ε4 amplifies the amyloid-to-p-tau step. Second, that ε4 amplifies the p-tau-to-tau-spread step. [1]
The first idea did not hold. APOE4 did not steepen the relationship between amyloid-PET and plasma p-tau217. Amyloid still drove p-tau into the blood. Genotype did not appear to turn that tap further open.
The second idea held in both groups, and it scaled with copy number. Higher plasma p-tau217 predicted faster tau-PET spread from each person’s local starting region across connected areas. That link was stronger with one ε4 copy than with none, and stronger still with two, even after accounting for amyloid. A spinal-fluid p-tau181 subset in ADNI (115 people) pointed the same way. [1]
Then they asked the practical question: at what plasma p-tau217 level does early abnormal tau-PET show up, split by allele count?
| Assay and cohort | 0 ε4 copies | 1 copy | 2 copies | Drop vs non-carriers |
|---|---|---|---|---|
| Fujirebio p-tau217, ADNI | 0.62 pg/mL | 0.34 pg/mL | 0.15 pg/mL | ~45% / ~76% |
| Eli Lilly p-tau217, A4-LEARN | 0.31 pg/mL | 0.23 pg/mL | 0.18 pg/mL | ~26% / ~42% |
Those two rows are not interchangeable. Different assays, different groups, different absolute numbers. The pattern is the news: more ε4 copies, lower the bar at which tau starts to seed.
This is the human imaging counterpart of a story the site already tells with mice. In how APOE4 affects the brain, the Huang lab finding is that adding ε4 to a tau-tangle mouse made neurodegeneration worse, and stripping ε4 out of astrocytes or neurons protected the brain. Steward and colleagues are saying, in people, that ε4 looks less like a p-tau factory and more like a tau-spread accelerant.
Why the threshold idea matters
p-tau217 has become the standout blood marker because it classifies amyloid and tau status with accuracy in the 90 to 95% range, close to spinal-fluid tests. [2] In May 2025 the FDA cleared a plasma p-tau217/Aβ42 ratio for adults 55 and older who already have cognitive symptoms. [3] Those tools use population cut-offs. If ε4 carriers tip into tangle spread at a lower plasma level, a “negative” or “low” result that is reassuring in a non-carrier may be late in someone with two copies.
That is a hypothesis about screening math, not a license to lower the cut-off in your patient portal. Assay-specific numbers from 200-person research groups do not transfer to a commercial lab. They do change how a specialist should think about a carrier’s result: the protein level is a risk context, not a universal traffic light.
The authors also float a treatment implication: ε4 carriers might need anti-amyloid drugs earlier, at lower p-tau217, to intercept the handoff to tangle spread. That is a trial-design idea. It is not evidence that starting lecanemab or donanemab at a lower blood number helps, and those antibodies already carry extra ARIA risk in ε4 carriers, especially people with two copies. ARIA is amyloid-related imaging abnormalities: swelling or bleeding on MRI. See anti-amyloid drugs and APOE4.
Thresholds by assay and copy number
| What | Figure | Context |
|---|---|---|
| Samples | 201 ADNI + 200 A4-LEARN | imaging, not a treatment trial |
| Amyloid → p-tau217, by genotype | no ε4 amplification | the secretion step looks similar |
| p-tau217 → tau-PET spread | stronger with more ε4 copies | both groups; after accounting for amyloid |
| Two-copy threshold vs non-carrier | ~76% lower (Fujirebio) / ~42% lower (Lilly) | assay-specific, not a lab cut-off |
| FDA-cleared use today | symptomatic workup, age ≥55 | not healthy-carrier screening |
What two imaging studies cannot do
Two observational imaging studies, about 400 people in total, cannot tell you when you will decline. They cannot tell a lab which number to print on a report. They cannot tell a clinician to start an antibody at a lower plasma value. The thresholds are statistical estimates inside specific assays. Mix assays and the milliliters stop meaning the same thing.
p-tau217 is also not a crystal ball in people without symptoms. Current guidance supports it in a workup of cognitive complaints, not as a direct-to-consumer screen. A “positive biology” result without a plan has a real psychological cost, which is why mild cognitive impairment and early detection puts symptoms and clinical context first. For the media-literacy version of “threshold” headlines, see reading a study like a skeptic.
How to use a p-tau217 result
- If you have no symptoms, this paper is not a reason to buy a blood test. Keep pulling the levers that do not wait on tau-PET: blood pressure, exercise, sleep, and metabolic health.
- If you are in a memory-clinic workup, ask whether the lab’s p-tau217 assay has a published cut-off, and mention your APOE genotype so the result is not read as if you were ε3/ε3.
- If treatment is on the table, the ARIA conversation still dominates for antibodies. A lower research threshold does not erase that safety issue.
- If you like research, prevention trials are increasingly using plasma p-tau217 to find people in the amyloid-positive, still-thinking window. Finding a trial is the constructive version of “get tested.”
Common questions
Does a low p-tau217 mean I am safe if I carry two copies? Safer than a high one, not safe. This paper says the warning may come at a lower number. It does not turn a low number into a lifetime guarantee.
Can I use the 0.15 pg/mL figure at my lab? No. That was a Fujirebio research estimate in ADNI. Your lab may use a different assay with a different scale.
Should carriers get anti-amyloid drugs earlier because of this? That is a question for a trial, not a clinic protocol. Earlier also means more years of ARIA risk.
APOE4 looks less like a p-tau factory and more like a lower tripwire for tau spread. That is a reason for specialists to interpret blood markers in genotype context, not a reason to screen yourself.