The Alzheimer Blood Test, and What the Anti-Amyloid Drugs Actually Measured

Your doctor mentions a blood test. Not a brain scan, not a lumbar puncture — a tube of blood that is supposed to say something about Alzheimer's disease. Since May 2025 such a test exists with a regulator's clearance behind it, and in the same conversation two drugs usually come up. What each of those three things actually measured is worth knowing, because the words in the headlines are not the words in the trial reports.

This is an informational article. It is not a diagnosis, not a treatment recommendation, and it does not replace a consultation with your doctor.

What does the blood test measure?

The line on the lab form reads "p-tau217" — a fragment of the tau protein, measured in plasma. It is compared against amyloid pathology in the brain, as seen on an amyloid PET scan or in cerebrospinal fluid, at the same moment. It is not a memory test and not a forecast.

On 16 May 2025 the FDA cleared the first blood test intended to aid the diagnosis of amyloid plaques: a plasma ratio of p-tau217 to beta-amyloid 1-42. That was a device clearance, not the approval of any treatment. In the study the agency described, 499 plasma samples came from cognitively impaired adults aged 55 and over, in specialised care. A positive result agreed with PET or spinal fluid in 91.7% of cases, a negative result in 97.3%, and fewer than one result in five was indeterminate. It is not a screening test for people without symptoms.

A larger peer-reviewed study (Nature Medicine, 2025) ran an automated p-tau217 assay in 1,767 people who already had cognitive symptoms — 1,219 in specialist clinics in four countries, 548 in Swedish primary care. With a single cut-off, accuracy was 89–91% in specialist care and 85% in primary care, where the positive predictive value was 82% and the negative 88%. Using two cut-offs raised accuracy to 92–94%, but only by setting aside the 12–17% of results that fell between them as intermediate. The endpoint was a spinal-fluid biomarker: not memory scores, and not whether a drug would help.

A third test works in the other direction. Its manufacturer announced a clearance on 13 October 2025 for an assay of p-tau181 — a different protein — meant to help rule out amyloid pathology in symptomatic people aged 55 and over. The published figure is a negative predictive value of 97.9% in 312 people in a low-prevalence setting: that answers "who probably does not have amyloid", not "who has Alzheimer's disease". It is the company's own announcement, with no matching figure given for a positive result. Further clearances reported during 2026 appeared only in secondary news coverage, so their performance figures are not quoted here.

What did the two drugs measure?

Lecanemab and donanemab are infused antibodies that clear amyloid from the brain. Both were tested only in people who already had symptoms and whose amyloid was confirmed.

Lecanemab (Clarity AD, phase 3, placebo-controlled, double-blind). 1,795 people aged 50–90 with mild cognitive impairment or mild dementia due to Alzheimer's disease, 898 on drug and 897 on placebo, 10 mg/kg intravenously every two weeks, primary endpoint at 18 months. On the CDR-SB scale (0–18, higher is worse) the drug group changed by +1.21 and placebo by +1.66 — a difference of −0.45 points (95% CI −0.67 to −0.23). Both groups got worse. That gap is under half a point on a scale whose early-disease range runs roughly 0.5 to 6. The "27% less worsening" figure in circulation is 0.45 divided by 1.66: arithmetic from the published means, not a separate endpoint. In the PET substudy of 698 people amyloid fell by 59.1 centiloids more than on placebo — a large change in plaque beside a small one on the clinical scale.

Donanemab (TRAILBLAZER-ALZ 2, phase 3, placebo-controlled, double-blind). 1,736 randomised (860 on drug, 876 on placebo), mean age 73.0, with both amyloid and tau on PET; 1,182 had low or medium tau and 552 had high tau. Infusions every four weeks for up to 72 weeks. The primary endpoint was the iADRS scale (0–144, lower is worse) at 76 weeks. In the low/medium-tau group, which the statistical plan weighted most, the change was −6.02 on drug versus −9.27 on placebo: a difference of 3.25 points (95% CI 1.88 to 4.62), presented in the paper as 35.1% less decline. In the full population the same scale gave 2.92 points (95% CI 1.51 to 4.33), or 22.3%. A 2026 re-analysis of the trial also reported fewer transitions to a worse stage, but that was not the primary endpoint and the paper mixes prespecified with post-hoc analyses.

How often did the brain-imaging side effect happen?

ARIA — amyloid-related imaging abnormalities — means swelling (ARIA-E) or bleeding and iron deposits (ARIA-H) seen on MRI. It is usually asymptomatic, and sometimes it is not. Lecanemab carries a boxed warning for it; the label states that ARIA can be fatal, and that serious intracerebral haemorrhages larger than 1 cm have occurred with this class of medicines.

In Clarity AD: ARIA-E in 12.6% on lecanemab versus 1.7% on placebo; ARIA-H in 17.3% versus 9.0%; infusion reactions in 26.4% versus 7.4%.

Genotype matters here more than anything else on this page. In the lecanemab arm, 31% carried no APOE ε4 copy, 53% carried one and 16% carried two. The label reports the incidence of ARIA as 13% in non-carriers (4% on placebo), 19% with one copy (9%) and 45% with two copies (22%). Symptomatic ARIA-E occurred in 1%, 2% and 9% respectively. This is why the label says APOE ε4 status should be tested before treatment starts. Anticoagulant use was linked to more brain haemorrhages than placebo.

In TRAILBLAZER-ALZ 2: ARIA-E in 205 of 860 on donanemab (24.0%), of whom 52 had symptoms, versus 18 (2.1%) on placebo, none of them symptomatic during the study. Infusion reactions 8.7% versus 0.5%. Three deaths on donanemab and one on placebo were judged treatment-related. By genotype, as stated in a JAMA letter quoting the trial, ARIA-E was 15.7% in non-carriers, 22.8% with one copy and 40.6% with two. An ARIA-H rate circulating from a conference presentation of the same trial is not quoted here: it was not in the peer-reviewed abstract. And the two drugs' ARIA figures cannot be set against each other — different patients, different MRI schedules, different genotype mixes.

What none of this shows

A blood test does not measure memory. It does not say who will later develop dementia, and it does not say that a drug will help the person tested. False positives are not a paperwork problem: a wrong Alzheimer label can lead to an anti-amyloid antibody, whose characteristic harm is ARIA.

Over 18 months and 76 weeks respectively, both trials measured less worsening on drug than on placebo — and worsening in both. Neither showed that decline stopped or reversed; neither reported a difference in nursing-home entry or death. The "35% slowing" figure belongs to the low/medium-tau subgroup; in the full donanemab population the same scale gave 22%, and high-tau patients did worse as a group. Both trials enrolled mild cognitive impairment or mild dementia with biomarker proof, so they say nothing about moderate or severe dementia, and nothing about memory complaints in people without amyloid. A subcutaneous at-home starting dose of lecanemab was approved in 2026 according to the manufacturers' announcement; per the regulator's notice it was cleared on drug levels and plaque reduction comparable to the infusion, not on a new 18-month cognitive trial of the injection.

Who are these drugs for?

Both labels say the same narrow thing: treatment is started at the mild cognitive impairment or mild dementia stage, and amyloid pathology must be confirmed first. Lecanemab moved from accelerated to traditional US approval on 6 July 2023; donanemab's label records an initial US approval in 2024, and its EU authorisation is dated 24 September 2025. Neither is a prevention drug, neither is cleared for people without symptoms, and a positive blood test is not by itself a reason to start either one.

What to do with a result

If a blood test is offered, the useful questions are plain ones. What is it for — ruling amyloid out, or confirming it? Does it apply to someone with symptoms, the only population these tests were studied in? What happens if the result comes back intermediate, as it did for roughly one person in six in the larger study?

If an anti-amyloid antibody is on the table, two things belong in that conversation before anything is scheduled: APOE ε4 testing, because the figures above differ several-fold between genotypes: symptomatic ARIA-E was 1% in non-carriers against 9% in those with two copies, and any blood thinner being taken. The size of the measured difference — well under a point on the scales used — belongs there too, alongside MRI monitoring and the infusion schedule. All of it is a decision for a specialist who has seen that person's own scans and records.

Sources

• FDA. FDA Clears First Blood Test Used in Diagnosing Alzheimer's Disease. 16 May 2025. FDA news release

• Palmqvist S et al. Plasma phospho-tau217 for Alzheimer's disease diagnosis in primary and secondary care using a fully automated platform. Nat Med. 2025. PMID 40205199 · doi:10.1038/s41591-025-03622-w

• van Dyck CH et al. Lecanemab in Early Alzheimer's Disease (Clarity AD). N Engl J Med. 2023. PMID 36449413 · NCT03887455

• Sims JR et al. Donanemab in Early Symptomatic Alzheimer Disease: the TRAILBLAZER-ALZ 2 randomized clinical trial. JAMA. 2023. PMID 37459141 · NCT04437511

• Pomara N et al. Use of Donanemab in Early Symptomatic Alzheimer Disease (letter; ARIA-E by APOE ε4). JAMA. 2023. doi:10.1001/jama.2023.21106

• Atri A et al. Clinical Meaningfulness of Donanemab in Early Symptomatic Alzheimer Disease: Data From the Randomized Phase 3 TRAILBLAZER-ALZ 2 Trial. Neurol Clin Pract. 2026 Jun. PMID 42128444

• LEQEMBI (lecanemab-irmb) prescribing information — boxed warning, ARIA by APOE ε4, indication. DailyMed. FDA label

• KISUNLA (donanemab-azbt) prescribing information — indication, amyloid confirmation, initial US approval 2024. DailyMed. FDA label

• FDA. FDA Converts Novel Alzheimer's Disease Treatment to Traditional Approval. 6 July 2023. FDA news release

• EMA. Kisunla (donanemab) — EU marketing authorisation 24 September 2025. EMA product page

• Not peer-reviewed, and named as such above: the p-tau181 rule-out clearance (manufacturer's announcement, 13 October 2025), the subcutaneous lecanemab starting-dose approval (manufacturers' announcement, 2026), and clearances reported during 2026 in secondary news only, whose figures are not quoted. The current EU status of lecanemab was not checked for this article, so nothing is claimed about it.