An FDA-cleared Alzheimer's blood test missed badly in one memory clinic cohort
At a Mayo Clinic cohort of 252 patients, 40% of amyloid-negative people tested positive. The manufacturer says it has found a manufacturing problem and put the product on hold.
Data presented at last week's Clinical Trials on Alzheimer's Disease meeting suggest that the first FDA-cleared blood test for Alzheimer's amyloid pathology performed substantially worse in one real-world memory clinic population than in the studies that supported its clearance [s1].
The test is Fujirebio's Lumipulse G pTau 217/β-Amyloid 1-42 Plasma Ratio, cleared by the FDA on 16 May 2025 as an aid to identify amyloid pathology in adults aged 50 and over who present with symptoms of cognitive decline in specialised care settings [s2]. In the clinical study supporting clearance, which included 499 patients, the test reported a positive predictive value of 92% and a negative predictive value of 97%, with about 20% of results falling in an indeterminate range that required further testing [s2].
What was presented
Alicia Algeciras-Schimnich of the Mayo Clinic in Rochester, Minnesota, reported results in 252 Mayo Clinic patients with dementia, of whom 177 were amyloid-positive and 75 amyloid-negative as determined by PET imaging or cerebrospinal fluid testing [s1].
In that cohort, 30 of the 75 amyloid-negative people tested positive on the blood test — a false positive rate of 40% [s1]. Overall accuracy came in at 77%, against 94% reported previously, and positive predictive value fell to 75% from 92% [s1].
The most telling detail is which half of the assay failed. The p-tau217 component measured on its own performed well, with a positive predictive value of 93% [s1]. The problem appeared when Aβ42 was included to form the ratio [s1].
Two explanations on the table, and they are not exclusive
The first is analytic fragility. Eric Reiman, quoted in the conference coverage, noted that Aβ42 "appears to be vulnerable to pre-analytic sources of variability, such as binding to test tube surfaces" [s1]. If a fraction of the Aβ42 in a sample sticks to plastic before it is measured, the measured ratio shifts toward "positive" — and the size of that shift depends on collection tubes, handling and timing, none of which are identical across sites.
The second is where the line was drawn. Sebastian Palmqvist observed that the FDA-cleared cutpoints were notably lower than those used in prior research, which increases the false-positive risk [s1]. A lower threshold buys sensitivity at the cost of specificity; in a population where many patients with dementia do not have Alzheimer's pathology, that trade lands hard on positive predictive value.
The third explanation is the one the manufacturer offered. Fujirebio stated: "During the CTAD meeting, we identified a manufacturing issue with recent commercial lots of the Lumipulse pTau 217/AB 1-42 test and have placed the product on a quality hold" [s1]. Algeciras-Schimnich said she had "not received a formal notification of a manufacturing issue" [s1].
The result that argues against a simple explanation
The same test, applied in cohorts from the Wisconsin Alzheimer's Disease Research Center, performed as expected: 94% positive predictive value and 95% accuracy [s1].
That is genuinely informative, and it cuts both ways. It is consistent with a bad lot reaching one site and not another. It is also consistent with the difference between a memory clinic population and a research cohort — research cohorts are typically enriched for Alzheimer's risk, which raises pre-test probability and mechanically improves positive predictive value for any test, however good or bad the assay is. Nothing presented resolves which of these is operating.
Separately, Michael Schöll's group at the University of Gothenburg reported that roughly 25% of autopsy-confirmed cases with no Alzheimer's disease neuropathologic change tested plasma-positive — a finding based on four cases, which the coverage itself flags as a small sample [s1]. Four cases cannot establish a false-positive rate. It can only fail to reassure.
Why the ceiling on this matters clinically
Blood-based amyloid testing was supposed to be the step that made anti-amyloid therapy scalable — cheaper and faster than PET, available outside academic centres. The value of that depends entirely on whether a positive result can stand on its own.
Michael Weiner of UCSF made the practical point: "An individual patient who is positive, but close to the cut point is very different than a positive patient who has a very high value" [s1]. He recommended confirmatory PET before initiating monoclonal antibody therapy [s1].
That recommendation, if adopted broadly, changes the economics of the whole approach. A blood test that triages who gets a PET scan is useful. A blood test that determines who gets an amyloid- targeting antibody is a different product with a much higher bar.
What this does not establish
This is conference-presented data, not peer-reviewed publication. One cohort at one centre showed poor performance; another set of cohorts showed performance in line with clearance [s1]. A manufacturing issue has been asserted by the manufacturer but, as of the meeting, not confirmed to the investigator who generated the discrepant data [s1]. Whether the Mayo samples came from affected lots is not established.
None of this is guidance for any individual. It is a caution about how a cleared diagnostic behaves outside the population it was validated in.
What to watch
Whether Fujirebio publishes lot-level detail and whether a formal recall or correction follows. Whether the Mayo results replicate at other memory clinics using different collection protocols. And whether anyone re-examines the cleared cutpoints — because if the threshold, rather than the lot, is the problem, no manufacturing fix will address it.
Sources
- [s1] Alzforum, "Trouble with Fujirebio's FDA-Cleared Blood Test? Or a Lousy Lot?", 6 December
- [s2] Fujirebio, "Fujirebio Receives Marketing Clearance for Lumipulse G pTau 217/β-Amyloid 1-42 Plasma Ratio In-Vitro Diagnostic Test," 16 May 2025. https://www.fujirebio.com/en-us/news-events/fujirebio-receives-marketing-clearance-for-lumipulser-g-ptau-217bamyloid-142-plasma-0
Sources
- Trouble with Fujirebio's FDA-Cleared Blood Test? Or a Lousy Lot? — Alzforum , December 6, 2025
- Fujirebio Receives Marketing Clearance for Lumipulse G pTau 217/β-Amyloid 1-42 Plasma Ratio In-Vitro Diagnostic Test — Fujirebio , May 16, 2025
More on
Alzheimer's blood tests exist. In people with no symptoms, they answer the wrong question.
The approved tests are for patients already showing cognitive decline. Used on the worried well, a positive result mostly signals a risk that may never become disease — and no treatment follows from it.
Alzheimer's blood tests are moving from "whether" to "when"
Three February papers push plasma p-tau217 past detection: clock models estimating years to symptom onset, four assays compared head to head, and a real-world audit of what the result changes.
A tau antibody drove a tangle marker down 90% — in seven people
The biomarker effect reported at CTAD is large and sustained over two years. The sample is seven, the design is open dose-escalation, and imaging was available for three.
A sugar-processing defect in the brain, and a supplement that made it worse in mice
University of Florida researchers identify excess glycan production as a driver of Alzheimer's pathology, and find oral glucosamine worsened cognition in mice and tracked with faster decline.