Verify any claim · lenz.io
Claim analyzed
Health“A blood test developed by researchers at the University of East Anglia can detect early-stage dementia with 79% accuracy.”
The conclusion
The 79% accuracy figure is real but significantly mischaracterized. UEA researchers developed a preliminary machine-learning model using six gut-derived blood metabolites that classifies participants across three study groups — healthy, mild cognitive impairment (MCI), and impaired — with 79% accuracy. This is not a validated clinical blood test, and MCI is a precursor state, not equivalent to "early-stage dementia." Even UEA-affiliated coverage describes the work as research that "could pave the way" for a future test, not a deployable diagnostic tool.
Based on 15 sources: 7 supporting, 1 refuting, 7 neutral.
Caveats
- The 79% accuracy applies to a three-way research classification model, not a validated diagnostic test for detecting dementia in clinical settings.
- Mild cognitive impairment (MCI) is a risk factor and precursor state that does not always progress to dementia — equating MCI detection with 'early-stage dementia detection' is a significant overstatement.
- Multiple outlets reporting the same 79% figure all derive from the same UEA press release, providing no independent corroboration of the claim's framing.
Sources
Sources used in the analysis
A blood test identified Alzheimer's disease correctly in older adults with about 90% accuracy. This test, called PrecivityAD2, measures amyloid beta and p-tau217 and was developed by researchers from Lund University in Sweden, not the University of East Anglia.
A machine-learning model built on just six of these metabolites was able to classify people into the three groups with 79 per cent accuracy and could distinguish healthy adults from those with mild cognitive impairment with over 80 per cent accuracy.
A machine-learning model built on just six of these metabolites was able to classify people into the three groups with 79 per cent accuracy and could distinguish healthy adults from those with mild cognitive impairment with over 80 per cent accuracy.
A machine-learning model built on six of these metabolites was able to classify participants into the three study groups with 79% accuracy. Additionally, the model distinguished healthy adults from those with MCI with more than 80% accuracy.
Research from Rutgers Health indicates that blood tests for Alzheimer's disease need to be interpreted with caution – particularly for Black patients. 'Poor transfer of these proteins from spinal fluid to blood means many patients will go undiagnosed, especially if their disease is mild, while the ability of other ailments to increase these proteins in the blood mean tests produce false positives.'
This test now showing 90% accuracy actually performing better than primary care physicians, better than neurologists in terms of actually detecting the disease. So really a lot of hope on the horizon there. Yeah, it's pretty head turning these tests have not been approved. I should mention by the FDA. So I think a big question right now is which tests can we trust.
In healthy adults versus MCI, predictions increased to a maximum of 0.84 based upon the RF model. In addition to being measured via AUC, the classification was characterized by 79% accuracy, which distinguishes healthy adults from MCI subjects, as noted by an additional finding by the University of Liverpool.
The team then built a machine-learning model which was able to distinguish healthy adults from those with mild cognitive impairment with over 80% accuracy.
A new study found that it is possible to identify early signs of dementia many years before symptoms appear using a simple blood test. This research, conducted by the University of California, San Diego, focused on p-tau217 levels and is separate from the University of East Anglia's work.
Scientists have created a blood test that can estimate when Alzheimer's symptoms are likely to begin. By measuring a protein called p-tau217, the model predicts symptom onset within roughly three to four years.
Declining blood levels of two molecules that occur naturally in the body track closely with worsening Alzheimer's disease, particularly in women. The research team's accuracy in diagnosing the severity of Alzheimer's disease rose from more than 80 percent—when using either amyloid beta and tangled tau protein levels collected from cerebrospinal fluid or the two blood molecules—to 93 percent when using both.
Researchers at the University of East Anglia have discovered that subtle changes in the blood may reveal the earliest signs of cognitive decline long before symptoms become obvious. The changes are caused by chemicals produced by gut bacteria - reinforcing the idea that the gut–brain connection plays an important role in early memory changes.
Researchers at the University of East Anglia have discovered that subtle changes in the blood may reveal the earliest signs of cognitive decline long before symptoms become obvious. The changes are caused by chemicals produced by gut bacteria - reinforcing the idea that the gut–brain connection plays an important role in early memory changes.
Mild Cognitive Impairment (MCI) is a stage between the expected cognitive decline of normal aging and the more serious decline of dementia. It can involve problems with memory, language, thinking, or judgment that are greater than normal age-related changes. While not everyone with MCI will develop dementia, it is considered a significant risk factor and often represents an early stage of cognitive decline that may progress to dementia.
The FDA approved two blood-based biomarker tests in 2025 for people 55 and older who show signs or symptoms of cognitive impairment. One, designed for primary care settings, can rule out Alzheimer's in those who test negative with a 90% negative predictive value. A second, used in specialty settings, can help confirm a diagnosis, reducing the need for spinal taps or PET scans that cost upward of $10,000.
What do you think of the claim?
Your challenge will appear immediately.
Challenge submitted!
Expert review
How each expert evaluated the evidence and arguments
Expert 1 — The Logic Examiner
The logical chain from evidence to claim contains several meaningful inferential gaps. Sources 2, 3, 4, and 12 confirm that UEA researchers did develop a blood-based machine-learning model using six gut-derived metabolites that achieves 79% accuracy — but this accuracy applies to a three-way classification task (healthy vs. MCI vs. impaired), not a validated diagnostic test for "early-stage dementia." The claim's language ("can detect early-stage dementia") implies a deployable diagnostic tool, whereas Sources 12 and 13 explicitly frame the UEA work as research that "could pave the way" for such a test, and Source 7 partially attributes the 79% figure to the University of Liverpool rather than UEA alone. Additionally, the opponent correctly identifies that the multiple supporting sources (3, 4, 12) all derive from the same UEA press release, meaning their convergence is not independent corroboration but repetition — a valid logical point the proponent fails to adequately rebut. The proponent's rebuttal that MCI is "an early dementia-linked stage" (Source 14) is factually grounded but does not fully bridge the gap between a research classification model and a clinically deployable diagnostic blood test, making the claim's framing misleading rather than outright false — the core finding (79% accuracy, UEA researchers, blood metabolites, early cognitive decline) is real, but the claim overstates its diagnostic maturity and precision of attribution.
Expert 2 — The Context Analyst
The claim omits that UEA's reported “79% accuracy” is for a machine‑learning research model using six blood metabolites to classify participants into study groups (including mild cognitive impairment), not a clinically validated, deployable diagnostic blood test for “early-stage dementia,” and even UEA-linked coverage frames it as something that “could pave the way” rather than an established test (Sources 2, 12–13). With that context restored—and noting that other high-accuracy blood tests discussed in the wider landscape are different assays from other institutions (Source 1)—the claim gives a misleading impression and is effectively false as stated.
Expert 3 — The Source Auditor
The highest-authority source (Source 1, NIH) is neutral-to-refuting on the specific claim, clarifying that the most accurate Alzheimer's blood test (90%, PrecivityAD2) was developed by Lund University, not UEA — though it does not directly address the UEA metabolite study. The UEA's own press release (Source 2, high-authority institutional source) does confirm that UEA researchers built a machine-learning model using six blood metabolites achieving 79% accuracy in classifying three groups, but critically frames this as research that "could" lead to a test, not a validated diagnostic tool; Sources 3, 4, 7, 8, 12, and 13 all appear to derive from the same UEA press release, making them circular rather than independently verified, and Source 7 (low-authority) even partially attributes the 79% figure to the University of Liverpool rather than solely UEA. The claim as stated — that UEA "developed a blood test" that "can detect early-stage dementia" with 79% accuracy — overstates what the most reliable evidence supports: the UEA work is a research-stage machine-learning model on gut-derived metabolites, not a validated clinical blood test for dementia, and the 79% figure applies to a three-group classification task, not a binary dementia detection test, making the claim misleading rather than outright false.
Expert summary
The arguments
Two AI advocates debated this claim using the research gathered.
Argument for
The University of East Anglia reports that its researchers developed a simple blood-test approach using a machine-learning model based on six blood metabolites that classifies people across healthy, mild cognitive impairment (an early-stage cognitive decline linked to dementia), and impairment groups with 79% accuracy (Source 2, UEA; Source 14). This 79% figure is independently echoed by multiple outlets covering the same UEA work (Source 3, Technology Networks; Source 4, clpmag.com; Source 12, Norwich Evening News), supporting the claim that a UEA-developed blood test can detect early-stage dementia-related decline at 79% accuracy.
Your argument conflates "classifying people across three groups" with "detecting early-stage dementia" — a critical logical leap, because as Source 2 (UEA) itself confirms, the 79% accuracy applies to a three-way classification model, not a validated diagnostic test for dementia, and Source 12 (Norwich Evening News) explicitly states the UEA work only "could pave the way" for such a test, meaning no actual diagnostic blood test yet exists. Furthermore, your appeal to multiple outlets echoing the 79% figure is a mere repetition fallacy — Sources 3, 4, and 12 all draw from the same UEA press release, so their agreement adds no independent evidentiary weight and does nothing to resolve the core misrepresentation that a research model is being falsely characterized as a deployable blood test for detecting early-stage dementia.
Argument against
The claim is critically misleading because the 79% accuracy figure from Source 2 (UEA) and Source 3 (Technology Networks) does not describe detecting "early-stage dementia" — it describes classifying participants across three groups using a machine-learning model built on gut-derived metabolites, while Source 7 (The Brighter Side of News) explicitly attributes the 79% accuracy to distinguishing healthy adults from MCI subjects, a finding additionally credited to the University of Liverpool, not solely the University of East Anglia. Furthermore, the claim's framing of a definitive "blood test" is contradicted by Source 12 (Norwich Evening News) and Source 13 (Eastern Daily Press), which describe the UEA work merely as research that "could pave the way" for a dementia blood test — meaning no validated diagnostic test yet exists — while Source 1 (NIH) identifies a separate, more accurate 90% blood test developed by Lund University in Sweden, undermining the premise that UEA's work represents a meaningful diagnostic breakthrough.
You're equivocating over wording to dodge what Source 2 (UEA) actually reports: a blood-based, six‑metabolite machine‑learning model that separates healthy people from mild cognitive impairment—an early dementia-linked stage (Source 14)—with overall 79% accuracy (and even higher in the healthy vs MCI comparison), which is exactly the claim's substance. And your appeal to “could pave the way” (Sources 12–13) plus a completely different Lund/PrecivityAD2 Alzheimer's assay (Source 1) is a red herring: neither source disputes that UEA researchers developed this blood-test approach nor that its reported accuracy is 79% for early cognitive decline detection (Sources 2–4).