Picture the moment that has become routine in memory clinics: a senior sits across from a neurologist, a blood test has come back positive for p-tau217, and the amyloid question is settled. What is not settled is the one the family actually asked. Is this Alzheimer's disease doing the damage, and how far along is it? Amyloid plaques can sit quietly in the brain for years. Tangles of tau are the pathology that tracks most closely with symptoms, and until recently the only way to see them was a tau PET scan, available at a handful of specialist centres.
On March 31, 2025, Nature Medicine published work from Kanta Horie, Gemma Salvado, Oskar Hansson, Randall Bateman and colleagues at Washington University in St Louis and Lund University describing a plasma species they call eMTBR-tau243: an endogenously cleaved fragment from tau's microtubule-binding region, measured by mass spectrometry. Across three cohorts (108 and 55 people in pilot samples, then 739 in a validation cohort) levels were already raised at the mild cognitive impairment stage and higher again in dementia, and they tracked tau PET binding closely (beta = 0.72, R2 = 0.56) and cognitive performance (beta = 0.60, R2 = 0.40), outperforming %p-tau217 and %p-tau205. Two negatives matter as much as the positives. Levels were no different in cognitively unimpaired people with amyloid than in those without, so plaques alone do not push the marker up; and levels were not elevated in progressive supranuclear palsy, corticobasal syndrome or frontotemporal dementia, the tauopathies that are not Alzheimer's. This is an observational biomarker study, not a trial of anything.
Which brings us to the treatment question, and to a correction worth making plainly. The hope is not that a tangle test finds patients with enough tau to respond to anti-amyloid drugs. It is close to the opposite. In TRAILBLAZER-ALZ 2, the phase 3 donanemab trial of 1,736 people with early symptomatic Alzheimer's published in JAMA on July 17, 2023, the low or medium tau population was the primary analysis population, and the treatment difference on the integrated Alzheimer's Disease Rating Scale at 76 weeks was larger there (a change of -6.02 with donanemab against -9.27 with placebo) than in the combined population that included people with high tau (-10.19 against -13.11). Horie put the logic to reporters when the Nature Medicine paper appeared: "For early stages with low tau tangles, anti-amyloid therapies could be more efficacious." A cheap tangle test, on that reading, is a way of asking whether the window is still open.
What has happened since is not a treatment trial but a pair of larger validation papers. In The Lancet Neurology in April 2026, Niklas Mattsson-Carlgren and colleagues reported on 572 patients recruited prospectively at two Swedish memory clinics (142 with subjective cognitive decline, 259 with mild cognitive impairment, 171 with dementia), followed to September 2024. Of the 350 who were positive on plasma %p-tau217, 341 were amyloid-positive, yet only 199, that is 57 percent, met the International Working Group definition of established Alzheimer's disease. Adding eMTBR-tau243 improved that classification and predicted both cognitive decline and tangle accumulation. Then in JAMA Neurology on May 26, 2026, Salvado and colleagues built a two-marker plasma staging model in 872 BioFINDER-2 participants and validated it in 156 from the Knight Alzheimer Disease Research Center; concordance with PET-based biological stages was high (C index 0.91 in both), and in the subsample with autopsy data the plasma stages lined up with neuropathology at an area under the curve of 0.96.
For a family, the honest translation is modest and still useful. A positive amyloid blood test answers a narrower question than it appears to; a second marker aimed at tangles would tell a clinician not just whether Alzheimer's biology is present but roughly how far it has travelled, and that is the axis on which the current drugs, their modest benefits and their real risks, are weighed. What none of this evidence does is hand anyone a threshold. No published study has randomised people to treatment on the basis of eMTBR-tau243, the measurements were made on research mass spectrometry platforms in academic cohorts, and the marker's own authors describe it as promising for staging and for tracking tau-targeted drugs in trials, which is a sentence about research infrastructure before it is a sentence about clinic appointments.
The frontier here is quieter than the headlines around blood tests suggest. It is the slow business of turning a decade of tau PET into something drawable from an arm, and then, only then, finding out whether knowing the stage changes the outcome.