A daughter in a clinic waiting room holds a printout with a single line on it: a protein name, a decimal number, and a word. Positive. Two years ago, reaching that word meant a needle between the vertebrae or an hour inside a PET scanner, with a wait and a bill to match. Now it takes a tube of blood drawn in the same visit where her mother was asked to remember three words and draw a clock. The speed is real. The question underneath it is harder: what does that number actually say about the woman sitting beside her?
Start with what the test measures. Phosphorylated tau 217, usually reported as a ratio with amyloid beta 42, rises in blood when amyloid plaques accumulate in the brain. In a study published in JAMA in July 2024, Sebastian Palmqvist and colleagues tested 1,213 people in Sweden who had come in with cognitive symptoms, and compared a p-tau217 blood measure against the established reference of amyloid PET or spinal fluid. The blood test was right about 90 percent of the time. Over the same patients, primary care physicians working with their usual tools reached the correct answer 61 percent of the time, and dementia specialists 73 percent. A follow-up in Nature Medicine in April 2025 ran a fully automated version of the assay through both primary and secondary care, the kind of boring engineering step that decides whether a finding survives contact with a hospital laboratory. That gap, 90 against 61, is the whole reason these tests moved so fast from research freezers into clinics.
Regulators have drawn the lines narrowly. In May 2025 the FDA cleared Fujirebio's Lumipulse G pTau 217/Beta-Amyloid 1-42 plasma ratio as an aid to identify amyloid pathology in adults already showing signs and symptoms, evaluated in specialized care. In October 2025 Roche's Elecsys pTau181 was cleared for use in primary care with a different job description: helping to rule out amyloid pathology in symptomatic adults. The Alzheimer's Association's first clinical practice guideline on these tests, published in Alzheimer's and Dementia in July 2025, set the arithmetic plainly. An assay with at least 90 percent sensitivity and 75 percent specificity can be used to triage, with positives sent on to spinal fluid or PET for confirmation. Only an assay hitting 90 percent on both counts can stand in for those older tests. The guideline covers people with cognitive impairment seen in specialty care, and declines to endorse the tests in people without symptoms.
The places where the number bends deserve the same plain treatment. Kidney function moves it: a 2025 analysis in Alzheimer's and Dementia quantified how chronic kidney disease raises plasma p-tau217 concentrations, which matters because reduced kidney function is common in the exact age group being tested. Thresholds move too. A December 2025 preprint from a French group, not yet peer reviewed, noted that published Lumipulse cutoffs for amyloid status span a wide range across studies, and tested whether thresholds derived in one cohort transfer to memory clinics elsewhere. And in October 2026, a team led by Doris Molina-Henry at the Keck School of Medicine of USC reported in JAMA Neurology on 15,612 cognitively unimpaired adults screened for a prevention trial, finding that baseline levels of the percentage p-tau217 measure ran lower in Asian and Black participants than in White participants, and lower in Hispanic than in non-Hispanic participants, while the relationship between higher values and brain amyloid held across groups. The biology travels. The cut point used to call a result positive may need local calibration before it does.
So what changes in the exam room. A positive result in a person with memory symptoms says amyloid is very likely present, which supports an Alzheimer's diagnosis and opens a conversation about whether anti-amyloid treatment is appropriate, a conversation that belongs to the treating clinician and the family. It says nothing about how fast decline will come, and it leaves open whether vascular damage or Lewy body pathology is also in the mix, since mixed pathology is ordinary in senior brains. A negative result in a symptomatic person is arguably the more useful document, because it redirects the search: toward vascular disease, frontotemporal degeneration, thyroid problems, sleep apnea, medication effects, depression, each with its own management. In someone with no symptoms at all, the result sits outside what the guideline covers, and the evidence for what to do with it does not yet exist.
The wonder here is quiet and worth naming. A protein fragment shed by neurons under stress crosses into the bloodstream in quantities measured in picograms per milliliter, and a machine can count them well enough to describe what is happening inside a skull it will never open. That is a genuine achievement of twenty years of work. It is also a measurement of one process, amyloid, in a disease that is never only one process.
The daughter will take the printout home and read it again at the kitchen table. The number on it is a fact about amyloid, dated and specific, and it belongs in a conversation with the clinician who knows the rest of her mother's chart: the kidney labs, the blood pressure years, the sleep, the medications. A good test narrows the question. The answer still gets built in a room with people in it.