Blood proteins can flag dementia risk years — even decades — before memory falters

A single blood draw, cheaper and far simpler than a brain scan, may one day tell a healthy middle-aged person whether dementia is likely decades down the road. That possibility has moved measurably closer in 2026, as several large independent studies converged on the same finding: proteins already circulating in the blood can predict cognitive decline years, and in some cases more than two decades, before the first symptom appears.

The most immediately practical result comes from a Washington University School of Medicine study published in Nature Medicine. Researchers there built a model using a blood protein called p-tau217 — a form of the brain protein tau that rises as Alzheimer's pathology accumulates — to estimate the age at which a person will begin experiencing Alzheimer's symptoms, with a margin of error of just three to four years.

Our work shows the feasibility of using blood tests, which are substantially cheaper and more accessible than brain imaging scans or spinal fluid tests, for predicting the onset of Alzheimer's symptoms. In the near term, these models will accelerate our research and clinical trials.— Suzanne E. Schindler, MD, PhD, associate professor of neurology, WashU Medicine

The WashU team analyzed data from 603 older adults enrolled in two long-running research initiatives — the Knight Alzheimer Disease Research Center and the Alzheimer's Disease Neuroimaging Initiative — all of whom were living independently at the time. They measured p-tau217 using PrecivityAD2, a clinically available diagnostic test from C2N Diagnostics, as well as several other p-tau217 tests including one cleared by the U.S. Food and Drug Administration. The model held up regardless of which test was used, a sign of robustness the researchers highlighted.

Amyloid and tau levels are similar to tree rings — if we know how many rings a tree has, we know how many years old it is. It turns out that amyloid and tau also accumulate in a consistent pattern and the age they become positive strongly predicts when someone is going to develop Alzheimer's symptoms. We found this is also true of plasma p-tau217, which reflects both amyloid and tau levels.— Kellen K. Petersen, PhD, instructor in neurology, WashU Medicine

The model also revealed an age effect: a person whose p-tau217 became elevated at 60 went on to develop symptoms roughly 20 years later, while someone whose levels didn't rise until 80 developed symptoms only about 11 years after that, according to WashU Medicine. The researchers interpret this as evidence that younger brains may be more resilient to early Alzheimer's changes.

A separate large study, published in JAMA and led by Dr. Rachel F. Buckley at Mass General Brigham in Boston, took a different angle on the same protein. Rather than predicting a specific age of onset, it put concrete five- and ten-year risk percentages to p-tau217 readings in nearly 2,700 healthy older adults pooled from six long-running studies across North America, Japan, and Australia. Over roughly five years of follow-up, 478 participants developed cognitive impairment. People in the lowest p-tau217 band faced about a 12 percent five-year risk; those in the highest band faced nearly 40 percent. Stretched to ten years, the highest band's estimated risk climbed toward 80 percent, though the researchers flagged those longer-range figures as uncertain given limited follow-up data.

That study also produced an unexpected finding on sex: men declined faster and more often than women at equivalent protein levels, even though Alzheimer's is more common in women over a lifetime. The researchers suggested women may hold cognitive ground longer in the earliest phase, with their rise in cases arriving later — a question the team plans to investigate further.

The p-tau217 signal also proved more predictive than amyloid brain scans. When the JAMA researchers accounted for each participant's amyloid load — the sticky plaque typically measured with a costly PET scan — the blood protein still predicted who would decline, and carried more predictive weight than the scan alone. Even among people whose scans showed little amyloid, those with elevated p-tau217 still faced higher odds of decline.

A third study, from the University of California, San Diego, published in JAMA Network Open, extended the prediction window dramatically. Researchers tested stored blood samples from 2,766 women enrolled in the Women's Health Initiative Memory Study in the late 1990s and followed for up to 25 years. Women with higher p-tau217 levels at enrollment — when all had normal cognition — were significantly more likely to develop dementia later. In some cases the signal appeared up to 25 years before diagnosis.

Our study suggests we may be able to identify women at elevated risk for dementia decades before symptoms emerge. That kind of long lead time opens the door to earlier prevention strategies and more targeted monitoring, rather than waiting until memory problems are already affecting daily life.— Aladdin H. Shadyab, PhD, MPH, associate professor of public health and medicine, UC San Diego

The UC San Diego team found the association was stronger in women over 70 than in younger participants, and stronger in women carrying the APOE ε4 genetic variant, which raises Alzheimer's risk. Differences were also observed between white and Black women, though combining p-tau217 levels with age improved prediction accuracy in both groups.

While those studies focused on p-tau217, a parallel line of research points to a different protein entirely. A study tracking more than half a million people for 15 to 25 years, published in Science Advances, found that elevated blood levels of GDF15 — a stress-responsive protein linked to aging, inflammation, and metabolism — before age 55 were significantly associated with developing dementia later in life. The association was particularly strong for vascular dementia rather than Alzheimer's. Elevated GDF15 was also linked to brain shrinkage and small vessel damage on neuroimaging, even in people who still had normal memory and thinking skills. Lab tests showed GDF15 altered the metabolic pathway of human immune cells and weakened their antiviral response, hinting at a biological mechanism rather than mere correlation.

Taken together, the studies represent a maturing field rather than a single breakthrough. Protein biomarkers in blood have gained significant attention as potential tools for diagnosing Alzheimer's and other dementias, with disruptions in proteins associated with metabolism, synaptic communication, and the immune system measurable as early as 20 years before diagnosis. The appeal is practical: blood tests cost a fraction of a PET brain scan and could reach far more patients, including the many who will never undergo neuroimaging.

None of these tests are ready for routine clinical use in people without symptoms. Current guidelines advise against screening asymptomatic individuals, partly because most study populations have been predominantly white and highly educated, and because risk estimates reflect broad probabilities rather than individual certainty. Researchers across all the studies stress that further validation in more diverse populations is needed before any of these markers enter standard care.

Eventually, the goal is to be able to tell individual patients when they are likely to develop symptoms, which will help them and their doctors to develop a plan to prevent or slow symptoms.— Suzanne E. Schindler, MD, PhD, associate professor of neurology, WashU Medicine

Ultimately, the goal is not just prediction, but using that knowledge to delay or prevent dementia altogether.— Aladdin H. Shadyab, PhD, MPH, associate professor of public health and medicine, UC San Diego

The most immediate payoff is for clinical trials. Sharper, time-specific risk estimates allow researchers to enroll the right people faster and run smaller, cheaper studies of drugs aimed at Alzheimer's silent phase — the long window before any symptom appears. More than 7 million Americans currently live with Alzheimer's disease, and the Alzheimer's Association projects national care costs will reach nearly $400 billion in 2025.

Why it matters — If validated across diverse populations, routine blood tests for dementia-linked proteins could replace costly brain scans as the first line of risk detection, giving patients and doctors years — or decades — to act before memory loss begins.

Reported by sciencealert.com, newsweek.com, scitechdaily.com, nejm.org, medicine.washu.edu, medicalxpress.com, earth.com