Blood p-tau217 levels predict cognitive impairment risk in older adults years ahead

A blood measure of phosphorylated tau 217 (p-tau217) can help estimate the chance that an older adult with normal cognition will develop cognitive impairment, according to a multi-cohort study led by the Mass General Brigham Neuroscience Institute. Among people with very high levels, the estimated risk was 38% at five years and 78% at ten years, although the longer estimate rests on much less data. The work was presented at the Alzheimer's Association International Conference and published in JAMA.

Pooling six cohorts across three continents

Blood tests for p-tau217 have already gained federal clearance in the United States, but it has been unclear how well they forecast symptoms in people who currently have none. To address this, the team harmonised data from six observational studies and clinical trials in North America, Japan and Australia. Together they included 2,684 older adults with no cognitive impairment at entry.

At enrolment, participants gave blood samples for p-tau217 analysis and underwent positron emission tomography (PET) brain scans. Annual assessments then tracked memory, thinking and other cognitive abilities. Recruitment began as early as 2004, and the latest follow-up took place in 2025. Around 478 participants went on to develop cognitive impairment.

Risk rose with biomarker level

Higher baseline p-tau217 was strongly associated with a greater likelihood of later impairment. The researchers translated this into time-specific probabilities rather than simply flagging a biological abnormality. For the highest levels, the five-year estimate was 38% and the ten-year estimate 78%. The authors stress that far fewer observations were available beyond ten years, so the longer-range figures are less certain.

Information beyond amyloid and APOE4

The biomarker stayed significantly associated with impairment after accounting for amyloid-beta plaques measured by PET. Its predictive value also held beyond known genetic risk factors such as APOE4. This suggests that plasma p-tau217 adds prognostic information to measures that are more costly, invasive or less widely available.

Lead author Rachel F. Buckley said the study is distinctive because it estimates an individual's level of risk, and that results were consistent across the harmonised cohorts.

Limits and clinical caution

The authors do not recommend currently available blood tests for people without symptoms. No disease-modifying treatments have been established for those found to be at high risk, so advice would be unchanged whatever the result: regular exercise, a healthy diet and good sleep. Senior author Reisa Sperling noted that this could change if ongoing trials of such therapies prove beneficial.

Several limitations apply:

• Participants may not represent the wider population, so selection bias could affect how far the findings generalise.
• The analysis predicted impairment over limited periods, not lifetime risk.
• Data thinned out at longer follow-up, widening uncertainty in the ten-year estimates.

Implications for prevention research

For now, the most immediate use is in research. According to Sperling, p-tau217 can help identify people at high risk of future Alzheimer's dementia for enrolment in prevention trials. Several trials are already testing whether disease-modifying therapies benefit people with biological signs of Alzheimer's but no impairment. If these succeed, individualised risk estimates could inform who is treated and monitored, in a manner the authors compare with cholesterol testing for heart attack risk.

The next steps are larger and more representative cohorts, longer follow-up and refinement of individual-level predictions. The funding came from several sources, including the US National Institutes of Health, the Alzheimer's Association and Eli Lilly; the source article lists author disclosures, which are available in full in the JAMA paper.

Source: A simple blood test could reveal Alzheimer’s risk 10 years before symptoms (ScienceDaily: top science, 31 October 2026).

Image: AI-generated illustration.