Design: Population-based cohort with stored-sample biomarker assay
Population: 5225 community-dwelling adults, mean age 69, UK civil servants
Primary outcome: Incident all-cause dementia over 12 years
Effect: p-tau217 worst quartile HR 3.74 (95% CI 3.12-4.50); C-index 0.69

Stored blood from 5225 Whitehall II participants (mean age 69, mostly men) was assayed for amyloid-beta 42/40, p-tau217, neurofilament light (NfL) and glial fibrillary acidic protein (GFAP). Participants were followed for a median of 12.2 years, during which 516 developed dementia. Published in Neurology on 23 October 2026.

The worst quartile of p-tau217 carried a hazard ratio of 3.74 for dementia, the strongest of the markers. As predictors, though, all were modest: p-tau217 alone gave a C-index of 0.69, all markers together 0.76, about the same as a model built from ordinary risk factors (0.77). Combining the two reached 0.81. Sensitivity was 42% to 60% and positive predictive value 16% to 24%, while negative predictive value exceeded 92%.

Blood tests for Alzheimer pathology are entering clinics. In a cognitively normal population, a positive result will usually not be followed by dementia, and a negative one is more informative than a positive.

Whitehall II participants were civil servants, mainly White men, so results may not transfer to Indian populations.

Do not order blood Alzheimer biomarkers to predict dementia in people without symptoms
Read a negative result as more reassuring than a positive result is alarming
Combine any biomarker result with age, vascular risk and cognitive testing
Explain to patients that a raised p-tau217 does not mean dementia is certain

Why it matters

Direct-to-consumer and clinic blood tests are spreading faster than evidence on what they mean in healthy people.

The statistics, in plain English

A C-index of 0.5 is a coin toss and 1.0 is perfect; 0.69 to 0.76 is moderate. A positive predictive value of about 20% means four in five people flagged as high risk did not develop dementia within 12 years.