Design: Prospective multicentre cross-sectional multitracer study
Population: 681 participants aged 50–90, cognitively unimpaired to dementia
Primary outcome: Inter-rater and inter-tracer agreement; tau positivity v approved read
Effect: κ 0.77–0.85; cognitively unimpaired tau-positive 26 → 49–71 of 364 (PR 1.89–2.73)

A prospective multicentre study in JAMA Neurology (21 September) took 681 people, from cognitively normal to dementia, who had head-to-head scans with up to four tau PET tracers, and tested a harmonised visual read with five classes: negative, low, moderate, high and atypical.

Agreement between readers was high (kappa 0.77 to 0.85) and agreement between the two main tracers was excellent (kappa 0.88). Among cognitively unimpaired people, the approved binary read called 26 of 364 tau positive; the new framework called 49 with flortaucipir and 71 with MK-6240. In MCI the gain was smaller, and in dementia the approaches converged. Higher tau classes tracked with worse cognition, more amyloid and higher plasma p-tau217.

This matters as anti-amyloid treatment shifts attention to early disease, where the current read is least sensitive. It is a diagnostic framework study; whether early tau staging changes treatment decisions or outcomes is not yet known, and tau PET access remains limited, including in India.

The approved binary tau PET read is designed for advanced disease and misses early tau.
A graded read may help place patients in the early window relevant to anti-amyloid therapy.
Tau class tracked with cognition, amyloid burden and plasma p-tau217.
Treat this as a research framework until reporting standards adopt it.

Why it matters

Early disease is where disease-modifying treatment decisions are made, and the current read is least informative there.

The statistics, in plain English

Kappa measures agreement beyond chance; above 0.8 is usually called excellent. A prevalence ratio of 2.73 means the new framework with one tracer found nearly three times as many tau-positive cognitively normal people. More positives are only useful if they are true positives, which this cross-sectional study cannot fully test.