Higher plasma phosphorylated tau 217 (p-tau217) in older adults who were cognitively unimpaired at baseline was associated with a hazard ratio of 1.38 (95% CI 1.30 to 1.46) for progression to cognitive impairment per 1 standard deviation increase, according to an observational longitudinal study by Rachel F. Buckley and colleagues published online in JAMA on 14 July 2026 (doi:10.1001/jama.2026.12556) and presented the following day at the Alzheimer's Association International Conference in London. The analysis harmonised data from 2,684 participants across six observational and clinical trial cohorts, including ADNI, the Harvard Aging Brain Study, the A4 prevention trial and its amyloid-negative companion study LEARN, the Wisconsin Registry for Alzheimer's Prevention and the Health and Aging Brain Study-Health Disparities. Median age was 69.6 years, 63% were women, and 478 participants progressed to mild cognitive impairment or dementia over a median 5.4 years of follow-up, with some followed up to 13.5 years. The association held after adjustment for amyloid PET (HR 1.32; 95% CI 1.24 to 1.41).

The consequential shift is not the hazard ratio but the format: the paper reports model-estimated absolute risk over 2, 5 and 10 years for four p-tau217 strata, defined in standard deviations above a cohort-specific reference drawn from amyloid-negative participants under 70. At two years, estimated risk was low across every stratum, roughly 1% to 4%. At five years, coverage of the paper reports estimates of about 12% in the low stratum, 15% intermediate, 24% high and 38% in the very high stratum; at ten years, about 40%, 45%, 62% and 78% respectively, with the very high ten-year figure carrying a confidence interval of 69% to 84%. Fewer participants had a decade of follow-up, which makes the ten-year numbers the least certain in the paper. Cognitive trajectories moved in the same direction: annualised five-year change on a latent Preclinical Alzheimer Cognitive Composite was -0.07 units per year in the very high group against +0.03 units per year in the low group.

The numbers also moved through peer review, which is visible because an earlier version is public. The medRxiv preprint posted on 9 February 2026 reported 2,705 participants and put ten-year absolute risk in the very high group at about 61% (95% CI 53% to 68%); the published version reports about 78% for the same stratum. Readers comparing the two are comparing successive models of the same pooled data, not two independent studies. The strata themselves are anchored to research assay distributions rather than to a commercial cutoff, so a laboratory report giving a p-tau217 concentration does not translate directly into any of these percentages.

The authors and the accompanying JAMA editorial set clear boundaries. The contributing cohorts are research volunteer samples, more educated and less racially, ethnically and socioeconomically diverse than the general senior population, and one contributing cohort required amyloid positivity for entry, so the observed distribution of strata is not a population prevalence. Assay platforms, recruitment and the way cognitive impairment was diagnosed differed between cohorts, and the authors call for validation in unselected populations that also accounts for other illnesses and death as a competing risk. Current practice guidance does not recommend p-tau217 testing in people without cognitive symptoms outside research or trials. Buckley told Medscape Medical News that the findings "move us toward individualized p-tau217-based risk estimates" for future impairment, framing them as an input to prevention trial screening rather than a clinical test.