Alzheimer’s disease is the most common cause of dementia and is characterized by the accumulation of amyloid-β plaques and tau neurofibrillary tangles in the brain. These pathological changes may begin 10–20 years before the onset of clinical symptoms. Although amyloid and tau positron emission tomography (PET) scans have been used to model disease progression and predict symptom onset, these methods are costly and not widely available.
Blood-based biomarkers, in contrast, are less costly and more scalable. The measurement involved the percentage of plasma tau217 phosphorylated at threonine 217 relative to non-phosphorylated tau at the same position (%p-tau217). This biomarker increases during the preclinical and early symptomatic phases of Alzheimer’s disease and shows a strong correlation with amyloid PET, tau PET, brain atrophy, and cognitive impairment. However, prior to this study, it had not yet been used to estimate the timing of symptom onset.
The researchers used longitudinal plasma percentage p-tau217 levels from two independent cohorts, consisting of 258 participants from the Knight Alzheimer Disease Research Centre (Knight ADRC) and 345 participants from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Repeated blood samples of subjects were taken over a period of years. Using these longitudinal data, the investigators constructed “clock” models to estimate the age at which individuals became plasma positive for %p-tau217, defined as levels exceeding 4.06%, corresponding to an amyloid PET Centiloid value of 20.
Temporal Integration of Rate Accumulation (TIRA) and Sampled Iterative Local Approximation (SILA), where two mathematical methods were applied. Both approaches corresponded to individual biomarker trajectories according to the estimated time of biomarker positivity. Despite demographic and clinical differences between the two cohorts, there was strong agreement across datasets and modeling approaches.
Importantly, the estimated age at plasma-%p-tau217 was strongly related to the age at which the symptoms of Alzheimer’s developed. Adjusted R2 values ranged between 0.337 and 0.612, depending on cohort and method of modeling. The mean absolute error in predicting the onset of the symptoms ranged from 3.0 to 3.7 years. Paired samples t-test t-values ranged between 0.771 and 0.839, indicating a good concordance between the cohorts.
One of the most notable findings was the influence of age on biomarker positivity. Individuals who became plasma positive at age 60 had a substantially longer median time to symptom onset compared with those who converted at age 80. For example, TIRA-based estimates indicated a median positivity at age 60, corresponding to a median of 20.5 years until symptom onset, whereas positivity at age 80 corresponded to 11.4 years.
Additional analyses showed that individuals who were biomarker-positive by age 80 could develop symptoms within as few as 6.2 years. This suggests that progression from biomarker positivity to clinical impairment accelerates with advancing age. The researchers also examined whether similar clock models could be constructed using other plasma assays, including an FDA-cleared p-tau217/Aβ42 ratio assay and four commercially available p-tau217 immunoassays.
Although results varied across assays, several demonstrated moderate correlations with symptom onset, supporting the broader applicability of the clock-modeling framework. While the prediction error margin of 3-4 years currently restricts its use for individual clinical decision-making, the approach has significant implications for clinical trials. Identifying individuals at high risk of developing symptoms within a defined time window could shorten trial duration, improve efficiency, and increase statistical power.
The authors caution that biomarker testing for Alzheimer’s disease in cognitively unimpaired individuals is not currently recommended outside of research settings. They also highlighted several limitations, including the restricted biomarker range applicability (1.06%-10.45%), limited population diversity, and potential survival bias. Overall, the findings suggest that a single percentage plasma-based PCp-tau217 measurement can estimate years to symptom onset with a clinically meaningful accuracy that potentially transforms preclinical disease staging and the design of prevention trials.
Reference: Petersen KK, Milà-Alomà M, Li Y, et al. Predicting onset of symptomatic Alzheimer’s disease with plasma p-tau217 clocks. Nat Med. 2026. doi:10.1038/s41591-026-04206-y
