← Back
NeuroTrials.ai
Neurology Clinical Trial Database

Plasma p-tau217 Clock

Predicting onset of symptomatic Alzheimer's disease with plasma p-tau217 clocks

Share Share Copied! Compare

Year of Publication: 2026

Authors: Kellen K. Petersen, Marta Milà-Alomà, Yan Li, ..., Suzanne E. Schindler and Alzheimer's Disease Neuroimaging Initiative (ADNI)

Journal: Nature Medicine

Citation: Nature Medicine volume 32, pages 1085–1094 (2026)

Link: https://doi.org/10.1038/s41591-026-04206-y

Bottom Line

Plasma %p-tau217 clock models can estimate the age at AD symptom onset with a median absolute error of 3.0-3.7 years, providing a blood-based tool for predicting not just if, but when, cognitively unimpaired individuals will develop symptomatic Alzheimer's disease, with older age at biomarker positivity associated with shorter time to symptom onset.

Major Points

  • Clock models using longitudinal plasma %p-tau217 data from two independent cohorts (Knight ADRC n=258, ADNI n=345) estimated age at biomarker positivity with high accuracy
  • Estimated age at %p-tau217 positivity predicted age at AD symptom onset with adjusted R² of 0.337-0.612 and median absolute error of 3.0-3.7 years
  • Time from %p-tau217 positivity to symptom onset was dramatically shorter in older individuals: 20.5 years for those positive at age 60 versus 11.4 years at age 80
  • Cox models demonstrated excellent discriminative ability with C-index of 0.784-0.790 (Knight ADRC) and 0.730-0.750 (ADNI) for ranking individuals by risk of developing symptoms
  • Cross-cohort validation showed high correlation of age estimates: adjusted R² of 0.978 for TIRA and 0.999 for SILA models
  • Secondary analyses with five different plasma biomarker assays (Fujirebio Lumipulse p-tau217/Aβ42, C2N Diagnostics, Janssen LucentAD Quanterix, ALZpath Quanterix, Fujirebio Lumipulse) demonstrated generalizability of the approach

Design

Study Type: Prospective observational cohort study with biomarker validation

Randomization:

Follow-up Duration: Median 6.5 years (Knight ADRC) and 4.5 years (ADNI) between first and last plasma collection

Centers: 0

Countries: United States

Sample Size: 912

Analyzed: 603

Analysis: Generalized additive models (GAMs) for rate of change modeling; Two clock model approaches: Temporal Integration of Rate Accumulation (TIRA) and Sampled Iterative Local Approximation (SILA); Cox proportional hazards models for symptom onset probability; Linear regression for age at symptom onset prediction; Interval-censored survival analysis

Inclusion Criteria

  • Longitudinal plasma %p-tau217 measurements collected at least 1 year apart
  • For clock model development: %p-tau217 values within interval of consistent change (1.06-10.45%)
  • For symptom onset models: initially cognitively unimpaired (CDR=0) with positive AD biomarkers

Exclusion Criteria

  • Insufficient longitudinal data
  • For some analyses: cognitive impairment at baseline
  • For symptom onset models: individuals who developed cognitive impairment before %p-tau217 positivity (in some analyses)

Arms

FieldKnight ADRC Clock CohortADNI Clock Cohort
N258345
InterventionLongitudinal plasma %p-tau217 measurement with C2N Diagnostics assayLongitudinal plasma %p-tau217 measurement with C2N Diagnostics assay
DurationMedian 6.5 years (IQR 3.9-9.8) between first and last plasma collectionMedian 4.5 years (IQR 4.0-6.3) between first and last plasma collection

Outcomes

OutcomeTypeControlInterventionHR / OR / RRP-value
Association between estimated age at plasma %p-tau217 positivity and age at onset of AD symptomsPrimary<0.05 for all models
Model performance - Knight ADRC TIRASecondaryAdjusted R² 0.599, MdAE 3.0 years, CCC 0.771
Model performance - Knight ADRC SILASecondaryAdjusted R² 0.612, MdAE 3.5 years, CCC 0.839
Model performance - ADNI TIRASecondaryAdjusted R² 0.337, MdAE 3.2 years, CCC 0.801
Model performance - ADNI SILASecondaryAdjusted R² 0.470, MdAE 3.0 years, CCC 0.805
Cox model C-index - Knight ADRC TIRASecondary0.784 (95% CI 0.720-0.843)
Cox model C-index - Knight ADRC SILASecondary0.790 (95% CI 0.728-0.847)
Cox model C-index - ADNI TIRASecondary0.730 (95% CI 0.622-0.834)
Cox model C-index - ADNI SILASecondary0.750 (95% CI 0.636-0.853)
Median time to symptom onset for %p-tau217 positivity at age 60Secondary20.5 years
Median time to symptom onset for %p-tau217 positivity at age 80Secondary11.4 years
Cross-cohort correlation - TIRA modelsSecondaryAdjusted R² 0.978
Cross-cohort correlation - SILA modelsSecondaryAdjusted R² 0.999
Clock model accuracy - Knight ADRC TIRA vs observed conversionSecondaryAdjusted R² 0.733
Clock model accuracy - ADNI TIRA vs observed conversionSecondaryAdjusted R² 0.815
Clock model accuracy - Knight ADRC SILA vs observed conversionSecondaryAdjusted R² 0.506
Clock model accuracy - ADNI SILA vs observed conversionSecondaryAdjusted R² 0.801

Subgroup Analysis

Age at %p-tau217 positivity was the primary effect modifier: older age at positivity associated with shorter time to symptom onset. APOE ε4 carrier status, sex, and years of education were examined but were either not significant or had minimal effects and were not included in final models.

Criticisms

  • Relatively small sample sizes for symptom onset models (59-61 in Knight ADRC, 20-22 in ADNI)
  • Variability in adjusted R² values between cohorts, particularly lower performance in ADNI cohort for some models
  • Limited generalizability beyond research cohorts to general population
  • Clock models less stable at very low (<1.06%) and very high (>10.45%) %p-tau217 values
  • Sparse longitudinal data at high %p-tau217 values reduces certainty of estimates
  • Time between clinical assessments introduces interval censoring that may affect precision of symptom onset determination
  • Potential survivor bias in analysis of initially cognitively unimpaired individuals
  • Models may not account for all factors influencing cognitive reserve and rate of decline
  • Limited racial and ethnic diversity in cohorts may affect generalizability

Funding

Foundation for the National Institutes of Health (FNIH) Biomarkers Consortium; multiple NIH grants including NIA, NINDS; Alzheimer's Association; various pharmaceutical and diagnostic companies

Based on: Plasma p-tau217 Clock (Nature Medicine, 2026)

Authors: Kellen K. Petersen, Marta Milà-Alomà, Yan Li, ..., Suzanne E. Schindler and Alzheimer's Disease Neuroimaging Initiative (ADNI)

Citation: Nature Medicine volume 32, pages 1085–1094 (2026)

Content summarized and formatted by NeuroTrials.ai.