A single blood test that measures plasma phosphorylated tau at position 217 (p-tau217) was able to estimate when cognitively unimpaired individuals were likely to develop Alzheimer’s disease (AD) symptoms, a new study showed.

In analyses of longitudinal plasma p-tau217 data from more than 900 participants enrolled in two large observational cohorts, investigators found that biomarker-based “clock” models using %p-tau217 predicted the age of symptom onset with a median absolute error of 3-4 years.

The findings showed that the timing of %p-tau217 abnormality helped predict symptom onset, with older age shortening the time from biomarker positivity to clinical disease.

“The current estimate is not yet accurate enough for clinical use, but we expect that it will be possible to create more accurate models,” senior author Suzanne E. Schindler, MD, PhD, associate professor in the Department of Neurology, School of Medicine, Washington University in St. Louis, St. Louis, told Medscape Medical News.

The study was published online on February 19 in Nature Medicine.

A Plasma-Based ‘Clock’

Blood-based biomarkers provide a practical alternative to amyloid and tau PET imaging yet predicting when symptoms will emerge has remained challenging.

Since plasma p-tau217 levels increase steadily from preclinical through early symptomatic disease, investigators believe it can track overall AD pathology, while %p-tau217 is particularly useful for modeling the timing of symptom onset.

For the current study, investigators analyzed longitudinal plasma %p-tau217 trajectories from participants in the Knight Alzheimer’s Disease Research Center (Knight ADRC; n = 506; median age, 67.7 years; women, 54.2%) and the Alzheimer’s Disease Neuroimaging Initiative (ADNI; n = 406; median age, 72.7 years; women, 49.3%). They applied statistical models to estimate when %p-tau217 became abnormal and predict the timing of subsequent symptom onset.

In the Knight ADRC cohort, 35.8% of participants were APOE epsilon 4 carriers, and 8.5% were cognitively impaired at baseline (Clinical Dementia Rating score > 0); the median follow-up was 7.1 years, and 61.3% contributed three or more plasma samples.

In the ADNI cohort, 34.2% of participants were APOE epsilon 4 carriers, and 48.3% were cognitively impaired at baseline, with a median follow-up of 5.0 years, and 99.5% of participants provided three or more plasma samples.

Across both cohorts, plasma %p-tau217 positivity was defined as greater than 4.06%, a threshold aligned with an amyloid PET Centiloid value of 20. Using %p-tau217, the ratio of phosphorylated to nonphosphorylated tau improves analytic stability and reduces variability across assays, the investigators noted.

Modeling the AD Timeline

To estimate the age at which plasma %p-tau217 crossed the abnormal threshold and to predict symptom onset, the investigators used two complementary modeling approaches: Temporal Integration of Rate Accumulation (TIRA) and Sampled Iterative Local Approximation (SILA).

In cohort-specific analyses, TIRA models achieved adjusted R² values of 0.733 in Knight ADRC and 0.815 in ADNI, whereas SILA models yielded adjusted R² values of 0.506 and 0.801, respectively. When evaluated across cohorts, model performance remained high, with adjusted R² values of 0.978 for TIRA and 0.999 for SILA, which showed strong generalizability.

Cox proportional hazards models were used to estimate the probability that initially cognitively unimpaired individuals would develop symptomatic AD. Across cohorts and modeling approaches, the median absolute error for predicting age at symptom onset ranged from 3.0 to 3.7 years, with nonparametric concordance correlation coefficients between 0.771 and 0.839. This demonstrated that predicted ages closely matched the observed onset ages.

Age Matters

The estimated age at plasma %p-tau217 positivity was strongly associated with observed age at symptom onset, but the interval between biomarker abnormality and clinical disease varied substantially by age.

Older individuals progressed more rapidly after biomarker positivity than younger individuals.

“We found that older adults developed symptoms much more rapidly after %p-tau217 became abnormal,” Schindler said. “For example, people who first had abnormal %p-tau217 levels around age 60 didn’t develop Alzheimer’s symptoms for about 20 years, whereas those who first had abnormal %p-tau217 levels around age 80 developed symptoms after only about 10 years.”

She noted that age- and disease-related changes in the brain appear to influence how quickly AD symptoms emerge.

The clock models performed consistently across multiple commercially available plasma p-tau217 assays, including C2N Diagnostics, Janssen, ALZpath, and Fujirebio. While some assays captured slightly less of the variability in age at symptom onset, the small differences in accuracy did not affect overall model reliability.

Limitations of the study included that the clock models applied only to participants with %p-tau217 values within a defined range, so very high or very low values could not reliably predict time to symptom onset. In addition, the predominantly non-Hispanic White cohort with varied clinical presentations may limit generalizability.

Still in the Research Domain

The investigators emphasized that these plasma %p-tau217-based clock models were developed primarily as research tools rather than routine clinical tests. Their main use lies in identifying participants who are likely to develop AD symptoms within a specific timeframe, which could help make clinical trials more efficient.

“If we are able to predict onset of Alzheimer’s symptoms with high enough accuracy, these models could be useful in planning or considering different interventions on an individual level,” Schindler said.

However, routine testing in cognitively unimpaired individuals is not currently recommended.

“Currently, we do not recommend that cognitively unimpaired individuals have %p-tau217 blood tests because of potential legal and ethical issues,” said Schindler. She added that this guidance could change depending on the outcomes of ongoing clinical trials.

Future work will focus on improving accuracy by integrating %p-tau217 with additional blood-based or imaging biomarkers.

“There are many other blood biomarkers and imaging biomarkers that we can combine with plasma %p-tau217 to improve the accuracy of predicting symptom onset,” Schindler said.

Improving the accuracy of these models may require incorporating factors that reflect the complexity of individual patients, such as coexisting brain diseases, medical comorbidities, and social influences, she added.

The study was funded by the National Institute on Aging for ADNI4 and the University of Pennsylvania ADRC P30 Biomarker Core. Schindler reported serving on scientific advisory boards for Eisai and Novo Nordisk and receiving speaking fees from Eisai, Eli Lilly, Novo Nordisk, Medscape, and PeerView. She reported providing unpaid scientific advising to Eisai, Johnson & Johnson Innovative Medicine, Eli Lilly, Biogen, Acumen, Cognito Therapeutics, and Danaher.