Two-step blood and scan strategy delivers near-perfect accuracy for early Alzheimer's signs

By Vijay Kumar MalesuReviewed by Susha Cheriyedath, M.Sc.Sep 16 2025

A simple blood test for p-tau217, when combined with confirmatory brain scans or CSF analysis, could transform early Alzheimer's detection and cut the burden of invasive procedures.

Study: Plasma Phosphorylated Tau 217 to Identify Preclinical Alzheimer Disease. Image Credit: Atthapon Raksthaput / Shutterstock

In a recent study published in JAMA Neurology, an international team of researchers evaluated whether plasma phosphorylated tau 217 (p-tau217) could classify amyloid β (Aβ) status in cognitively unimpaired adults as a stand-alone test and in a two-step workflow with confirmatory positron emission tomography (PET) or cerebrospinal fluid (CSF) testing.

Background

Alzheimer's disease accounts for 60% to 70% of the world’s 55 million dementia cases, placing a substantial burden on patients, families, healthcare systems, and memory clinics that rely on resource-intensive scans or lumbar punctures for diagnosis. Blood-based biomarkers offer access through a simple blood draw, eliminating the need for a hospital visit. Aβ plaques and tau tangles emerge years before symptoms, so detecting pathology in cognitively unimpaired adults could improve outcomes. Plasma p-tau217 appears promising; however, screening must strike a balance between accuracy, cost, and burden by age. Further research is needed to define blood-first workflows and thresholds for screening.

About the study

This cross-sectional cohort analysis pooled data from 12 cohorts in the United States, Europe, Australia, and Canada. Cognitively unimpaired adults provided plasma for immunoassay measurement of p-tau217; a subset also underwent mass spectrometry of the %p-tau217 ratio for comparison. Aβ status was determined using CSF biomarkers or amyloid PET. PET data were harmonized to the Centiloid scale, and a Centiloid threshold greater than 25 or a positive visual read defined Aβ positivity. Within cohorts, plasma p-tau217 values were also log-transformed, standardized, and modeled in logistic regression with age to estimate the probability of Aβ positivity.

Cut points for plasma p-tau217 were selected in a training split targeting 95% specificity, then evaluated in held-out data to estimate positive predictive value (PPV), negative predictive value (NPV), accuracy, and the proportion of Aβ-positive cases detected. Uncertainty was quantified with bootstrap resampling. Investigators compared a plasma-only strategy with a two-step strategy that confirmed plasma-positive cases by CSF or PET. Recruitment simulations estimated the numbers initially screened, the confirmatory tests required, and the participants enrolled to obtain 100 Aβ-positive individuals under various plasma:CSF:PET cost ratios. Analyses evaluated performance by method and age, and assessed alternative PET thresholds to examine the trade-offs between sensitivity and specificity.

Study results

Among 2916 cognitively unimpaired adults from 12 cohorts, 33.3% were amyloid positive by either CSF or PET. Women comprised 57.2% of the sample, and 38.1% carried at least one apolipoprotein E epsilon 4 (APOE ε4) allele. Using a p-tau217 threshold chosen for 95% specificity and modeling age as a covariate, stand-alone blood testing achieved 81% overall accuracy and a PPV of 79% for classifying amyloid status defined by CSF or PET. At this 95% specificity operating point, the blood test identified about 46% of true amyloid-positive cases, reflecting the challenge of lower prevalence and lower pathology burden in unimpaired populations. More stringent cutoffs further increased PPV but reduced sensitivity, consistent with the expected trade-off.

To raise certainty without sacrificing detection, a two-step workflow was evaluated: individuals positive on plasma proceeded to a confirmatory CSF or PET test. This two-step approach increased PPV to ~91% when CSF confirmed plasma positives against PET as the reference, and up to ~99% when PET confirmed plasma positives against CSF as the reference, and substantially reduced false positives compared with plasma alone. Because CSF markers become abnormal earlier than PET, CSF identified more true positives, whereas PET delivered higher PPV, reflecting different disease phases captured by each modality.

Recruitment simulations translated these metrics into operational impact for a hypothetical trial needing 100 amyloid-positive participants. Compared with imaging alone, using plasma first significantly reduced confirmatory procedures: approximately 139 PET scans were needed after plasma prescreening, versus 536 when PET alone was used; plasma followed by CSF required approximately 124 lumbar punctures. Although this two-step approach reduced false positives and raised PPV, it also detected fewer true positives than using CSF or PET alone, illustrating the deliberate trade-off between certainty and sensitivity. 

In sensitivity analyses, using a more sensitive plasma threshold only when a confirmatory CSF or PET test follows increased the proportion of positives selected while preserving high PPV at the end of the two-step pathway. These differences imply sizable savings in scanner time, radiotracer exposure, and participant burden.

Head-to-head analyses indicated that mass spectrometry of the %p-tau217 ratio provided higher overall accuracy and detected more amyloid-positive cases than immunoassays, while PPVs were similar. Performance varied with age, as PPV rose from approximately 38% in individuals younger than 60 years to approximately 93% in those 80 years or older, whereas the NPV value declined with age, a pattern driven by changes in amyloid prevalence.

Sensitivity analyses supported robustness. Alternate Centiloid thresholds for PET shifted accuracy and PPV in expected directions, and restricting to a single immunoassay platform improved accuracy and detection rates, while leaving PPV essentially unchanged. Collectively, the results show that a blood-first strategy can effectively triage who needs confirmatory testing, striking a practical balance between scale and certainty for the identification of preclinical Alzheimer's disease.

Conclusions

Plasma p-tau217 offers a practical, minimally burdensome entry point for identifying preclinical Alzheimer's disease in cognitively unimpaired adults. As a stand-alone test, it delivers moderate accuracy; pairing plasma screening with confirmatory CSF or PET testing raises PPV, sharply reduces false positives, and optimizes use of invasive or high-cost procedures. Mass spectrometry yields higher accuracy than immunoassays but may be less scalable; standardization and quality control can narrow this gap.

Programs should select thresholds and workflows based on setting, age mix, and intervention risk. Adopting blood-first workflows can accelerate trials and, as therapies expand, support safe, equitable early access to disease-modifying care.