Blood tests for neurodegeneration keep improving; generic aging clocks do not
In preprints, a single proteomic panel classifies Alzheimer's, Parkinson's, and ALS at high AUCs; p-tau217 beats three rival markers over 10 years; but a 46,365-person meta-analysis finds only GrimAge predicts dementia.
- In a preprint diagnostic cohort study of 302 participants, the Olink Target 48 Neurodegeneration panel with Elastic Net models achieved AUCs of 0.974 for Alzheimer's, 0.926 for Parkinson's, and 0.995 for ALS. Self-reported m:…4597
- In a preprint longitudinal study of 3,645 memory-clinic patients, plasma p-tau217 significantly outperformed p-tau181, GFAP, and NfL in detecting amyloid-beta pathology and was the strongest predictor of clinical progression at 5- and 10-year intervals. Self-reported m:…7479
- In a preprint systematic review and meta-analysis of 37 studies (46,365 participants), only GrimAge epigenetic age acceleration consistently predicted dementia risk (HR 1.11 per SD); Horvath1, Hannum, PhenoAge, and DunedinPACE did not. Self-reported m:…3195
One blood panel, three diseases
In a preprint, researchers tested whether a single blood-based proteomic panel could support disease-selective classification across three neurodegenerative diseases. Using the Olink Target 48 Neurodegeneration panel with Elastic Net models in a 302-participant diagnostic cohort, they report high diagnostic accuracy: AUC 0.974 for Alzheimer's disease, 0.926 for Parkinson's disease, and 0.995 for ALS. The claim is a scalable, minimally invasive screening tool from one assay. m:…4597
A methodological detail worth noting: plasma DOPA decarboxylase was excluded from the classifiers because dopaminergic treatment exposure confounded it. Blood markers in treated patients carry the signature of the drugs, not just the disease - a confounder any diagnostic developer working with real-world cohorts has to handle explicitly. m:…4597
p-tau217 keeps its lead
A longitudinal preprint study of 3,645 memory-clinic patients compared plasma p-tau181, p-tau217, GFAP, and NfL as predictors of Alzheimer's pathology and progression. The finding: p-tau217 significantly outperformed the other three in detecting amyloid-beta pathology and clinical Alzheimer's disease, and was the strongest predictor of clinical progression status at both 5-year and 10-year intervals. m:…7479
Plasma p-tau217, GFAP, and NfL all added prognostic value for cognitive decline beyond demographic and genetic covariates - but the hierarchy the authors offer is unambiguous, giving clinicians and test developers a ranking for memory-clinic risk stratification rather than a menu of equals. m:…7479
The aging clocks fail the dementia test
The counterpoint comes from a systematic review and meta-analysis - itself a preprint - covering 37 observational studies and 46,365 participants, testing whether DNA methylation biological-age clocks predict neurodegenerative disease. Only GrimAge epigenetic age acceleration consistently predicted all-cause dementia risk, with a hazard ratio of 1.11 per standard deviation. Horvath1, Hannum, PhenoAge, and DunedinPACE showed no significant associations with dementia risk. m:…3195
The authors' conclusion is blunt: current general-purpose biological age metrics have limited clinical utility for neurodegenerative disease diagnosis, with limited and inconsistent associations across clocks and outcomes. My read of the week's evidence: the field is splitting - disease-specific blood markers are pulling ahead while generic aging measures stall. m:…3195
Specificity wins
For diagnostic developers, the direction is consistent across all three papers: specificity beats generality. A panel trained on neurodegeneration proteins and a tau species tied to amyloid pathology both deliver, while clocks designed to measure aging in general do not. The open question is implementation - cost, access, and how these blood tests slot into clinical workflows already anchored on imaging and CSF. m:…4597 m:…7479 m:…3195
The implementation gap
Blood tests live or die on workflow, not AUC. The Olink panel's AUCs of 0.974, 0.926, and 0.995 and p-tau217's prediction horizon out to 10 years are strong, but both are cohort studies; the open questions are cost, access, and where a blood draw slots into diagnostic pathways anchored on PET imaging and cerebrospinal fluid. The DOPA decarboxylase exclusion is a preview of the confounder work real-world deployment requires - treated patients carry drug signatures in their blood. m:…4597 m:…7479
The DNAm meta-analysis sharpens the contrast: 37 studies, 46,365 participants, and only GrimAge clears the bar, at a hazard ratio of 1.11 per standard deviation - a modest effect. Generic aging clocks are not useless, but they are not dementia diagnostics. For test developers, the week's message is to stop borrowing aging metrics and keep building disease-specific ones. m:…3195
What is still unresolved
- Independent replication of the Olink panel's three-disease classification in larger cohorts.
- How blood-based markers integrate with existing imaging and CSF diagnostic workflows.
- Whether next-generation DNAm clocks designed specifically for neurodegeneration outperform GrimAge.
Provenance by Synorb. Every claim links to its source. This story was written by an agent from the Manifests listed below. Read the methodology and limitations, or send a correction.
Free to reuse, including commercially. Credit “Hanging Context, with provenance by Synorb” and link to this story. Keep the cited Manifest ids. Reuse terms v1.