Comparison of brain-derived versus conventional pTau biomarkers in CSF and plasma for detecting Alzheimer’s disease

Abstract Brain-derived phosphorylated tau species are postulated to improve the detection of Alzheimer’s disease pathology by mitigating non-specific increases. However, their true diagnostic performance, relative to conventional CSF and plasma phosphorylated tau assays, remains uncharacterized. The objective was to compare the diagnostic performance of brain-derived and conventional phosphorylated tau assays across multiple analytes (phosphorylated tau 181, phosphorylated tau 217, phosphorylated tau 231) in CSF and plasma, and to examine biomarker differences across amyloid-β status and cognitive groups. This study included 1,098 participants with amyloid-PET imaging and CSF samples from the Alzheimer’s Disease Neuroimaging Initiative, and a subset of 925 individuals with plasma samples. Diagnostic performance of brain-derived and conventional phosphorylated tau assays was assessed by receiver operating characteristic curves, with calculation of area under the curve, sensitivity, specificity, and optimal cutpoints. Associations with amyloid-β burden were evaluated using Spearman correlations, and fold-changes between amyloid-PET-positive and amyloid-PET-negative participants were computed in cognitively unimpaired and cognitively impaired subgroups with 95% confidence intervals from 2,000 bootstrap resamples. Brain-derived phosphorylated tau assays showed significantly higher diagnostic performance than conventional phosphorylated tau assays at the same phosphorylation site in both CSF and plasma in the overall sample and in cognitively impaired individuals, except for CSF phosphorylated tau 217. While CSF brain-derived phosphorylated tau correlations with amyloid-PET were similar to conventional CSF assays, pairwise comparisons using DeLong’s test revealed significantly higher diagnostic accuracy for brain-derived assays. Receiver operating characteristic performance was generally better in cognitively impaired participants than in cognitively unimpaired participants. Additionally, using a three-range method with cutpoints corresponding to 90% sensitivity and 90% specificity, plasma brain-derived phosphorylated tau 181, phosphorylated tau 217, and brain-derived phosphorylated tau 217 had <20% of participants in the intermediate range overall and in cognitively impaired individuals. In summary, brain-derived phosphorylated tau assays improve detection of amyloid-β positivity relative to conventional assays, especially in plasma, and provide robust differentiation of amyloid-PET-positive and amyloid-PET-negative participants. If these findings are confirmed using fully automated diagnostic platforms that are intended for use in clinical settings, brain-derived phosphorylated tau assays may support future blood-based biomarker triage workflows for identifying Alzheimer’s disease.