Robustness of plasma p-Tau217 diagnostic thresholds for Alzheimer's disease across various clinical populations and laboratory environments

Background and ObjectivesBlood-based biomarkers (BBMs), especially tau phosphorylated at threonine 217 (p-tau217), offer a minimally invasive approach to diagnose Alzheimers disease (AD) with strong potential for clinical implementation. However, the robustness and transferability of diagnostic thresholds across sites remain uncertain. This study aimed to evaluate BBMs ability to predict amyloid and tau status in a deeply phenotyped AD research cohort and to test p-tau217 cutoffs across diverse clinical populations and analytical instruments. MethodsWe included four Western European cohorts (total n=411): an exploration cohort from Brussels with amyloid and tau-positron emission tomography (PET) data (n=215), and clinical validation cohorts from Paris (memory clinic, n=117, and monogenic FTLD, n=43) and Bordeaux (early-onset neurodegenerative disorders, n=36). BBMs concentrations (p-tau217, p-tau181, NfL, BD-tau, A{beta}42, A{beta}40) were measured. Of these, p-tau217 was measured at each site using three Lumipulse analyzers (G600II in Brussels and Bordeaux; G1200 in Paris). Amyloid status was determined by PET or CSF A{beta}42, and tau status by tau-PET or CSF p-tau181. ResultsPlasma p-tau217 outperformed other BBMs, and their related ratios. In the exploration cohort, p-tau217 closely related to amyloid and tau PET, clearly separating A{beta}+ from A{beta}- individuals (AUC=0.96; optimal cutoff 0.193 pg/mL with [~]92% sensitivity/specificity) and tracking tau-PET Braak stages (AUC=0.93-0.98). Using amyloid-derived thresholds, p-tau217 detected medial temporal and neocortical tauopathy with >90% sensitivity and specificity at stricter dual 95-97.5% sensitivity/specificity cutpoints (0.142/0.256 pg/mL and 0.110/0.319 pg/mL). In validation cohorts, the 0.193 pg/mL cutoff accurately discriminated clinical-biological AD from non-AD in routine care (AUC=0.98; [~]93% sensitivity/specificity), in early-onset dementia (AUC=0.94; >92% sensitivity/specificity), and in the monogenic FTLD cohort (89% specificity). In routine care, gray zones (8-25%) were largely resolved with second-line CSF testing (>75%). Elevated p-tau217 was rare in non-AD, mainly in FTLD-ALS and some older individuals with possible AD copathology. p-tau217/A{beta}42 and p-tau217/A{beta}42/A{beta}40 ratios added no diagnostic value. Discussionp-tau217 thresholds maintained high diagnostic accuracy (>90%) across independent sites, analytical platforms, and various clinical situations, supporting their robustness and transferability. These observations support implementing p-tau217 as a reliable, scalable test for detecting AD pathology, for diagnostic work-ups of patients with objective cognitive impairment, not for general population screening.