Plasma phosphorylated tau 217 in preclinical Alzheimer’s disease

An accurate blood test for Alzheimer’s disease that is sensitive to preclinical proteinopathy and cognitive decline has clear implications for early detection and secondary prevention. We assessed the performance of plasma phosphorylated tau 217 ([Formula: see text]) against brain PET markers of amyloid [[Formula: see text]-labelled Pittsburgh compound B (PiB)] and tau ([Formula: see text] MK-6240) and its utility for predicting longitudinal cognition. Samples were analysed from a subset of participants with up to 8 years follow-up in the Wisconsin Registry for Alzheimer’s Prevention (WRAP; 2001–present; plasma 2011–present), a longitudinal cohort study of adults from midlife, enriched for parental history of Alzheimer’s disease. Participants were a convenience sample who volunteered for at least one PiB scan, had usable banked plasma and were cognitively unimpaired at first plasma collection. Study personnel who interacted with participants or samples were blind to amyloid status. We used mixed effects models and receiver–operator characteristic curves to assess concordance between plasma [Formula: see text] and PET biomarkers of Alzheimer’s disease and mixed effects models to understand the ability of plasma [Formula: see text] to predict longitudinal performance on WRAP’s preclinical Alzheimer’s cognitive composite (PACC-3). The primary analysis included 165 people (108 women; mean age = 62.9 [Formula: see text] 6.06; 160 still enrolled; 2 deceased; 3 discontinued). Plasma [Formula: see text] was strongly related to PET-based estimates of concurrent brain amyloid ([Formula: see text] = 0.83 (0.75, 0.90), P < 0.001). Concordance was high between plasma [Formula: see text] and both amyloid PET (area under the curve = 0.91, specificity = 0.80, sensitivity = 0.85, positive predictive value = 0.58, negative predictive value = 0.94) and tau PET (area under the curve = 0.95, specificity = 1, sensitivity = 0.85, positive predictive value = 1, negative predictive value = 0.98). Higher baseline [Formula: see text] levels were associated with worse cognitive trajectories ([Formula: see text] = −0.07 (−0.09, −0.06), P < 0.001). In a convenience sample of unimpaired adults, plasma [Formula: see text] levels correlate well with concurrent brain Alzheimer’s disease pathophysiology and with prospective cognitive performance. These data indicate that this marker can detect disease before clinical signs and thus may disambiguate presymptomatic Alzheimer’s disease from normal cognitive ageing.