The potential of key Alzheimer’s plasma biomarkers to mimic tau PET MUBADA-based disease staging

Abstract Background Patients with Alzheimer’s disease (AD) with a low to intermediate tau-PET burden might benefit most from anti-amyloid treatment. Staging tau burden with plasma biomarkers would offer a scalable alternative to staging with tau-PET. This study investigated whether key plasma biomarkers P-tau217, P-tau181, Aβ42/40, GFAP and NfL can be used to accurately stage amyloid status (A-/A+) and tau-PET burden, and evaluated the relation of such a plasma-based staging system with cognitive outcomes over time. Methods We included 105 participants with subjective cognitive decline ( n = 27 A-, n = 18 A+), A+ mild cognitive impairment ( n = 10) or A + AD-dementia ( n = 50) from the Amsterdam Dementia Cohort who underwent [ 18 F]flortaucipir PET-burden assessment (T low , T intermediate or T high ; based on MUBADA SUVr) and longitudinal cognitive assessment (average follow-up: 4.1 ± 3.5 years). AD-related plasma biomarkers were measured with Simoa. Discriminative performance (AUC) of each marker was compared using ROC analysis, and combined utility was assessed with logistic regression. Subsequently, cutoffs were established aiming for 90%-specificity, to regroup participants into a plasma-based staging scheme. Age-, sex- and education-adjusted linear mixed models (LMM) were performed to compare associations of plasma versus PET-based staging with longitudinal cognition. Results 27 participants were A-T PET_low , 22 A+T PET_low , 27 A+T PET_int and 29 A+T PET_high . To discriminate A-T PET_low participants from A+T PET_low/int participants, P-tau217 performed best among all measured markers P-tau217, P-tau181, Aβ42/40, GFAP and NfL (AUC = 0.92 [95% CI: 0.856–0.985]). To discriminate A+T PET_high participants from A+T PET_low/int participants, also P-tau217 performed best among all markers (AUC = 0.74 [95% CI: 0.618–0.862]). A combination of Wald’s backward-selected plasma markers did not statistically improve discriminative performance (DeLong’s p > 0.05; two-marker combinations selected). Applying two cutoffs for P-tau217 as well as for the two-marker combinations, at 90% specificity to discriminate subsequent groups, we derived two plasma-based staging schemes. While the tau-PET staging scheme significantly and consistently associated with cognitive performance and decline across cognitive domains in LMMs, the plasma staging schemes did not. Conclusions Performance of plasma-based staging approaches developed in this study were high when discriminating individuals without amyloid pathology, while this was moderate when discriminating amyloid positive individuals with a high tau-PET burden. Our LMM findings visualize that tau staging in amyloid-positive individuals remains optimally performed with tau-PET scans.