Targeting pre-symptomatic AD individuals: Deep cognitiveassessment and plasma biomarkers to predict tau aggregation

BACKGROUND: Cognitively unimpaired (CU) individuals with both elevated brain amyloid load and tau burden in the medial temporal (MTL) and temporal neocortex (NEO) face high risk of short-term cognitive decline (≤5 years; risk=50%). However, identifying these individuals in the population remains challenging due to their low prevalence (8-10%), the cost and invasiveness of validated biomarkers. Cognitive measures and blood-based biomarkers offer promising scalable alternatives, but most plasma biomarkers are more closely associated with amyloid than tau aggregates, and tau-specific measures remain poorly defined. This study assessed whether specific cognitive tasks, including tasks targeting the functions of the first affected regions by tauopathy, and blood-based biomarkers can predict early tau aggregation. METHOD: Seventy-seven CU participants completed the Visual Short-Term Binding Test (VSTMBT), the Conceptual Matching Task (CMT), the cognitive tests required for the Preclinical Alzheimer's Cognitive Composite (PACC5), a blood-test, [18F]-MK6240 tau-PET imaging, 3T-MRI, and amyloid (A) status determination (A+ for Centiloid≥ 20 or cerebrospinal fluid amyloid-beta42≤437 pg/mL). The VSTMBT and CMT (Figure 1) involve fined-grained perceptual and conceptual discrimination, respectively, supposedly relying on the transentorhinal cortex. The sample included 55 A- CU and 22 A+ CU (Table 1). Standard Uptake Value ratios (SUVr) were computed for MTL and temporal NEO region of interests (ROI; Ossenkoppele et al., 2022; reference=grey cerebellar). Plasma p-tau217 and p-tau181 levels were quantified using Lumipulse and SIMOA. Univariate regression models predicting ROI tau burden based on demographics (age, sex, education), cognitive performance (VSTMBT, CMT, PACC5), and plasma p-tau species (2 ROIs x 8 predictors) were conducted to select contributing predictors (highlighted in green in Table 2) for further stepwise regression analyses (both directions). RESULT: For MTL tau burden, optimal model fit (initial AIC=4.88, final AIC=3.55) was found with the VSTMBT (b = -0.01, SE=0.004, p = .004) and plasma p-tau217 level (b = 1.18, SE=0.276, p <.001) as predictors. For temporal NEO tau burden (initial AIC=41.04, AIC=-44.92), best fit was found with the PACC5 (b = -0.09, SE=0.04, p = .027) plasma p-tau217 (b = 0.77, SE=0.15, p <.001; b = 0.76) and p-tau181 (b = -0.003, SE=0.002, p = .125) levels as predictors. CONCLUSION: Plasma p-tau217 predicted tau burden across both ROIs, alongside different cognitive tasks depending on the ROI, likely reflecting their associated cognitive functions.