Plasma Alzheimer biomarkers in the 1946 British birth cohort: predicting conversion to amyloid PET positivity over a two‐year follow‐up period

Abstract

AbstractBackgroundIn the Insight 46 sub‐study of the 1946 British birth cohort we investigated concordance of baseline plasma biomarkers (age 71) with cerebral amyloid PET status at two time‐points (ages 71 and 73).MethodBaseline plasma samples were tested using Simoa (Quanterix) commercial (NfL, GFAP, Aβ42, Aβ40, p‐tau181V2) and homebrew digital immunoassays (p‐tau181, p‐tau231: University of Gothenburg; NT1 tau: Harvard University). Logistic regression and ROC analysis were used to assess concordance of single biomarkers with cerebral amyloid PET status at each timepoint, defined as 18F‐florbetapir SUVR> = 1.071 normalized to whole cerebellum without partial volume correction (13 centiloids). Areas under the curve were compared to that from the base model (age, sex, APOE ε4 carrier status) using De Long tests. Combinations of the top five biomarkers that had singly significantly improved on the base model were compared using Akaike Information Criteria (AIC). Differences between biomarker values across the three amyloid PET groups (both scans negative, baseline negative and follow‐up positive, both positive) were assessed using Kruskal‐Wallis ANOVA and post hoc Dunn’s tests with Bonferroni correction.ResultBaseline assessment of 502 individuals (mean age 70.7 years (s.d. 0.7), 51% male, 29.6% APOE ε4 carriers) yielded 432 individuals with plasma and amyloid PET available at the first timepoint (19.4% PET‐positive) and 336 at the second timepoint (25.3% PET‐positive). No single biomarker outperformed the base model in predicting amyloid PET status at either timepoint. However, each of Aβ40/42, all the p‐tau biomarkers, NT1 tau, and GFAP improved base model performance at both timepoints (Table 1). A model combining base model covariates with Aβ40/42, p‐tau181 homebrew and GFAP gave the best fit for follow‐up amyloid PET status, (AIC 292, compared with the base model AIC of 355 ‐ Table 2). Individuals who converted from PET‐negative to positive had higher median values of p‐tau181 (both assays) and p‐tau231 at baseline than those who remained amyloid PET‐negative (Figure 1).ConclusionCombining baseline plasma Aβ40/42, p‐tau181 and GFAP gave the best prediction of amyloid PET status over 2‐year follow‐up. Individuals who converted from amyloid PET‐negative to positive had higher values of p‐tau181 and p‐tau231 than those who remained PET‐negative.