31 Prediction of emerging amyloid and tau pathology in cognitively healthy individuals: a longitudinal CSF study

Abstract

Objectives/aims The importance of identifying emerging rather than widespread amyloid pathology is highlighted by evidence showing that rapid amyloid accumulation associates with cortical atrophy, cognitive deficits and tau deposition even where individuals are in the ‘amyloid negative’ range. The study aimed to test the hypothesis that among cognitively healthy individuals distinct groups exist in terms of amyloid and phosphorylated tau accumulation rates. We also hypothesised that membership to the faster accumulator group can be predicted (using age, genetics, cognition and hippocampal volume). Finally we aimed to identify time points of significant increase in the rate of AD protein accumulation and the associated predictors. Methods The analysis reports data from 263 individuals from the BIOCARD NIH-funded study who had a mean of 2.23 cerebrospinal fluid (CSF) measurements since the study began in 1995. We used latent class mixed-effect models to identify distinct classes in terms of amyloid and p-tau accumulation rates. To investigate non-linear changes of AD-related CSF biomarkers we used generalised additive modelling. For both analyses demographic, genetic (APOE4 genotype) as well as repeated cognitive and structural MRI data were explored as predictors. Results For both amyloid and p-tau latent class models we confirmed the existence of two separate classes: accumulators and non-accumulators. The accumulator and non-accumulator groups differed significantly in terms of baseline AD protein levels (mean age 60 at study inclusion) and slope. APOE4 homozygosity and cognitive reserve predicted amyloid class membership while APOE4 genotype only predicted p-tau class. The non-linear models revealed significant increase in the rate of amyloid and p-tau accumulation at ages 54 and age 55 respectively. APOE4 carrier status associated with earlier age of rapid accumulation start (ages 53 and 50 for amyloid p-tau respectively). Higher hippocampal volume but not cognitive ability was protective for amyloid accumulation. Neither hippocampal volume nor cognition were significant in the p-tau non-linear models. Conclusions The current analysis demonstrates the existence of distinct classes of amyloid and p-Tau accumulators. Predictors for class membership were identified but the overall accuracy of the models was modest highlighting the need for biomarkers that are sensitive to early disease phenotypes. That amyloid and p-tau concentrations remain largely static until the mid-50s with a simultaneous sharp increase in accumulation thereafter also suggests that prediction of these points of inflection can be valuable in targeting specific time periods in disease modification trials.