Ability of tau‐PET, phospho‐tau217, NfL and cortical thickness to predict short‐term cognitive decline in early symptomatic Alzheimer’s disease

Abstract

AbstractBackgroundBiomarkers for the prediction of cognitive decline in patients with mild cognitive impairment (MCI) due to Alzheimer’s disease (AD) and mild AD dementia are needed for inclusion of suitable patients in clinical trials. We therefore assessed the ability of tau‐PET, blood and cerebrospinal fluid (CSF) phospho‐Tau217 (pTau217), CSF neurofilament light (NfL) and cortical thickness to predict cognitive decline early symptomatic AD.Method73 amyloid‐β positive (CSF Aβ42/Aβ40<0.752) participants with MCI and mild AD dementia (MMSE ≥ 22) were assessed at baseline with cognitive testing, blood and cerebrospinal fluid (CSF)‐sampling, MRI and [18F]RO948 PET and were followed longitudinally with annual cognitive testing over 24 months. Slopes of change in MMSE and preclinical Alzheimer cognitive composite (PACC) were calculated over two years. We used linear regression models to assess the ability of biomarkers (plasma pTau217, CSF‐pTau217, temporal meta‐ROI [18F]RO948 PET standardized uptake value ratios (SUVR), CSF‐NfL and MRI “AD‐signature” cortical thickness) to predict cognitive decline both individually and in combination. The linear regression models were adjusted for age, sex, education, and baseline MMSE.ResultAll tested biomarkers individually showed significant associations with cognitive decline using both MMSE and PACC as outcome variables. For predicting longitudinal change in MMSE using individual biomarkers, [18F]RO948 SUVR showed the highest R2‐value (0.39, 95% CI [0.28‐0.67]) and best model fit (lowest Akaike Information Criterion score)(Table 1). Adding other biomarkers yielded a slightly higher R2‐value (0.44 [0.33‐0.63]), but did not improve the model fit. In a sensitivity analysis we performed a similar analysis in a subset of participants having undergone longitudinal testing with a modified PACC test battery (n=53). Results again showed the best performance for [18F]RO948 SUVRs, when tested alone, but with improved model fit and R2‐values with the addition of cortical thickness and CSF‐NfL data (Table 2).ConclusionTau‐PET imaging outperformed CSF, plasma and MRI measures when predicting cognitive decline in prodromal AD and mild AD dementia. Specifically, tau‐PET provided the best prediction alone for change in MMSE and in combination with NfL and cortical thickness for change in PACC, indicating that it would be a suitable biomarker for inclusion in clinical trials.