Year: 2020
Type: article
Abstract: Background Analyses from the Swedish BioFINDER‐2 study have shown that tau burden change seen via 18F‐RO‐948 tau PET tracer replicates the post‐mortem spreading pattern of tau. The greatest longitudinal changes have thus far been seen in Braak regions I/II for amyloid positive (Aβ+) cognitively unimpaired (CU) subjects, and Braak regions III/IV and V/VI for Aβ+ subjects with mild cognitive impairment (MCI) and AD dementia, respectively. In the design of clinical trials, selecting a composite cortical region with the greatest longitudinal effect size (LES) is a key objective in keeping trial sample size practical. In this study, we employed a Lasso optimization method to determine a combination of brain subregions optimized for maximal effect size in the pre‐symptomatic and early AD populations. Method Study participants were scanned at baseline and a targeted first follow‐up scan at 12 months with 18F‐RO‐948 tau PET. Annualized change of tau burden for 116 subregions was computed using FreeSurfer v6.0 and an SUVR method with an inferior cerebellar grey reference region. Subjects were analysed separately based Aβ‐status (using CSF Aβ42/Aβ40) and baseline clinical diagnosis: CU, MCI and dementia due to AD. A Lasso optimization with 5‐fold cross validation was used to explore the optimum combination and weighting of ROIs to maximize LES. Result Longitudinal scans from 14 Aβ+ CU subjects and 18 Aβ+ MCI were included in this analysis. Maximum LES via Lasso for the Aβ+ CU group was 1.01, and 0.81 for the Aβ+ MCI group. Both sets of composite ROIs included three subregions, listed as follows with decreased Lasso weighting: (left‐entorhinal, right‐amygdala, right‐hippocampus) for Aβ+ CU and (left‐fusiform, left‐temporal pole, right‐superiorparietal) for the Aβ+ MCI group. Lasso LES was larger than LES found with Braak staging ROIs: 0.87 for the Aβ+ CU group using Braak I/II, and 0.62 for the Aβ+ MCI group using Braak III/IV. Conclusion A Lasso method determined composite target ROIs with greater effect size than ROIs selected using Braak staging models. Little overlap exists in region sets optimized for pre‐symptomatic and early AD populations, indicating that careful endpoint design is required to detect change in longitudinal studies.
Source: Alzheimer s & Dementia
DOI: 10.1002/alz.045765
Language: English
Subfield: Psychiatry and Mental health
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W3110929904