Differential network‐based propagation of cortical thinning in atrophy subtypes in non‐demented individuals

• Published in: Alzheimer's & dementia 2024-12, Vol.20 (S9), p.n/a
• Main Authors: Lorenzini, Luigi, Tranfa, Mario, Pieperhoff, Leonard, ten Kate, Mara, Pontillo, Giuseppe, Alves, Isadora Lopes, Shekari, Mahnaz, Altomare, Daniele, den Braber, Anouk, Ritchie, Craig W, Boada, Mercè, Marquié, Marta, Vandenberghe, Rik, Luckett, Emma S., Hanseeuw, Bernard J, Visser, Pieter Jelle, Schöll, Michael, Frisoni, Giovanni B., Birck, Cindy, Mett, Anja, Stephens, Andrew W., Gismondi, Rossella, Buckley, Christopher, Farrar, Gill, Jessen, Frank, Gispert, Juan Domingo, Mutsaerts, Henk‐Jan, García, David Vállez, Wink, Alle Meije, Collij, Lyduine E., Barkhof, Frederik
• Format: Article
• Language: English
• Subjects: Alzheimer's Imaging Consortium
• ISSN: 1552-5260; 1552-5279
• DOI: 10.1002/alz.093975

Background Different patterns of atrophy exist in the dementia stage of AD. However, little is known about the heterogeneity of atrophy patterns and the mechanisms that drive subsequent propagation of the disease in the preclinical stages. Method From the AMYPAD‐PNHS cohort, we included a total of 1323 non‐demented individuals, including 1094 amyloid‐negative, and 229 amyloid‐positive participants (Table 1). Gray matter thickness in 100 regions of interest was measured using FreeSurfer. Global cortical amyloid burden was derived from amyloid PET scans in Centiloids (CL). Individuals were assigned to atrophy subtypes based on cortical thickness measures using Non‐Negative Matrix Factorization (Figure 1). Multinomial regression models were used to study differences between subtypes in age, sex, global CL and APOE E4 carriership. Linear regression models were used to investigate the effect of time on atrophy within each subtype. Coordinated deformation models (Figure 1, Shafiei et al. 2023) were used to estimate network‐based thickness changes in each region, by multiplying the effect of time in connected regions by the strengths of their connections. Connectivity strength was based on templates of functional, structural and morphometric similarity networks, and allowed us to evaluate differential mechanisms. Significance was tested using a permutation approach. Results Three significant subtypes of regional GM atrophy were identified:‘limbic predominant’, ‘typical AD’, and ‘diffuse cortical’ (Figure 2). Individuals assigned to the limbic predominant subtype were significantly older, while the typical AD subtype had more APOE E4 carriers compared to the others (all p