A data‐driven latent atrophy factor model reveals differential associations between relative frontal atrophy patterns and specific neuropsychiatric symptoms in Alzheimer’s disease

Abstract

AbstractBackgroundto examine the relative role of the frontal cortex in neuropsychiatric symptoms in Alzheimer’s disease (AD), we assessed associations between latent atrophy factors and neuropsychiatric symptoms in AD.MethodWe employed a data‐driven Bayesian modelling framework based on Latent Dirichlet Allocation (LDA) to identify latent atrophy factors in a cohort of 1135 Aβ‐positive MCI and AD patients from the Amsterdam Dementia Cohort (age: 66.4±7.44, 50% male, MMSE: 22.3±0.1). The model uses standardized gray matter density images as input, adjusted for age, sex, intracranial volume and scanner field strength, and yields voxelwise probabilistic maps for a predetermined number of atrophy factors, providing data–driven factor expression loadings per individual. These factor expressions were correlated to scores on the Neuropsychiatric Inventory (NPI) using general linear models, adjusting for sex, age, education and MMSE scores.Resultthe LDA model revealed a frontal (dorsolateral and insular), a medial temporal lobe (hippocampal and entorhinal), a temporoparietal and a late‐stage (sensorimotor and occipital) factor (Figure 1). The highest percentage of individual expressions was found in the late‐stage factor (34.0±0.6), followed by the MTL (27.3±0.6), TPC (22.0±0.5) and frontal (16.7±0.5) factors (Figure 2). Higher frontal expression relative to both the MTL and late‐stage factors was associated with a higher odds for apathy (MTL; OR=1.012, p=0.010, late‐stage; OR=1.011, p=0.046; Figure 3). Negative associations, however, were found between relative frontal atrophy and disinhibition, anxiety and nighttime behaviors (Figure 3); a higher frontal load relative to each reference factor related to a lower odds for disinhibition (frontal‐MTL: OR=0.978, p=0.035; frontal‐TPC: OR=0.969, p=0.042; frontal‐‘late‐stage’: OR=0.976, p=0.029), a higher frontal load relative to MTL and late‐stage factors were associated with a lower odd for anxiety (frontal‐MTL: OR=0.984, p=0.015; frontal‐‘late‐stage’: OR=0.983, p=0.003) and night‐time behaviors (frontal‐MTL: OR=0.982, p=0.023; frontal‐‘late‐stage’: OR=0.980, p=0.005).ConclusionRelative frontal atrophy was related to more frequent apathy and less frequent disinhibition, anxiety and night‐time behaviors and may point towards more AD‐typical regions as determining neurobiological regions for disinhibition, anxiety and night‐time behaviors than frontal regions in AD. Alternatively, disinhibition may be more dependent on the right frontal pole, less prominently involved in the current frontal factor.