Relationship between white matter connectivity loss and cortical thinning in cerebral amyloid angiopathy

Abstract

Patients with cerebral amyloid angiopathy (CAA) show loss of white matter connectivity and cortical thinning on MRI, primarily in posterior brain regions. Here we examined whether a potential causal relationship exists between these markers of subcortical and cortical brain injury by examining whether changes in cortical thickness progress in tandem with changes in their underlying connections. Thirty-one patients with probable CAA with brain MRI at two time points were included (follow-up time: 1.3 ± 0.4 years). Brain networks were reconstructed using diffusion MRI-based fiber tractography. Of each network node, we calculated the change in fractional anisotropy-weighted connectivity strength over time and the change in cortical thickness. The association between change in connectivity strength and cortical thickness was assessed with (hierarchical) linear regression models. Our results showed that decline in posterior network connectivity over time was strongly related to thinning of the occipital cortex (β = 0.65 (0.35-0.94), P < 0.001), but not to thinning of the other posterior or frontal cortices. However, at the level of individual network nodes, we found no association between connectivity strength and cortical thinning of the corresponding node (β = 0.009 ± 0.04, P = 0.80). Associations were independent of age, sex, and other brain MRI markers of CAA. To conclude, CAA patients with greater progressive loss of posterior white matter connectivity also have greater progression of occipital cortical thinning, but our results do not support a direct causal relationship between them. The association can be better explained by a shared underlying mechanism, which may form a potential target for future treatments. Hum Brain Mapp 38:3723-3731, 2017. © 2017 Wiley Periodicals, Inc.

Keywords: atrophy; cerebral small vessel diseases; connectome; diffusion tensor imaging; diffusion tractography.

Figure 1

Flowchart of the brain network reconstruction. For each subject, whole‐brain fiber tract reconstructions (A) were parcellated into 78 cortical regions according to a standard grey matter atlas (B). Only the network of the hemisphere free of intracerebral hemorrhages was examined. Each cortical region represents a network node (red spheres). The connectivity strength (i.e., the mean FA of all connected white matter tracts) and the mean cortical thickness of each node were calculated. Two types of analyses were performed: mean change in posterior network connectivity strength was related to change in lobar cortical thickness (C) and change in connectivity strength of each individual posterior node was related to change in cortical thickness of the corresponding node (D). [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 2

(A) Longitudinal association between change in mean connectivity strength of all posterior network nodes (follow‐up–baseline) and change in cortical thickness of the occipital lobe (β = 0.63, P < 0.001). Units are adjusted for age, sex, and follow‐up time. (B) Heat map of the association between decline in posterior white matter connectivity strength and cortical thinning of each network node. The strongest positive correlations are found for nodes within the occipital cortex (red/orange nodes indicating the occipital superior, occipital middle, calcarine, cuneus, precuneus, and fusiform cortex; all P < 0.05 uncorrected for multiple comparisons). An association with increased cortical thickness was found for the superior temporal pole (black node). [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 3

No longitudinal association between change in FA‐weighted connectivity strength of each individual network node and change in cortical thickness of the corresponding node. Pooled analysis of all 22 posterior nodes of all subjects (n = 682). Units are standardized z‐scores, adjusted for age, sex, and intrasubject variance in hierarchical mixed models analysis; the mean of each axis is therefore zero. No association was found (β = 0.009 ± 0.04, P = 0.80).