Evidence for Dissociable Linkage of Dimensions of Psychopathology to Brain Structure in Youths

Abstract

Objective: High comorbidity among psychiatric disorders suggests that they may share underlying neurobiological deficits. Abnormalities in cortical thickness and volume have been demonstrated in clinical samples of adults, but less is known when these structural differences emerge in youths. The purpose of this study was to examine the association between dimensions of psychopathology and brain structure.

Methods: The authors studied 1,394 youths who underwent brain imaging as part of the Philadelphia Neurodevelopmental Cohort. Dimensions of psychopathology were constructed using a bifactor model of symptoms. Cortical thickness and volume were quantified using high-resolution 3-T MRI. Structural covariance networks were derived using nonnegative matrix factorization and analyzed using generalized additive models with penalized splines to capture both linear and nonlinear age-related effects.

Results: Fear symptoms were associated with reduced cortical thickness in most networks, and overall psychopathology was associated with globally reduced gray matter volume across all networks. Structural covariance networks predicted psychopathology symptoms above and beyond demographic characteristics and cognitive performance.

Conclusions: The results suggest a dissociable relationship whereby fear is most strongly linked to reduced cortical thickness and overall psychopathology is most strongly linked to global reductions in gray matter volume. Such results have implications for understanding how abnormalities of brain development may be associated with divergent dimensions of psychopathology.

Keywords: Adolescents; Brain; Child Psychiatry; Cortical Thickness; Development; MRI.

Figure 1.. Correlated dimensions of psychopathology show a high degree of overlapping symptoms.

A) An exploratory factor analysis of 112 psychiatric symptoms identified four correlated dimensions of psychopathology: anxious-misery, psychosis, behavioral, and fear, which show a high degree of overlap across dimensions and diagnostic screening categories. Here we show the mean factor scores of each dimension (anxious-misery, psychosis, behavioral, and fear) in the related screening diagnoses. B) A confirmatory bifactor analysis constrained the dimensions of psychopathology (anxious-misery, psychosis, behavioral, and fear) to be orthogonal, and extracted a common factor (overall psychopathology). C) The orthogonal factors load more specifically onto the relevant disorders. Sample sizes for each diagnostic screening category are shown in parentheses. GAD = generalized anxiety disorder; Depress = depressive disorders; Psych = psychosis; ODD = oppositional defiant disorder; ADHD = attention-deficit/hyperactivity disorder; Con = conduct disorder; PTSD = posttraumatic stress disorder; Agora = agoraphobia; Soc Anx = social anxiety disorder; Spec Ph = specific phobia; Sep Anx = separation anxiety disorder; TD = typically developing.

Figure 2.. Structural covariance networks delineated by NMF.

Structural covariance networks are shown for the 18-network solution, with the spatial distribution of each network indicated by loadings at each voxel in arbitrary units (shown with the color bar, where warmer colors correspond to higher values). High symmetry can be seen between the left (L) and right (R) hemispheres. The anatomical coverage of each structural covariance network was as follows: 1) cingulate cortex; 2) medial temporal cortex; 3) temporal pole; 4) dorsolateral prefrontal cortex; 5) posterior cingulate cortex; 6) superior parietal cortex; 7) superior temporal cortex; 8) dorsal prefrontal cortex; 9) insular cortex; 10) fusiform cortex; 11) inferior temporal cortex; 12) right lateral occipital cortex; 13) subgenual cingulate, anterior cingulate, and anterior insula; 14) inferior prefrontal cortex; 15) inferior parietal cortex; 16a) precuneus and 16b) temporoparietal junction; 17) lingual gyrus; 18) medial occipital cortex.

Figure 3.. Fear is associated with reduced cortical thickness in multiple structural covariance networks.

Mass-univariate analyses using GAMs that controlled for linear and nonlinear age, sex, and image quality revealed that fear symptoms were associated with reduced cortical thickness in multiple networks. This association was maximal in networks such as the temporal-parietal junction and posterior cingulate cortex (Figure 3A; network 16). Significant associations were also present in networks that included the anterior cingulate, anterior insula, and subgenual cingulate cortex (Figure 3C; network 13). Composite network boundaries were obtained by assigning each voxel to one of the 18 networks with the highest loading for that voxel. Multiple comparisons were accounted for using the False Discovery Rate (Q<0.05). Dotted lines on scatterplots represent the 95% confidence interval.

Figure 4.. Overall psychopathology is associated with reduced volume globally, while anxious-misery is associated with greater volume in multiple structural covariance networks.

Mass-univariate analyses using GAMs that controlled for linear and nonlinear age, sex, and image quality revealed that overall psychopathology was associated with reduced volume across the brain (Figure 4A and 4B). Conversely, anxious-misery symptoms were associated with increased volume in most networks (Figure 4C and 4D). Composite network boundaries were obtained by assigning each voxel to one of the 18 networks with the highest loading for that voxel. Multiple comparisons were accounted for using the False Discovery Rate (Q<0.05). Dotted lines on scatterplots represent the 95% confidence interval.