Structural Covariance of Cortical Gyrification at Illness Onset in Treatment Resistance: A Longitudinal Study of First-Episode Psychoses

Abstract

Treatment resistance (TR) in patients with first-episode psychosis (FEP) is a major cause of disability and functional impairment, yet mechanisms underlying this severe disorder are poorly understood. As one view is that TR has neurodevelopmental roots, we investigated whether its emergence relates to disruptions in synchronized cortical maturation quantified using gyrification-based connectomes. Seventy patients with FEP evaluated at their first presentation to psychiatric services were followed up using clinical records for 4 years; of these, 17 (24.3%) met the definition of TR and 53 (75.7%) remained non-TR at 4 years. Structural MRI images were obtained within 5 weeks from first exposure to antipsychotics. Local gyrification indices were computed for 148 contiguous cortical regions using FreeSurfer; each subject's contribution to group-based structural covariance was quantified using a jack-knife procedure, providing a single deviation matrix for each subject. The latter was used to derive topological properties that were compared between TR and non-TR patients using a Functional Data Analysis approach. Compared to the non-TR patients, TR patients showed a significant reduction in small-worldness (Hedges's g = 2.09, P < .001) and a reduced clustering coefficient (Hedges's g = 1.07, P < .001) with increased length (Hedges's g = -2.17, P < .001), indicating a disruption in the organizing principles of cortical folding. The positive symptom burden was higher in patients with more pronounced small-worldness (r = .41, P = .001) across the entire sample. The trajectory of synchronized cortical development inferred from baseline MRI-based structural covariance highlights the possibility of identifying patients at high-risk of TR prospectively, based on individualized gyrification-based connectomes.

Keywords: MRI; clozapine; first-episode psychosis; gyrification; longitudinal; schizophrenia; treatment-resistant.

Fig. 1.

Depicts 3 steps (referred to as A, B, and C) followed to construct an individual-specific LGI network in the present study. (A) For a single group of (N-1) subjects (treatment resistance [TR] or non-TR), a specific group-based network is constructed by the correlations between LGI values based on regional LGI measures from 148 parcellations in this group, with the exclusion of a single subject i. This group network (based on N-1 matrix) has the “normative” covariance structure of that group’s gyrification pattern. (B) A new subject i belonging to that patient group is added to the group, and the perturbed network with this additional individual is constructed in the same way as the (N-1) matrix. The difference between the (N) and the (N-1) network is due to the individual i (or j or k…). (C) An individual contribution-based network is constructed using the difference of the corresponding edge between the (N) and (N-1) matrix. For illustrative purposes, only one group (TR) and only 3 nodes are shown. LGI-based networks in this study were made of 148 nodes, each node representing a single region of the parcellation scheme.

Fig. 2.

Plot depicting the relationship, adjusted for TR status, age and sex and the positive symptom dimension, between the residuals of small-worldness index (σ) and positive symptom dimension in the entire sample. Also, see supplementary figure 2.