Structural correlates of psychopathological symptom dimensions in schizophrenia: A voxel-based morphometric study

Abstract

Structural neuroimaging has substantially advanced the neurobiological research of schizophrenia by describing a range of focal brain alterations as possible neuroanatomical underpinnings of the disease. Despite this progress, a considerable heterogeneity of structural findings persists that may reflect the phenomenological diversity of schizophrenia. It is unclear whether the range of possible clinical disease manifestations relates to a core structural brain deficit or to distinct structural correlates. Therefore, gray matter density (GMD) differences between 175 schizophrenic patients (SZ) and 177 matched healthy control subjects (HC) were examined in a three-step approach using cross-sectional and conjunctional voxel-based morphometry (VBM): (1) analysis of structural alterations irrespective of symptomatology; (2) subdivision of the patient sample according to a three-dimensional factor model of the PANSS and investigation of structural differences between these subsamples and healthy controls; (3) analysis of a common pattern of structural alterations present in all patient subsamples compared to healthy controls. Significant GMD reductions in patients compared to controls were identified within the prefrontal, limbic, paralimbic, temporal and thalamic regions. The disorganized symptom dimension was associated with bilateral alterations in temporal, insular and medial prefrontal cortices. Positive symptoms were associated with left-pronounced alterations in perisylvian regions and extended thalamic GMD losses. Negative symptoms were linked to the most extended alterations within orbitofrontal, medial prefrontal, lateral prefrontal and temporal cortices as well as limbic and subcortical structures. Thus, structural heterogeneity in schizophrenia may relate to specific patterns of GMD reductions that possibly share a common prefrontal-perisylvian pattern of structural brain alterations.

Introduction

Over the last decade, neuroimaging revealed a growing number of structurally altered brain regions in schizophrenic patients. Recent meta-analyses (Honea et al., 2005, Wright et al., 2000) identified the main foci of these alterations in frontal, temporal and limbic regions as well as the ventricular system. Despite this substantial progress, evidence for neuroanatomical abnormalities remains considerably heterogeneous.

The debate about structural heterogeneity in schizophrenia has been inspired by the disease’s psychopathological diversity and has contributed to many different disease models, representing either a more “monistic” or “pluralistic” neurobiological perspective. One important concept is the hypothesis of a single, unifying pathophysiological process that underlies different disease phenotypes. Based on this concept, Andreasen (1999) integrated the diverse clinical phenomenology into the heuristic model of “cognitive dysmetria” by suggesting a disruption of the cortico-cerebellar-thalamo-cortical circuit (CCTCC) as the main cause for the disassociation of mental activity. This model has received support from previous MRI studies of schizophrenic patients that detected alterations in the three key nodes (prefrontal cortex, thalamus, cerebellum) of the CCTCC (Schlösser et al., 2003, Volz et al., 2000, Gaser et al., 1999).

Beside these brain regions, further structural alterations were identified in the temporal, limbic and paralimbic areas of schizophrenic patients by means of Region-of-Interest (ROI) techniques (Shenton et al., 2001, Wright et al., 2000). Recently, whole-brain VBM and deformation-based morphometric studies (Davatzikos et al., 2005, Gaser et al., 2004, Gaser et al., 1999, Hulshoff Pol et al., 2001) confirmed and extended these findings, thus providing an insight into the complex, distributional nature of cortical alterations involved in the neurobiology of the disorder. These results may strengthen the hypothesis of a core fronto-temporo-limbic disconnectivity as the neural substrate of the diverse schizophrenic symptomatology.

In contrast to these models, the “pluralistic” perspective suggests that the broad clinical syndrome of schizophrenia is subdivided into different nosological entities. Carpenter et al. (1988) identified a “separate disease” within the syndrome of schizophrenia characterized by enduring negative and less severe positive symptoms. Sigmundsson et al. (2001) applied the VBM methodology to a homogeneous sample of schizophrenic patients selected for these “deficit symptoms” and identified two clusters of left-lateralized GMD reductions in the perisylvian/opercular and the medial temporal region, as well as bilateral reductions in the medial prefrontal cortices.

Exploring the associations between clinical phenotypes and the underlying neurobiology may be an appropriate strategy for testing the validity of these different disease models. In this context, several authors observed correlations between positive symptoms and morphometric alterations in the orbitofrontal cortex, the insula, the temporal pole and superior temporal gyrus using ROI or VBM techniques (Makris et al., 2006, Pressler et al., 2005, Gaser et al., 2004, Crespo-Facorro et al., 2004, Crespo-Facorro et al., 2000, Matsumoto et al., 2001, Flaum et al., 1995). Associations between disorganized symptoms and volume reductions have been described for the amygdala–hippocampus–complex (AHC), the parahippocampus and the STG (Suzuki et al., 2005, Rajarethinam et al., 2001, Shenton et al., 1992). Interestingly, many of these cortical structures belong to the “paralimbic brain”—a higher-order, multimodal association network potentially involved in the pathophysiology of schizophrenia (Mesulam and Mufson, 1986).

A sensible strategy to study the links between psychopathology and neurobiology assumes that specific pathophysiological processes are associated with clusters of psychopathological items, that can be described as patterns of co-occurring symptoms revealed by means of factor analysis. This technique has been consistently applied in order to study the phenotypical heterogeneity of schizophrenia. Using factor analysis of the PANSS and SANS/SAPS scales, a large number of studies have provided strong evidence for a dimensional model of schizophrenic psychopathology with (at least) three symptom dimensions (Grube et al., 1998).

In contrast to earlier phenomenological distinctions of paranoid, hebephrenic or katatonic schizophrenia, factorial models emphasize that different symptom dimensions are independent, but also share overlapping clinical features, and, therefore, may not be regarded as clinical subtypes or syndromes of the disease (Andreasen et al., 1995). These models enabled researchers to study the relationships between specific symptom constellations and their underlying neural substrates. However, up to now only few authors (Pressler et al., 2005, Crespo-Facorro et al., 2004, Flaum et al., 1995) have applied this promising approach to structural neuroimaging in schizophrenia. First, Flaum et al. (1995) described associations between ventricular and temporal lobe volumes and positive and negative symptom clusters derived from a three-factor model of the SANS/SAPS consisting of negative, positive and disorganized symptom dimensions (Arndt et al., 1991).

Based on a factorial model of schizophrenic psychopathology, the present VBM study explored the phenotypical and structural heterogeneity of schizophrenia by examining the associations between different symptom dimensions and possible underlying structural brain correlates in a large, cross-sectionally recruited database of patients compared to matched healthy controls. The following hypotheses were tested: (1) Different symptom dimensions relate to selective patterns of structural brain alterations: negative symptoms may be linked to prominent alterations of the fronto-temporo-limbic system. Positive symptoms may be associated with left-hemispheric alterations of the frontal and perisylvian structures, the limbic system and the thalamus, whereas disorganized symptoms may relate to alterations of the temporal, paralimbic and limbic structures. (2) A significant overlap between these patterns exists, possibly representing a core neurobiological substrate of all symptom dimensions. Therefore, a three-step approach was implemented. First, structural alterations were examined irrespective of symptomatology in the entire patient sample compared to healthy volunteers. Secondly, a factor analysis was employed for dividing the patients into three subsamples according to the prevailing individual symptom dimension. Structural alterations were examined within these subsamples compared to healthy controls. Thirdly, a conjunctional VBM analysis was performed in order to evaluate possible core alterations across all subsamples.