Voxel-based morphometry in eating disorders: Correlation of psychopathology with grey matter volume

Abstract

Twenty-nine adult female patients with eating disorders (17 with bulimia nervosa, 12 with restrictive anorexia nervosa) were compared with 18 age-matched female healthy controls, using voxel-based morphometry. Restrictive anorexia nervosa patients showed a decrease of grey matter, particularly affecting the anterior cingulate cortex, frontal operculum, temporoparietal regions and the precuneus. By contrast, patients with bulimia nervosa did not differ from healthy controls. A positive correlation of “drive for thinness” and grey matter volume of the right inferior parietal lobe was found for both eating disorder groups. The strong reduction of grey matter volume in adult patients with restrictive anorexia nervosa is in line with results of adolescent patients. Contrary to other studies, this first voxel-based morphometry report of bulimic patients did not find any structural abnormalities. The inferior parietal cortex is a critical region for sensory integration of body and spatial perception, and the correlation of “drive for thinness” with grey matter volume of this region points to a neural correlate of this core psychopathological feature of eating disorders.

Introduction

The lifetime prevalence of anorexia nervosa (AN) is 0.9%, and that of bulimia nervosa (BN) is 1.5% in women (Hudson et al., 2007). The core psychopathology of these eating disorders (ED) consists of preoccupation with food and body shape, drive for thinness (DT) and disturbances of eating behaviour. It typically affects female adolescents and young adults in a critical phase of psychosexual development. The aetiology is multidimensional and only partially understood. AN is divided into a restrictive type (AN-R) and a binge-eating/purging type.

Brain imaging has disclosed functional and structural abnormalities in patients with ED (Frank et al., 2004), the most consistent findings being a decreased brain mass and increased cerebrospinal fluid (CSF) space in AN (Frank et al., 2004). A large cranial computed tomography (CCT) study demonstrated increases in CSF spaces in AN, and to a lesser extent in BN patients (Krieg et al., 1989). The latter finding was substantiated by further investigations (Hoffman et al., 1989, Kiriike et al., 1990).

Subsequent structural imaging studies in AN mostly used a region of interest (ROI) approach and found abnormalities in a variety of areas, including the mesial temporal lobe, thalamus and brain stem (Neumarker et al., 2000, Giordano et al., 2001, Connan et al., 2006, Husain et al., 1992). A study of adolescent AN patients described whole brain grey matter (GM) and white matter (WM) decreases (Katzman et al., 1996). Adult AN patients showed decreased WM in temporoparietal regions and trends for GM decrease, when investigated with a semiautomated tissue segmentation method (Swayze et al., 2003). WM and GM correlated with the body mass index (BMI). Castro-Fornieles et al. (2009) used a voxel-based morphometrical (VBM) approach in 12 adolescent AN patients and reported structural changes of temporoparietal regions, including the precuneus, and the midcingulate cortex (MCC). In this patient group, whole brain GM and CSF showed significant deviations, whereas WM did not (Castro-Fornieles et al., 2009). With weight restoration these deviations returned to normal — though less in the MCC (Castro-Fornieles et al., 2009). Other authors (Kerem and Katzman, 2003) also reported GM decreases in adolescent AN patients and incomplete reversibility following weight normalization. In adult ED populations, VBM analyses have been reported for recovered patients only (Muhlau et al., 2007, Wagner et al., 2006).

This study investigated symptomatic adult patients with ED using VBM — which allows whole brain analyses without a priori decisions on ROIs. AN-R and BN patients were chosen in order to study two poles of ED (Fairburn et al., 2003). These subgroups share core features like a strong DT and preoccupation with shape and weight, but differ in the degree of malnutrition, hormonal disturbances, impulsive eating and purging behaviour.

Based on the findings summarized above, we expected an increase of CSF in the AN-R group paralleled by a decrease in global GM and WM, the former being more affected. We hypothesized the GM reduction to be more pronounced in temporoparietal and cingular regions (Castro-Fornieles et al., 2009). We assumed GM and WM deviations to be stronger in AN-R compared to BN, and correlated with BMIs.

In addition, we correlated central psychopathological features of eating disorder pathology with local GM volume, focussing on DT and bulimia, which can be assessed by subscales of the Eating Disorder Inventory (EDI) (Paul and Thiel, 2005). DT is a core and transdiagnostic symptom of ED, which can be viewed as a continuum (Fairburn and Harrison, 2003). DT sharply distinguishes ED groups from healthy controls (HC), more than body dissatisfaction does (Garner, 1991). Furthermore, AN patients are underweight, and body dissatisfaction is (therefore) less pronounced, compared to BN patients (Garner, 1991, Ruuska et al., 2005). Therefore, we focussed the analysis on DT. Associations of the subscale “bulimia” with GM volume concerned the BN group only.

Based on previous reports of deviations of the parietal cortex in ED (Frank et al., 2004, Wagner et al., 2003, Uher et al., 2004, Bailer et al., 2004, Goethals et al., 2007) and its association with body image distortion (Frank et al., 2004, Wagner et al., 2003, Delvenne et al., 1999, Castro-Fornieles et al., 2009), we hypothesized a correlation of DT with parietal GM volume in the whole ED group. Based on functional magnetic resonance imaging (fMRI) data (Uher et al., 2004), we hypothesized a correlation of DT with GM volume of the anterior cingulate cortex (ACC) in both ED groups. Thirdly, due to experimental data (Uher et al., 2004, Uher et al., 2005), we expected a negative correlation of GM of the dorsolateral prefrontal cortex and bulimic symptoms.