Introduction
Anorexia nervosa (AN) is an eating disorder characterized by a refusal to maintain normal body weight, an intense fear of gaining weight despite being underweight, a disturbance in the perception of one's own body weight or shape and, in postmenarcheal women, an absence of at least three consecutive menstrual cycles. AN is further divided into a binge-eating/purging (AN-BP) subtype and a restricting (AN-R) subtype. The former is defined by regular binge-eating or purging behaviour, which may include vomiting, use of diuretics, laxative and enemas, whereas the latter is defined by a lack of this behaviour (APA, 1994).
Brain alterations have previously been associated with AN as a consequence of starvation, with underweight patients showing a reduction of grey matter volume (GMV) and an increase in cerebrospinal fluid (CSF) volume (Kerem & Katzman, Reference Kerem and Katzman2003; Van Den Eynde et al. Reference Van Den Eynde, Suda, Broadbent, Guillaume, Van Den Eynde, Steiger, Israel, Berlim, Giampietro, Simmons, Treasure, Campbell and Schmidt2012). Additionally, some studies have reported a decrease in white matter volume (WMV) (Swayze et al. Reference Swayze, Andersen, Andreasen, Arndt, Sato and Ziebell2003; Boghi et al. Reference Boghi, Sterpone, Sales, D'Agata, Bradac, Zullo and Munno2011; Roberto et al. Reference Roberto, Mayer, Brickman, Barnes, Muraskin, Yeung, Steffener, Sy, Hirsch, Stern and Walsh2011) and a reduction of WM integrity (Kazlouski et al. Reference Kazlouski, Rollin, Tregellas, Shott, Jappe, Hagman, Pryor, Yang and Frank2011), although the literature regarding WM alterations is comparatively lacking. Some studies have suggested complete reversibility of these brain alterations with weight recovery (Swayze et al. Reference Swayze, Andersen, Andreasen, Arndt, Sato and Ziebell2003; Wagner et al. Reference Wagner, Greer, Bailer, Frank, Henry, Putnam, Meltzer, Ziolko, Hoge, McConaha and Kaye2006; Mainz et al. Reference Mainz, Schulte-Rüther, Fink, Herpertz-Dahlmann and Konrad2012), while others have concluded that weight-recovered patients display an intermediate profile somewhere between healthy controls (HC) and currently ill patients (Katzman et al. Reference Katzman, Zipursky, Lambe and Mikulis1997; Lambe et al. Reference Lambe, Katzman, Mikulis, Kennedy and Zipursky1997; Neumärker et al. Reference Neumärker, Bzufka, Dudeck, Hein and Neumärker2000; Joos et al. Reference Joos, Hartmann, Glauche, Perlov, Unterbrink, Saum, Tüscher, Tebartz Van Elst and Zeeck2011; Roberto et al. Reference Roberto, Mayer, Brickman, Barnes, Muraskin, Yeung, Steffener, Sy, Hirsch, Stern and Walsh2011; Friederich et al. Reference Friederich, Walther, Bendszus, Biller, Thomann, Zeigermann, Katus, Brunner, Zastrow and Herzog2012).
Structural imaging studies employing a region of interest (ROI) approach have reported decreased GMV in subcortical regions in AN. An early study by Husain et al. (Reference Husain, Black, Doraiswamy, Shah, Rockwell, Ellinwood and Krishnan1992) found both thalamus and midbrain volumes to be significantly smaller. Connan et al. (Reference Connan, Murphy, Connor, Rich, Murphy, Bara-Carill, Landau, Krljes, Ng, Williams, Morris, Campbell and Treasure2006) reported a decrease in hippocampal volume in underweight patients with AN. Similarly, Giordano et al. (Reference Giordano, Renzetti, Parodi, Foppiani, Zandrino, Giordano and Sardanelli2001) found a significant volume reduction in the hippocampus–amygdala formation. Finally, Neumärker et al. (Reference Neumärker, Bzufka, Dudeck, Hein and Neumärker2000) reported a volume decrease in the mesencephalon and pons in patients that persisted after weight recovery.
Whilst early magnetic resonance studies relied on visual assessment and manual or semi-automated measurements of ROI, the development of automated techniques has allowed for the analysis of large samples without requiring the time-consuming measurements associated with manual quantification (Whitwell, Reference Whitwell2009). Voxel-based morphometry (VBM) (Wright et al. Reference Wright, McGuire, Poline, Travere, Murray, Frith, Frackowiak and Friston1995; Ashburner & Friston, Reference Ashburner and Friston2000) is an automated technique based on voxel-wise statistical analysis of pre-processed structural magnetic resonance images (Ridgway et al. Reference Ridgway, Henley, Rohrer, Scahill, Warren and Fox2008). This technique overcomes the limitations of ROI approaches in that it does not require for a priori brain areas to be selected for analysis.
This approach has led to a steady increase in studies assessing brain abnormalities in AN; however, findings from these studies have been inconsistent (Van Den Eynde et al. Reference Van Den Eynde, Suda, Broadbent, Guillaume, Van Den Eynde, Steiger, Israel, Berlim, Giampietro, Simmons, Treasure, Campbell and Schmidt2012; Supplementary Table 1). Some studies report global GMV loss (Joos et al. Reference Joos, Klöppel, Hartmann, Glauche, Tüscher, Perlov, Saum, Freyer, Zeeck and van Elst2010; Suchan et al. Reference Suchan, Busch, Schulte, Grönermeyer, Herpertz and Vocks2010; Gaudio et al. Reference Gaudio, Nocchi, Franchin, Genovese, Cannatà, Longo and Fariello2011; Friederich et al. Reference Friederich, Walther, Bendszus, Biller, Thomann, Zeigermann, Katus, Brunner, Zastrow and Herzog2012; Mainz et al. Reference Mainz, Schulte-Rüther, Fink, Herpertz-Dahlmann and Konrad2012), whereas other studies have failed to replicate this finding (Boghi et al. Reference Boghi, Sterpone, Sales, D'Agata, Bradac, Zullo and Munno2011; Brooks et al. Reference Brooks, Barker, O'Daly, Brammer, Williams, Benedict, Schiöth, Treasure and Campbell2011a ). Additionally, those studies looking at focal GMV loss in AN implicate a wide range of brain regions with only a few being replicated. Castro-Fornieles et al. (Reference Castro-Fornieles, Bargalló, Lázaro, Andrés, Falcon, Plana and Junqué2009) reported a decrease in GMV in the precuneus, inferior parietal lobe and superior temporal gyrus. Alongside a global decrease in GMV of 1%, Mühlau et al. (Reference Mühlau, Gaser, Ilg, Conrad, Leibl, Cebulla, Backmund, Gerlinghoff, Lommer, Schnebel, Wohlschläger, Zimmer and Nunnemann2007) also reported a 5% decrease in the anterior cingulate cortex (ACC). Similarly, Joos et al. (Reference Joos, Klöppel, Hartmann, Glauche, Tüscher, Perlov, Saum, Freyer, Zeeck and van Elst2010) reported GMV decreases in the ACC as well as decreases in the cerebellum, frontal operculum, precuneus, temporoparietal cortex and parietal cortex in AN-R patients. While Friederich et al. (Reference Friederich, Walther, Bendszus, Biller, Thomann, Zeigermann, Katus, Brunner, Zastrow and Herzog2012) only found a trend for a global decrease in GMV, they did report significant regional volume decreases in the insula, hippocampus, ACC and in the amygdala. Suchan et al. (Reference Suchan, Busch, Schulte, Grönermeyer, Herpertz and Vocks2010) reported significant volume loss in the extrastriate body area (EBA) and superior temporal gyrus. A recent meta-analysis of seven VBM studies in AN revealed a decrease in regional GMV in the hypothalamus, caudate nucleus, lentiform nucleus and the inferior parietal lobe (Titova et al. Reference Titova, Hjorth, Schioth and Brooks2013). The only study reporting an increase in regional GMV was performed by Brooks et al. (Reference Brooks, Barker, O'Daly, Brammer, Williams, Benedict, Schiöth, Treasure and Campbell2011 a) who showed less age-related GM atrophy in the dorsolateral prefrontal cortex, alongside a decrease of GMV in the cerebellum, parahippocampal gyrus, fusiform gyrus, insular cortex and posterior cingulate gyrus.
Several studies have explored associations between GMV loss and clinical characteristics of AN, but again with mixed results. Both current and lowest lifetime body mass index (BMI) have been shown to correlate with global (Katzman et al. Reference Katzman, Lambe, Mikulis, Ridgley, Goldbloom and Zipursky1996) and regional (Mühlau et al. Reference Mühlau, Gaser, Ilg, Conrad, Leibl, Cebulla, Backmund, Gerlinghoff, Lommer, Schnebel, Wohlschläger, Zimmer and Nunnemann2007; Boghi et al. Reference Boghi, Sterpone, Sales, D'Agata, Bradac, Zullo and Munno2011) GMV loss. A study by Brooks et al. (Reference Brooks, Barker, O'Daly, Brammer, Williams, Benedict, Schiöth, Treasure and Campbell2011a ) found that BMI and dietary restraint were correlated with preservation of dorsolateral prefrontal cortex volume with ageing compared with HC. Joos et al. (Reference Joos, Klöppel, Hartmann, Glauche, Tüscher, Perlov, Saum, Freyer, Zeeck and van Elst2010) found no correlation with BMI, but did report a positive correlation between GMV in the right inferior parietal lobe and a ‘drive for thinness’, as measured by the Eating Disorder Inventory (Garner et al. Reference Garner, Olmstead and Polivy1983). With regards to behavioural correlations, Neumärker et al. (Reference Neumärker, Bzufka, Dudeck, Hein and Neumärker2000) reported a decrease in volume in the mesencephalon that was significantly correlated with lower performance on vocabulary tests. Castro-Fornieles et al. (Reference Castro-Fornieles, Bargalló, Lázaro, Andrés, Falcon, Plana and Junqué2009) reported a negative correlation between copy time on the Rey Complex Figure Test and global GMV in adolescents with AN.
The inconsistent findings in the literature could be attributed to differences in methodology and the recruitment of relatively small sample sizes and to the heterogeneity within AN (Van Den Eynde et al. Reference Van Den Eynde, Suda, Broadbent, Guillaume, Van Den Eynde, Steiger, Israel, Berlim, Giampietro, Simmons, Treasure, Campbell and Schmidt2012). As co-morbidity is the rule rather than the exception, brain abnormalities found in people with AN may, at least in part, reflect the influence of co-morbid disorders, such as anxiety or depression (Lorenzetti et al. Reference Lorenzetti, Allen, Fornito and Yücel2009; Bellani et al. Reference Bellani, Baiano and Brambilla2010, Reference Bellani, Baiano and Brambilla2011; Martin et al. Reference Martin, Ressler, Binder and Nemeroff2010; Radua et al. Reference Radua, van den Heuvel, Surguladze and Mataix-Cols2010; Bora et al. Reference Bora, Fornito, Pantelis and Yucel2012). To date, there has not been a study that has assessed the role of co-morbidities on brain structure in AN. Onset of AN often occurs during adolescence and some studies have reported the presence of these alterations when the brain is not yet fully matured (Katzman et al. Reference Katzman, Lambe, Mikulis, Ridgley, Goldbloom and Zipursky1996; Castro-Fornieles et al. Reference Castro-Fornieles, Bargalló, Lázaro, Andrés, Falcon, Plana and Junqué2009; Mainz et al. Reference Mainz, Schulte-Rüther, Fink, Herpertz-Dahlmann and Konrad2012). During normal development, it is assumed that there are region-specific, age-dependent, non-linear decreases in GMV associated with increases in WMV (Zielinski et al. Reference Zielinski, Gennatas, Zhou and Seeley2010; Raznahan et al. Reference Raznahan, Lerch, Lee, Greenstein, Wallace, Stockman, Clasen, Shaw and Giedd2011; Franke et al. Reference Franke, Luders, May, Wilke and Gaser2012). These changes may reflect regional alterations in synaptic density or may be due to intra-cortical myelination (Giorgio et al. Reference Giorgio, Watkins, Chadwick, James, Winmill, Douaud, De Stefano, Matthews, Smith, Johansen-Berg and James2010; Silk & Wood, Reference Silk and Wood2011). However, irregular food intake leads to changes in physiology during which the body adapts to the starvation. The relatively small amount of energy that is available is conserved by, amongst other mechanisms, delaying growth and lowering metabolic rate (Hasan & Hasan, Reference Hasan and Hasan2011). While the brain is comparatively spared, or at least catabolized more slowly than other tissues (McCue, as cited in Hasan & Hasan, Reference Hasan and Hasan2011), very little is known about the effect of starvation on the brain during adolescence.
The first aim of the study was to correlate clinical measures with brain tissue volume (whole-brain volume, GMV and WMV) in AN. The second aim, as very little is known about the impact of early onset of AN or about the effects of chronic duration on brain structure, was to explore the impact of the age of onset of illness and the duration of illness on these volumetric measures.
Method
Participants
A total of 35 individuals with a current diagnosis of AN by a certified clinician, fulfilling Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV) criteria, were recruited from the hospital and community services of the South London and Maudsley NHS Trust and from an online advertisement on the website of beat™ (beating eating disorders; http://www.b-eat.co.uk), the UK's largest eating disorder charity. However, three participants were excluded due to a lack of completion of all measures and one was excluded due to a BMI falling within the normal range. Out of the remaining 31, 12 (39%) reported taking antidepressant medication; 22 (71%) were diagnosed as AN-R subtype, six (19%) as AN-BP subtype and three (10%) were unspecified. A total of 37 healthy individuals with no personal or family history of eating disorders were recruited from the community and staff of the Institute of Psychiatry, King's College London. However, three healthy participants were excluded from further analysis due to currently taking antidepressant medication and three participants were excluded after optimal matching of the two groups in terms of age, group size and estimated intelligence quotient (IQ). Both groups consisted predominantly of right-handed participants (AN 25, HC 27). Age, medication and BMI scores were obtained as well as both the age of onset and the duration of illness for participants with AN. The screening module of the research version of the Structured Clinical Interview for DSM Disorders (SCID-I/P; First et al. Reference First, Gibbon, Spitzer and Williams1997) was used as a screening tool for the healthy controls. The National Adult Reading Test (Nelson & Willison, Reference Nelson and Willison1991) was used to estimate IQ. Participants' consent was obtained according to the Declaration of Helsinki and was approved by the local research ethics committee (no. 11-LO-0952).
Clinical measures
All participants completed the questionnaires before attending the scanning session to measure their anxiety, depression, eating behaviour and obsessionality. The Hospital Anxiety and Depression Scale (HADS; Zigmond & Snaith, Reference Zigmond and Snaith1983) is a measure consisting of 14 items to assess overall severity of depression and anxiety (clinical threshold total = 10). The Eating Disorders Examination Questionnaire (EDE-Q; Fairburn & Beglin, Reference Fairburn and Beglin1994) is a self-report questionnaire consisting of 36 items that look at a participant's eating behaviour over the past 28 days. The Obsessive–Compulsive Inventory – Revised (OCI-R; Foa et al. Reference Foa, Huppert, Leiberg, Langner, Kichic, Hajcak and Salkovskis2002) is an 18-item list of first-person statements describing experiences that the subject has to rate according to the level of distress they felt when they experienced those statements over the past month (clinical threshold = 25).
Image acquisition
Magnetic resonance imaging (MRI) was performed using a 1.5 T GE Signa HDx TwinSpeed MRI scanner (GE-Medical Systems, USA) at the Centre for Neuroimaging Sciences, Institute of Psychiatry, King's College London. The body coil was used for radio frequency (RF) transmission, with an eight-channel head coil for RF reception. Whole-brain scans were obtained as high-resolution T1-weighted MPRAGE (magnetization prepared rapid acquisition gradient echo) acquired in the sagittal plane (repetition time = 8.592 ms, echo time = 3.8 ms, inversion time = 1000 ms, flip angle = 8°, field of view = 240 mm, matrix = 192 × 192, voxel size = 1.2 × 1.2 × 1.2, 180 slices). Full brain and skull coverage was required for the MRI datasets and detailed quality control carried out on all MR images, according to previously published quality-control criteria (Simmons et al. Reference Simmons, Westman, Muehlboeck, Mecocci, Vellas, Tsolaki, Kloszewska, Wahlund, Soininen, Lovestone, Evans and Spenger2009, Reference Simmons, Westman, Muehlboeck, Mecocci, Vellas, Tsolaki, Kloszewska, Wahlund, Soininen, Lovestone, Evans and Spenger2011).
Structural image evaluation, using normalization, of atrophy (SIENAX)
Brain tissue volume, normalized for subject head size, was estimated with SIENAX (Smith, Reference Smith2002), part of FSL (Smith et al. Reference Smith, Jenkinson, Woolrich, Beckmann, Behrens, Johansen-Berg, Bannister, De Luca, Drobnjak, Flitney, Niazy, Saunders, Vickers, Zhang, De Stefano, Brady and Matthews2004). SIENAX starts by extracting brain and skull images from the single whole-head input data (Smith, Reference Smith2002). The brain image is then affine-registered to Montreal Neurological Institute (MNI) 152 standard space (Jenkinson & Smith, Reference Jenkinson and Smith2001; Jenkinson et al. Reference Jenkinson, Bannister, Brady and Smith2002), using the skull image to determine the registration scaling. This is primarily in order to obtain the volumetric scaling factor, to be used as a normalization for head size. Next, tissue-type segmentation with partial volume estimation was carried out (Zhang et al. Reference Zhang, Brady and Smith2001) in order to calculate total volume of brain tissue (including separate estimates of GMV, WMV and ventricular CSF volume).
Statistical analysis
An assessment of normality was performed for all measures using the Shapiro–Wilks test in SPSS Statistics for Windows 20 (IBM Corp., USA). When the assumption of normality was violated, the non-parametric Mann–Whitney U test was used to assess differences. When scores were normally distributed, Student's t test was used. The non-parametric Spearman's rho (ρ) was used to assess correlations between volumetric measurements and clinical measures.
VBM
To assess voxel-wise changes in GMV, structural data were analysed with FSL-VBM (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSLVBM) (Douaud et al. Reference Douaud, Smith, Jenkinson, Behrens, Johansen-Berg, Vickers, James, Voets, Watkins, Matthews and James2007), an optimized VBM protocol (Good et al. Reference Good, Johnsrude, Ashburner, Henson, Friston and Frackowiak2001) carried out with FSL tools (Smith et al. Reference Smith, Jenkinson, Woolrich, Beckmann, Behrens, Johansen-Berg, Bannister, De Luca, Drobnjak, Flitney, Niazy, Saunders, Vickers, Zhang, De Stefano, Brady and Matthews2004). This approach is unbiased, in that it requires no a priori information about the location of possible differences in the grey matter, and is not operator-dependent. First, structural images were brain-extracted and grey matter-segmented before being registered to the MNI 152 standard space using non-linear registration (Andersson et al. Reference Andersson, Jenkinson and Smith2007). The resulting images were averaged and flipped along the x-axis to create a left–right symmetric, study-specific grey matter template. Second, all native grey matter images were non-linearly registered to this study-specific template and modulated to correct for local expansion (or contraction) due to the non-linear component of the spatial transformation. The modulated grey matter images were then smoothed with an isotropic Gaussian kernel with a sigma of 3 mm (7 mm FWHM). Finally, a voxel-wise general linear model (GLM) was applied using permutation-based non-parametric testing, correcting for multiple comparisons across space, to search for regional differences in GMV between patients and controls. Threshold-free cluster enhancement (Smith & Nichols, Reference Smith and Nichols2009) was used to detect significant clusters of activation and family-wise error correction was used to correct for multiple comparisons across space.
Two separate analyses were performed to assess voxel-wise changes in GMV. In the first analysis only age, IQ and years of education were entered as confounding covariates. Because global changes in GMV have been reported in AN, the second analysis added global GMV as an additional confounding covariate to look for regional changes in GMV that cannot be explained by a global change in GMV.
Additional correlational analyses were then performed by adding a separate regressor per group to assess associations previously reported between GMV and BMI (Katzman et al. Reference Katzman, Lambe, Mikulis, Ridgley, Goldbloom and Zipursky1996; Mühlau et al. Reference Mühlau, Gaser, Ilg, Conrad, Leibl, Cebulla, Backmund, Gerlinghoff, Lommer, Schnebel, Wohlschläger, Zimmer and Nunnemann2007; Boghi et al. Reference Boghi, Sterpone, Sales, D'Agata, Bradac, Zullo and Munno2011; Brooks et al. Reference Brooks, Barker, O'Daly, Brammer, Williams, Benedict, Schiöth, Treasure and Campbell2011a ) and eating behaviour (Joos et al. Reference Joos, Klöppel, Hartmann, Glauche, Tüscher, Perlov, Saum, Freyer, Zeeck and van Elst2010), as well as to explore possible associations with depression, anxiety and obsessive–compulsive symptoms. In the AN group correlations with duration of illness (Boghi et al. Reference Boghi, Sterpone, Sales, D'Agata, Bradac, Zullo and Munno2011) and age of onset were also explored. To assess any effects of medication on GMV, patients reporting antidepressant medication were compared with a matched group of patients without medication.
Results
Group characteristics
Demographic information is summarized in Table 1. Patients and controls were similar in terms of age, but there was a significant difference between groups in terms of IQ, years of education and BMI.
Table 1. Demographic characteristics
AN, Anorexia nervosa; HC, healthy controls; IQ, intelligence quotient; BMI, body mass index.
a Scores are not normally distributed and values are given as median (interquartile range).
b Scores are normally distributed and values are given as mean (standard deviation).
c One participant did not report years of education and scores are calculated based on 61 participants.
d Age of onset and duration of illness were not reported by all participants and the score is calculated based on 28 participants instead of 31.
Clinical measures
The screening module of the SCID-I/P revealed that there was no indication of psychiatric disorders in the HC subjects. The AN group scored significantly higher on all self-report measures compared with HC, as illustrated in Table 2. Scores on the HADS in the AN group indicate clinical levels of anxiety and depression (Bjelland et al. Reference Bjelland, Dahl, Haug and Neckelmann2002). OCI-R scores illustrate a high proportion of obsessionality in AN and a likely presence of obsessive–compulsive disorder symptomatology (Foa et al. Reference Foa, Huppert, Leiberg, Langner, Kichic, Hajcak and Salkovskis2002).
Table 2. Clinical characteristics
AN, Anorexia nervosa; HC, healthy controls; EDE-Q, Eating Disorders Examination Questionnaire; HADS, Hospital Anxiety and Depression Scale; OCI-R, Obsessive–Compulsive Inventory-Revised.
a Scores are not normally distributed and values are given as median (interquartile range).
b Scores are normally distributed and values are given as mean (standard deviation).
Global differences in brain volume
Using SIENAX, whole-brain volume, global GMV and WMV and ventricular CSF volume were estimated and the results for both groups are summarized in Table 3. Whole-brain volume and global GMV were found to be significantly lower in AN alongside enlargement of ventricular CSF. In AN, global GMV correlated negatively with age (ρ = − 0.62, p < 0.001), but no other significant correlations were found. In HC, global GMV correlated with age (ρ = − 0.47, p = 0.008) and with the EDE-Q (ρ = 0.48, p = 0.006), but these correlations did not survive Bonferroni correction. No other correlations with clinical measures were found.
Table 3. Whole-brain, grey matter, white matter and ventricular CSF volumes (mm3)
CSF, Cerebrospinal fluid; AN, anorexia nervosa; HC, healthy controls.
a Scores are not normally distributed and values are given as median (interquartile range).
b Scores are normally distributed and values are given as mean (standard deviation).
Focal alterations in GMV
VBM analysis revealed regional atrophy in AN compared with HC, whilst controlling for estimated IQ, age and years of education (see Fig. 1). Results are summarized in Table 4. There was no indication of regions showing greater GMV in AN compared with HC. No significant difference in GMV was found between AN taking antidepressant medication and AN not taking medication. When further controlling for differences in global GMV, no local GMV reductions were found in AN compared with HC.
Fig. 1. Grey matter volume loss in anorexia nervosa (n = 31) compared with healthy controls (n = 31) using FSL-VBM (threshold-free cluster enhancement, p < 0.05, family-wise error corrected), whilst controlling for age, intelligence quotient and years of education. L, Left; R, right; P, posterior; A, anterior.
Table 4. Voxel-based morphometry results showing grey matter volume loss in anorexia nervosa patients (n = 31) compared with healthy controls (n = 31) whilst controlling for intelligence quotient, age and years of education
MNI, Montreal Neurological Institute.
Correlational analyses revealed a significant negative correlation between duration of illness and GMV (see Fig. 2 and Table 5). There was no significant correlation of GMV with age of onset. No significant correlations were found with clinical measures.
Fig. 2. Regions showing a negative correlation with duration of illness in anorexia nervosa (n = 28) using FSL-VBM (threshold-free cluster enhancement, p < 0.05, family-wise error corrected), whilst controlling for age, intelligence quotient, years of education and global grey matter volume. L, Left; R, right; P, posterior; A, anterior.
Table 5. Regions showing a negative correlation between grey matter volume and duration of illness in anorexia nervosa patients (n = 28) whilst controlling for age, intelligence quotient, years of education and global grey matter volume
MNI, Montreal Neurological Institute.
Discussion
Our findings support studies that have reported a global GMV loss (Katzman et al. Reference Katzman, Lambe, Mikulis, Ridgley, Goldbloom and Zipursky1996, Reference Katzman, Zipursky, Lambe and Mikulis1997; Lambe et al. Reference Lambe, Katzman, Mikulis, Kennedy and Zipursky1997; Joos et al. Reference Joos, Klöppel, Hartmann, Glauche, Tüscher, Perlov, Saum, Freyer, Zeeck and van Elst2010; Suchan et al. Reference Suchan, Busch, Schulte, Grönermeyer, Herpertz and Vocks2010; Gaudio et al. Reference Gaudio, Nocchi, Franchin, Genovese, Cannatà, Longo and Fariello2011; Roberto et al. Reference Roberto, Mayer, Brickman, Barnes, Muraskin, Yeung, Steffener, Sy, Hirsch, Stern and Walsh2011; Friederich et al. Reference Friederich, Walther, Bendszus, Biller, Thomann, Zeigermann, Katus, Brunner, Zastrow and Herzog2012; Mainz et al. Reference Mainz, Schulte-Rüther, Fink, Herpertz-Dahlmann and Konrad2012; Titova et al. Reference Titova, Hjorth, Schioth and Brooks2013), but there was no indication of differences in global WMV (Swayze et al. Reference Swayze, Andersen, Andreasen, Arndt, Sato and Ziebell2003; Roberto et al. Reference Roberto, Mayer, Brickman, Barnes, Muraskin, Yeung, Steffener, Sy, Hirsch, Stern and Walsh2011). However, the methods used in this study do not allow for a proper measurement of WM integrity or connectivity and therefore it is not possible to conclude that there are no alterations in WM in AN. Unlike previous studies, no correlations with BMI or eating disorder-related symptoms were found in AN (Katzman et al. Reference Katzman, Lambe, Mikulis, Ridgley, Goldbloom and Zipursky1996; Joos et al. Reference Joos, Klöppel, Hartmann, Glauche, Tüscher, Perlov, Saum, Freyer, Zeeck and van Elst2010); however, this study does find that current age correlates negatively with global GMV. This is not uncommon, as studies have shown that in normal ageing there is a region-specific and non-linear pattern of age-related changes in GMV (Resnick et al. Reference Resnick, Pham, Kraut, Zonderman and Davatzikos2003; Terribilli et al. Reference Terribilli, Schaufelberger, Duran, Zanetti, Curiati, Menezes, Scazufca, Amaro, Leite and Busatto2011).
Using VBM, GM atrophy was found in regions previously reported to be affected in AN using a whole-brain approach. Among these regions, the precuneus has been suggested to play a central role in a wide range of processes, including visuospatial processing, attention, episodic memory, self-consciousness and self-related mental representations (Cavanna & Trimble, Reference Cavanna and Trimble2006). Previous studies have also found GM atrophy in temporoparietal and occipital regions in combination with the precuneus (Castro-Fornieles et al. Reference Castro-Fornieles, Bargalló, Lázaro, Andrés, Falcon, Plana and Junqué2009; Joos et al. Reference Joos, Klöppel, Hartmann, Glauche, Tüscher, Perlov, Saum, Freyer, Zeeck and van Elst2010; Favaro et al. Reference Favaro, Santonastaso, Manara, Bosello, Bommarito, Tenconi and Di Salle2012). Body image distortions have been postulated to reflect dysfunctional processing of information (Williamson et al. Reference Williamson, White, York-Crowe and Stewart2004) and both lesion and functional studies have reported alterations in regions associated with visuospatial processing (Uher & Treasure, Reference Uher and Treasure2005; Gaudio et al. Reference Gaudio, Nocchi, Franchin, Genovese, Cannatà, Longo and Fariello2011). More specifically, recent investigations have linked body image distortions in AN to the EBA and the posterior part of the fusiform gyrus, as both GMV of the EBA and functional connectivity from the fusiform gyrus to the EBA have been found to negatively correlated with body image distortion (Suchan et al. Reference Suchan, Busch, Schulte, Grönermeyer, Herpertz and Vocks2010, Reference Suchan, Bauser, Busch, Schulte, Gronemeyer, Herpertz and Vocks2013). Furthermore, Vocks et al. (Reference Vocks, Schulte, Busch, Gronemeyer, Herpertz and Suchan2011) reported a significant increase in neural activation in the EBA as well as frontal regions in patients with eating disorders following cognitive–behavioural body image therapy; taken together, this suggests that body image distortions are related to both functional and structural changes within these areas. The superior frontal gyrus has been suggested to play a role in introspection (Goldberg et al. Reference Goldberg, Harel and Malach2006) and reduced activation was previously reported in AN during processing of self-images (Vocks et al. Reference Vocks, Busch, Gronemeyer, Schulte, Herpertz and Suchan2010). Zhu et al. (Reference Zhu, Hu, Wang, Chen, Guo, Li and Enck2012) found reduced activation in the precuneus and parietal regions during presentation of food stimuli as well as an increase in activation in the medial frontal cortex, anterior cingulate, caudate nucleus and lentiform nucleus. It has been suggested that those with AN utilize cognitive strategies that restrict appetitive responses in the brain to reduce the strength of the saliency of food stimuli (Brooks et al. Reference Brooks, Prince, Stahl, Campbell and Treasure2011b ). This notion is further supported by previous studies that reported atrophy within the fusiform gyrus and posterior cingulate gyrus (Brooks et al. Reference Brooks, Barker, O'Daly, Brammer, Williams, Benedict, Schiöth, Treasure and Campbell2011a ; Favaro et al. Reference Favaro, Santonastaso, Manara, Bosello, Bommarito, Tenconi and Di Salle2012), regions associated with recognition of familiar objects and spatial awareness. In addition, Brooks et al. (Reference Brooks, Barker, O'Daly, Brammer, Williams, Benedict, Schiöth, Treasure and Campbell2011a ) suggested that the cognitive bias for food stimuli in AN could, in part, be related to an imbalance in accurate object recognition due to hypoactive fusiform-driven spatial recognition systems.
As the current study demonstrated that the AN group has a smaller brain volume and less global GMV, it is possible that these effects are not region-specific but instead part of a global decrease in GMV. Indeed, when global GMV is controlled for there are no region-specific alterations on top of this global decrease. It is important to note that different adjustments for global effects in VBM can lead to different results and it is possible that employing a different approach would lead to significant findings (Peelle et al. Reference Peelle, Cusack and Henson2012). Previous VBM studies in AN employing different adjustments for global effects have reported local effects in the regions we report in this study (Brooks et al. Reference Brooks, Barker, O'Daly, Brammer, Williams, Benedict, Schiöth, Treasure and Campbell2011a ; Gaudio et al. Reference Gaudio, Nocchi, Franchin, Genovese, Cannatà, Longo and Fariello2011). Finally, a recent meta-analysis of seven studies did not find regional differences in these areas at all (Titova et al. Reference Titova, Hjorth, Schioth and Brooks2013).
A correlation was found between cerebellar atrophy and duration of illness that could not be explained by global changes in GMV and which replicates previous findings of greater GMV loss in more chronic cases (Boghi et al. Reference Boghi, Sterpone, Sales, D'Agata, Bradac, Zullo and Munno2011). The cerebellum has been thought to play a modulatory role in higher-order neural functions and its dysfunction may be related to behaviours evident in people with AN, such as anxiety, ritualistic and stereotypical behaviours as well as dysphoria, depression, and ruminative and obsessive behaviours (Schmahmann et al. Reference Schmahmann, Weilburg and Sherman2007). Animal studies have also shown disruptions in feeding behaviour and weight loss caused by cerebellar lesions (Zhu & Wang, Reference Zhu and Wang2008), which suggests that reductions in cerebellar volume may contribute to successfully maintaining low weight and therefore the duration of the illness.
Chronic cases also demonstrated greater GMV loss bilaterally in the mesencephalon, at the substantia nigra and ventral tegmental area. Animal studies in AN have suggested that areas such as the substantia nigra and ventral tegmental area play a prominent role in motivation and anticipation of food reward, food intake and over-exercising through projections to different cortico-limbic structures (Narayanan et al. Reference Narayanan, Guarnieri and DiLeone2010; Adan et al. Reference Adan, Hillebrand, Danner, Cardona Cano, Kas and Verhagen2011; Verhagen et al. Reference Verhagen, Luijendijk and Adan2011; van Zessen et al. Reference van Zessen, van der Plasse and Adan2012). Furthermore, the brainstem plays a role in autonomic and visceral control and has been implicated in abnormalities in taste and pain perception in individuals with AN (Kaye et al. Reference Kaye, Fudge and Paulus2009). This may suggest that the core symptoms of AN become subjectively easier to maintain in chronic cases due to subcortical changes in brain structure. Additionally, we have shown that this effect cannot be explained due to the global decrease in GMV in AN.
No correlations were found with clinical measures in AN, suggesting that the observed differences in GMV are not associated with depression, anxiety or obsessive–compulsive symptoms. Additionally, unlike previous studies there was no correlation between GMV and BMI (Katzman et al. Reference Katzman, Lambe, Mikulis, Ridgley, Goldbloom and Zipursky1996; Joos et al. Reference Joos, Klöppel, Hartmann, Glauche, Tüscher, Perlov, Saum, Freyer, Zeeck and van Elst2010; Boghi et al. Reference Boghi, Sterpone, Sales, D'Agata, Bradac, Zullo and Munno2011; Brooks et al. Reference Brooks, Barker, O'Daly, Brammer, Williams, Benedict, Schiöth, Treasure and Campbell2011a ). This is surprising given that those studies recruited patient cohorts with a similar average BMI to the current cohort. Samples in previous studies did all differ in terms of AN subtype ratios, presence of medication and age. One other possibility could be the low variance of BMI within the AN group in the current study. No correlations with eating disorder symptomatology were found, though the current study did not use ‘drive for thinness’ separately as a covariate (Joos et al. Reference Joos, Klöppel, Hartmann, Glauche, Tüscher, Perlov, Saum, Freyer, Zeeck and van Elst2010).
Automated techniques such as VBM and SIENAX allow for relatively fast assessments of brain structure in large groups. The results presented here are consistent with the majority of the literature and indicate GM atrophy in people with AN. It is unlikely that these alterations are the cause of AN, as previous studies have shown either full recovery (Swayze et al. Reference Swayze, Andersen, Andreasen, Arndt, Sato and Ziebell2003; Wagner et al. Reference Wagner, Greer, Bailer, Frank, Henry, Putnam, Meltzer, Ziolko, Hoge, McConaha and Kaye2006; Mainz et al. Reference Mainz, Schulte-Rüther, Fink, Herpertz-Dahlmann and Konrad2012) or an intermediate profile between controls and currently ill (Katzman et al. Reference Katzman, Zipursky, Lambe and Mikulis1997; Lambe et al. Reference Lambe, Katzman, Mikulis, Kennedy and Zipursky1997; Neumärker et al. Reference Neumärker, Bzufka, Dudeck, Hein and Neumärker2000; Joos et al. Reference Joos, Hartmann, Glauche, Perlov, Unterbrink, Saum, Tüscher, Tebartz Van Elst and Zeeck2011; Roberto et al. Reference Roberto, Mayer, Brickman, Barnes, Muraskin, Yeung, Steffener, Sy, Hirsch, Stern and Walsh2011; Friederich et al. Reference Friederich, Walther, Bendszus, Biller, Thomann, Zeigermann, Katus, Brunner, Zastrow and Herzog2012). Instead, these changes are more likely to be the result of starvation and in turn facilitate the persistence of psychopathological features found in AN. This is further supported by the finding of greater atrophy in more chronic cases. Age of onset of AN does not seem to play a role in the observed differences in brain structure, suggesting that brain maturation is not as severely hindered by starvation as the rest of the body. Despite the heterogeneity within AN and the multitude of co-morbid disorders, there was no indication of a relationship between co-morbid disorders and alterations in brain structure in this study. This would suggest that the changes are due to starvation and independent of co-morbid disorders that have been implicated in alterations in brain structure.
To date, this is the largest study that has assessed GMV alterations in AN using brain tissue volume estimation and VBM. The statistical methods used in this study avoid a priori assumptions by using permutation-based non-parametric testing and provide a more stringent and correct approach to neuroimaging analysis. Further studies should aim to validate the findings presented here and longitudinally assess whether full recovery occurs with weight restoration. The majority of the current sample was comprised of AN-R subtype patients; therefore it is uncertain whether there is a significant difference in GM alterations between subtypes of AN. It is important to note that there are difficulties in establishing the exact age of onset or duration of AN and more studies are required to assess these effects from early on in adolescence or those at risk of developing AN. Further longitudinal investigations of recovered AN are required to ensure that these effects are in fact state-related phenomena and not present prior to the onset of the illness. Finally, there is a lack of studies that have assessed changes in WM in AN and as of yet it is unknown if there is any change related to GM alterations.
Supplementary material
For supplementary material accompanying this paper visit http://dx.doi.org/10.1017/S0033291713002389.
Acknowledgements
This work was supported by the Swiss Anorexia Foundation and Psychiatry Research Trust, and by the National Institute for Health Research (NIHR) Biomedical Research Centre for Mental Health at South London and Maudsley NHS Foundation Trust and Institute of Psychiatry, King's College London. We would like to thank Dr Helen Davies and Naima Lounes for data collection, Dr Owen O'Daly and Dr Matthew Kempton for feedback on statistical analysis as well as Nick Lao-Kaim for helpful suggestions in preparation of this paper.
Declaration of Interest
None.
