Brain responses to body image stimuli but not food are altered in women with bulimia nervosa

People with bulimia nervosa (BN) engage in binge-eating (i.e. over-eating with a feeling of losing control) and compensatory behaviors such as self-induced vomiting, the use of laxatives or diuretics or excessive exercising. It affects approximately 1% of young women [1] and treatment outcomes such as remission rates are modest [2, 3]. The neurobiological underpinnings of BN are poorly understood; Functional magnetic resonance imaging (fMRI) studies focused on self-regulatory [4, 5] and reward processes [6‐8] have identified differences in fronto-striatal functioning in people with BN and these may underlie core BN symptomatology such as binge eating, purging and undue body weight and shape preoccupation and dissatisfaction. However, fMRI research in BN is still at an early stage [9]; for example, replication of findings is rare and knowledge on the processing of visual food and body image cues is based on different patient cohorts.

Processing of food images in people with BN has been examined in 5 fMRI studies (n_range = 8–20) by comparing responses to food and non-food images [10‐14]. However, there is a lack of consistency in the findings. Uher et al. (2004) reported that, compared to healthy controls (HCs), BN patients showed increased left medial prefrontal cortex activation, combined with a decreased left (dorso)lateral prefrontal cortex (DLPFC) activation in response to visual food cues [10]. Joos et al. (2011) also found lower activation levels in people with BN in a larger region comprising parts of the anterior cingulate cortex (ACC) and lateral prefrontal cortex [12]. Weak activation of the DLPFC activation in response to food stimuli has been interpreted as poor ‘top-down’ self-regulatory control in people with BN [10, 12]. A number of findings have not yet been replicated, e.g. reduced activation in the right middle temporal lobe and mid-cingulate cortex [12], reduced left visual cortex activation [15], and reduced inferior parietal lobe and post-central gyrus activations [14]. Other findings appear to be in ‘opposite directions’, e.g. Brooks et al. (2011) reported decreased bilateral insula activation in people with BN, while Schienle et al. (2009) reported increased right insula activation [11]. Furthermore, the increased bilateral ACC activation in BN seen by Schienle et al. (2009) contrasts with the finding by Joos et al. (2011) in this region. Thus, there is insufficient evidence to state that fronto-striatal circuit dysfunctions are associated with food/binge-eating related BN psychopathology. Part of the problem may relate to the fact that all previous studies used a non-food (e.g. stationary) control condition rather than a contrast of the food stimuli with a ‘low level baseline’ (LLB) condition, e.g. looking at a cross, which has been used in HC studies [16]. It is possible that women with BN already differentially respond to certain control conditions, i.e. non-food [or non-body image [17]], the contrast between the food condition and a LLB may clarify the differential involvement of fronto-striatal areas in food processing between women with and without BN.

Undue body image concerns are another major pathological feature of BN and hence brain activation patterns associated with the processing of visual presentations of one’s own and other women’s bodies have been studied using various fMRI paradigms. One repeated finding is that women with BN engage the medial prefrontal cortex less than HCs when presented with body stimuli [18‐21]. However, differences in brain activation patterns between women with and without BN have also been seen in several regions (temporal, parietal and occipital) [18‐22]. Methodological variations such as the use of different stimuli (e.g. own or others’ bodies) and different verbal instructions associated with these visual stimuli [20] may have contributed to these divergent findings. Responses to stimuli involving evaluation of one’s own or comparative evaluation of another body involve both interoceptive awareness and emotional elements. The insula, which is involved in interoceptive awareness [23, 24], has been implicated in the pathophysiology of eating disorders [25, 26]. In people with BN, the insula has been reported to be involved in the processing of thinner own-body representations [20]. Evaluation of own physical appearance in comparison to ‘slim ideals’ is common in women and in those with BN it induces high levels of anxiety [27, 28]. However, no fMRI study has investigated whether or which alterations in brain function (e.g. in the insula), are associated with this phenomenological distinction between women with and without BN using self-schematic processing of other slim bodies.

In summary, to obtain information on the neurobiological underpinnings of binge-eating and body weight and shape concerns in BN, the neural correlates of processing of food and body image stimuli require further investigation. Accordingly, this fMRI study investigated brain activation patterns associated with processing of food stimuli and body image. We hypothesized that, compared to HCs, people with BN would 1) activate the medial prefrontal cortex and ACC more, and the DLPFC less in response to food stimuli; and 2) show more insular cortex and less medial prefrontal cortex activation when they compared their own appearance to that of slim women.

The current study differs from previous research in this area in two ways. First, we used a written instruction before each block of stimuli that aimed to engage participants to think about eating the food in the picture, or to stimulate self-schematic processing of their own body in comparison to pictures of thin women. Secondly, we used a baseline condition that consisted of looking at a cross (referred to here as low level baseline; LLB) instead of a reference condition (i.e. looking at another type of stimulus). Additionally, we recruited a large sample of patients before the start of treatment.

This fMRI study investigated both food preoccupation/eating problems and body image concerns in the same clinical group. Specifically, in women with BN and in HCs, we investigated brain activation patterns associated with cognitive processing of both high caloric food images and pictures of other women’s thin bodies. Overall, our data suggest that the neural correlates of self-schematic processing of slim other women’s bodies differ between women with BN and HCs; in contrast, the neural correlates of processing visual food stimuli do not differ substantially.

With respect to the processing of food stimuli, difference between people with BN and HCs were limited. The bilateral cuneus was more activated in BN in response to the food stimuli. While little is known about the role of the cuneus in eating or other psychiatric disorders, there is indirect evidence that it may be involved in food associated reward, i.e. in a decision-making task with food cues, cuneus activity was related to both value and saliency [45]. Successful treatment may have an effect on the cuneus as recovered BN patients reportedly show less activation when given a taste of glucose [46]. Longitudinal studies with repeated assessments are required to establish whether treatment can have an effect on the role the cuneus plays in appreciation of food in people with BN.

The notion that dysfunctional fronto-striatal circuits underlie food and eating related psychopathology in people with BN, is not supported by our findings as we did not replicate previous findings of decreased DLPFC activity [10, 12]. Two possible explanations for this relate to methodological aspects. First, the instructions on how to engage with food stimuli [those in the current study are the same as in other studies [10, 12, 15]] may impact on brain activation patterns. In the present study, instructions were adapted to the type of stimulus (food/body, control or LLB) and were repeated prior to each block to ensure continuous engagement throughout. In two previous studies, participants were instructed once, prior to the start of the experiment, with the following wordings “You will be shown pictures of food and other objects. Look at each picture and think how hungry it makes you feel” [10] and “You will be shown pictures. Look at each picture attentively” [12], respectively. It is conceivable that in the present study, the processing of stimuli was more consistent throughout the paradigm. Our instructions were the same as used by Brooks et al. (2011), but they were audio-recorded in that study. As in the present study, Brooks et al. (2011) did not report frontal or striatal activation differences between people with and without BN. While no study has directly investigated this ‘instruction’ effect in a within-subject comparison in people with an eating disorder, Siep et al. (2012) have demonstrated that brain activation in frontal and striatal regions can differ based on the instructions given, e.g. 1) to passively view foods, 2) to up-regulate food palatability thoughts, 3) to apply cognitive reappraisal (e.g., thinking about health consequences), or 4) suppress food palatability thoughts and cravings [47]. An indirect comparison with their findings suggests that the left medial prefrontal cortex activation seen in both our HC and BN group (Table 2) is in accord with the task to ‘up-regulate’ food palatability. However, our protocol most likely combines aspects of more than one of their instructions and this may explain why we do not find a difference in areas such as the DLPFC. Thus, the way the participants engaged with the visual stimuli in our and the Brooks et al. (2011) study may have been different from other studies that reported on altered frontal functioning [10, 12].

A second explanation for the difference between our and previous results relates to the analysis. The current study includes a LLB to which the food condition is contrasted; others used food vs non-food contrasts. Our post-hoc food vs non-food contrast analyses yielded no clusters of brain activation differences between people with BN and HCs, despite activity alterations in several frontal areas within the individual (BN and HC) groups (Table 4). Thus, this does not replicate any of the previous findings and hence use of a LLB control condition does not explain differences between reports.

It is important to interpret the current data in the context of the subjective experiences which indicate that anxiety, and not craving levels, are higher in the BN group, compared to the HC group. These within-paradigm findings correspond with post-scanning assessment of the stimuli in previous studies [10, 12, 14]; craving was not assessed by Brooks et al. (2011). This is the first study in BN to assess craving and anxiety during the paradigm and it appears that food provokes more anxiety rather than more craving in people with BN (compared to HCs). It is possible that visual presentation of food stimuli is less salient than other forms of exposure such as real food or virtual reality [48]. For example, exposure to real food results in more craving in people with bulimic disorders than HCs [49]. Within-paradigm assessments of food craving in HCs showed increased craving only when people were instructed to think about the palatability of the food, but not in other conditions [47]; this supports the idea that the instruction that we used was more complex and may have resulted in suppression of urges to eat the food in both healthy and BN participants.

In contrast to the processing of food stimuli, an evaluative comparison of own body against slim women is associated with distinct brain activation patterns in women with and without BN. People with BN activated the insula more and the fusiform gyrus less; this indicates that–when comparing themselves to slim women–they focus more on their self/own body [i.e. heightened self-referencing [50]] and less on the actual ‘other’ body shape/contours. Our data are in accord with findings of increased insula activation in people with BN when rating satisfaction of ‘thin self-body images’ [20] and increased anterior insula activation for the desired low body size in patients with AN [17, 51]. These data support hypotheses that propose that there is altered insula functioning during the integration of interoceptive information and emotion processing in people with eating disorders [26]. Less engagement of the fusiform gyrus, a key region in visual processing of bodies [52], in people with BN was also reported when line drawings of bodies were used [22]. Also, the left fusiform gyrus and the middle temporal gyrus may reflect a certain degree of body-related avoidance; interventions focused on body image do indeed show an increase in the Extrastriate Body Area (EBA) following successful treatment when patients look at other slim women [53]. We did not replicate the decreased medial prefrontal activation reported by others [18‐21]; this is most likely due to the methodological differences related to the stimulus type [54] and to the instruction. Furthermore, as the medial prefrontal cortex is involved in self-referential processing and participants were requested to compare themselves to the women in the pictures, this may have led to an increased activity in the medial prefrontal cortex that counteracted the expected effect. Similar to the food paradigm, anxiety during the task was more prominent in the BN group and may have accentuated insula activation.

It is possible that anxiety related to the study procedures plays a role in participants’ behavior. People with BN reported higher anxiety levels following exposure to visual food and slim body stimuli, although craving levels did not differ. Participants with BN were more anxious at the start of the scanning session, but, it remains unclear whether this relates to the fMRI procedure [55, 56] or to higher baseline stress and (anticipatory) anxiety.

To investigate the potential effects of antidepressant intake on the results, we analyzed brain activation patterns between participants on medication and those not on medication. The differences in the food (fusiform and lingual gyrus, and posterior cerebellum) and body image (lingual gyrus and anterior cerebellum) paradigm do not explain the main results of the current study. Detailed information is available from the corresponding author.

Strengths of this study are the use of two paradigms that relate to core BN psychopathology within the same group of participants and the assessment of subjective experiences in response to stimuli during the scanning. Secondly, the use of a LLB is beneficial; in its absence, the possibility that group differences already occur in the control rather than the active condition cannot be ruled out [17]. We note that Mohr et al. (2010) also used a fixation cross as baseline in their fMRI investigation of body image issues in people with BN and they also reported increased insula activation. Limitations include the lack of DSM Axis II assessment and, in the food paradigm, the use of various food types rather than an individualised set of stimuli. In addition, the current paradigm does not allow investigation of a contrast between high and low calorie foods: however, electrophysiological data show that people with BN have a high attentional bias towards food regardless of the caloric value [57]. For practical reasons, it was not possible to standardize the food intake prior to scanning. We did not study participants in the same menstrual phase but, an equal proportion in each group was on an oral contraceptive or in the follicular phase. It is also of note that the fMRI parameters may have resulted in only partial coverage of the cerebellum; hence these data should be interpreted with caution. ROIs in both paradigms were based on reported coordinates and were relatively small (in light of a conservative approach). Finally, this cross-sectional study is unable to differentiate between the ‘state’ or ‘trait’ nature of the findings.

In summary, processing visual food stimuli and comparing oneself to other slim women elicits more anxiety, but not craving, in women with BN compared to HCs. Women with and without BN use similar brain structures to process food stimuli. When comparing their body against slim women, women with BN use the insula more (i.e. reflect more on themselves) and the fusiform gyrus less (i.e. look less at the other’s actual shape). This supports the idea that psychotherapy for BN should have a particular focus on body image and not solely focus on food and eating related issues. There is evidence that psychotherapy can alter brain activation patterns in response to visual body stimuli in people with eating disorders [53]. Another clinical implication of our findings is that they can guide the development of future directed interventions such as transcranial or deep brain stimulation (e.g. low frequency, i.e. inhibitory, transcranial magnetic stimulation to the insula may reduce body image concerns).

FVDE was a research fellow in the Marie Curie Research Training Network INTACT (MRTN-CT-2006-035988) from 08/2007 to 07/2010.

RU consults for the World Health Organization. RU is supported by the European Commission Innovative Medicine Initiative Joint Undertaking (IMI-JU) grant number 115008.

US and AS receive salary support from the National Institute for Health Research (NIHR) [Mental Health Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King's College London]. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.