Anorexia Nervosa (AN) is a prevalent psychiatric disorder among adolescents, the majority of whom are girls with a peak age of onset at ~15 years of age (Bühren et al., 2017; Solmi et al., 2022). AN has a poor treatment prognosis, resulting in it being the mental illness with the highest mortality rate (Smink et al., 2013). According to ICD-11 criteria, AN is characterized by a significant loss of weight, leading to a body mass index (BMI) that is at least < 18.5 kg²/m² for adults or < 5th BMI age percentile for children and adolescents, respectively [World Health Organization (WHO, 2019/2021)]. Furthermore, individuals experience an extreme fear of weight gain, body image disturbance and high body dissatisfaction, which shows the high relevance of shape and weight in AN (Cash and Deagle, 1997; Fairburn and Harrison, 2003). In line with this, patients with AN show a particularly strong reaction to being confronted with their own bodies in front of a mirror, i.e., they show a strong increase in negative emotions as well as in negative cognitions (Vocks et al., 2007). This negative arousal upon being confronted with the own body can lead to both body avoidance and body checking behaviors (Glashouwer et al., 2019). While the former shows in behaviors such as avoiding mirrors, scales and situations in which oneself or others would see one's body (Rosen et al., 1991), the latter may include repeatedly measuring body parts and weighing oneself, as well as critically examining one's shape in the mirror. Body avoidance and body checking can be interpreted as being the behavioral implementation of negative body image (Glashouwer et al., 2019).

Another mitigating role in perpetuating negative body image in AN seems to be an attentional bias (AB), i.e., a skewed perception of the body, which occurs due to dysfunctional cognitive schemata (Aspen et al., 2013). A possible explanation for particularly strong ABs in patients with AN is the assumption of these patients having a highly developed self-schema based on body size and shape cues (Bauer et al., 2017). Hence, the self-schema distorts information processing in a way that results in dysfunctional visual scanning behaviors, which then perpetuates negative body image, as mentioned above. Corresponding to this, Kerr-Gaffney et al. (2018) state that individuals with AN have an AB for body parts they categorize as being unattractive and that the more one dislikes one's/their body, the stronger this AB is. Furthermore, Toh et al. (2020) found that AN patients have a particular AB to body parts which display information about weight status (i.e., lower and mid-torso regions, including the waist, stomach, buttocks, hips and thighs; see also Horndasch et al., 2012). Put together, the combination of high body dissatisfaction, a bias toward looking at body zones that display information about weight status, as well as having an AB toward body parts that are judged as being most unattractive, may lead to a vicious cycle that patients are unable to get out of.

Furthermore, viewing their own body (body exposure) evokes negative emotions, such as anxiety and disgust, in patients with AN. In this context, findings from Melles and Jansen (2023) support the assumed usefulness of exposures to reduce anorexic fears. More generally, exposure therapy has been described as the “gold standard” for the treatment of anxiety disorders (Van Loenen et al., 2022). It aims to reduce fear responses by altering pathological fear structures through activating the respective structure and subsequently providing information that is incongruent with it (Foa and McLean, 2016). It is important for the fear structure to be activated, as it cannot be modified otherwise. The same principle can be applied to body exposures, i.e., the body is the fear-inducing stimulus, in relation to which the fear structure needs to be modified. Interestingly, state anxiety also seems to be a stable predictor for the strength of the AB (Radix et al., 2023). While specific phobias have been shown to be treatable even in a single exposure session (Öst, 1989), more complex, fear-related disorders need several exposure sessions to modify existing fear structures [e.g., 10-12 exposure sessions for posttraumatic stress disorder (Sripada and Rauch, 2015) or 9–10 exposure sessions for social anxiety disorder (Jeong et al., 2021)]. Additionally, disgust might play a particularly relevant role in the maintenance of AN, as proposed by Glashouwer and de Jong (2021). Their proposition stems from the idea that individuals with AN internalize moral ideas (such as a societal notion of having to be thin) to a very high degree, so that when these are incongruent with their own feelings (e.g. “feeling fat”), a (self-) disgust response is evoked (Glashouwer and de Jong, 2021). This often results in body image avoidance, as individuals with AN try to avoid feelings of disgust (Espeset et al., 2012).

As Vocks et al. (2018) report, there have been multiple findings that body exposure therapy yields positive results on behavior associated with body image disturbance. Indeed, body exposure sessions are part of many structured programmes for the treatment of body image disturbance in AN, which have been developed and tested (Herpertz et al., 2018). Nonetheless, it should not be neglected that studies, such as one by Böse (2002), could not find significant improvements in all areas related to body image, despite finding generally positive effects of a body image group therapy. Tuschen-Caffier et al. (2015) state that—because individuals with AN focus more on negatively evaluated parts of their bodies, which perpetuates negative body image—attention should be directed toward neutrally and/or positively evaluated body parts, which are otherwise neglected. However, it needs to be highlighted that the number of studies on adolescent AN is scarce, while there are more studies on such programmes for adult AN. Nevertheless, there are a few studies on adolescent AN, such as one by Biney et al. (2021) who tested “practical body image” (PBI) therapy for children and adolescents with AN. PBI therapy consists of 14 sessions of talking about body perception, addressing challenges therein, as well as six sessions of mirror exposure. For our study, we used a modified version of the body exposure guidance from the manual “Body image therapy in Anorexia and Bulimia Nervosa: a cognitive-behavioral treatment program” by Vocks et al. (2018) to carry out the body exposure with our participants. Vocks et al. (2018) evaluated their own manual for body image therapy in several pilot studies, showing significant improvements in participants' body dissatisfaction, body checking behavior, a reduction of the intensity of negative cognitions in relation to their own bodies and an overall reduction of ED symptomatology (Legenbauer et al., 2011). Intrasession effects were analyzed in another evaluation (Vocks et al., 2009), using the body image states scale (Cash et al., 2002), which showed a decrease of body dissatisfaction within session. Hence, the authors of the manual (Vocks et al., 2018) came to the conclusion that their manual is a suitable instrument to treat body image disturbance in individuals with EDs. Nevertheless, they acknowledge that it is unclear which part of their intervention is the crucial one, i.e., the one responsible for improvements in participants' body image disturbance. Hence, it is unclear what role the body exposure sessions, and particularly the AB, play in the treatment of body image disturbance in adolescent AN. Our study aimed to move away from traditional mirror exposures, instead employing a photo-based approach. On one hand, this may ease the process of looking at one's body in a gentler way. On the other hand, it might have the advantage that looking at photos rather than a mirror is more similar to how adolescents see bodies on social media, hence aligning it to a growing part of virtual adolescent life. To distinguish the possible effects of the intervention from the regular treatment effects, we also employed a treatment-as-usual group (TAU).

Our hypotheses for this study were:

A study protocol has previously been published; please see Stonawski et al. (2022) for more detailed information on the methods. Not all parameters from the original study were taken into account for the analyses in this paper. Ethical approval for the study was granted by the medical faculty of Friedrich-Alexander-University Erlangen, Germany. The approval is in line with the declaration of Helsinki.

The study compared adolescents with AN in an intervention (INT) group to a treatment-as-usual (TAU) group. The INT group received four sessions of computer-based body exposures, while the TAU group did not. As recommended by Tuschen-Caffier et al. (2015), attention was directed to all body parts, so that participants were put off focusing solely on their shape and weight-displaying body parts. We were not able to find any significant improvements in body dissatisfaction or drive for thinness (EDI-2 subscales); body image avoidance (BIAQ); or body checking (BCQ) behaviors in either group (hypothesis 1). There are other studies, in which no changes in ED symptomatology after exposure training (e.g., ABMT) were found (Smeets et al., 2011; Porras-García et al., 2021). For example, Porras-García et al. (2021) also used a subscale of the EDI. It is unclear whether the previously mentioned constructs, which are recorded in the questionnaires, can be changed within a period of 2–3 weeks or whether the used questionnaires are sensitive enough to track small, short-term changes in participants' psychopathology, respectively. Perhaps, using different questionnaires, such as the body image states scale (Cash et al., 2002) which was used by Vocks et al. (2009), would be better to identify such changes.

We were able to replicate findings of (adolescent) patients with AN exhibiting a strong AB toward disliked and weight-displaying body parts (Kerr-Gaffney et al., 2018; Toh et al., 2020; Horndasch et al., 2012). However, upon analyzing our data in terms of a potential reduction of the AB in the INT group, no significant change pre to post intervention, nor an interaction with the TAU group, could be found (hypothesis 2). Nevertheless, a small trend was found for a reduction of the AB in frontal view in the pre to post intervention comparison for both groups. These findings make sense for the INT group, as the body exposures of the intervention are completed in frontal view. However, this trend was also present in the TAU group, which suggests that this slight change might not be due to the body exposures in the intervention, but rather a general treatment effect. This was previously also observed by Kampmann et al. (2018) in a study on social anxiety disorder when comparing participants in an exposure therapy condition to participants in a waiting-list condition.

Analyses of the AB scores from the INT group within and between sessions revealed no significant reduction of the AB (hypothesis 3). We also found no correlation between AN-psychopathology (as measured by the abovementioned questionnaires) and AB scores, i.e., a stronger AB did not correspond to higher AN-psychopathology (hypothesis 4). Ferrer-Garcia et al. (2021) mention that a stronger AB pre exposure therapy is often connected to a lower decrease of ED symptoms during it. Since participants mostly exhibited a strong AB, it may be that we found no changes in the strength of the AB or the questionnaires because of this characteristic. Nevertheless, it should be mentioned that, in some instances, it was found that participants' body dissatisfaction was reduced during body exposure, even though their AB did not change (Vocks et al., 2007; Smeets et al., 2011; Porras-García et al., 2021). Vocks et al. (2018) themselves state that their intervention had the least effect on participants' body perception, while improving cognitive-affective as well as behavioral aspects of EDs in participants. Also, during specific ED therapy, different levels of emotion and attention processing might show a different temporal trajectory with one system changing earlier or later than the other. Attentional biases to social affective pictorial stimuli appear to be more “trait-like” and associated with a lifetime history of AN, whereas emotion regulation difficulties appear to remit during treatment (Harrison et al., 2010). More research is needed to clarify these findings. There was a significant between-session decline of anxiety scores (hypothesis 5), which is in line with, e.g., Melles and Jansen (2023). On the other hand, disgust scores seem to have been activated within each body exposure, as pre to post session analyses consistently showed higher scores at the post session ratings, but which had then decreased again until the next pre session rating (hypothesis 5). Nevertheless, the activation pattern reduced across sessions so that the disgust response was activated less in session four compared to session one, for example. Based on the analyses of the anxiety and disgust scores, it can be said that there was a subjective benefit of the intervention for participants, even though this effect does not seem to have translated to bigger changes in ED pathology or a reduction of the AB. Especially the anxiety reduction should not be disregarded as a small benefit, as this would be a primary outcome for, e.g., anxiety-related exposure therapy (see Meyerbröker and Emmelkamp, 2010; Opriş et al., 2012).

The finding of reduced subjective emotional load but no changes in AB is consistent with reports on a disconnection between different emotion processing levels in AN. A distinct pattern of higher subjective emotion ratings without corresponding physiological responses has been found (Nandrino et al., 2012) e.g., for skin conductance responses, but also for visual stimuli more generally across different processing levels—albeit to a greater extent for food than body stimuli (Burmester et al., 2021), so more research into different processing levels during the specific intervention of body exposure is needed to determine the possible interactions between these. Strengths of our study included the randomized groups, which matched each other in terms of disorders, psychopathology and other demographic features (please see Table 1). Another strength was the standardized procedure of the intervention sessions, which included set time frames and the adherence to a pre-defined protocol according to which the sessions took place. All study attendants used the same pre-recorded audio guidance to facilitate the exposures and the large majority of participants always saw the same study attendant, while a few participants saw a maximum of two study attendants. Unfortunately, we were unable to control for instances in which sessions could not take place or had to be postponed due to illness, participants' distress (e.g., suicidal thoughts) and such. Limitations of our study included small sample size, because of which we were unable to, e.g., analyze subgroups of participants, such as comparing those with typical AN to those with atypical AN, as has been called for, e.g., by Meneguzzo et al. (2023); comparing those with restrictive AN to those with binge/purge-type AN; or analyzing different age groups. Also, participants were not evaluated for alexithymia prior to participation, the reduced ability to describe emotional states often found in EDs might have influenced the results. Participants were neither asked about their sexual orientation nor explicitly about their gender; however, none of the participants showed gender incongruence in the clinical setting in the course of a detailed child and adolescent psychiatric diagnosis. Unfortunately, some of the eyetracking data was of poor quality. Furthermore, the INT group did not record a viewing pattern at the post session, which would have made their data more comparable to that of the TAU group. Moreover, it would have been useful to provide participants with more than four intervention session, which was beyond the scope of our intervention. Subsequently evaluating our study, we agree with Vocks et al. (2007) that more exposures would likely be more effective. In comparison with Biney et al. (2021) whose PBI study comprised of fourteen session, targeting cognitive, affective, and behavioral aspects, including six sessions of mirror exposures, our study seems fairly short. The same applies in regard to the teenBodyWise study by Rosewall et al. (2020), which consisted of eight sessions focused on psycho-educative content about body image. Ascione et al. (2023), who combined an attentional bias modification treatment (ABMT) with an eyetracking paradigm in a virtual reality (VR) setting, showed that it is possible to change the AB in adolescents with AN, even in a single session setting. They highlight the usefulness of ABMT in this context, which seems to be a promising instrument for future treatment options, especially when combined with VR (see also Porras-García et al., 2021).

Finally, we can say that our intervention was not able to change participants' ED symptomatology such as body dissatisfaction, drive for thinness, body image avoidance or body checking behaviors. We were able to confirm the existence of an AB toward disliked and weight-related areas of the body in adolescents with AN; however, our intervention did not significantly reduce this AB. For future research, it might be useful to carry out more body exposure sessions and to incorporate ABMT into an intervention paradigm.