Differences in demographic characteristics and study measures among DSM-5 bulimic-type ED groups are summarized in Table 1.

An initial test of the measurement objectification model resulted in a good fit to the entire sample data (Model 1, Table 2), and all loadings (ranging from .75 to .94)7* were significant (p<.001). The results of the multiple-group analysis revealed measurement invariance across diagnoses, as the difference in fit between constrained and unconstrained models was non-significant (Models 2-3, Table 2). Thus, all latent variables were adequately operationalized by their respective measured/observed indicators, and the measurement model was equivalent for DSM-5 groups.

When the structural components of the objectification model were specified (Figure 1A), the results indicated that the model provided a good fit to the entire sample data (Model 4, Table 2), and all paths were significant (see Figure 1B). The model, controlling for covariates, accounted for 50.1% of the variance in body shame, 47.8% of the variance in appearance anxiety, 40% of the variance in dietary restraint, and 65.5% of the variance in binge eating. However, the multiple-group analysis revealed structural path differences (i.e., structural non-invariance) across the three DSM-5 bulimic-type ED groups, as the difference in fit between constrained and unconstrained models8 was significant (Models 5-6, Table 2). To pinpoint the source of the non-invariance (i.e., which structural path(s) was/were not invariant across groups), follow-up analyses were conducted by comparing the more restrictive (constrained) model (i.e., the model with structural path fixed to be equal) with seven different (unconstrained) models, each allowing only one of the structural paths (Figure 1B) to vary (Byrne, 2011). Chi-square difference tests indicated that the path from dietary restraint to binge eating was responsible for the non-invariance [Δχ2 (2)=28.24, p<.001]. Further comparisons of the constrained and unconstrained models in each pairs of diagnostic groups indicated that the dietary restraint-binge eating association was significantly greater in BN [Δχ2 (1)=17.45, p<.001] and bulimic-type EDNOS [Δχ2 (1)=12.02, p<.001] than in BED (see the structural coefficients in Figure 1B).

The bootstrap procedure showed that all indirect effects embedded within the structural objectification model (Figure 1B) were significant in each diagnostic group (Table 3). The type of mediation is also shown in Table 3.

Finally, as can been seen in Table 2, neither rival models provided a good fit to the entire bulimic-type ED sample data, and higher AIC and BIC values (Models 7-8) suggested that the objectification structural model (Model 4 in Table 2; Figure 1B) is preferable to the rival models. Thus, we did not conduct multi-group SEM (see data analysis).

The present study aimed at evaluating in a large sample of individuals with bulimic-type EDs the objectification model of eating pathology (Calogero et al., 2011), in an attempt to advance our knowledge about the underlying mechanisms of influence of body checking on binge eating. We found that (a) body checking behaviours were associated with increased body shame and appearance anxiety, which, in turn, were associated with increased dietary restraint and binge eating; (b) there was a direct path from dietary restraint to binge eating; (c) the objectification model fitted the data well and its robustness is further supported by the fact that the configuration of the rival models are not equally plausible and less ideal (Byrne, 2011); (d) with one exception (i.e., the direct path from dietary restraint to binge eating) differences across DSM-5 diagnostic groups on the strength of the associations amongst model variables (Figure 1B) were not observed. This last finding is highly important given increasing efforts to identify underlying aetiological and maintenance processes of eating pathology regardless of specific diagnostic categorisation (e.g., Dakanalis, Timko, Clerici et al., 2014), and to develop common psychological assessment and interventions (Dakanalis, Timko, Zanetti et al., 2014).

The finding indicating that body checking behaviours impacted binge eating only through body shame and appearance anxiety amongst bulimic-type ED patients is in accordance with the objectification model of eating pathology, which postulates that binge eating has a regulatory function and occurs in order to reduce or cope with negative emotional experiences surrounding the body and physical appearance (Calogero et al., 2011). Our data are also consistent with extant studies amongst bulimic-type ED patients indicating the lack of a significant association between body checking and binge eating (e.g., Calugi et al., 2006; Dakanalis, 2014; Mountford et al., 2007; Reas et al., 2005). However, this study expanded them, as it contributes to gaining insight into how body checking behaviours may act in indirect ways to predict binge eating, highlighting the need for targeting the intervening variables in treatment programmes.

Body checking behaviours have also been found to affect dietary restraint (through body shame and appearance anxiety) which, in turn, is directly linked to binge eating. These results are in accordance with the objectification model of eating pathology (Calogero et al., 2011) and with previous, experimental studies showing that monitoring of one's body leads to increased negative emotional experiences towards body and physical appearance, predictive of later dieting practices (see Tiggemann, 2013). This pattern has also been confirmed in longitudinal studies (e.g., Dakanalis, Carrà, Calogero, Fida et al., 2014). Moreover, there is evidence that some bulimic-type ED patients deliberately monitored their body in order to induce a state of stress, anxiety and body discontent, increasing motivation to engage in dietary restraint, regardless of a net loss, gain, or stable weight (Dakanalis, 2014; Reas et al., 2005).

Overall, our findings on body shame and appearance anxiety in the association between body checking, dietary restraint and binge eating lend support to evidence showing that binge eating may be rooted in dietary restraint, negative affect, or some combination of the two (e.g., Crowther & Williams, 2011; Dakanalis, Timko, Carrà et al., 2014; Stice & Presnell, 2010). Although significant differences among DSM-5 diagnostic groups were observed in direct association between dietary restraint and binge eating, the weak, but significant, association found in the BED group (Figure 1B) is consistent with other studies indicating that moderate dietary restraint attempts may be a characteristic of BED (see Stice & Presnell, 2010). Nevertheless, the potential contribution of dietary restraint attempts to the maintenance of BED is, not fully understood, so far (Stice & Presnell, 2010).

To our knowledge, this was the first study evaluating in a large treatment-seeking sample for bulimic-type EDs the mechanisms by which body checking may impact binge eating. However, some limitations of the study should be acknowledged. First, despite complex data analytic procedures, the inclusion of certain covariates within SEM analyses, and the comparison of objectification model to two rival models, our findings need to be interpreted with caution, as the cross-sectional nature of our data precludes any firm conclusions about the sequence of model variables and does not allow examination of causal relationships and feedback maintenance loops within the model (i.e., if binge eating encourages further dietary restraint). Empirical testing of these loops would need an enhanced experimental and longitudinal design (Byrne, 2011). Second, although the inclusion of semi-structured interviews would reduce social desirability, the use of self-report measures of some constructs of interest is prone to relevant biases, and replication with other methods of data collection (e.g., ecological momentary assessment) would be beneficial. Finally, as noted in the method section, we re-categorized bulimic-type EDNOS cases previously diagnosed following DSM-IV criteria, before evaluating the mediating links between body checking and binge eating assumed in the objectification model of eating pathology. Although this is a standard procedure (Machado et al., 2013), it should be noted that participants were not interviewed again and we only relied on notes made during previous clinical interviews.

The role of body checking in the maintenance of bulimic-type EDs is significant (e.g., Fitzsimmons-Craft et al., 2011; Williamson et al., 2004) and many patients may not disclose their control behaviours to clinicians (because of shame or lack of awareness). Thus, it is essential that clinicians explore body checking behaviours when assessing/treating individuals with EDs (Tiggemann, 2013). The results of this study also highlight the importance of assessing body shame and appearance anxiety. Cognitive-behavioural therapy (CBT) is considered the treatment of choice for binge eating problems (Hay & Claudino, 2010). Although CBT implements body checking interventions as a matter of course, remissions involve only 40-50% of patients and it is more effective in reducing ED behavioural symptoms rather than producing changes in body image disturbance (e.g., Crowther & Williams, 2011; Ferrer-García & Gutiérrez-Maldonado, 2012; Hay & Claudino, 2010). However similar disturbances seem playing a major role in the persistence and relapse processes of bulimic-type EDs (e.g., Dakanalis, Carrà, Calogero, Zanetti et al., 2014; Hilbert & Hartmann, 2013). In particular CBT does not address the emotional experiences directed toward one's body and physical appearance (i.e., body shame, appearance anxiety) (Ahrberg et al., 2011; Crowther & Williams, 2011; Ferrer-García & Gutiérrez-Maldonado, 2012) that, according to our results, may be important targets. Therefore, the use of standard CBT interventions in combination with other promising approaches (i.e., virtual reality, exposure techniques) designed to address body image and weight-related problems, might be an important venue in order to decrease binge eating and improve treatment outcomes (Ahrberg et al., 2011; Crowther & Williams, 2011; Ferrer-García & Gutiérrez-Maldonado, 2012; Hilbert & Hartmann, 2013; Riva, Gaggioli, & Dakanalis, 2013; Riva et al., 2015).

The work was supported by a grant from the Onassis Foundation (O/RG 12410/GH23). The authors report no declarations of financial interest. Special appreciation is expressed to all participants and their parents.