1. Introduction
Childhood maltreatment (CM) has largely been recognized as an important non-specific risk factor for Eating Disorders (EDs) [Reference Jacobi, Hayward, de Zwaan, Kraemer and Agras1]. Compared to both healthy subjects and people suffering from other psychiatric disorders, people with EDs report a higher prevalence of CM, regardless of its nature and of the type of the ED, although the association between CM and EDs seems to be less pronounced for subjects with anorexia nervosa (AN) restricting type [Reference Caslini, Bartoli, Crocamo, Dakanalis, Clerici and Carrà2]. In addition, maltreated individuals with EDs were seen to be characterized by a more severe clinical presentation in terms of earlier age at onset, symptom severity, psychiatric comorbidities and suicidal/self-harm behaviors [Reference Molendijk, Hoek, Brewerton and Elzinga3].
Among the different types of CM emotional abuse has shown a direct connection with ED symptoms while sexual and physical abuse have been associated to ED psychopathology through the mediation of one or more psychiatric comorbidities. The association between emotional abuse and EDs is corroborated also by the findings of Molendijk et al [Reference Molendijk, Hoek, Brewerton and Elzinga3] who have reported that AN and BN are more strongly connected with emotional abuse than with other types of CM. The potential mechanisms sustaining the association between CM and the risk of developing an ED have been partially investigated. From a psychopathological perspective, Trottier and MacDonald [Reference Trottier and MacDonald4] suggested that emotion dysregulation or maladaptive beliefs promoted by early adverse experiences are the mechanisms more closely associated with the ED symptoms. From a neurobiological point of view, the endogenous stress response system seems a putative mediator between CM and the risk of developing an ED, as recently highlighted by two studies showing a reduced cortisol awakening response and an impaired response of the hypothalamus-pituitary-adrenal axis to an acute psychosocial stressor in maltreated people with EDs [Reference Monteleone, Monteleone, Volpe, De Riso, Fico and Giugliano5, Reference Monteleone, Patriciello, Ruzzi, Cimino, Del and Steardo6]. In addition, reduced grey matter volume in the right paracentral lobule and in the left inferior temporal gyrus as well as reduced white matter integrity seem to be potential biological scars of CM in EDs [Reference Monteleone, Monteleone, Esposito, Prinster, Ruzzi and Canna7]. Thus, psychopathological, biological and clinical data appear to suggest the presence of a maltreated ecophenotype in EDs like in other psychiatric conditions [Reference Teicher and Samson8]. Despite this evidence, it is not clear whether each type of CM is a risk factor for specific or for all ED symptoms nor which clinical pathways account for these associations.
In the last few years, the network analysis (NA) approach has been changing the conceptualization of psychopathology and also providing prognostic and therapeutic implications [Reference Borsboom9, Reference McNally10]. Indeed, the network theory considers mental disorders as the product of symptom interactions. Their interplay constitutes feedback loops, which maintain a state of prolonged activation of symptoms that is phenomenologically recognized as the disorder [Reference Cramer, Waldorp, van der Maas and Borsboom11]. However, the network theory does not rule out that common causes may induce some symptoms that, in turn, promote the onset of others and together contribute to the maintenance of the disorder [Reference Levinson, Vanzhula, Brosof and Forbush12]. In these so-called “hybrid models”, a common cause accounts for certain symptoms, which then promote other symptoms through a network process [Reference Bringmann and Eronen13]. Thus, symptoms may be influenced by external factors, which are conditions outside the psychopathological network but not outside the person [Reference Borsboom9]. A NA investigates the relationships between symptoms, identifies the strongest connections as patterns of co-activation between symptoms, and explores the associations of symptoms with variables in the external field [Reference Borsboom9]. The subjective experience of CM is a potential variable belonging to the external field. Therefore, the use of NA may allow us to identify the possible pathways between each CM type and ED symptoms. A possible limitation of this approach is the inability to draw causal inference from the interactions among variables assessed as cross-sectional data [Reference Levinson, Vanzhula, Brosof and Forbush12]. Thus, in this study, we have integrated the NA with the mediation analysis approach [Reference Hayes and Rockwood14] in order to build up a comprehensive investigation of the pathways through which CM experiences operate to promote the activation of ED symptoms. Indeed, the mediation analysis is a highly validated method, which makes possible to elucidate how the potential causal effect of CM experiences on ED symptoms operates through the psychopathological nodes identified in these pathways by the network analysis. Given the lack of agreement in literature data regarding the association between CM types and ED-specific symptoms and/or diagnoses [Reference Caslini, Bartoli, Crocamo, Dakanalis, Clerici and Carrà2, Reference Molendijk, Hoek, Brewerton and Elzinga3], we have considered our study exploratory and no a priori hypotheses were formulated.
2. Methods
2.1 Participants
Consecutive patients attending the Eating Disorder Center of the Department of Psychiatry at the University of Campania “Luigi Vanvitelli” were included in the study if they met the following criteria: a) female gender; b) age ≥ 18; c) current diagnosis of AN, atypical AN or bulimia nervosa (BN), according to DSM-5 and confirmed by the Structured Clinical Interview for DSM-5 Disorders–Research Version (SCID) [Reference First and Williams15]; d) absence of current/lifetime comorbid diagnosis of schizophrenia, bipolar disorder, substance abuse disorder; e) willingness to cooperate in the experimental procedures and to sign a written informed consent. All patients were tested before entering specific treatment programs. Diagnostic assessment was made at the moment of referral to the outpatient unit by a trained psychiatrist (A.M.M.) who first made the diagnosis through a routine clinical interview and then used the SCID to confirm the ED diagnosis and, eventually, psychiatric comorbidity.
The ED sample was composed of 228 women: 80 suffering from AN restrictive type (ANR), 14 with atypical AN, 23 with AN purging type and 111 suffering from BN. Given the reported less pronounced prevalence of CM in patients with AN restricting type in comparison to other EDs [Reference Caslini, Bartoli, Crocamo, Dakanalis, Clerici and Carrà2, Reference Molendijk, Hoek, Brewerton and Elzinga3], we have distinguished participants with ANR from those with AN purging type, and because of the low number of the latter in our sample, we have merged these patients with those suffering from BN in a single “bingeing-purging” (BP) group. This is reasonable on the light of the evidence that people with AN purging type have a pattern of association to CM more similar to people with BN than to those with AN restricting type [Reference Molendijk, Hoek, Brewerton and Elzinga3].
2.2 Procedure
Sociodemographic, psychopathological and clinical data were collected through a face-to-face interview by expert psychiatrists.
Each participant in the study was asked to fill in the following questionnaires: 1) the Eating Disorders Inventory-2 (EDI-2) [Reference Garner16]; 2) the State-Trait Anxiety Inventory (STAI) [Reference Spielberger, Gorsuch, Lushene, Vagg and Jacobs17]; 3) the Childhood Trauma Questionnaire (CTQ) [Reference Bernstein, Stein, Newcomb, Walker, Pogge and Ahluvalia18].
The EDI-2 [Reference Garner16] evaluates ED symptomatology and psychopathology. The questionnaire includes 11 subscales: ineffectiveness, social insecurity, drive to thinness, interoceptive awareness, maturity fear, body dissatisfaction, perfectionism, interpersonal distrust, impulsivity, bulimia and ascetism. In our sample, Cronbach’s values ranged from 0.75 (maturity fears) to 0.94 (ineffectiveness).
The STAI [Reference Spielberger, Gorsuch, Lushene, Vagg and Jacobs17] is a self-report 40-item questionnaire which evaluates two types of anxiety: state anxiety, referring to temporary feelings perceived at the time of a threat, and trait anxiety, indicating how people feel in typical everyday situations. Cronbach’s values resulted to be 0.86 for STAI-state subscale and 0.91 for the trait subscale.
The CTQ [Reference Bernstein, Stein, Newcomb, Walker, Pogge and Ahluvalia18] explores a self-report recall of childhood trauma experiences. It is a 28-item questionnaire which identifies childhood experience across five types of CM: emotional neglect (EN) (cut-off score ≥15; Cronbach’s α = 0.88), emotional abuse (EA) (cut-off score ≥10; Cronbach’s α = 0.85), sexual abuse (SA) (cut-off score ≥ 8; Cronbach’s α = 0.74), physical neglect (PN) (cut-off score ≥ 8; Cronbach’s α = 0.78) and physical abuse (PA) (cut-off score ≥ 8; Cronbach’s α = 0.71). The reported cut-off scores indicate the occurrence of that type of maltreatment [Reference Walker, Gelfand, Katon, Koss, Von Korff and Bernstein19]. In our network we included a sum-score for each sub-scale [Reference Bernstein and Fink20] given that continuous scales are considered the best measures for network analyses [Reference Levinson, Vanzhula, Brosof and Forbush12]. The use of continuous measures to evaluate trauma is largely documented in literature studies [Reference Trottier and MacDonald4] and also employed in network analysis studies [Reference Epskamp, Cramer, Waldorp and Schmittmann21].
The study was approved by the Institutional Board of the University of Campania L. Vanvitelli.
2.3 Statistical analysis
An independent samples t-test was used to compare age, BMI, illness duration, age at onset, EDI-2, CTQ scores between the AN restricting group and the binge-purging group. The presence/absence of each type of CM was compared between the study groups by the Chi-square test.
2.4 Network analyses
NA was performed through R, version 3.4.4, using qgraph package [Reference Epskamp and Fried22]. A network is composed of nodes, which represent the measured variables, and edges, which are the connections among them. We have included the following nodes in our network: the EDI-2 subscores, the CTQ items and the state and trait scores of the STAI. Network associations are undirected, i.e. there is no direction in the association. The thickness of an edge graphically represents the magnitude of the association. In this study, we have estimated partial-correlation networks, where the association between two nodes is controlled for the influence of all other variables [Reference Friedman, Hastie and Tibshirani23]. We have built two separate networks: one for the sample with ANR and one for the BP group. In order to retain only meaningful associations, we applied a “least absolute shrinkage and selection operator” (LASSO) regularization [Reference Chen and Chen24], which shrinks small partial correlations and sets them to zero [Reference McNally10]. The Extended Bayesan Information Criterion (EBIC) [Reference Epskamp, Borsboom and Fried25], a parameter that sets the degree of regularization/penalty applied to sparse correlations, was set to 0.5 in these analyses.
We then estimated the accuracy of edge-weights, by drawing bootstrapped confidence intervals calculated through “nonparametric” bootstrapping (nboots = 2500) [Reference Epskamp, Maris, Waldorp, Borsboom, Irwing and Hughes26], by means of the bootnet package [Reference Brandes27].
Next, we computed two networks (one for each study group) showing the shortest paths between each childhood trauma node, as assessed by the CTQ, and the ED core symptoms (drive to thinness, body dissatisfaction and bulimia) evaluated by means of the EDI-2. The shortest path between two nodes represents the minimum number of steps needed to go from one node to the other [Reference Dijkstra28], computed using Dijkstra’s algorithm [Reference Isvoranu, van Borkulo, Boyette, Wigman, Vinkers and Borsboom29]. This kind of network allows us to identify the mediating variables between early traumatic experiences and specific symptoms of EDs. Indeed, it detects the shortest path between two nodes, i.e. the quickest out of all the routes connecting these two nodes. The undirected edges used in this network indicate conditional dependence between two variables: the edge-weight parameters reflect the strength of unique associations between variables, which in turn may highlight potential causal relationships [Reference Epskamp, Maris, Waldorp, Borsboom, Irwing and Hughes26, Reference Epskamp, Cramer, Waldorp and Schmittmann21].
2.5 Mediation analyses
Mediation analyses were conducted using the PROCESS macro for SPSS version 3.1 [Reference Hayes and Rockwood14]. We followed the Baron and Kenny method [Reference Baron and Kenny30] to confirm whether the CM experiences (the independent variables) predict ED core symptoms (body dissatisfaction, drive to thinness and bulimia; the dependent variables) through the mediation of those nodes identified in our network analyses as those included in the shortest pathways between CM and ED core symptoms. Indeed, mediation analysis focuses on specific pathways of influence on one or more dependent variables that need to be identified a priori. Since we had not a robust priori hypothesis from literature data, we used the network analysis to identify those symptoms that, in the global structure of ED psychopathology, convey the association between CM experiences and ED core symptoms. The statistical significances of the mediating and indirect effects were assessed using bootstrapped bias-corrected percentile based confidence interval of 5000 bootstrap draws [Reference Preacher and Hayes31]. All analyses were performed using SPSS version 25 (SPSS Inc, Chicago, IL).
3. Results
3.1 General results
Clinical characteristics of the study samples are reported in Table 1. A comorbid anxiety disorder was detected in 61 subjects while 68 individuals were diagnosed with a comorbid depressive disorder. The independent samples t-test (Table 1) showed that, compared to women with ANR, people with BP behaviors reported higher age and BMI and scored higher in the body dissatisfaction, drive to thinness, bulimia, interoceptive awareness and ascetism sub-scores of the EDI-2 questionnaire. Higher values were also detected in the BP sample for CTQ emotional neglect and physical abuse scores and for STAI trait value.
BMI, Body Mass Index; EDI, Eating Disorders Inventory; IN, Ineffectiveness; MF, Maturity Fear; SI, Social Insecurity; BD, Body Dissatisfaction; P, Perfectionism; ID, Interpersonal Distrust; IR, Impulse Regulation; DT, Drive for Thinness; BU, Bulimia; IA, Interoceptive Awareness; ASC, Asceticism; CTQ, Childhood Trauma Questionnaire; EN, Emotional Neglect; EA, Emotional Abuse; SA, Sexual Abuse; PN, Physical Neglect; PA, Physical Abuse; STAI-S, State and Trait Anxiety Inventory – State; STAI-T, State and Trait Anxiety Inventory – Trait.
The Chi-square results showed that the distribution of each CM type was not statistically different between the ANR and the BP groups, with the exception of emotional neglect and sexual abuse, which exhibited, respectively, a higher (X2 = 4.38, p = 0.03) and a trend towards higher prevalence in the BP group (X2 = 3.52, p = 0.06) (Table 2).
3.2 Network analyses
The network in Fig. 1 shows the connections between childhood trauma experiences and the EDI-2 and STAI scores in the ANR group. From the analysis of the network, it emerged that there were no direct connections between CM nodes and ED symptoms, except for the connection between emotional abuse and interoceptive awareness. This node was also directly connected with impulse regulations and with the ED core symptom drive to thinness. Moreover, all CM nodes were highly interconnected between each other, while the STAI nodes had weak connections only with EDI-2 ineffectiveness and social insecurity nodes.
* Pearson's chi-squared test.
Restrictive Group Network.
Estimated network of the restrictive AN group depicting the 5 dimensions of Childhood Trauma Questionnaire, the Eating Disorders Inventory-2 subscores and State and Trait Anxiety Inventory scores. Each edge within the network corresponds to a partial correlation between two nodes. For color version, see this figure online.
The network in Fig. 2 illustrates the connections between childhood trauma experiences and EDI-2 and STAI scores in the BP group. As in the ANR network, CM nodes were highly connected between each other and not connected with ED symptoms, except for emotional abuse that was connected with ineffectiveness. This node, in turn, was positively associated with interoceptive awareness, body dissatisfaction and social insecurity.
The shortest pathways analysis identified the shortest route between the CTQ nodes and the EDI-2 core symptoms. Fig. 3 illustrates these pathways in the ANR group. In this network, the shortest path to reach drive to thinness from each type of CM was via emotional abuse and interoceptive awareness. Interoceptive awareness was also in the shortest path between emotional abuse and ineffectiveness or interpersonal distrust.
Fig. 4 depicts pathways between trauma and ED psychopathology in the BP group. The shortest path to reach body dissatisfaction and drive to thinness from CM nodes included emotional abuse and ineffectiveness, while the route to reach bulimic symptoms was via emotional abuse, ineffectiveness and interoceptive awareness. Similar to findings in the ANR group, all the pathways from CM nodes to ED psychopathology passed through emotional abuse. Contrary to the ANR group, ineffectiveness, instead of interoceptive awareness, was included in all the pathways from CM to ED symptoms.
In both study groups, anxiety nodes were not in the pathways between CM and ED core symptoms.
The bootstrapped confidence intervals of estimated edge-weights are reported in Supplementary Fig. 1 and 2.
3.3 Mediation analyses
The network analyses conducted in both ED samples pointed to emotional abuse as the node from which all other CM types can evoke ED symptoms.
In the ANR group interoceptive awareness was in the shortest route between childhood emotional abuse and AN core-symptoms (Fig. 3). Thus, we assessed whether the association between emotional abuse and drive to thinness was explained by an indirect effect of interoceptive awareness. A total mediation effect was revealed pointing to the association between emotional abuse and drive to thinness as completely mediated by the interoceptive awareness (Indirect Effect: b = 0.28, 95% C.I.: 0.11 – 0.54).
Binge-purging Group Network.
Estimated network of the binge-purging group depicting the 5 dimensions of Childhood Trauma Questionnaire, the Eating Disorders Inventory-2 subscores and State and Trait Anxiety Inventory scores. Each edge within the network corresponds to a partial correlation between two nodes. For color version, see this figure online.
Restricting Group Shortest Pathways.
Network illustrating the shortest path between childhood trauma dimensions and eating disorder symptoms in the restrictive group. For color version, see this figure online.
Binge-purging Group Shortest Pathways.
Network illustrating the shortest path between childhood trauma dimensions and eating disorder symptoms in the binge-purging group. For color version, see this figure online.
In the BP group (Fig. 4) the shortest pathway between emotional abuse and body dissatisfaction included ineffectiveness while both ineffectiveness and interoceptive awareness were in the shortest path between emotional abuse and bulimic symptoms. Thus, two mediation models were run. In the first, ineffectiveness totally mediated the relationship between emotional abuse and body dissatisfaction (Indirect Effect: b = 0.29, 95% C.I.: 0.13 – 0.49). The second one was a serial multiple mediator model [Reference Spielberger, Gorsuch, Lushene, Vagg and Jacobs17]: emotional abuse was found to predict bulimic symptoms trough the contemporary mediation of ineffectiveness first and, then, interoceptive awareness (Indirect Effect: b = 0.09, 95% C.I.: 0.03 – 0.17). Ineffectiveness and interoceptive awareness were not significant independent mediators since the 95% CI for their indirect effect (between emotional abuse and bulimic symptoms) included zero (Indirect Effect for Ineffectiveness: b = -0.01, 95% C.I.: -0.1 – 0.05; Indirect Effect for Interoceptive Awareness: b = 0.001, 95% C.I.: -0.07 – 0.08).
4. Discussion
The present study provides, for the first time, network-based evidence of the associations between CM and ED psychopathology. By mean of the shortest pathways analysis, we have identified the childhood emotional abuse experience as the common node mediating the association between each type of CM and ED core symptoms. Moreover, in the ANR group interoceptive awareness was included in the shortest pathway between emotional abuse and drive to thinness, while in the BP group ineffectiveness was the psychopathological variable in the shortest route from emotional abuse to body dissatisfaction, and both ineffectiveness and interoceptive awareness were in the shortest pathway between emotional abuse and bulimic symptoms. The possible mediating role of these clinical variables (ineffectiveness and interoceptive awareness) between childhood adverse experiences and ED core symptoms has been confirmed through mediation analysis. By combining network and mediation analysis approaches we have potentially constructed a hybrid model, which accounts for the role of emotional abuse as a variable mediating between different early adverse experiences and some ED symptoms whose interplay may, in turn, maintain a state of prolonged activation of further ED symptoms.
In our pathway analyses we have observed that the effect of each CM type can propagate to ED core symptoms exclusively through emotional abuse. This is in line with previous studies suggesting that different types of CM do not have the same impact on ED symptoms [32–Reference Monteleone, Ruzzi, Patriciello, Pellegrino, Cascino and Castellini34]. In particular, a recent study indicated a direct association between emotional abuse and ED core symptoms while the associations of other CM types with ED symptoms were dependent on co-occurring psychiatric comorbidities [Reference Guillaume, Jaussent, Maimoun, Ryst, Seneque and Villain35]. In addition, other studies [Reference Groleau, Steiger, Bruce, Israel, Sycz and Ouellette36, Reference Racine and Wildes37] have hypothesized that emotional abuse promotes emotional deregulation, which significantly contributes to ED symptoms onset [Reference Aldao, Nolen-Hoeksema and Schweizer38].
Present data do not show any difference between ANR and BP groups in the pathways from each CM type and ED psychopathology. In particular, sexual abuse was associated with emotional abuse through physical abuse in both groups, which is surprising given the literature data [Reference Caslini, Bartoli, Crocamo, Dakanalis, Clerici and Carrà2] pointing to a higher prevalence of sexual abuse in subjects with BP symptoms, as we have also observed in our sample. Therefore, although ED groups may differ in terms of the prevalence of some CM types [Reference Caslini, Bartoli, Crocamo, Dakanalis, Clerici and Carrà2], our study points out that the experience of early adversities, assessed from a dimensional perspective, may generate similar pathways accounting for the associations to ED psychopathology.
Our findings also show that some differences occur between people with ANR type and those with BP symptoms. In the former group, the shortest pathway between CM (emotional abuse) and ED core symptoms (drive to thinness) included interoceptive awareness while in the latter group ineffectiveness alone mediated the effects of emotional abuse on body dissatisfaction and both ineffectiveness and interoceptive awareness were in the pathway leading to bulimic symptoms. Moreover, in the BP group, the mediation analyses showed that ineffectiveness mediated the relationship between emotional abuse and bulimic symptoms only through the mediation of interoceptive awareness. Therefore, we can suggest that interoceptive awareness plays a central role in the complex interactions between CM and ED core symptoms. This is in line with Bruch’s model [Reference Bruch39] of AN, which has been corroborated both in a recent review [Reference Treasure and Cardi40] and in a previous study from our group [Reference Monteleone, Castellini, Ricca, Volpe, De Riso and Nigro41]. Indeed, according to Bruch’s theory, it seems plausible that early humiliating and self-demeaning behaviors as well as not responding to the needs of the children may compromise their ability to discriminate body sensations and feelings such as hunger and satiety. Moreover, interoceptive awareness reflects the ability to discriminate not only between hunger and satiety perception but also different body sensations and feelings [Reference Pollatos, Kirsch and Schandry42]. This is particularly relevant on the light of the somatic markers theory, which claims that the recognition of the internal state is the foundation of our emotional feelings and is associated with the capacity to regulate behaviors [Reference Craig, Lewis, Haviland-Jones and Barrett43, Reference Damasio44]. In line with this, a wide range of studies has suggested that early trauma impact on emotion regulation skills and, in turn, may contribute to the development and maintenance of eating symptoms [Reference Trottier and MacDonald4]. Conversely, unlike in the ANR group, in the BP group the central role of ineffectiveness provides a slightly different route to bulimic symptoms suggesting an interplay between emotional abuse, feelings of inadequacy and interoceptive recognition ability in this clinical condition.
The central role played by interoceptive awareness in the pathways from CM to ED core symptoms in both the ANR group and the BP group allows us to rule out, at least in part, a previous hypothesis suggesting that the link between certain types of CM and EDs is mediated by the occurrence of psychiatric comorbidities [Reference Guillaume, Jaussent, Maimoun, Ryst, Seneque and Villain35], since interoceptive awareness is a specific ED psychopathological dimension [Reference Brown, Berner, Jones, Reilly, Cusack and Anderson45] and, in our NA, anxiety symptoms were not in the pathways from CM to ED core symptoms. However, our network did not include other general psychiatric symptoms (i.e. depressive or post-traumatic stress symptoms) which may also be involved in these pathways. Although this limitation, it is interesting to remark that a general psychopathological symptom, ineffectiveness, was found to be involved in the relationships between CM and ED core symptoms in BP individuals. In line with this evidence, previous studies showed that self-criticism or inadequacy, but not depression or anxiety, mediated between early emotional abuse and ED symptoms [Reference Feinson and Hornik-Lurie46, Reference Groleau, Steiger, Bruce, Israel, Sycz and Ouellette36]. Moreover, our finding may partially explain the nature of CM as a non-specific risk factor for different psychopathological conditions [Reference Leverich, McElroy, Suppes, Keck, Denicoff and Nolen47, Reference Young, Abelson, Curtis and Nesse48, Reference Teicher and Samson8]. The associations between ineffectiveness and ED core symptoms is in accordance with findings from recent NA studies which showed that feelings of worthlessness span ED symptoms and comorbid psychopathological variables [Reference Elliott, Jones and Schmidt49, Reference Smith, Mason, Crosby, Cao, Leonard and Wetterneck50] and predict treatment outcomes [Reference Elliott, Jones and Schmidt49, Reference Olatunji, Levinson and Calebs51]. Our findings supplement this evidence with the possibility to consider early emotional abuse as a possible variable associated with self-negative evaluation in both clinical assessment and therapeutic interventions.
Some limitations of the present study need to be acknowledged. First, the use of cross-sectional data outlines the need to exercise caution when interpreting conclusions regarding the direction of causality effects although, as explained below, this limitation has been partially overcome by our statistical model. Second, in our study we were not able to include general psychiatric symptoms, such as depressive symptoms, which are frequent in EDs [Reference Spindler and Milos52] and may be possible mediators of the connections between CM experiences and ED core symptoms. Third, a larger sample would have been appropriate to improve the strength of our findings in each ED group and in particular subsamples such as, for example, people with defined past event of sexual abuse or those who had a transdiagnostic cross-over or a diagnostic stability. Fourth, other variables, such as self-harm behaviors or quality of life, which have been found to be important outcomes when assessing the effects of CM in EDs [Reference Guillaume, Jaussent, Maimoun, Ryst, Seneque and Villain35, Reference Molendijk, Hoek, Brewerton and Elzinga3], or PTSD symptoms, emotion deregulation or dissociation, which are often associated with early trauma, were not explored in our sample. Finally, the CTQ, that we used to assess early traumatic experiences, relies on the subjects’ recollections without specification of the context and of the time of trauma occurrence: as a result, we cannot rule out that some participants had distorted memories of their early environment or chose not to disclose their maltreatment. This may have been responsible for some bias, although this limitation is not thought to be as critical as some might expect [Reference Brewin, Andrews and Gotlib53].
Our work presents some strengths also. Above all, it is the first research assessing the potential pathways between CM and ED core symptoms through a NA combined with pathways analysis. In addition, on the basis of the NA results, we have also run a mediation analysis yielding a possible hybrid model for the first time. The latter identifies symptoms connected to a possible common cause (CM) and those which directly perpetuate other symptoms. Second, we have performed our analyses in both the ANR group and the BP group, in accordance with literature recommendations [Reference Caslini, Bartoli, Crocamo, Dakanalis, Clerici and Carrà2, Reference Molendijk, Hoek, Brewerton and Elzinga3], and we have highlighted similarities as well as differences between the two ED groups. Third, we share all the codes used so as to enable reproducibility of the study design (see the appendix).
To conclude, our study provides the first evidence for possible pathways mediating the relationships between childhood trauma exposure and ED core symptoms. Future studies are recommended to further explore these connections including data deriving from multiple levels of analyses (i.e. biological, behavioral and psychophysiological data) and putative resilience factors for CM, such as the quality of parental bonding [Reference Monteleone, Ruzzi, Patriciello, Pellegrino, Cascino and Castellini34]. From a clinical point of view, the present work provides significant therapeutic implications as it sheds light on putative interventions in the external field [Reference Borsboom9] which refers to the associations between CM experiences and specific features of ED psychopathology. Indeed, in a recent study maltreated and non-maltreated ED patients showed different outcomes, particularly in terms of comorbid psychiatric symptoms persistence, two years after the completion of a cognitive-behavioral therapy [Reference Castellini, Lelli, Cassioli, Ciampi, Zamponi and Campone54]. Furthermore, a meta-analysis [Reference Grenon, Carlucci, Brugnera, Schwartze, Hammond and Ivanova55] highlighted the lack of superiority for cognitive-behavioral treatment in comparison to other bona fide therapies when assessing outcomes such as self-esteem and interpersonal problems in people with EDs. Thus, our findings suggest that, in the presence of childhood maltreatment, it may be important to extend intervention strategies beyond ED core symptoms, focusing on the person’s self-esteem development, abilities to recognize inner body states and emotions and taking into account their putative mediating role between childhood adverse experiences and ED symptoms.
Funding sources
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Declaration of Competing Interest
None.
Acknowledgements
None.
Appendix A Supplementary data
Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.eurpsy.2019.08.002.
Comments
No Comments have been published for this article.