Indirect association between childhood maltreatment and food and alcohol disturbance through insecure attachment in university students

Abstract

Background

Food and Alcohol Disturbance (FAD) is characterized by a functional relationship between disordered eating behaviors and problematic alcohol use. Childhood adversity has been found to be associated with FAD, but its potential impacts on attachment profiles and FAD behaviors have never been examined. The present study tests a model in which insecure attachment mediates the relationship between childhood maltreatment and FAD.

Methods

University students (N = 629) participated in the study, completing questionnaires on FAD, childhood maltreatment history, and adult attachment. Structural equation modeling was used to test the mediated effect of childhood maltreatment through attachment on FAD behaviors.

Results

Insecure attachment and childhood maltreatment were related; however, only insecure attachment was directly related to FAD behaviors. A history of childhood maltreatment was indirectly related to FAD, mediated by insecure attachment. Attachment insecurity, particularly preoccupied and fearful patterns, was directly related to FAD behaviors in students and mediated the relationship between childhood maltreatment and FAD.

Conclusions

These findings may explain how early experiences of child maltreatment can be related to attachment style, and later, in adulthood, influence FAD behaviors in students. Clinicians should adapt treatment and preventions strategies to help patients transition from insecure attachment to a more secure pattern and regulate their negative emotions.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40337-025-01321-y.

Plain language summary

Food and Alcohol Disturbance (FAD) is characterized by disordered eating in combination with problematic alcohol use, to control weight or to enhance alcohol’s effects. This study reveals that early life experiences, particularly childhood maltreatment (sexual, physical and/or emotional abuse), may contribute to the development of FAD, but not directly. Instead, these adverse experiences are related to attachment styles -the way individuals form and maintain emotional relationships- which, in turn, is related to FAD behaviors. Individuals with insecure attachment (characterized by fear of rejection or excessive emotional dependence) are more likely to engage in FAD. This suggests that FAD may be a coping strategy for emotional distress rather than simply being a habit or lifestyle choice. These findings highlight the importance of emotional regulation and attachment-based interventions in addressing FAD. Supporting individuals in developing more secure attachment patterns could be key to reducing the impact of childhood adversity on FAD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40337-025-01321-y.

Background

Food and Alcohol Disturbance (FAD) refers to a range of eating disorder behaviors that occur before, during, and/or after alcohol use. These behaviors aim to compensate for alcohol-related calorie intake and/or maximize the psychoactive effects of alcohol [1]. FAD is not a comorbid condition of alcohol use and disordered eating, but rather describes a functional relationship in which disordered eating behaviors occur in the context of alcohol consumption, with motives involving the functional relationship of the behaviors (alcohol-enhancement or alcohol-related compensation for caloric intake) [2]. It represents a distinct phenomenon from the co-occurrence of alcohol use and eating disorders, which exist independently for distinct purposes [1]. FAD is relatively prevalent in university students, with approximately 50% of students reporting engagement in this behavior [3–5].

Numerous studies have examined the correlates of FAD (for review see 2,5). In addition to the alcohol and disordered eating-related behaviors primarily described in the literature, the association between psychological factors and FAD has also been reported. Early maladaptive schemas, such as emotional deprivation, subjugation, failure, and social isolation, have been associated with the frequency and motives of FAD [6, 7]. Similarly, avoidance strategies intended to avoid emotional distress arising from stressful situations are a key predictor of FAD behaviors among young adults [8]. These findings suggest that early maladaptive schemas may be a risk factor for the development of FAD behaviors. These studies also support the hypothesis that traumatic events in childhood may play a critical role in FAD in adulthood. Indeed, adverse childhood experiences [9] and body image victimization (referring to the humiliation about a person’s body shape) perpetrated by peers and parents during childhood and adolescence have been associated with FAD behaviors [10]. Psychological distress was a mediator in the relationship between body image victimization and FAD, such that more frequent past body-shaming experiences predicted higher levels of psychological distress, which, in turn, predicted higher levels of FAD in young adults [10].

Childhood adversity has been reported to be associated with harmful drinking in adults [11] and eating disorder symptoms in a clinical sample of individuals with eating disorders [12]. In these studies, attachment insecurity was identified as a significant mediator, supporting the idea that insecure attachment may explain how childhood maltreatment contributes to alcohol use and eating disorders in adulthood. Despite numerous studies on attachment and childhood maltreatment in students with heavy alcohol use or eating disorders symptoms [13–16], the relations between attachment profiles, childhood maltreatment, and FAD behaviors have not been explored to date.

Attachment patterns develop through repeated interactions between infants and caregivers during times of perceived threat or distress, shaping expectations and behaviors towards others [17]. Attachment relationships remain significant throughout life [17, 18] and can function as either a risk or protective factor against emotional problems when other risks are present [19], such as mental health issues. Secure attachment is associated with comfort in intimacy, relative autonomy in relationships [18], satisfying relationships, emotion regulation, and a stable sense of self. In contrast, insecure attachment often leads to heightened sensitivity to relationship loss, an incoherent sense of self, preoccupation, leading to anger or passivity, and difficulties in emotion regulation [18]. Preoccupied individuals feel unworthy but maintain a positive view of others [20]. Fearful individuals have a sense of unworthiness combined with an expectation that others will hold negative feelings, perceiving them as untrustworthy and rejecting. This attachment style is characterized by the tendency to avoid close relationships as a means of self-preservation, anticipating potential rejection. Although fearful and preoccupied attachment styles differ in interpersonal behaviour (fearful individuals tend to avoid intimacy due to distrust, while preoccupied individuals are emotionally dependent), both share a negative internal working model of the self, high attachment anxiety, and difficulties regulating affects in close relationships [18, 21]. These shared characteristics are core components of the construct of insecure attachment, which may be particularly relevant when studying maladaptive behaviours such as FAD, which are associated with emotional distress and relational vulnerability.

From a developmental perspective, insecure attachment may be one of the pathways through which early adversity, such as childhood maltreatment, shapes maladaptive behaviours later in life. Childhood maltreatment disrupts the formation of secure attachment patterns, leading to negative internal working models of the self, impaired emotion regulation, and increased sensitivity to rejection. This may in turn increase vulnerability to coping strategies such as FAD that are used to manage distress or cope with feelings of rejection or emotional insecurity. Thus, attachment theory provides a valuable framework for understanding how early adversity may contribute to emotional and behavioural dysregulation in adulthood.

Insecure attachment has been associated with increased alcohol use, alcohol-related problems, and disordered eating among students [22, 23]. Specifically, preoccupied individuals are more likely to engage in alcohol consumption as a means of coping with negative emotions [24], alleviating anxiety related to social situations, and facilitating social relationships [25]. In contrast, fearful individuals use alcohol as a coping mechanism for stress [26], to avoid social interactions, and to suppress emotions [25]. Similarly, studies have demonstrated a correlation between the preoccupied and fearful attachment dimensions and disordered eating behaviours [14], including bulimia and binge eating [27, 28]. This was explained by emotional dysregulation, which is considered a key factor underlying this association [29, 30].

On the other hand, several studies indicate that childhood maltreatment is associated with more severe alcohol use [16, 31] and disordered eating among students [13]. Childhood maltreatment encompasses a wide range of interpersonal victimization experienced during childhood, including sexual abuse, often perpetrated by someone in a position of authority; physical abuse, involving interpersonal violence typically inflicted by a parent or authority figure; and emotional or psychological abuse, which includes instances of criticism, rejection, and humiliation [32]. The effects of childhood abuse can be multifaceted, including an inability to tolerate emotional states, particularly those of a negative nature. This can lead to psychological distress and have an impact well into adulthood [33–35]. Physical, sexual, and emotional abuse have been reported as associated with higher alcohol use [11] among high-risk youths [34, 36] and disordered eating in university students [13]. A study also reported that a history of traumatic events (e.g., physical or sexual assault, child abuse) and post-traumatic stress disorder symptoms were positively associated with FAD behaviors [37], supporting the link between trauma and FAD.

Study overview

Despite the abundant literature on attachment and childhood maltreatment in students with heavy alcohol use or disordered eating, further investigation is needed to better understand how childhood maltreatment and attachment may contribute to FAD behaviors. As noted in previous studies examining early maladaptive schemas, avoidance strategies, and trauma, FAD may serve as a means of escaping negative emotional states stemming from adverse childhood experiences and insecure attachment.

Childhood maltreatment, by disrupting the ability to form secure attachments, impairs emotional regulation and leads to maladaptive coping strategies. In the context of FAD, these maladaptive strategies may manifest as a functional link between alcohol use and disordered eating, rather than their independent co-occurrence. Specifically, insecure attachment, characterized by heightened sensitivity to social rejection and difficulties in managing psychological distress, may lead individuals to engage in FAD behaviors as a means of both emotional escape and achieving a sense of control. As supported by previous studies, FAD has been associated with avoidance-based coping strategies, such as denial or disengagement from stressors [8]. In this context, FAD may represent a maladaptive strategy aimed at suppressing or avoiding intense negative emotions [38, 39] and temporarily disengage from distress. This pathway highlights how the interplay between early adverse experiences and attachment insecurity may uniquely contribute to FAD behaviors. Unlike a history of childhood maltreatment, attachment insecurity may be amenable to change and could therefore serve as a target for prevention and support for individuals with FAD. Identifying attachment insecurity as a mediator between childhood maltreatment and FAD could enable personalized interventions based on attachment profiles and assist individuals in regulating FAD behaviors. Although no study has directly examined this pathway in FAD, previous studies have demonstrated the mediating role of attachment insecurity in the relationships between childhood maltreatment and heavy alcohol use [11] as well as eating disorder symptoms [12] in adulthood. Starting from this evidence and considering the shared characteristics between FAD, alcohol use, and eating disorders, investigating a potential indirect effect of childhood maltreatment on FAD via the mediating role of insecure attachment is theoretically justified.

Thus, the current study aims to investigate the potential mediating effect of attachment on the relationship between childhood maltreatment and FAD behaviors in university students. It is hypothesized that insecure attachment, conceptualized through preoccupied and fearful profiles, mediates the association between childhood maltreatment (including sexual, physical, and emotional abuse) and FAD.

Method

This study is part of a larger research program devoted to examining substance consumption among University students (ADUC; Alcohol and Drugs at University of Caen).

Participants

Six hundred and fifty-six volunteer students at the University of Caen Normandy (France) participated in the present study between 2022, 2023 and 2024. They were recruited based on their willingness to participate (through flyer posted on campus) and were individually invited to a laboratory room to complete paper-and-pencil questionnaires. Twenty-seven participants were excluded as they did not complete all the questionnaires, resulting in a final sample of 629 participants. All participants were aged between 18 and 36 years (20.0 ± 1.70), included both sexes (men N = 202, women N = 427), and were native French speakers. All participants were alcohol consumers (i.e., they had consumed alcohol in the last 12 months), but they were asked to abstain for at least 12 h before the examination (verified using a breath analyzer) to ensure they were not under the acute effects of alcohol when completing the study. This precaution was taken to increase the reliability of self-reported responses. None of the participants had a history of severe traumatic brain injury or neurological diseases. Sample characteristics of the participants are presented in Table 1.

Table 1.

Sample characteristics of the participants (N = 629)

Socio-demographic variables
Age Range	20.00 ± 1.70 18–30
Sex (Men/Women)	202/427
Alcohol variables
AUDIT Range	9.06 ± 5.87 0–31
Eating disorders variables
SCOFF Range	1.30 ± 1.29 0–5
Food and Alcohol Disturbance (FAD)
CEBRACS total score Range	23.5 ± 6.17 21–76
CEBRACS “alcohol enhancement” subscale Range	9.43 ± 4.51 7–35
CEBRACS “dietary restraint and exercising” subscale Range	8.93 ± 3.98 7–35
CEBRACS “purging” subscale Range	5.07 ± 0.56 5–14
CEBRACS “extreme fasting and vomiting” subscale Range	2.18 ± 0.76 2–9

Except for sex, data are shown as mean ± standard deviation. AUDIT: Alcohol Use Disorders Identification Test; CEBRACS: Compensatory Eating and Behaviors in Response to Alcohol Consumption Scale

All participants were informed about the study prior to their inclusion and provided written informed consent. The study adhered to the Ethical Principles of Psychologists and Code of Conduct of the American Psychological Association [40] for the ethical treatment of human participants. It was registered with and authorized by the “National Commission for Information Technology and Civil Liberties” under the number u24-20171109-01R1.

Measures

Descriptive statistics for all the measures listed below are presented in Tables 1 and 2.

Table 2.

Descriptive statistics and intercorrelations among variables included in the structural equation modeling analysis

		Mean ± SD	1	2	3	4	5	6	7	8	9
Food and Alcohol Disturbance	1. CEBRACS « alcohol »	9.43 ± 4.51	-
	2. CEBRACS « dietary »	8.93 ± 4.51	0.40***	-
	3. CEBRACS « purge »	5.07 ± 0.56	0.23***	0.34***	-
	4. CEBRACS « extreme fasting »	2.18 ± 0.76	0.36***	0.60***	0.37***	-
Attachment (RSQ scale)	5. Fearful attachment	3.16 ± 0.85	0.14***	0.09*	0.01	0.09*	-
	6. Preoccupied attachment	3.19 ± 0.89	0.15***	0.08*	0.03	0.11**	0.33***	-
Child maltreatment (CTQ scale)	7. Physical abuse	5.81 ± 2.37	0.06	-0.01	0.02	0.03	0.11**	0.05	-
	8. Emotional abuse	7.83 ± 3.93	0.09*	0.09*	0.04	0.09*	0.24***	0.22***	0.57***	-
	9. Sexual abuse	6.14 ± 2.78	0.11**	0.09*	0.06	0.12**	0.17***	0.14***	0.31***	0.39***	-

* p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001 (two-tailed)

The data are presented as Spearman’s rho coefficients (correlation coefficients for non-normally distributed data)

To minimize social desirability bias, the questionnaires were explained verbally to participants, who completed them independently without the experimenter’s presence.

Food and alcohol disturbance (FAD)

FAD was assessed using the French version of the Compensatory Eating and Behaviors in Response to Alcohol Consumption Scale (CEBRACS; [41]. The CEBRACS is designed to investigate compensatory eating behaviors related to alcohol consumption over the past three months, including behaviors aimed at compensating for alcohol-related calorie intake and/or enhancing the intoxicating effects of alcohol. Participants were asked to rate the frequency of each behavior (1 = never; 2 = rarely (approximately 25% on occasions); 3 = sometimes (approximately 50% on occasions); 4 = often (approximately 75% on occasions); 5 = nearly always) for three time periods: before drinking, while under the effects of alcohol (during drinking), and after the effects of alcohol wore off (after drinking). The CEBRACS yields a total score ranging from 21 to 105 points (Cronbach’s α = 0.91) and consists of four subscales: “alcohol effects” (7 items (CEBRACS 1;3;6;7;9;12;14); Cronbach’s α = 0.94); “dietary restraint and exercising” (7 items (CEBRACS 2;4;10;11;16;18;20); Cronbach’s α = 0.90), “purging” (5 items (CEBRACS 5;8;13;15;17); Cronbach’s α = 0.85) and “extreme fasting and vomiting” (2 items (CEBRACS 19;21); Cronbach’s α = 0.68). A confirmatory factor analysis conducted on our sample supported the factor structure of the French version (CFI = 0.990; RMSEA = 0.041, 95% CI [0.035–0.046]; SRMR = 0.056).

Using the established CEBRACS cutoff score of > 21 points [41–46], 55% of participants in the study were identified as engaging in FAD behaviors.

Alcohol and eating disorders measures

The French version of the Alcohol Use Disorders Identification Test (AUDIT; [47] which consists of 10 items, was used to assess the risk of alcohol-related problems. Each question is scored from 0 to 4, with a maximum AUDIT score of 40. The AUDIT has been validated and recommended as an effective measure of alcohol use disorder in university students [48, 49].

The SCOFF questionnaire was used to assess the risk of developing eating disorders in at-risk and student populations [50, 51]. The questionnaire consists of five dichotomous questions (“yes” or “no” answers), with a total score ranging from 0 to 5.

Attachment

The participants’ attachment style was assessed using the French version of the Relationship Scales Questionnaire (RSQ; [52, 53]. Adapted from the Relationship Questionnaire [18] and the Adult attachment Scale [54], the RSQ is the most commonly used questionnaire for assessing different attachment styles [53]. The RSQ consists of short statements, and individuals rated their responses on a five-point Likert scale (“not at all like me” = 1 point to “exactly like me” = 5 points). The preoccupied attachment subscale (3 items (RSQ 8;16;25); Cronbach’s α = 0.52) assesses feelings of insecurity, anxiety, emotional instability, worry about being abandoned by significant others, and a tendency to perceive all situations as threatening. The fearful attachment subscale (4 items (RSQ 1;5;12;24); Cronbach’s α = 0.60) assesses a desire for intimacy with significant others, along with perceived difficulty in depending on and trusting them due to a negative self-image and perception of others. Item-level analysis indicated that no single item significantly improved internal consistency, suggesting that the observed reliability reflects the conceptual heterogeneity rather than measurement error. Scores for each attachment pattern are derived by calculating the mean score of the corresponding items, with higher scores indicating higher levels of attachment.

We chose to include only the fearful and preoccupied subscales, as these profiles are theoretically and empirically associated with emotional dysregulation and compensatory behaviors such as FAD. These two styles reflect high levels of attachment anxiety and interpersonal distress, in contrast to the dismissive style, which is associated with emotional deactivation and was not theoretically aligned with the emotional and social regulation functions underlying FAD behaviors (see Introduction).

Childhood maltreatment

Childhood maltreatment was assessed using the Childhood Trauma Questionnaire (CTQ; [55]. The CTQ is a questionnaire with subscales that measure different types of childhood trauma experiences: emotional abuse (5 items (CTQ 3;8;14;17;22); Cronbach’s α = 0.81), physical abuse (5 items (CTQ 9;10;12;15;16); Cronbach’s α = 0.81), and sexual abuse (5 items (CTQ 11;19;20;21;24); Cronbach’s α = 0.77). All items are 5-point Likert-type questions (from “never true” to “very often true”), with higher scores indicating greater levels of childhood trauma experiences. The French version of the CTQ has been demonstrated to have good reliability and validity [56].

Statistical analyses

For all participants, the associations among FAD, attachment, and childhood maltreatment measures were assessed using Spearman’s rho coefficient (two-tailed), as the data were not normally distributed (skewness and kurtosis values all >|2|). Subsequent analyses were conducted to examine associations between childhood maltreatment, attachment insecurity, and FAD. Analyses tested whether maltreatment was associated with both attachment insecurity and FAD, whether attachment was associated with FAD, and whether an indirect association between maltreatment and FAD via attachment emerged. The CEBRACS subscales were treated as the latent FAD factor (from the items of the four subscales described in the Measures section), while the fearful attachment subscale and preoccupied attachment subscale of the RSQ were treated as the latent attachment insecurity factor (also described in the Measures section). The sexual abuse, physical abuse, and emotional abuse subscales of the CTQ were considered as the latent factors of childhood maltreatment (from items of the three subscales described in the Measures section). The definition and coding of all variables used to construct these latent factors are detailed in the publicly available OSF materials associated with the project (see OSF: https://osf.io/4h9tj/).

Confirmatory factor analysis (CFA) was used to test the fit of the measurement model and assess associations with FAD, attachment, and childhood maltreatment factors. Criteria for acceptable model fit included Chi², a Comparative Fit index (CFI) greater than 0.95, and standardized root mean squared residual (SRMSR) of 0.08 or less. Variables contributing significantly to each factor were identified with a p value of ≤ 0.001.

Structural Equation Modeling (SEM) was then applied to all participants because the CEBRACS, RSQ, and CTQ scales produced continuously distributed scores with sufficient variability to support dimensional modelling across the entire sample. For each scale, a higher score indicates greater frequency or severity of the behavior. According to Bartholomew and Horowitz [18], the attachment subscales were not designed to be used categorically; rather, each sub-score reflects the extent to which the individual perceives themselves as corresponding to a particular attachment style. SEM was conducted using the Diagonally Weighted Least Squares (DWLS) method because the measures were obtained from ordinal scales and were non-normally distributed [57, 58]. Although the CEBRACS subscales were computed as composite scores from 5-point Likert items, several of them showed significant deviation from normality (with skewness kurtosis values ranging from 0.07 to 134), which justified the use of DWLS to ensure robust parameter estimation. FAD measures were specified as the latent endogenous variable and attachment and childhood maltreatment as the latent exogenous variables. Alcohol and eating disorders measures were included in the analysis to account for their potential confounding effects. The SCOFF score was used as a continuous variable, because previous studies have demonstrated that it can validly reflect the severity of disordered eating symptoms when treated dimensionally [46, 59]. All variables entered into the model were standardized before estimation to ensure equal weighting. Criteria for acceptable model fit included Chi², a CFI greater than 0.95, a SRMR of 0.08 or less, a root mean square error of approximation (RMSEA) of 0.08 or less, and a Bentler-Bonett Non-Normed Fit Index (NNFI) greater than 0.60. The indirect effect (i.e., mediated effects) between childhood maltreatment and attachment on FAD was tested using the following defined path: childhood maltreatment * attachment ~ FAD. To obtain more accurate parameters, bootstrapping (1000 replicates) was used with an adjusted bias-corrected method [60]. The indirect effect is statistically significant at p ≤ 0.05 if the 95% CI for the estimate does not include zero [61].

Results

Associations between FAD, attachment, and childhood maltreatment measures, assessed using Spearman’s rho coefficients, are presented in Table 2. The subscales within each measure (CEBRACS, RSQ, and CTQ) were correlated with one another. The RSQ subscales were positively correlated with the CTQ subscales, except for the associations between the RSQ preoccupied subscale and CTQ physical abuse subscale. The CEBRACS “alcohol” subscale was positively correlated with both RSQ subscales. The CEBRACS “dietary restraint and exercising” subscale was positively correlated with the RSQ subscales, and both the CTQ emotional abuse and sexual abuse subscales. The CEBRACS “purging” subscale was not correlated with the RSQ or CTQ subscales. The CEBRACS “extreme” subscale was positively correlated with the RSQ subscales and both the CTQ emotional abuse and sexual abuse subscales. The AUDIT score and the SCOFF score were positively correlated with the CEBRACS subscales (all p values ≤ 0.001).

The CFA conducted to test the fit of the measurement model and assess relationships with FAD, attachment, and childhood maltreatment factors showed that the measurement model provided an excellent fit to the data (Chi² ₍₂₄₎ = 36.51; p = 0.06; CFI = 0.99 and SRMR = 0.03 (90% CI [0.01–0.05]). All the CEBRACS subscales, the attachment subscales, and the childhood maltreatment subscales were highly correlated with the FAD, attachment, and childhood maltreatment factors, respectively (all p values < 0.001). Figure 1 shows the latent factors with standardized coefficients for the variables. The results of the CFA also showed that the FAD factor was positively correlated with both the attachment factor (β = 0.22, p ≤ 0.001) and the childhood maltreatment factor (β = 0.14, p = 0.004), and the attachment factor was positively correlated with the childhood maltreatment factor (β = 0.43, p < 0.001). Results are presented in the Supplementary Table S1.

Fig. 1.

Structural model and path coefficients illustrating the relations between attachment, childhood maltreatment, and FAD. FAD: Food and Alcohol Disturbance, assessed with the CEBRACS subscales. Alcohol was assessed with the AUDIT questionnaire; Eating disorders were assessed with the SCOFF questionnaire. The data are presented as standardized β coefficients. * significant at p ≤ 0.05; ** significant at p ≤ 0.01; *** significant at p ≤ 0.001

Results of the SEM showed a good fit for the data (Chi² ₍₃₈₎ = 41.0; p = 0.340; SRMR = 0.06; RMSEA = 0.01; 95% IC [0.01–0.03]; CFI = 0.995 and NNFI = 0.993), and accounted for 34.5% of the variance in FAD and 27.3% in attachment. To address whether fearful and preoccupied attachment should be modeled separately or conjointly, we compared two structural equation models: our initially proposed second-order factor labeled “insecure attachment,” and an alternative model considering fearful and preoccupied attachment as distinct latent predictors of FAD. Both models demonstrated good overall fit, but our initially second-order model was more parsimonious [62], with Parsimonious Normed Fit Index (PNFI) values of 0.791, and Chi²/ddl ratio of 1.08, compared to 0.622 and 1.69 in the two-factor solution. This supported our decision to retain the unified construct of insecure attachment in the final model.

To assess the robustness of the model estimation, an alternative model was run using a Robust Maximum Likelihood (MLR) estimator. However, the fit indices were less satisfactory (both CFI and TLI = 0.886; RMSEA = 0.059, 95% IC [0.05–0.06], p < 0.001), which further supported the appropriateness of the DWLS estimator given the ordinal scaling of the majority of variables. Results of the SEM with MLR estimator are presented in the Supplementary Table S2.

As shown in Fig. 1, the paths from childhood maltreatment to attachment insecurity (β = 0.25 (95% CI [0.07–0.43]; p 0.01) and from attachment insecurity to FAD (β = 0.52 (95% CI [0.39–0.65]; p < 0.001) were both significant, whereas the direct path from childhood maltreatment to FAD was not significant (β= -0.10 (95% CI [-0.26-0.06]; p = 0.228). Given the absence of a direct association between childhood maltreatment and FAD in our data, a mediation model is more appropriate as it allows for the investigation of indirect pathways, whereas a moderation model, which requires a direct association to justify interaction effects, is not theoretically or statistically viable. Also depicted in Fig. 1, the path coefficients from the alcohol and eating disorder covariates included in the model to FAD were statistically significant (β = 0.30 (95% CI [0.20–0.40]; p ≤ 0.001; β = 0.41 (95% CI [0.27–0.55]; p ≤ 0.001 respectively). The Tucker-Lewis Index (TLI), an index that accounts for the number of parameters and sample size [63], in contrast to the NNFI [64], also showed an excellent fit (value = 0.993). The bootstrap method (1000 replicates) with an adjusted bias-corrected method [60] was used to test the indirect effect in the model with the path: childhood maltreatment * attachment ~ FAD. The findings showed that the indirect effect of attachment on FAD, mediated by childhood maltreatment, was significant (β = 0.13 (95% CI [0.02–0.22], p = 0.036).

Monte Carlo power analysis showed that the power to detect both direct and indirect effects exceeded 0.995 across 1000 replications with a sample size of 629.

Mann–Whitney U tests showed a significant sex effect on the CEBRACS total score and on the “Alcohol Enhancement” and “Extreme Fasting and Vomiting” subscales (p < 0.005, Bonferroni corrected), with small effect sizes (r ≤ 0.16). Females scored higher than males. Due to the unequal sex ratio, sex was included as a covariate in the SEM model (male = 1; female = 2). A significant positive effect of sex on FAD was found (β = 0.37, 95% CI [0.14–0.26], p < 0.001), with higher levels in females. Including sex did not affect the significance or strength of other paths in the model. Age was not included as a covariate in the SEM due to its lack of association with key variables (all p values > 0.05) and low variability in the sample (see Table 1).

Discussion

The aim of the present study was to investigate the potential mediating effect of insecure attachment upon the association between childhood maltreatment and FAD behaviors. The results of the SEM conducted with all participants, using measures of attachment conceptualized within the framework of the fearful and preoccupied profiles, childhood maltreatment (including sexual, physical, and emotional abuse), and FAD, indicate an indirect association with childhood maltreatment on FAD through insecure attachment.

The results demonstrated that a history of child abuse may not be directly associated with FAD, contrasting with the existing literature, which indicates an association between adverse childhood experiences and FAD in university students [9, 10]. However, while our CFA analysis revealed a correlation between the child abuse factor and the FAD factor, the direct path between these variables was not significant in the SEM model. This suggests that the association is fully mediated by attachment, as the inclusion of attachment in the model accounts for the association. Indeed, early experiences of emotional, physical, and sexual abuse may alter the developmental trajectory and compromise attachment quality. This observation is consistent with evidence suggesting that childhood maltreatment and early dysfunctional family functioning impact adult attachment later in life [65].

The present study makes a novel contribution to the field by demonstrating an association between attachment insecurity and FAD behaviors. Prior research has established an association between insecure attachment and alcohol use, alcohol-related problems [22, 23], as well as disordered eating [14] among students. However, this is the first study to examine the association between insecure attachment and FAD. It has been demonstrated that preoccupied and fearful individuals are more likely to drink or to adopt disordered eating as a means of coping with negative emotions [14, 24–26]. Therefore, it can be posited that university students with an insecure attachment style may engage in FAD behaviours to facilitate social relationships [25], minimize fear of intimacy, and cope with anxiety-provoking situations. Additionally, these behaviours may be employed as a means of fostering a sense of belonging within a social group, driven by underlying concerns about abandonment and the need to navigate the fear of intimacy in relationships more effectively [66]. FAD behaviors have been shown to be associated with difficulties in emotion regulation [39, 67–69], including the use of these behaviors to enhance positive mood or reduce negative mood [70]. In addition, FAD may serve interpersonal functions, such as seeking social approval or fostering a sense of connectedness. For individuals with insecure attachment, who often struggle with emotional dysregulation and heightened sensitivity to social rejection, FAD behaviors may provide a maladaptive but functional coping strategy to address both emotional and interpersonal challenges. This functional specificity supports the theoretical association between insecure attachment and FAD behaviors.

The main result of the present study is the indirect association between childhood maltreatment and FAD through insecure attachment. This finding supports a model in which insecure attachment is not only directly related with FAD, but childhood maltreatment also partly explains its association through attachment on adult behaviors. This model may explain how early experiences of child maltreatment can be associated with attachment profiles and subsequent behaviors related to FAD. Our results align with studies demonstrating that insecure attachment mediates the association between childhood adversity and harmful drinking [11] or symptoms of eating disorders [12]. One possible mechanism suggested by the authors to explain this indirect effect is affect dysregulation occurring in the context of child abuse and its impact on attachment quality. Emotional dysregulation may lead individuals to engage in FAD behaviors [38, 68, 69] as a way to reduce negative affect in the absence of more adaptive emotion regulation strategies.

Students with a history of childhood maltreatment and insecure attachment may therefore develop FAD as a way to cope with distorted emotional perceptions. On the other hand, children who experience abuse may come to believe that they cannot reliably depend on their caregivers for support, which affects their relationships with partners in adulthood. This is in agreement with previous studies that reported associations between avoidance strategies [8], history of trauma [37], poor self-control, emotional deprivation, social isolation [6, 7], and FAD behaviors. According to the authors, FAD may be used to escape negative affect and cope with emotions induced by stressful situations. Overall, these previous studies support our results, demonstrating the indirect association between childhood maltreatment and FAD through attachment insecurity, which is characterized by individuals experiencing difficulties in regulating their affective states [18].

While this study provides valuable insights into the potential associations between childhood maltreatment, attachment insecurity, and FAD behaviors, the cross-sectional design limits the ability to establish causal inferences. Although our findings are consistent with theoretical models and previous studies in other populations suggesting an indirect pathway, it is important to emphasize that the directionality of these associations cannot be definitively determined. This limitation is particularly relevant given the novelty of the research on FAD and the absence of prior studies directly exploring its association with childhood maltreatment and attachment. Future research employing longitudinal designs is necessary to confirm temporal precedence and validate the proposed mediational model. Additionally, experimental studies targeting attachment insecurity could help to further clarify its causal role in the development and maintenance of FAD behaviors. Despite these limitations, this study represents an important step toward understanding the potential mechanisms underlying FAD and provides a basis for further exploration in this area.

Clinical implications

These findings encourage the assessment of attachment styles and inquiry into adverse childhood experiences to tailor support for students engaged in FAD behaviors. Schema-focused therapies targeting early maladaptive schemas [71] and trauma may provide essential support. Interventions should emphasize emotional regulation [72–74], foster self-esteem ( [71], and modify negative self-beliefs [72]. Additionally, the therapeutic alliance plays a critical role in creating a secure environment for gradual and appropriate emotional expression [72]. Since attachment patterns are dynamic and can change in adolescence [75] and adulthood [72], therapeutic strategies aimed at fostering a shift from an insecure to a more secure pattern should be considered [76].

Limitations

This study is limited to a single sample of French university students and should be replicated in other student populations as well as in non-student populations, where FAD behaviors have also been observed [42, 77]. The poor Cronbach alphas for the attachment measure may be due to the limited number of items used to qualify the attachment style. The low reliability observed in RSQ subscales is consistent with prior research highlighting the multidimensional nature of attachment constructs [21]. Each RSQ subscale assesses distinct but interrelated facets of attachment styles, which may explain the moderate inter-item correlations. For example, in the fearful RSQ subscale, items capture a range of experiences from discomfort in intimacy to distrust, potentially leading to lower consistency. However, previous studies have shown that attachment dimensions can be reliability assessed with self-reported measures [53, 54] and the RSQ is a better questionnaire to capture the multidimensional nature of attachment styles rather than the Relationship Questionnaire [53]. Moreover, the RSQ remains the most commonly used questionnaire for assessing different attachment styles [54]. This limited internal consistency may affect the precision of estimates in mediation models, particularly those involving latent constructs such as attachment. Therefore, future studies should consider using attachment measures with stronger psychometric properties. Although participants were asked retrospectively about their history of childhood abuse, attachment style and FAD behaviors were assessed concurrently in adulthood. As already mentioned, longitudinal studies assessing these factors across development are needed to clarify causality, but they are methodologically challenging. Finally, although DWLS estimation is appropriate given the data structure, it differs from the method used in previous studies and may limit direct comparability. However, re-analysis with MLR produced similar results.

Conclusion

In conclusion, the present study demonstrates an indirect association between childhood maltreatment and FAD behaviors through attachment insecurity in adult students. The findings highlight how early experiences of child maltreatment, particularly sexual, physical, and emotional abuse, may be negatively related to attachment styles characterized by fearful or preoccupied attachment, subsequently contributing to FAD behaviors in adulthood among students. From a clinical perspective, these findings suggest that assessing attachment functioning and childhood maltreatment is crucial for individuals exhibiting FAD, in order to tailor appropriate treatment and prevention strategies. Further studies are needed to more precisely characterize the role of a history of childhood abuse and the attachment insecurity profile in relation to the various motives of FAD.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

L. Ritz: data analyses, writing the original draft, formal analysis; N. Mauny: project administration, investigation, review and editing; C. Montcharmont: project administration, investigation; N. Dessommes: contributed to writing the introduction; D. Jacquet: conceptualization, review and editing of the original draft, supervision; M. Lemercier-Dugarin: review and editing of the original draft, supervision.

Funding

Non founding.

Data availability

Data and measurements are available at: https://osf.io/4h9tj Identifier: 10.17605/OSF.IO/4H9TJ.

Declarations

Ethical approval and consent to participate

All participants were informed about the study prior to their inclusion and provided written informed consent. The study adhered to the Ethical Principles of Psychologists and Code of Conduct of the American Psychological Association [40] for the ethical treatment of human participants. It was registered with and authorized by the “National Commission for Information Technology and Civil Liberties” under the number u24-20171109-01R1.

Competing interests

The authors declare no competing interests.