In recent years several studies have explored the potential relation between personality variables and MDD. [6] proposed, from a psychoanalytic perspective, a model in which personality evolves through the dialectic transaction between two personality configurations, the anaclitic and the introjective, each of which arises from disruptions in the developmental process of establishing interpersonal relatedness and attaining self-definition, a positive sense of self, respectively ([10]). [8] postulated that the normal developmental process can lead to the formation of mature, satisfying relationships involving reciprocity and mutuality. It can also give rise to the establishment of an integrated, consolidated, and differentiated self-identity and sense of individuality. This theoretical formulation has been validated in studies of MDD ([12]; [55]; [62]). Hence, disruptions in the aforementioned process provoke psychological disturbances or else vulnerabilities to MDD, that is dependency and self-criticism. The former personality type seeks for approval of others and becomes depressed when important social ties are thwarted. The latter gains approval through academic or work success. The distorted preoccupation with issues of relations and the neglect of issues of self-definition can be a source of distress. Conversely, the exaggerated worry about accomplishments and achievement omitting the concern about relations can also cause depressed mood. The two configurations, apparently, were conceptualized into measurable depressogenic dimensions through the Depressive Experiences Questionnaire, dependency and self-criticism ([11]).

Attachment has been associated with adult depressive symptomatology ([52]). Childhood experiences become encoded into relationship patterns named internal working models ([13]) that may resemble a mental strategy. Three adult attachment categories have been delineated by [41]. The autonomous includes people who are objective and self-confident, and lack defensive behavior; while those with elevated scores in the preoccupied/enmeshed attachment tend to feel overwhelmed in their close relationships, they are afraid of abandonment, and are not sure about the commitment of those proximate to them. Finally, people with detached/dismissing attachment reject relationships, do not trust their family members, appreciate self-sufficiency, independence, and emotional control, and their idealized self-image can be overturned.

Along with the previously stated associations, literature encompasses studies regarding correlations between personality vulnerabilities, attachment patterns, and MDD ([18]; [45]; [59]). [6] descriptions of the two personality constructs share similarities with the two main insecure attachment patterns, the enmeshed/preoccupied and the avoidant/ dismissing. Common to both frameworks is the combination of two paths, of self and others, pertaining to MDD, namely the focus on the development of the self in a relational context. On the one hand, the fear of losing acceptance and love or being abandoned by the rejecting important others is the area of agreement between anaclitic dimension and preoccupied attachment. On the other hand, the mental representation of self as weak and incompetent and of others as critical and, subsequently, the ambivalence about one’s self-image, when someone realizes the difference between the ideal and the real, seems to be the common ground between introjective dimension and detached attachment ([19]). The relationship between vulnerability types and attachment in MDD has been emphasized by researchers ([36]; [39]).

Another relationship that has received attention is the association of attachment with self-esteem and MDD ([29]; [35]). Self-esteem refers to how people approve of and value themselves ([5]). The relationships throughout one’s life are transformed into models, representations of self and others. These early transactions become working models in the adult life that function as rules concerning the achievement of self-worth and the availability and responsiveness of the others. Insecure attachment may lead to poor self-perception and to negative view of oneself and important others and contribute to MDD ([14]). [60] have indicated the mediational role of self-image in the association between insecure attachment and MDD. Self-esteem has been studied both as a causal risk factor for and as a consequence of MDD ([47]). Nevertheless, relatively little work has been conducted regarding the potential connection between self-esteem and depressive vulnerabilities. Dependency may involve reciprocity and at the same time a clear and well-articulated sense of self, because the development of an individual as a demarcated entity is defined by the increasing capacity for mature interactions with significant others. The differentiation, integrity, and autonomy of one’s self emerge, are established, and maintained through interpersonal relationships ([8]). Self-criticism has also been linked with self-esteem ([43]).

Overall, in line with the two-polarities theoretically and empirically supported model of personality developed by [7], that was used as a valid framework in order to formulate our hypothetical model, this study focused on the intermediate factors between attachment and MDD. More precisely, in keeping with previous research, preoccupied attachment accords with the dependent vulnerability type ([61]) and is expected to connect, through the loss of object love, with low-self-esteem. On the other hand, detached attachment is associated with self-critical vulnerability type ([61]) and is expected to link, via the loss of ideal self-image, with low self-esteem. In light of the above-mentioned hypotheses, we explored the, mediated by personality vulnerability, relation between insecure attachment and depressive symptoms, via the moderation of self-esteem (Figure 1).

Figure 1. Hypothetical model.

Consistently with this aim, the following hypotheses were tested: Dependency will significantly mediate the association between preoccupied attachment and MDD, while self-criticism will significantly mediate the association between detached attachment and MDD. Simultaneously, self-esteem is expected to regulate the association between attachment, depressive vulnerability types, and depressive symptoms ([4]; [34]; [51]). Researchers have investigated depression with other variables, for instance [21] examined the connection between the need for affiliation and depression in university students, taking into consideration the mediating and moderating role of self-esteem and insecure attachment. Moreover, a meta-analysis ([54]) of longitudinal studies has explored the relationship between depressive vulnerability and depressive symptoms. Nevertheless, to the best of our knowledge, our study is the first to determine the jointly role of these psychological traits in MDD.

A convenience sample of 323 adult outpatients, aged between 18 and 56 years, participated in the research. This study primarily received permission from the Institutional Review Board of the University of Athens and approval from the National Health Operations Center and the directorate of each of 15 outpatient clinics of mental health centers and psychiatric departments of general hospitals in Athens, Greece, which implement a protocolized diagnostic process. Then the psychiatrists of each setting were contacted and 18 psychiatrists agreed to cooperate. During the study, the patients visiting these settings were attending either their first appointment or a subsequent, re-examination, one. The diagnostic procedure included the psychiatric interview, that lasted almost an hour, whereas the subsequent appointments were shorter in duration. The individuals who were approached and agreed to participate in the study numbered 345. They were recruited for the research voluntarily, on the basis of their medical file according to which, those who met the DSM diagnostic criteria for major depressive disorder, single episode or recurrent, joined the study group. The exclusion criteria comprised additional psychiatric disorders, such as bipolar disorders, organic brain syndrome, severe alcohol or drug addiction, mental retardation or other cognitive impairment that might create impediments in the accuracy of the assessment. The comorbidities permitted were dysthymic disorder, anxiety and personality disorders. All the selected subjects were administered the BDI and did not take part if they had a score of ≤10, that is, the cut-off score of mild depression according to the US and the Greek standardization samples ([57]). Subsequently, 19 patients were excluded due to their BDI score, as well as three questionnaires with missing values, that reached or exceeded 2% of the items. The remaining 323 individuals received a consent form with information about the purpose of the study, its confidential nature, their ensured anonymity, and the ability to communicate with the researchers. The completion of the questionnaires took place at the waiting areas of the settings, under surveillance, in individual sessions, after the appointment with their psychiatrist and responding to BDI; and it lasted approximately an hour and a half. The procedures followed were in accordance with the Declaration of Helsinki ethical principles. No reward or compensation were offered.

Vulnerability was measured by the Depressive Experiences Questionnaire-DEQ ([11]). It includes 66 items and three factors, dependency, and self-criticism, corresponding to the anaclitic and introjective type, respectively, and efficacy, that was excluded from the present study, because it was outside of its scope and it is not included in Blatt’s theoretical concept. Participants rate the statements on a 7-point Likert scale from 1 (strongly disagree) to 7 (strongly agree). Since there is no Greek version, the multiple forward backward approach, suggested by [28]), was applied. Confirmatory factor analysis was carried out with AMOS (version 21) to corroborate the factor structure. It showed a good fit to the data according to the fit indices, χ² (271) = 617.541, p < .001, χ²/df = 2.279, GFI = .94, TLI = .94, CFI = .95, RMSEA = .04, SRMR = .05. The Cronbach’s alphas in the current study were .86 and .81 for dependency and self-criticism, correspondingly.

Insecure attachment was evaluated by Cartes de Modèles Individuels de Relations-CAMIR ([49]). It is a 72-item scale that describes the attachment strategies and provides information on the quality of an adult’s attachment through internal working models. Responses are given on a 5-point Likert scale ranging from 1 (totally true for me) to 5 (totally untrue for me). It includes four templates, the secure and the insecure, the preoccupied, the detached, and the unresolved that was out of the current study’s aim. Based on the research of Bowlby and Ainsworth, CAMIR has acceptable psychometric properties with the coefficient alpha ranging from .68 to .95 and has been associated with MDD, suicide attempts, and schizophrenia ([48]). In its Greek standardization, coefficient alpha values ranged from .73 to .80 ([56]). In the current study the coefficient alphas were .77 and .90 for preoccupied and detached attachment, respectively.

The Rosenberg Self-Esteem Scale-RSES ([53]) was administered to document general self-worth. It comprises 10 items and uses a 5-point scale that ranges from 1 (totally disagree) to 5 (totally agree). The internal consistency was good with alpha coefficients ranging from .77 to .88. The test-retest reliability index has ranged from .85 to .88. RSES has also exhibited good concurrent, predictive, and construct validity ([5]). In its Greek validation the coefficient alpha was .80 ([26]). The coefficient alpha in the current study was .77.

Beck Depression Inventory-BDI ([2]) was used to measure the severity of recent depressed mood and has been validated in the Greek language ([57]). It is an instrument containing 21 items that reflect symptoms’ intensity rating from 0 to 3. It has showed good reliability with an average coefficient alpha of .81 and .86 in non-clinical and psychiatric populations, correspondingly. The test-retest reliability has ranged from .48 to .86 in clinical populations and from .60 to .90 in non-clinical populations. High construct, discriminant, and criterion validity have also been demonstrated with BDI ([1]). The coefficient alpha in this study was .75.

Initially, means, standard deviations, and correlations (Pearson’s r) between all study measures were computed. In order to test the hypothesis that the variables play an important role in predicting MDD, a hierarchical multiple regression analysis was carried out controlling for confounding demographic variables, such as comorbid disorder, hospitalization in a psychiatric department, psychotherapy, and pharmacotherapy, that were not used in the moderated mediation analyses. We finally performed mediation and moderation analyses into a unified model ([31]; [32]). The bootstrapping technique with bias-corrected 95% confidence intervals was taken with 5,000 resampling ([50]). This approach was used to assess the statistical significance of the indirect effect, from the independent variable to the dependent through the mediator, across differing levels of the moderator ([30]; [40]). In particular, insecure preoccupied attachment was the predictor variable, with dependent vulnerability as the mediator in the first case, whereas insecure detached attachment was the predictor with self-critical vulnerability as the mediator in the second case. The outcome variable was depressive symptoms and self-esteem was the proposed moderator. Hence, we tested the conditional indirect effect of self-esteem on the relationship between insecure attachment and depressive symptoms via depressogenic vulnerability to depressive symptoms. Figure 2 presents the hypothetical conceptual, moderated mediation, model that tests whether self-esteem moderates the direct effect c’ (prime) and the effect of path b. If the resampling distribution confidence interval of the indirect effect does not include zero, the indirect effect is considered statistically significant ([50]). All analyses were attained by SPSS-26 and PROCESS, a regression-based modeling tool designed to analyze effects in multiple mediation and moderation models ([30]).

Figure 2. Moderated mediation model: Relationship between insecure attachment and depressive symptoms mediated by vulnerability and moderated by self-esteem.

The sample of the study consisted of 323 MDD outpatients, with a mean age of 37.4 years (SD = 11.7), 67 (20.7%) males and 256 (79.3%) females, 22 (6.8%) with elementary, 21 (6.5%) junior high-school, 127 (39.3%) high-school, 44 (13.6%) college, and 109 (33.8%) university graduate or post-graduate education. Among them 96 (29.7%) had comorbidity with dysthymic disorder, anxiety and personality disorders whereas 227 (70.3%) did not, 173 (53.6%) had received or received pharmacotherapy while 150 (46.4%) had never taken medication, 173 (53.6%) had been or were treated with psychotherapy whereas 150 (46.4%) had never undergone treatment, and 291 (90.1%) had never been hospitalized while 32 (9.9%) had been hospitalized in the past. Participants were predominantly (94.4%) of Greek origin.

The means and standard deviations are presented in Table 1. Most of the correlations were small to medium. The two insecure attachment patterns demonstrated the strongest association. Dependency showed a higher positive correlation with MDD than self-criticism. Self-esteem had negative correlations with all the measures. Subsequently, according to the Beta values of the hierarchical multiple regression all the variables were significant predictors of MDD (Table 2). More specifically, the control variables predicted 9.8% of the variance, F (4, 318) = 9.777, p < .001 (step 1), preoccupied and detached attachment (step 2 and 3) predicted 14.9% of the variance, F (5, 317) = 12.294, p < .001, and 16.2% of the variance, F (6, 316) = 11.370, p < .001, of MDD, respectively. When dependency and self-criticism (steps 4 and 5) were added, the variance of MDD changed 17.4%, F (7, 315) = 10.714, p < .001, and 24.7%, F (8, 314) = 14.174, p < .001, correspondingly. In the final step 6, self-esteem predicted 39.1%, F (9, 313) = 24.001,p < .001, of the variance of the outcome variable.

Table 1. Correlations, means, and standard deviations of preoccupied and detached attachment, dependency, self-criticism, self-esteem, and depression.

Note: **p < .01, BDI: Beck Depression Inventory, CAMIR: Cartes de Modèles Individuels de Relations, DEQ: Depressive Experiences Questionnaire, RSES: Rosenberg Self-Esteem Scale.

Table 2. Hierarchical regression analysis for the prediction of depression.

Note: **p < .01, ***p < .001. BDI: Beck Depression Inventory, CAMIR: Cartes de Modèles Individuels de Relations, DEQ: Depressive Experiences Questionnaire, RSES: Rosenberg Self-Esteem Scale.

3.2.1. Anaclitic Moderated Mediation

As mentioned earlier, we tested the effect of self-esteem on the relationship between preoccupied attachment and depression via dependent vulnerability to depressive symptoms. The effect of dependent vulnerability on MDD was not significantly moderated by self-esteem (Interaction M × W coefficient = −.032, SE (HC4) = .023, t = −1.354, ns) while self-esteem was found to significantly moderate the effect of preoccupied attachment on MDD (Interaction X × W coefficient = −.069, SE (HC4) = .032, t = −2.169, p < .05). The interaction of the b path explains .4% while the interaction of the direct effect explains .8% of the variance of depressive symptoms. Higher dependent vulnerability was associated with higher depressive symptoms, B = .043, SE (HC4) = .021, t = 2.035, p < .05 (Table 3). The overall moderated mediation model was not supported with the index of moderated mediation = −.018 (SE = .014) (95% CI = −.047; .008). In particular, as zero is within the CI, it is indicated a non-significant moderating effect of self-esteem on the indirect effect, via dependent vulnerability, on depressive symptoms, although the analysis provided direct evidence of moderation of the indirect effect of preoccupied attachment by self-esteem. Subsequently, we can say with 95% confidence that the indirect effect of preoccupied attachment on depressive symptoms through dependent vulnerability does not depend on self-esteem. The indirect effects are not all significantly different from each other, as displayed in the significant index of moderated mediation. The conditional indirect effect was strongest in those low in self-esteem (1 SD below the mean; effect = .139, SE = .042, 95% CI = .056; .222) and mean self-esteem (effect = .085, SE = .028, 95% CI = .031; .140) and, not significantly, lowest in those high in self-esteem (1 SD above the mean, effect = .031, SE = .032, 95% CI = −.031; .094). Hence, we have an overall non-significant indirect effect.

Table 3. Unstandardized OLS regression coefficients with confidence intervals estimating preoccupied attachment, dependent vulnerability, self-esteem, and depressive symptoms.

Note: *p < .05, **p < .01, ***p < .001, ns: non-significant.

3.2.2. Introjective Moderated Mediation

As previously mentioned, we also tested the effect of self-esteem on the relationship between detached attachment and depression via self-critical vulnerability to depressive symptoms. Self-esteem was found to significantly moderate the effect of self-critical vulnerability on MDD (Interaction M × W coefficient = −.080, SE (HC4) = .025, t = −3.229, p < .01), but not the effect of detached attachment on MDD (Interaction X × W coefficient = −.032, SE (HC4) = .026, t = −1.236, ns). The interaction of the b path explains 2.3% while the interaction of the direct effect explains .3% of the variance of depressive symptoms. Higher self-critical vulnerability was associated with higher depressive symptoms, B = .063, SE (HC4) = .022, t = 2.946, p < .01 (Table 4). The 95% bootstrap confidence interval for the index of moderated mediation, −.030 (SE = .010), did not include zero (−.053 to −.012) and provides direct evidence of moderation of the indirect effect of detached attachment by self-esteem. As zero is not within the CI, it is indicated a significant moderating effect of self-esteem on the indirect effect, via self-critical vulnerability, on depressive symptoms. The indirect effects are significantly different from each other, as displayed in the significant index of moderated mediation. The conditional indirect effect was strongest in those with low self-esteem (1 SD below the mean; effect = .047, SE = .013, 95% CI = .023; .075) and mean self-esteem (effect = .023, SE = .008, 95% CI = .008; .040) and, not significantly, lowest in those high in self-esteem (1 SD above the mean, effect = .001, SE = .010, 95% CI = −.021; .017). Hence, we have a significant indirect effect for only two values of the moderator, 1 SD below and the mean but not 1 SD above.

Table 4. Unstandardized OLS regression coefficients with confidence intervals estimating detached attachment, self-critical vulnerability, self-esteem, and depressive symptoms.

Note: *p < .05, **p < .01, ***p < .001, ns: non-significant.