Introduction

Self-reported low self-esteem has consistently been observed in episodes of major depressive disorder(MDD; [1]) and in episodes of anxiety disorders (AD; [2]) like social anxiety disorder, panic disorder, and generalized anxiety disorder. It is unsurprising that low self-esteem is prevalent in MDD given that it is one of the possible symptoms as outlined in the most recent version of the Diagnostic and Statistical Manual of Mental Disorders (“feelings of worthlessness”; [3]). In a meta-analysis [4], low self-esteem was shown to predict prospective symptoms of depression and anxiety, highlighting a potential causal role in MDD and AD etiology. The authors argued that low self-esteem might lead to social disruptions, increased self-focus, and rumination, which would in turn lead to the development (and maintenance) of depressive symptoms. For symptoms of anxiety, high self-esteem has been argued to act as an anxiety buffer [5], and therefore low levels of self-esteem would make an individual more vulnerable to anxious thoughts and feelings. Indeed, those who had been manipulated with a self-esteem boost reported less anxiety following a traumatic video than those who had received a neutral self-esteem manipulation [5]. As such, the presence and relevance of low self-esteem in both AD and MDD is fairly indisputable. In the present article, we use the term “episode” also for the duration that a person meets the criteria for an AD without recovery.

Self-esteem has been predominantly indexed by self-report measures, which are limited by what the respondent is willing and able to disclose (e.g., [6,7]). Meanwhile, relevant self-associations may not necessarily be accessible for conscious introspection, and individuals may dismiss certain initial associations as irrelevant when asked to verbalize a global affective self-evaluation. This points to the importance of complementing self-report measures reflecting the “self-endorsed”, deliberate self-evaluations (explicit self-esteem; [8]) with measures of automatic associations that require neither verbalization nor introspection (implicit self-esteem; cf. [9]). The relevance of differentiating between explicit (ESE) and implicit self-esteem (ISE) is further emphasized by the view that both facets of self-esteem are differentially involved in more controlled/strategic versus more automatic/spontaneous behaviors. ESE is considered to be especially relevant in the context of more deliberative/reflective behavior (e.g., choosing not to speak up in a group discussion), while ISE is argued to be critically involved in more reflexive, spontaneous behaviors (e.g., blushing during public speaking; [10]). In support of this, Spalding and Hardin [11] found that those low in ISE, as indexed with an affective priming task, displayed more observer-rated anxious behaviors during an interview, specifically when the interview involved self-related questions. A more recent study found that ESE was related to self-reported measures of anxiety/nervousness, and controlled nonverbal behaviors during a speech (e.g., hand gestures to emphasize the point verbally made), while ISE, as indexed by both a cognitive load task and the implicit association test (IAT), was related to general observer-rated anxiety and spontaneous nonverbal behaviors during a speech (e.g., nervous mouth movements; [10]). Given the distinct roles of ISE and ESE in automatic and controlled dysfunctional (pathogenic) behaviors, it is feasible that these two components of self-esteem are (partly) differentially involved in AD and MDD. As measures of ISE are thought to capture different aspects of the construct [12], it is important to note which ISE measure has been used previously as this may explain inconsistencies in findings.

Empirical evidence for the presence of low ISE in AD and MDD is mixed. Studies looking at ISE in AD samples are relatively few and seem to have focused almost exclusively on social anxiety disorder (symptoms). Two analogue studies using an IAT as a measure of ISE found relatively low ISE in females scoring high on symptoms of social anxiety [13,14]. Consistent with this, a correlational study focusing on adolescents showed a negative relationship between symptoms of social anxiety and ISE measured with an IAT [15]. A more recent small-scale clinical study found that patients with social anxiety disorder (n = 45) had lower ISE, as indexed by an IAT, than healthy controls [16]. Patients with panic disorder (n = 24) showed a similar tendency that was not significant, possibly due to low statistical power. Concerning MDD, a meta-analysis of 25 studies (n = 2831, of which 77% non-clinical) indicated that there is an association between low ISE and symptoms of depression [17]. However, when focusing only on studies using clinical, adult depressed samples, many have failed to find lower ISE in comparison to those who have never been depressed [18–21]. Risch et al. [22] did, however, find lower ISE in current MDD patients (both first-onset and recurrent) compared to healthy controls, when using an IAT. Franck, De Raedt, Dereu and van den Abbeele [23] also found evidence for lowered ISE in current MDD patients using an IAT but only in those without suicidal ideation, while MDD patients with suicidal ideation did not differ from never-depressed controls. As such, despite the strong theoretical grounds for anticipating low ISE in MDD and AD, the evidence thus far, particularly for MDD, is not very convincing, and suggest that dysfunctional self-related thoughts and behaviors may be most pronounced at the deliberate, explicit level.

Previous studies involving clinical samples may simply not have had enough power to detect an effect of ISE. Further, as little is known about ISE in comorbid MDD and AD, and many studies looking at MDD do not have anxiety disorder diagnosis as an exclusion criterion (e.g., [21,22]), it is important to look at ISE not only in comorbidity, but in (relatively) pure forms of MDD and AD as well. This may elucidate the apparently inconsistent earlier findings regarding ISE in MDD and AD. Therefore, the present study was designed in the context of a large-scale, national study: the Netherlands Study of Depression and Anxiety (NESDA; www.nesda.nl/). Specifically, ISE (and ESE) in (relatively) pure forms of MDD and AD, as well as in individuals with comorbid MDD/AD were analyzed. Given previous reports of relatively high reliability [12] with strong evidence of validity [10], the IAT was used to measure ISE. In addition, we took suicidal ideation into account to test the robustness of the earlier finding that only in the absence of suicidal ideation is MDD associated with lowered ISE [23].

Given that low self-esteem is prevalent in both depression and anxiety, it seems feasible that self-esteem is a transdiagnostic factor which may explain the high rates of comorbidity between the two [24,25]. Indeed, low self-esteem may increase vulnerability for both depression and anxiety, and other factors may then determine, specifically, which of the two develops. Consistent with this view, a relatively large scale study among non-referred adolescents found that the association between symptoms of depression and social anxiety could largely be explained by adolescents’ explicit self-esteem [15]. This earlier study provided no evidence for a unique relationship between low self-esteem and the covariation of (social) anxiety and depressive symptoms, but rather independent relationships between low self-esteem and the severity of depression and anxiety symptoms. This suggests that low self-esteem may increase the chance of concurrent symptoms of anxiety and depression by independently increasing the chance of developing both types of symptoms. Given that this earlier study was limited to a non-clinical sample of adolescents, it remains to be seen whether self-esteem can (partly) explain the high rates of comorbidity in an adult sample containing many who meet the clinical criteria for a depression and/or anxiety disorder. Therefore, the relationship between depressive and anxious symptoms are analyzed in the present study, once controlling for self-esteem.

It is important to see to what extent low self-esteem will be normalized when remitted/recovered from a depressive episode or anxiety disorder. Given the high rate of recurrence that is typical of both MDD and AD, identifying possible “scars” that remain following episodes (e.g., lowered ISE) may be crucial in identifying those who will relapse [26]. Even when AD and MDD enters remission, or recovery, it is important to differentiate between ISE and ESE. It is feasible that those who are in remission/recovery are able to address negative self-related thoughts, and are motivated to reappraise a situation in order to derive at a more positive self-evaluation. Indeed, self-awareness and addressing negative thoughts effectively is the core aim of most therapies (e.g., cognitive behavioral therapy). However, given the lack of control over ISE, spontaneous (pathogenic) behaviors may continue that are not necessarily within the realm of awareness. Consistent with this, Vasey, Harbaugh, Buffington, Jones and Fazio [27] found that implicit associations, as measured with an IAT, following an exposure therapy predicted return of fear in public-speaking phobia. Despite this, two studies reported that ISE and ESE in remitted MDD did not differ from never-depressed controls, both of which had used the IAT [18,22]. Given the small sample sizes used, it is important to test the hypothesis in larger samples, and also take AD and comorbid MDD and AD into consideration as distinct disorders with distinct etiologies. As ISE may improve over time through consistent high ESE that becomes overlearned and the default reaction [28], it is also important to differentiate between those whose symptoms have recently remitted and those who no longer meet the criteria for MDD and AD for some time (recovered). The present study therefore included remitted and recovered MDD, AD and comorbid MDD/AD groups.

For some individuals, ESE and ISE may differ considerably (i.e., discrepant self-esteem; [29]). While fragile self-esteem refers to the pattern of low ISE and high ESE, and has been linked to narcissistic tendencies [30], damaged self-esteem refers to the pattern of high ISE and low ESE, and has been linked to symptoms of depression [31]. While some studies looked at discrepancy by including the interaction between ISE and ESE into the model [32], others argued that this fails to acknowledge the potential influence of the direction of the discrepancy [33]. That is, the extent that ISE and ESE differ may only be related to symptoms of depression or anxiety when, for example, ISE is higher than ESE. Without distinguishing the direction of discrepancy, the interaction may appear statistically non-significant. Other studies looking at self-esteem discrepancies adopted analyses that allowed for differentiating between the direction of the discrepancy (i.e., fragile or damaged), but did not allow for the inclusion of the main effects (i.e., ESE and ISE; [34]). The inclusion of ESE into the model resulted in an issue of multicollinearity. Given strong relationships between ESE and psychopathology are often reported, previously observed associations between damaged self-esteem and depression may have simply been an artefact of ESE, regardless of ISE. The final aim of the present study is therefore to explore an alternative method to analyse the role of discrepant self-esteem in MDD and AD which allows for differentiating between fragile and damaged self-esteem while statistically controlling for the potential main effect of ESE.

All in all, the present study tested the following hypotheses: i) those without a lifetime diagnosis of depression or anxiety (i.e., the comparison group) will show higher ESE and ISE than current MDD, AD, and comorbid MDD/AD; ii) ISE and ESE in the comparison group will also be higher than those who were recovered or remitted from AD, MDD, or comorbid MDD/AD; iii) those who have recovered or remitted from AD, MDD or comorbid both, will have higher ISE and ESE than those who currently meet the diagnosis; iv) self-esteem is a transdiagnostic factor, and as such, levels between MDD, AD and comorbid MDD/AD will not differ at the remitted level, the recovered level and at the current diagnosis level. Further, a novel way of testing the presence of self-esteem discrepancies in those with a current depression or current anxiety disorder is explored that allows for the inclusion of ESE as a main effect in the model.

Materials and Methods

Participants

The Netherlands Study of Depression and Anxiety (NESDA; www.nesda.nl/) is an ongoing longitudinal cohort study that, at baseline (2004–2007), involved 2981 participants who have been followed-up biannually across a number of measures. In order to follow the long-term course of depression and anxiety, 1701 participants with a current depressive or anxiety disorder, and 907 participants with a life-time diagnosis or at-risk (e.g., subthreshold symptoms), were recruited from the community, primary care, and mental health organisations. A further 373 participants with no current or history of any depressive disorder or AD were recruited as controls. There were two exclusion criteria in the NESDA: 1) A primary, clinically overt diagnosis of a psychotic disorder, an obsessive-compulsive disorder, a bipolar disorder, or a severe addiction; 2) Non-fluent command of the Dutch language. A thorough overview of the recruitment process, design, and overarching aims of NESDA are published elsewhere [35]. All participants provided written consent, and ethical approval was granted by all ethical committees of participating universities (VU University Medical Center, Leiden University Medical Center and University Medical Center Groningen).

The present study makes use of data collected in the most recent wave at the time of writing, which is approximately 6 years since baseline, and the fourth biannual measurement. At this wave, 2256 participants were measured (24% attrition since baseline), where 1799 (80%) received the measures relevant for the present analysis. The remaining 457 did not receive all the measures for various technical reasons (e.g., completing measures at home or over the phone prohibiting computer-based measures). Participants who were given all relevant measures were aged between 23 and 72 (M = 48.05, SD = 13.18; 63.6% female). Participants were excluded from the present analysis if they had developed a bipolar disorder at some point during the study, or reported an alcohol dependence since the last interview (approximately 2 years ago; n = 83).

Clinical groups were formed for MDD and AD, and split by those currently in an episode (diagnosis in past month), those in remission (an episode that had ended in the last six—one month), and those recovered (an episode in the last seven years–six months). We used these cut-offs as these were more readily available within the study. It should be noted that what defines, for example, a depression in remission varies across studies. Frank et al. [36] recommends that remission be considered as a depression-free period of 2–6 months, with longer than 6 months considered a recovery. Our cut-offs are not too far from this. Cut-offs for ADs are dependent on the type of AD, however we apply the same cut-offs as used for MDD for consistency when comparing the groups and creating comorbid groups. To create a more homogeneous group, participants were excluded from the current and remitted AD groups if they had also met the criteria for MDD (or dysthymia; n = 135 & 27, respectively) since the last interview. Likewise, those with an AD since the last interview were excluded from the current and remitted MDD groups (n = 78 & 45, respectively). Those in the recovered AD or MDD groups had no history of MDD (and dysthymia) or AD, respectively. Current and remitted comorbid AD and MDD (CM) groups were also formed based on the same criteria as the MDD and AD groups. Those who had comorbid dysthymia in the current CM and MDD groups were not excluded. A recovered CM group was not created given that the available information made it difficult to determine whether AD and MDD had occurred and ended at the same time. The comparison (control) group consisted of individuals without a history of AD, MDD or dysthymia. The upper half of Table 1 provides an overview of the demographics and size of each group.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Means (& standard deviations; unless stated otherwise) of demographics and variables per group.

https://doi.org/10.1371/journal.pone.0166116.t001

Results

Discussion

The key findings of the present study were: i) Implicit self-esteem was lower in current comorbid individuals compared to a healthy comparison group; ii) Explicit self-esteem was lower in all current, remitted, and recovered clinical groups than the healthy comparison group; iii) Explicit self-esteem in current MDD and AD was not lower than in remitted MDD and AD, respectively. However, self-esteem was lower than in recovered MDD and AD. iv) There was no support for discrepant self-esteem in current MDD and current AD, once statistically controlling for explicit self-esteem.

Previous studies have consistently found lowered ESE in individuals with MDD and AD [1,2]. It is not particularly surprising that current MDD had lower ESE in both previous studies and the present study given that low ESE is a possible symptom of MDD. Despite this, ESE was not lower in MDD than in AD, where low self-esteem is not a symptom per se. This may highlight low self-esteem as a transdiagnostic factor explaining why comorbidity between depression and anxiety is so high [24]. Some argue that depression and self-esteem have overlapping causal factors, which may explain some of the shared variance. For example, Neiss et al. [51] concluded that there was a common genetic and environmental influence on self-esteem, negative emotionality and depression, based on stronger associations in monozygotic twins, compared to dizygotic twins. In the present study, although self-esteem in current MDD and current AD was similarly low, the correlation between symptoms of depression and anxiety remained high when partialling out ESE. Although this is consistent with the findings from a previous cross-sectional study [15], it need not mean that self-esteem does not increase the risk of developing a comorbidity. A longitudinal design is required to see which aspects of a current depression or anxiety, like low self-esteem, is related to increased risk for comorbid anxiety and depression, respectively. Highlighting which symptoms to target first to prevent comorbid disorders from developing may be particularly important given that comorbid AD/MDD is more persistent and treatment-resistant than purer forms of MDD and AD [52].

Despite the indisputable presence of low ESE in AD and MDD, how it relates to anxiety and depression remains an important question. Low self-esteem has also been argued to have a causal role in the development of AD and MDD, as lower levels are often observed to precede symptomatology [4]. Through social disruptions, increased self-focus, rumination, and lack of ability to negate or minimize the effects of anxious or threatening thoughts and feelings, low self-esteem could make a person more vulnerable to the development of AD and MDD symptomatology [4,5]. This may explain why those with social anxiety disorder had lower ESE than both panic disorder and agoraphobia, given both the social-related aspects and feelings of anxiety that may arise in low self-esteem. Interventions that focus specifically on increasing self-esteem might prevent both MDD and AD from developing. For example, in competitive memory training, positive self-images are made more salient and therefore increases the likeliness that the positive self-image is activated when the self is brought to attention. This intervention was found to improve self-esteem and depressive symptomatology in addition to treatment as usual in patients with MDD [53], but it remains to be seen whether it can also be used in prevention. Given the nature of self-report measures (e.g., the awareness that answers will be seen by someone else), negative self-evaluations are purposefully self-endorsed in MDD and AD, self-esteem interventions should not only aim to change the content of self-related evaluations but also aim to learn how to oppose negative self-evaluations that arise.

ISE in relatively pure MDD and AD was found not to differ from controls. In other words, self-related negative thoughts occur when purposefully thought about, whereas negative self-associations do not already arise at the automatic and reflexive level. For MDD, this does not appear to be explained by the presence (or absence) of suicidal ideation as was found in a previous study [23]. The lack of low ISE in MDD is consistent with many findings (e.g., [20]) but conflicts with Risch and colleagues [22]. Given that the same measure of ISE was used, explaining the conflicted findings is not straightforward. As we found lower ISE in the current CM group, the presence of AD in the MDD group may account for the lower ISE observed in their study. In the previous study where low ISE was observed in individuals with social anxiety disorder (n = 33), 40% also reported a comorbid depressive disorder [16]. Although in this earlier study a comparison of those with and without a depressive disorder suggested no difference in ISE, the small sample sizes may have limited the power to detect an effect. Previously reported low ISE in AD as well as MDD might be explained by the unaccounted presence of comorbidity. Then together with the numerous previous cross-sectional studies using clinical populations where no difference in ISE was found [20], one would be tempted to conclude that low ISE is not a feature of pure AD and MDD. However, if ISE is formed following prolonged periods of ESE that becomes overlearned, as argued by many (e.g., [8]; although see [54]), the low ESE prevalent in MDD and AD should eventually manifest in lower ISE. Another assumption of this dual processing model, is that the level of ESE is stable, before it becomes overlearned and automatic [28]. However, highly unstable ESE has been argued to be a larger vulnerability factor for MDD and AD than consistently low ESE [55,56]. As such, if ESE is particularly unstable in MDD and AD, then this may also explain why the findings involving ISE are relatively inconsistent.

The role of prolonged low ESE in the development of low ISE may explain why low ISE was observed in comorbid AD and MDD. There is some indication that the duration of symptomatology in CM would be longer given that remission rates for CM between baseline and two-year follow-up in NESDA were a lot lower, and more months with symptomatology were reported, than those with either MDD or AD alone. This is despite those with CM were more likely to seek treatment.[52]. Prolonged periods of MDD and AD might give rise to longer periods of (stable) low ESE, and this in turn might explain why lower ISE was observed in CM only. Future studies looking at the duration, severity, and stability of low self-esteem, should look at whether the duration of symptoms in current clinical groups is associated with low ISE. Alternatively, more frequent measures of ESE over a period of time might be a more precise way of looking at the role of ESE in ISE, and could also be used to look at the role of self-esteem stability in clinical disorders [57]. Untargeted negative self-evaluations at the implicit level may also explain why those with CM are more likely to show a poorer response to treatment (e.g., [58]). Concerning self-esteem scars in those with a previous MDD and/or AD, ISE was not lower in remitted and recovered AD/MDD in comparison to controls. However, this is perhaps unsurprising given that current AD and MDD also did not display lower ISE. ESE in participants with a previous MDD or AD was significantly lower than those who had never had a MDD or AD. This might partially explain the highly recurrent nature of both disorders given that low ESE is argued to increase the vulnerability for symptomatology [4], and may suggest the need for a self-esteem intervention in those who have recently recovered or remitted from AD or MDD. Interventions for remitted MDD, in general, have been shown to be effective in reducing relapse risk [59].

Damaged self-esteem increased, while fragile self-esteem decreased, both the likelihood of current MDD and current AD, compared to controls. This is consistent with previous findings [31], although we extend this by highlighting that discrepant self-esteem variables were no longer significant once controlling for the main effect of ESE. Previous methodology either did not allow for the inclusion of ESE in the model [31,33], or did not allow for the specification of the direction of discrepancy (i.e., using the interaction between ISE and ESE to represent discrepancy). Although the current methodology allows for the inclusion of one main effect, it does not allow the direction of the discrepancy to be taken into account (i.e., damaged vs. fragile) regardless of the extent of discrepancy which has been argued to be an important factor in discrepant self-esteem [33]. However, the relevance of such a variable is debatable. Even if one were to select only those with discrepant self-esteem (i.e., excluding those with fairly congruent self-esteem), it is theoretically still anticipated that a relatively large discrepancy in one specific direction would have a stronger association with symptomatology compared to a smaller discrepancy within the same direction. Therefore, the method used in the present study not only allows for the inclusion of one main effect, but also eradicates variables that are difficult to interpret.

Although we feel that the method used in the current study is an improvement, there are still a number of limitations. In using the IAT and RSES to derive at measures of discrepancy there is an assumption that the two measures only differ on the construct they tap into. However, the RSES refers to explicit self-esteem in the past seven days, while in the IAT self-esteem is measured at that specific moment. Further, although ISE is anticipated to be a trait rather than a state, it is also argued to be context-dependent. As such, the lab settings may have an influence. For ESE, participants are able to reflect to more natural, daily situations. As such, not just a difference in timing, but also a difference in context might be measured. Ideally, a measure would be developed that directly taps into self-esteem discrepancy. We do not reject the notion of discrepant self-esteem, necessarily, but current methods (present study including) of combining ISE and ESE measures may introduce too much noise for quantifying real discrepancies.

Conclusions

The present study was the first to compare levels of both ESE and ISE across various phases of MDD and AD groups and with careful consideration of comorbidity. We found that consistent across all clinical groups, at all phases, a more negative, self-evaluation was endorsed in comparison to the control group. In other words, conscious behaviour consistent with low ESE is more likely to manifest in MDD and AD, and remain during remittance and recovery. However, negative self-evaluations at the implicit level were only evident in individuals with current CM, while levels of ISE in relatively pure forms of MDD and AD were equivalent to the controls. This may not only highlight why CM is more persistent and treatment-resistant, but also emphasizes the need for future research to investigate whether the aetiology of CM differs from MDD or AD. As such, it is also important to control for the potential presence of a comorbid disorder when further examining the relevance of ISE (and ESE) in AD or MDD.

Acknowledgments

We would like to thank Marieke Timmerman for her support in the planning of this paper.