Skip to main content

Stressful life events and resilience in individuals with and without a history of eating disorders: a latent class analysis

Abstract

Background

Eating disorders (EDs) are associated with a range of stressful life events, but few have investigated protective factors that may affect these associations. The current study used mixture modelling to describe typologies in life stress exposure and availability of protective resources in individuals with and without eating disorders (EDs).

Methods

A case – control sample (n = 916) completed measures of stressful life events, resilience protective factors, emotion regulation, and symptoms of EDs, depression and anxiety. We conducted latent class analyses to identify subgroups of stress exposure and profile analyses of emotional regulation and resilience. The resulting two latent variables were combined to explore effects on ED status and symptomatology, depression, and anxiety as distal outcome variables.

Results

We identified four classes of stressful life events (generally low, some abuse/bullying, sexual/emotional assaults, and high adversity). For protective resources, we identified six profiles that ranged from low to higher levels of protection with variations in social/family resources. The latent protection variable contributed more strongly to the distal outcomes than the latent stress variable, but did not moderate the latent stress and distal outcome variable relationships. Profiles characterized by lower protective resources included higher proportions of individuals with a lifetime ED, and were associated with higher scores on all symptom measures.

Conclusions

Intra- and interpersonal protective resources were strongly associated with lifetime EDs and current mental health symptom burden after accounting for stressful event exposure, suggesting protective factors may be useful to target in the clinical treatment of patients with ED.

Plain English summary

Previous studies have highlighted risk factors for eating disorders such as experiencing stressful or traumatic events. Protective resources, on the other hand, have received less attention. Factors such as resilience and emotion regulation are associated with eating disorders and could be important protective factors against severe illness in the presence of before mentioned risk factors. This study investigated levels of both potential risk (stressful life events) and protective (resilience and emotion regulation) factors in individuals with and without eating disorders. We found that individuals with low levels of protective resources showed more symptoms related to eating disorders, depression, and anxiety, suggesting that promoting protective factors could be an important avenue for future research, and a potential target for prevention and intervention efforts.

Background

Among the factors found to increase risk of eating disorders EDs; [1, 2], are adverse events such as childhood trauma (e.g., abuse, neglect), sexual assault, war related trauma, and bullying [3,4,5,6,7]. Additionally, exposure to multiple or repeated trauma or adverse experiences is associated with a cumulative risk of negative health effects [8].

However, the individual differences in response to life stressors may be substantial. A concept of interest in this regard is resilience, representing a multi-dimensional construct denoting the ability to sustain relatively normal functioning despite exposures to significant adversity or trauma [9]. Resilient people often show increased flexibility and capacity to cope with life troubles, and have available protective resources that vulnerable individuals more often lack [10, 11]. In addition to resilience, emotion regulation is a multidimensional concept that plays an important role across the ED subtypes, and is associated both with disorder characteristics and prognosis [12,13,14]. Efficient emotion regulation is therefore another protective resource related to the other factors captured by the concept of resilience. Protective resources that promote adaption seem to cluster around three over-arching domains; a) psychological and dispositional attributes, b) family support and cohesion, and c) external support systems [11, 15].

Lower levels of resilience-related factors have been associated with increased risk of depression, anxiety, and other mental health indicators [16]. Good access to internal and external resilience factors may help mitigate risks, such as the effect of stressful life events (SLEs), by increasing the propensity to actively manage stressful incidents or new situations. Individuals with EDs often find it difficult to handle change and novelty, and are often socially isolated, maintain few friends, and in many cases family dynamics have been disrupted [17]. In addition, difficulties with inhibition, impulsivity, adapting to changes, and cognitive flexibility are common across the spectrum of EDs, and might therefore be related to the presence and the ability to utilize resilience factors when faced with psychological stressors or adversity in general [18,19,20].

Thus, examining specific resilience factors among individuals with EDs may aid in understanding contributions to risk and protection in this population, and it is of clinical interest to study factors that may influence the development or course of the illness within a transdiagnostic perspective. In line with this, exploring both ED specific symptoms and more general psychopathological features related to depression and anxiety furthers the transdiagnostic approach and explores differences in clinical presentations not confined solely to ED characteristics.

A few studies have attempted to characterize resilience among individuals with EDs. A recent study by Fergerson and Brausch [21] found that the effect of trauma on ED behaviors in women who had experienced sexual assault was significantly mediated by resilience as a measure of the ability to recover from adversity. In a 1-year longitudinal study of ED patients [22] examining resilience as self-acceptance and personal competence, improvements in quality of life and eating attitudes was seen among those with high resilience. Another recent study by Robert, Shankland [23] observed that resilience may be relevant for the prognosis of EDs, as higher resilience yielded a better chance of recovering from ED [23].

In contrast to the above studies that have used conventional regression analytic methods, few studies have adopted person-centered analytic approaches that explore how risk and protection factors may be related to EDs in disparate ways in different substrates of a heterogeneous sample. The use of latent class analysis LCA: e.g., [24], which is part of the broader family of latent mixture modeling approaches [25], have become increasingly popular for identifying such patterns. The LCA searches for attributes that subgroups of individuals share on selected indicator variables and make them alike. It estimates latent class parameters that describe the probability each individual has of belonging to a specific class (or group), and assign the individual to the class with the highest probability. This process identifies individuals that tend to cluster together, thus maximizing homogeneity within classes and heterogeneity between classes. By reporting the probabilities individuals in specific classes have of endorsing specific indicator values, the nature of the extracted classes may be described [26]. A LCA approach may be highly useful in studies on EDs for identifying and describing different patterns of symptom expressions that may be clinically meaningful within the transdiagnostic model. It also provides further opportunities for exploring if certain risk/protective patterns are more pronounced within certain substrates of individuals with EDs, and what the nature of these might be both within and between samples with ED pathology and healthy controls [27]. Moreover, it enables an analysis of how these patterns relate to disorder characteristics by adding “distal” outcome variables. In addition to ED symptoms and diagnosis, we also explore the relationship between risk/protective patterns and commonly co-morbid and more general features of depression and anxiety. Finally, this method allows for covariates both for conventional adjustment purposes, and, for predicting latent class memberships.

The objective of this study was to use mixture modeling to explore latent clustering in two sets of indicator variables: (1) history of exposure to stressful life events (SLEs), and (2) response patterns in protection/vulnerability data (resilience resources and emotion regulation) in a sample of individuals with and without a lifetime ED. We first examined the nature of the latent profiles related to the SLE and the protective data separately, and then how these two latent domains (SLE and protection/vulnerability) separately and in combination correlate with expressions of EDs, depression, and anxiety symptomatology. Finally, we tested if any latent protection/vulnerability classes moderated the relationship between the SLE classes and psychopathology scores.

Methods

Study setting and design

The current study was a part of the cross-sectional case control study Eating Disorders – Genes and Environment (EDGE), investigating risk and protective factors for EDs. Individuals (above 16 years) with and without a lifetime history of EDs were eligible, and the final study sample represents a convenience sample of cases and controls. All data were collected online between June 2019 and January 2020. The study and the procedures were approved by the Norwegian Regional Committee for Medical and Health Research Ethics (#2017/0606), and was conducted in accordance with ethical guidelines and regulations.

Participants and procedures

A total of 916 individuals (95% female, age M 29.6, SD 10.7 years) participated. Individuals were classified as either cases (n = 495) or controls (n = 395) according to lifetime history of EDs. We were unable to determine ED status for the remaining 26 participants, and these were included in the whole sample analyses but not the direct case – control comparisons. All participants completed an online assessment including the study measures and descriptive information, and provided informed consent electronically using the Norwegian secure login system BankID. Each item had to be completed in order to limit missing data. The complete study materials took between 20 and 60 min to finalize, and participants could enter in to win an iPad if they wished to do so. All data was stored on a platform for sensitive information hosted by the University of Oslo.

Recruitment of both case and control participants was achieved through online social media platforms (Facebook and Twitter), and flyers and posters at Norwegian universities. Posts on websites for ED user organizations and flyers at psychiatric clinics across the country specifically targeted individuals with an ED history. The study was advertised as an investigation of stressful life events and eating disorders. Further details on the recruitment and study procedures have been described previously [28,29,30]. Due to coding error, one participant was excluded from all LCA analyses resulting in a sample of 915 individuals for all LCA/LPA models.

Measures used to estimate latent classes/profiles (LCA/LPA)

Stressful Life Events (SLEs). Exposure to SLEs was recorded with the Stressful Life Events Screening Questionnaire SLESQ; [31] covering 12 events: disease (serious/life threatening), accident (serious/life threatening), assault (e.g., physical attack or robbery), bereavement (loss of a close relative, partner or friend), rape, other sexual assault (unwanted sexual contact/touching), childhood physical abuse (< 18 years of age), adult physical abuse (> 18 years of age), emotional abuse, threats with weapon or by force, witnessing violence (seeing another person being hurt, abused, or dying), or other events (representing a threat to life, health, or safety). We also included one item assessing exposure to bullying during school age (6–18 years). This was based on responses from the Retrospective Bullying Questionnaire RBQ; [32], and we coded bullying as present according to guidelines in the original measure and our previous publication [28, 32]. All measures of SLEs were thus based on retrospective recall of past events.

Resilience Scale for Adults (RSA). The RSA is a self-report measure of resilience resources covering two over-arching domains: intra-personal and inter-personal protective factors. The RSA uses 33 items that are scored on a seven-point semantic differential response format [33], and assesses six protective factors [15, 34]: perception of self, planned future, social competence, and structured style (intrapersonal domain), and family cohesion and social resources (interpersonal domain). Higher scores on the RSA predict less psychiatric symptoms following stressful exposures [35]. Subscale scores are calculated as the average of the subscale item scores. In the present study, subscale scores were transformed to a 0–100 range as the RSA was used in combination with the DERS-SF scale (also transformed to a 0–100 range) for conducting latent class analyses. The subscale “structured style” was not included in these analyses as it has consistently performed less well in terms of construct validity and item score reliability [36].

The difficulties in emotion regulation scale – short form (DERS-SF). The 18-item DERS-SF [37] was used to assess emotion regulation deficits. The items were scored on a Likert scale (1-almost never to 5-almost always) and summed to obtain a total score. The DERS has been translated and validated for use in Norwegian samples [38], and the Cronbach’s alpha was high in the current study (α = 0.93). The DERS represents a transdiagnostic vulnerability factor, as disordered emotional regulation is common across eating disorder diagnoses [39, 40]. Scores were transformed to a 0–100 range, as was done for the RSA, and scores were reversed so that high scores indicated better functioning to match the RSA. DERS in the current study was therefore used as a protection factor, as the absence of emotion regulation difficulties was interpreted as a positive resource.

Measures used for outcomes and covariates

ED100K. The self-report measure ED100K was used to assess lifetime history of the three main EDs anorexia nervosa (AN), bulimia nervosa (BN), and binge-eating disorder (BED) and was used to classify cases and controls [41]. The measure identified lifetime EDs based on presence and severity of symptoms and behaviors according to the DSM-5 diagnostic criteria [42]. Individuals who did not fulfil criteria for an ED at any point in their lifetime were classified as controls. Only criteria for AN, BN, and BED were used and therefore the presence of other EDs were not assessed in the sample. The measure has been previously validated and shown to provide accurate identification of EDs compared with diagnostic interviews with good positive (0.85–1) and negative (0.77–1) predictive validity [41].

Eating Disorder Examination-Questionnaire (EDE-Q). The EDE-Q is a 28-item scale measuring the presence and severity of ED symptoms and behaviors in the past 28 days [43]. A validated Norwegian translation was used [44]. Items are scored on a 7-point scale (from “0 – no days” to “6 – every day”). Scores from the individual items are summed and averaged to obtain a global score. In a Norwegian setting, a global EDE-Q cut-off score of > 2.5 has been found to successfully discriminate between clinical and non-clinical populations [45]. The EDE-Q had a satisfactory Cronbach’s alpha in our sample (α = 0.97).

Generalized anxiety disorder (GAD) scale 7. The 7-item GAD scale [46] was used to assess anxiety symptoms in the last 14 days. Each item was scored on a Likert scale (0 = not at all to 3 = nearly every day), and item scores summed up to achieve a total score. Scores ≥ 10 are considered to be in the clinical range [46]. A validated Norwegian version was used [47], and we obtained a satisfactory Cronbach’s alpha (α = 0.91).

Patient health questionnaire (PHQ-9). The PHQ-9 [48] was used to assess severity of depressive symptoms in the last 14 days. The scale consists of nine items scored on a Likert scale (0 = “not at all” to 3 = “nearly every day”) and summed to achieve a total sore, with scores ≥ 10 considered to be in clinical range. The PHQ-9 has been deemed appropriate for research purposes in Norway [49], and the Norwegian translation has acceptable psychometric properties [30]. The Cronbach’s alpha in our sample was satisfactory (α = 0.91).

Statistical analysis

Descriptive statistics, correlations between study measures, and comparisons of means using Welch t-tests were conducted in R version 4.1.3 [50]. We conducted all latent variable analyses in Mplus 8.7 [51].

Latent class analyses: To identify typical patterns of exposure to the stressful life events (SLEs), a LCA was conducted based on the 13 dichotomously scored SLE variables (12 SLESQ items and one RBQ item). Since these represented discrete risk factors (scored 0-no or 1-yes), a single latent threshold parameter was estimated for each SLE variable for expressing the probability in terms of log-odds of a case belonging to each of the latent classes given their indicator score. Based on these estimates, a posterior probability was estimated for assigning the case to the class with the highest probability [27]. The number of latent classes fit to the data were continuously increased until an optimal class structure could be decided.

Latent profile analyses: The protection/vulnerability indicators (RSA/DERS) were continuous variables (score range 0–100), thus requiring an additional variance parameter that may require constrictions (e.g., equal variance across classes) to avoid convergence problems. The five RSA subscale scores and the DERS total score were rescaled to a common 0–100 range, and the DERS was reversed so that a low versus high score on any scale indicated lower versus better functioning, respectively. For simplicity, we use the term “protective resources” consistently throughout the manuscript when referring to the analysis based on these variables. We extracted an increasing number of profiles until further improvements in model fit abated.

Model fit: Both mixture models were estimated based on the entire sample, thus maximizing heterogeneity and enabling extraction of classes/profiles that are more sensitive to clinical deviations from normality. The log-likelihood function was estimated with the maximum likelihood function using robust errors (MLR). To avoid converging on local maxima, the number of random starts was adjusted upwards to achieve replication of the lowest log likelihood value for the estimated parameters [24]. To decide on the most appropriate model, the Bayesian Information Criterion BIC; [52] and the sample-size adjusted BIC SABIC; [53] were examined with lower values indicating better fit. In addition, the Bootstrapped Likelihood Ratio test BLRT; [54] indicates whether a k-class model yields significantly better fit than the k-1 class (simpler) model according to the p-value. Entropy is reported as a measure of the precision of the latent classifications ranging from 0-low to 1-high. Simulation studies suggest that the BIC and BLRT perform best for deciding the number of latent classes or profiles to extract [55]. In addition, we considered interpretability and differentiation of the latent class/profile solutions.

After deciding the number of latent classes and profiles, both were included as predictors (see Fig. 1) of the distal outcome variables assessing ED diagnosis (yes/no), EDE-Q symptom score, PHQ-depression score, and GAD-generalized anxiety score. Age, gender and education were added as covariates. The final joint model with both latent variables included (SLE and protective resources) together with the covariates for predicting the distal outcome scores, were based on the logit parameters for the within-class separations as devised by Asparouhov and Muthén [56]. This avoids substantial re-estimation of the within/between class parameters conditioned upon these extra variables. To examine the statistical significance of adding the two latent class/profile variables separately as main effects, combined and as an interaction effect, we applied the chi-square difference testing method with robust errors according to Satorra and Bentler [57].

Fig. 1
figure 1

The Conceptual Latent Variable Model. Notes: DERS = Difficulties in Emotion Regulation Scale; EDE-Q = Eating Disorder Examination-Questionnaire; GAD = Generalized Anxiety Scale-7; PHQ = Patient Health Questionnaire-9; RSA = Resilience Scale for Adults

The expected mean score of the four distal outcome variables were examined separately; hence, we reduced the alpha level to < 0.01. We added both latent variables together with a distal outcome variable, first without covariates (crude model) and subsequently with covariates included (adjusted model). The joint mixture model represents a two-way full interaction model since the expected mean outcome values are estimated as free in all 24 latent group combinations (4 SLE × 6 protection classes). To examine the significance of SLE and RSA-DERS as separate fixed effects, we used model constraints. The baseline model had the outcome mean values constrained as equal across all 24 group combinations, thus representing a simple intercept model. To test the fixed effect of the RSA-DERS factor, we allowed the outcome scores to vary between the six RSA-DERS profiles while constraining them equal across the SLE classes, and vice versa when testing the fixed effect of the latent SLE variable. If the log likelihood of this model improved significantly as compared to the log likelihood of the intercept model according to the chi-square difference test [57], it was considered significant. The significance testing of the SLE latent class variable was done comparably. We then examined the combined fixed effect of both latent variables. Since the RSA-DERS variable improved model fit the most, this model functioned as a comparison model to the addition of both latent variables. The model constrictions were similar as above for the RSA-DERS fixed factor but allowed for an additive SLE effect. The interaction model was compared to the model specifying both latent variables as combined fixed effects, but had no constrictions, thus freely estimating all 24 combinations. The final adjusted model included the covariates.

Contingency analysis: Based on the two latent variables, the nominal latent variable membership values were saved and analysed in SPSS 28 [58] with regard to their categorical associations using the cross-tabulation function. In case of an overall significant chi-square test, differences between column frequencies were followed up using standardized adjusted residuals, i.e., N(0,1). This provides a z-test based value that needs to surpass the square root of the critical chi-square value for the degrees of freedom of the test in question, i.e., d.f. = 15 [59]. Due to the large number of comparisons (60 cells), these z-tests were Bonferroni adjusted [59].

Results

Descriptive statistics for the sample are reported in Table 1. In the whole sample, 35% met criteria for a current ED while 19% had a past history of an ED. The remaining 46% were treated as no-ED control cases. In the case group, 36% had a history of AN, 37% of BN/BED, and 27% of both AN and BN/BED. As expected, the ED group scored higher than the control group on all measures of psychopathology and emotion regulation difficulties (Table 1). The overall prevalence of stressful life events (81% vs 65%) and bullying (32% vs 19%) was higher in the ED compared to the control group, as previously reported [28, 29].

Table 1 Description of the Overall Sample (N = 916), and Separate by ED Case–Control Status

Differences in RSA scores for ED cases and controls

Cronbach’s alphas were satisfactory for most RSA subscales, except for “structured style” (Table 2), which had unsatisfactory values in both groups (cases and controls). The RSA global scores were significantly different between cases and controls, with a large effect size (g = − 0.87). For the subscales “perception of self”, “perception of future”, “social competence”, “family cohesion”, and “social resources”, the ED group scored significantly lower than the control group with medium to large effect sizes (g’s from − 0.5 to − 1.04). Table 2 shows the means, differences between groups, and effect sizes for all RSA scores.

Table 2 Mean Score Differences on the Resilience Scale for Adults with Respect to ED Case–Control Status

Correlations between RSA and other measures (EDE-Q, PHQ, GAD, and DERS)

All RSA scores except “structured style” were significantly negatively correlated with the ED symptom measure EDE-Q (range − 0.32 to − 0.60, p’s < 0.05). “Perception of self” and “planned future” showed the strongest correlations with EDE-Q total score. The RSA global score also correlated strongly negatively with the other symptom measures of PHQ depression (r = − 0.73, p < 0.001) and GAD anxiety (r = − 0.64, p < 0.001), as well as with DERS-SF emotion regulation difficulties (r = − 0.73, p < 0.001). Higher resilience protective scores thus implied a lower degree of current pathological symptoms. The RSA “structured style” subscale was weakly related to the symptom measures, which is in line with previous reports.

Latent class analyses of stressful life events

The fit indices were inconsistent regarding the preferred number of latent classes, with the BIC favouring a 3-class solution, the SABIC 4 classes, and the BLRT 5 classes (Table 3). We preferred the 4-class solution as it balanced parsimony with sufficient class differentiation and interpretability. The 4-class solution provided the best SABIC, as well as a BIC close to the 3-class solution. The entropy was also better for the 4-class solution.

Table 3 Model Fit Indices for the Latent Class Analysis of Stressful Life Event Exposures

Regarding class characterization, class #1 was the most prevalent (51%) and represented individuals reporting a low level of exposure to any kind of SLEs. Class #2 was less prevalent (27%) and was characterized by a heightened endorsement (compared to class #1) of SLEs related to childhood physical abuse, emotional abuse, molestation, and bullying, as well as a particularly high endorsement of unspecified events (76%). Class #3 (14%) mimicked class #2 but represented individuals who all (100%) had been exposed to sexual assaults. Class #4 (9%) describes a low prevalence but high adversity class representing individuals with a high probability of being exposed to a broad spectrum of adverse events, thus representing individuals with a considerable accumulation of stress burden. Table 4 shows the probabilities of endorsing the SLE items within each of the classes.

Table 4 Model Estimated Class-Specific Proportions of Stressful Life Event Exposure

Latent profile analyses of protective resources

The LPA models of the protection scores modelled the indicator variances as free across profiles rather than fixed (equal) as they fit consistently better. No further than seven profiles were modelled due to a local maxima that could not be resolved (lack of replication of the log-likelihood estimate). Since the BIC improvement decelerated substantially after the 6th profile, this was the candidate solution. A deceleration in fit improvement was also evident following extraction of 4 profiles; however, as the 6-profile solution contained some qualitative differences with regard to higher family and social resources in combination with lower personal resources that the 4-profile solution missed, the 6-profile solution was preferred due to best fit and a conceptual relevant differentiation (see Table 5 for fit indices).

Table 5 Model Fit indices for the Latent Profile Analysis of the RSA and DERS Indicator Variables

The descriptive nature of the protection profiles are presented in Table 6. Profiles #1 and #2 represented 30% of the sample and included individuals with good access to protective resources, of which the first profile had the highest probability for good adaptation capacity. Profile #3 and #4 constituted 42% of the sample representing individuals with a medium level of resilience resources, in which profile #3 was distinguished from profile #4 by better access to family and social protective resources. Profile #5 and #6 were less prevalent (27%) and were characterized by low availability of intrapersonal protective resources, of which profile #5 had better family and social resources than profile #6.

Table 6 Latent Profile Estimated Mean Scores for the Resilience Protection and DERS Emotional Regulation Scores

Associations between the latent SLE and protection variables

The nominal latent class and nominal latent profile categorizations were saved and subjected to a two-way contingency analysis, which was significant (χ2 df=15 = 113.2, p < 0.001; moderate effect size Cramer’s V = 0.20). See Table 7 for the observed and expected cell observations (‘contingency’). The adjusted standardized z-test residual values indicate if placements in specific SLE classes are significantly more or less frequent than expected. The associative pattern showed that individuals in the less well protected RSA-DERS profile groups #4–6 are significantly more often (positive z-values) exposed to adverse events than individuals in the better protected groups #1–3 (negative z-values). The z-values were also significantly different for individuals in RSA/DERS profile #3 and profile #4, which shows that individuals from a family characterized by less cohesion are more often exposed to adverse events of SLE class #2 (physical/emotional abuse, and unspecified events) and class #4 (broader and generally higher level of adversity) than individuals from a family of high cohesion despite both groups having relatively comparable intrapersonal resources.

Table 7 Categorical Associations (Contingency Tests) Between SLE and RSA-DERS Latent Variables

Significance testing of the latent variables and the predictors

The RSA-DERS latent variable significantly explained the mean scores of all distal outcome, whereas the SLE latent variable contributed significantly in three of the four distal outcome models (see Table 8). The exception was GAD-7 anxiety, in which the latent SLE variable turned non-significant after adding the RSA-DERS latent variable. Since the interaction model did not reach significance in any of the distal outcome models, implying that the latent RSA-DERS factor did not modify the relationship between the latent SLE factor and any of the distal outcome variables, the interaction effect was omitted in the final distal outcome results as presented in Table 9.

Table 8 Significance Tests of the Full Model Parameters Together with Covariates and Distal Outcomes
Table 9 Distal Outcomes Associated with the Latent Class/Profile Memberships of SLE and RSA-DERS Latent Variables

Covariate effects. The covariates were significantly associated with the distal outcome variables in the expected directions (Table 8, lower part). Females reported a higher symptom burden than males, with the caveat that this was a predominantly female sample (95%). Individuals with lower education had significantly more ED symptoms than those with higher education, whereas higher age implied significantly less symptoms of depression.

When examining PHQ depression and GAD anxiety as distal outcomes, the EDE-Q symptom score was included as a covariate in order to provide an adjustment in these analyses due to the case – control nature of the study sample. Having more eating disorder symptoms was positively associated with more depression and anxiety scores, as expected. Since EDE-Q also was correlated with gender, education, and age (higher scores for females, lower education, and younger age), these covariate effects canceled out and was overtaken by EDE-Q.

Final adjusted distal outcome mean scores

The adjusted mean values of the distal outcome variables are given in Table 9.

The SLE latent variable. The final adjusted mean scores for all distal outcome variables (Table 9) showed an increase in the symptom burden when moving from class #1 (low exposure) through to class #4 (broad and high adversity). Post-hoc testing (not part of Table 9) showed a significant difference (p < 0.001) between class #1 and #2, and class #1 and #3 for ED diagnosis; between class #1 and #3, and class #1 and #4 for EDE-Q; between class #1 and #4 for PHQ depression, whereas no significant mean class differences were observed for GAD anxiety. Calculation of standardized mean difference (effect size, M = 0, SD = 1) showed highest SMD between class #1 and #4 with SMD’s equaling 0.49 (EDE-Q), 0.34 (PHQ-9) and 0.19 (GAD-7). These effect sizes were in the medium to low range.

The latent protection variable (RSA/DERS). As for the SLE class variable, moving to a higher RSA-DERS profile number, from #1 to #6, implied an increasing symptom burden. Moreover, the increase in symptom burden followed a stepped curve characteristic with minor differences between profile #1 and #2 (the two best profiles in terms of protection), an increased but roughly comparable symptom burden for profile #3 and #4, and finally, a further increased symptom burden for profile #5 and #6. Confidence intervals (99%) and post-hoc tests of these differences are described in the cells and notes of Table 9, respectively. In addition, some of the differences within steps (as categorized above) were also significantly different, e.g., profile #3 and #4, and #5 and #6 for PHQ depression. The SMD differences between profile #1 and #2 were minor across all distal outcomes (average = 0.13, range 0.10–0.16), but high between the combined profile #1 and #2 and the combined profile #3 and #4 scores (average = 0.65, range 0.54–0.80) and very strong between the combined profile #1 and #2 and the combined #5 and #6 scores (average = 1.55, range 1.40–1.71). The effect sizes related to the latent RSA-DERS factor were thus strong and substantially higher than for the latent SLE factor.

Discussion

The current study examined how stressful life events (SLE) and protective resources (here measured as resilience and emotion regulation abilities), are expressed in a sample of individuals with and without a lifetime history of EDs. We used mixture modelling (latent class analysis) to identify how SLEs and protective resources, as well as their combination, are differently expressed in subgroups of the sample, as well as the associations of these classifications with lifetime ED diagnosis, ED symptoms, and associated symptoms of depression and anxiety. The LCA analyses revealed four classes based on participants’ exposure to SLEs. Although the majority of individuals belonged to classes with low to medium levels of exposure, around one quarter of the sample fell into classes characterized by high adversity or sexual assaults. The latent analysis of protective resources settled on 6 profiles differentiating individuals ranging from high to low levels of protection. Participants mainly differed in terms of quantitative levels of resources, except for some classes that had comparable levels of intrapersonal resources (e.g., personal and social competence) but various interpersonal levels of resources (i.e. family cohesion and social resources). The main finding from the final outcome model was that the latent variables for SLE exposure and protective resources significantly predicted levels of psychopathology and ED case status, with larger effects for protective resources. The relationship between SLEs and the psychopathology outcome data was not moderated by the protective resources classifications.

Investigating protective resources in relation to mental health outcomes provides an important addition to risk factor research. As there are likely both risk and protective factors influencing an individual’s vulnerability to develop psychopathology, the combinations of these factors are important to explore. To our knowledge, this is the first study to explore the nature of resilience factors using a mixture modelling approach in the context of EDs. A previous study investigating resilience among healthy adolescents with the same measure as in the current study mainly supported a four-profile solution that primarily differed in terms of quantity [60]. This is relatively comparable to the results of the present study as we also considered a four-profile solution because the addition of the two extra profiles offered minor improvements in model fit. However, they provided some extra differentiation with regard to family and social resources that we deemed substantial enough to warrant further exploration. Our analysis also included emotion regulation as part of the analysis. Similarly, recent studies conducting LPA analysis of emotion regulation profiles in individuals with or at risk for EDs have found three or four profiles clearly distinguishing emotion regulation and ED characteristics [14, 61]. Despite our study combining emotion regulation with resilience factors in the LPA, our prime finding was comparable to these previous studies by mainly supporting a quantitative differentiation with an increasing symptom burden for profiles characterized by lower protection. We also included individuals both with and without EDs, which might have influenced the class differentiation in our study. The extra differentiation we observed in terms of higher versus lower interpersonal and intrapersonal resilience factors is an interesting finding that calls for further scrutiny about how these factors are associated with ED pathology. In our study, individuals in the two profiles characterized by the lowest levels of protection had symptom levels that were within an ED clinical range regardless of SLE exposure, and within associated clinical range for depression and anxiety irrespective of ED pathology level. This indicates that individuals with low levels of protective resources, hence indicating a lower capacity for adaptation, more commonly have symptoms of mental disorders.

While the main focus of the current study was the associations between resilience factors and EDs, correlations were stronger between the RSA and measures of depression and anxiety symptoms than ED pathology. Consistent with this, the effect of the LPA variable for protective resources was also present in the models for depression and anxiety in addition to ED symptoms. The positive effects of having more protective resources available is thus not specific to ED symptomatology, but extends to a range of pathologies in line with previous studies showing that RSA is associated with both depression and anxiety across different contexts [36, 62]. While we cannot establish directional effects in a cross-sectional study, the results are consistent with previous empirical findings regarding resilience that suggest low protective resources as a vulnerability factor for developing maladaptive habits or cognitions that could translate into mental illness [35].

While the RSA defines resilience as not just an outcome, but a set of protective resources [63], other descriptions have defined it as an ability to «bounce back» after a trauma or stressful experience [64]. These different ways of conceptualizing resilience imply variations in how resilience is measured and interpreted. Despite having a good rationale for measuring resilience as protective resources, the understanding of which factors or resources to measure is far from clear-cut. Resilience is thus closely tied to the instruments that are used to measure it [65]. In relation to this, a systematic review by Windle, Bennett [66] identified no “gold standard” method for measuring resilience, but the RSA, as included in the current study, received high ratings in terms of adequately capturing the breadth of the construct, i.e., covering four intrapersonal domains (e.g., personal and social competence), as well as external inter-personal domains (i.e., family cohesion and social resources). The interpretations in the current study must thus be seen in light of the chosen measure and how it operationalizes the underlying concept.

In our study, we used a data-driven, person-centered approach to investigate the relationship between protective factors and SLEs in individuals with EDs. In previous conventional regression analytic studies, the RSA has been supported as a protective measure by moderating or dampening the negative effects of a stressor on an outcome [35, 67]. Similar findings have been reported in studies using other resilience measures assessing resilience as an outcome rather than a set of protective factors (i.e., the CD-RISC). For example, Yubero, de las Heras [68] reported that resilience moderated the relationship between chronic bullying and current well-being and Wingo, Wrenn [69] found that resilience moderated the association between trauma and depression. Thus, we had reason to expect comparable moderation effects, which we did not observe. The latent profiles of protective resources instead showed strong associations with the distal outcomes in this study; hence, the addition of a significant moderation effect would contribute less. Protective factors, such as personal competence and self-acceptance, have been associated with long-term quality of life in individuals with EDs [22], but are still relatively understudied within the ED literature. Since resilience has been highlighted as a possible important factor in ED recovery [70,71,72], this indicates a complex interplay between risk and protective factors that warrants further exploration in future longitudinal studies on the psychological functioning and outcome of ED patients.

Strengths of this study include the use of an exploratory person-centered approach. We incorporated latent variables of both potential risk and protective factors into the same model to explore relationships to EDs. We were able to extract separate classes/profiles of individuals, and all subsequent analyses using the latent variables were conducted within the LCA framework which does not overstate clustering accuracy by retaining measurement errors inherent in such classifications. This method allowed for individualized patterns of responses to be considered in the analysis. By including highly correlated SLE’s and protective factors in one model, which causes multi-collinearity problems in conventional regression analyses, we were able to shed light on individual differences in potential risk/protection profiles with regard to ED pathology and related symptom burden.

This study has some limitations. First, we did not measure protective resources prior to exposure to potential stressors, which means that resilience or emotion regulation resources may have been influenced by the participants’ current mental state and history of adversity. Longitudinal studies are needed to explicitly test the stability of these protective factors over time. However, given the high test–retest stability of the RSA in previous follow-up studies [15], the strong correlations with stable personality traits [34] and comparable findings in previous studies on stressful events or adversities [35, 67, 73], a similar protective role of these resources in the present study is likely. Second, our sample was predominantly female and we did not record information on race, ethnicity, or immigration status, precluding us from exploring these potential covariates further. Third, the sample sizes for some of the latent class within-group combinations were low and of possible low statistical power, which also prohibited stratified LCA analyses based on ED case or control status. However, having substantial heterogeneity in the sample data may also be considered an important premise [24], and contribute to identify latent classes that differentiate well between clinical and non-clinical cases. This was the case in the present study, showing a strong latent class differentiation for the ED case / non-case status variable (shown in Table 9). Fourth, all data was self-reported and relied on each individual’s memory of past events and accurate reporting of symptoms and descriptives. Finally, as the current study aimed to compare cases and controls with and without EDs, the sample is naturally biased towards specific subgroups and the sample must be viewed as a convenience sample not necessarily representative of the larger population.

Conclusions

In this study, we investigated both potential risk and protective factors within a latent variable model for individuals with eating disorders. Notably, protective factors had a large effect on the pathology measures whereas the contribution of stressful life events were minor. Individuals with low availability of protective resources may be at a higher risk of maintaining maladaptation or psychiatric symptoms following illness or other stressful events causing such problems. Expanding this knowledge could be used to target preventative measures to facilitate resilience and lessen the burden of EDs and other mental health difficulties.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Abbreviations

AN:

Anorexia nervosa

BED:

Binge-eating disorder

BN:

Bulimia nervosa

DERS-SF:

Difficulties in emotion regulation scale – Short form

ED:

Eating disorder

EDE-Q:

Eating disorder examination questionnaire

GAD:

Generalized anxiety disorder scale

LCA:

Latent class analysis

LPA:

Latent profile analysis

PHQ:

Patient health questionnaire

RSA:

Resilience scale for adults

SLE:

Stressful life event

References

  1. Jacobi C, Hayward C, de Zwaan M, Kraemer HC, Agras WS. Coming to terms with risk factors for eating disorders: application of risk terminology and suggestions for a general taxonomy. Psychol Bull. 2004;130(1):19–65.

    PubMed  Google Scholar 

  2. Solmi M, Radua J, Stubbs B, Ricca V, Moretti D, Busatta D, et al. Risk factors for eating disorders: an umbrella review of published meta-analyses. Braz J Psychiatry. 2020;43:314–23.

    PubMed Central  Google Scholar 

  3. Molendijk M, Hoek H, Brewerton T, Elzinga B. Childhood maltreatment and eating disorder pathology: a systematic review and dose-response meta-analysis. Psychol Med. 2017;47(8):1402–16.

    PubMed  Google Scholar 

  4. Wonderlich SA, Crosby RD, Mitchell JE, Thompson KM, Redlin J, Demuth G, et al. Eating disturbance and sexual trauma in childhood and adulthood. Int J Eat Disord. 2001;30(4):401–12.

    PubMed  Google Scholar 

  5. Arditte Hall KA, Bartlett BA, Iverson KM, Mitchell KS. Eating disorder symptoms in female veterans: the role of childhood, adult, and military trauma exposure. Psychol Trauma. 2018;10(3):345–51.

    PubMed  Google Scholar 

  6. Madowitz J, Matheson BE, Liang J. The relationship between eating disorders and sexual trauma. Eating Weight Disorders - Stud Anorexia, Bulimia Obesity. 2015;20(3):281–93.

    Google Scholar 

  7. Lie SØ, Rø Ø, Bang L. Is bullying and teasing associated with eating disorders? a systematic review and meta-analysis. Int J Eat Disord. 2019;52(5):497–514.

    PubMed  Google Scholar 

  8. Hughes K, Bellis MA, Hardcastle KA, Sethi D, Butchart A, Mikton C, et al. The effect of multiple adverse childhood experiences on health: a systematic review and meta-analysis. The Lancet Public Health. 2017;2(8):e356–66.

    PubMed  Google Scholar 

  9. Luthar SS, Cicchetti D, Becker B. The construct of resilience: a critical evaluation and guidelines for future work. Child Dev. 2000;71(3):543–62.

    PubMed  PubMed Central  Google Scholar 

  10. Werner EE. High-risk children in young adulthood: A longitudinal study from birth to 32 years. Am J Orthopsychiatry. 1989;59(1):72–81.

    PubMed  Google Scholar 

  11. Luthar SS. Resilience in development: A synthesis of research across five decades. In: Cicchetti D, Cohen D, editors. Developmental psychopathology, Vol 3: Risk, disorder, and adaptation. 2nd ed. Hoboken, US: John Wiley & Sons Inc; 2015. p. 739–95.

    Google Scholar 

  12. Monell E, Clinton D, Birgegård A. Emotion dysregulation and eating disorders—Associations with diagnostic presentation and key symptoms. Int J Eat Disord. 2018;51(8):921–30.

    PubMed  Google Scholar 

  13. Monell E, Clinton D, Birgegård A. Emotion dysregulation and eating disorder outcome: Prediction, change and contribution of self-image. Psychol Psychother Theory Res Pract. 2022;95(3):639–55.

    Google Scholar 

  14. Nordgren L, Ghaderi A, Ljótsson B, Hesser H. Identifying subgroups of patients with eating disorders based on emotion dysregulation profiles: A factor mixture modeling approach to classification. Psychological Assessment. 2021:No Pagination Specified-No Pagination Specified.

  15. Friborg O, Hjemdal O, Rosenvinge JH, Martinussen M. A new rating scale for adult resilience: what are the central protective resources behind healthy adjustment? Int J Methods Psychiatr Res. 2003;12(2):65–76.

    PubMed  Google Scholar 

  16. Hu T, Zhang D, Wang J. A meta-analysis of the trait resilience and mental health. Personality Individ Differ. 2015;76:18–27.

    Google Scholar 

  17. Treasure J, Sepulveda AR, MacDonald P, Whitaker W, Lopez C, Zabala M, et al. The assessment of the family of people with eating disorders. Eur Eat Disord Rev. 2008;16(4):247–55.

    PubMed  Google Scholar 

  18. Zakzanis KK, Campbell Z, Polsinelli A. Quantitative evidence for distinct cognitive impairment in anorexia nervosa and bulimia nervosa. J Neuropsychol. 2010;4(1):89–106.

    PubMed  Google Scholar 

  19. Stedal K, Broomfield C, Hay P, Touyz S, Scherer R. Neuropsychological functioning in adult anorexia nervosa: A meta-analysis. Neurosci Biobehav Rev. 2021;130:214–26.

    PubMed  Google Scholar 

  20. Aloi M, Rania M, Caroleo M, Bruni A, Palmieri A, Cauteruccio MA, et al. Decision making, central coherence and set-shifting: a comparison between binge eating disorder, anorexia nervosa and healthy controls. BMC Psychiatry. 2015;15(1):6.

    PubMed  PubMed Central  Google Scholar 

  21. Fergerson AK, Brausch AM. Resilience mediates the relationship between PTSD symptoms and disordered eating in college women who have experienced sexual victimization. J Interpers Violence. 2022;37:1–2.

    Google Scholar 

  22. Calvete E, las Hayas C, Gómez del Barrio A. Longitudinal associations between resilience and quality of life in eating disorders. Psychiatry Res. 2018;259:470–5.

    PubMed  Google Scholar 

  23. Robert M, Shankland R, Andreeva VA, Deschasaux-Tanguy M, Kesse-Guyot E, Bellicha A, et al. Resilience is associated with less eating disorder symptoms in the nutrinet-sant&eacute; cohort study. Int J Environ Res Public Health. 2022;19(3):1471.

    PubMed  PubMed Central  Google Scholar 

  24. Nylund-Gibson K, Choi AY. Ten frequently asked questions about latent class analysis. Transl Issues Psychol Sci. 2018;4(4):440–61.

    Google Scholar 

  25. Bauer J. A primer to latent profile and latent class analysis. In: Goller M, Kyndt E, Paloniemi S, Damşa C, editors. Methods for. Researching Professional Learning and Development: Springer, Cham; 2022.

    Google Scholar 

  26. Sterba SK. Understanding linkages among mixture models. Multivar Behav Res. 2013;48:775–815.

    Google Scholar 

  27. Ferguson SL, Moore EW, Hull DM. Finding latent groups in observed data: A primer on latent profile analysis in Mplus for applied researchers. Int J Behav Dev. 2020;44(5):458–68.

    Google Scholar 

  28. Lie SØ, Bulik CM, Andreassen OA, Rø Ø, Bang L. The association between bullying and eating disorders: A case–control study. Int J Eating Disord. 2021;54(8):1405–14.

    Google Scholar 

  29. Lie SØ, Bulik CM, Andreassen OA, Rø Ø, Bang L. Stressful life events among individuals with a history of eating disorders: a case-control comparison. BMC Psychiatry. 2021;21(1):501.

    PubMed  PubMed Central  Google Scholar 

  30. Wisting L, Johnson SU, Bulik CM, Andreassen OA, Rø Ø, Bang L. Psychometric properties of the Norwegian version of the Patient Health Questionnaire-9 (PHQ-9) in a large female sample of adults with and without eating disorders. BMC Psychiatry. 2021;21(1):6.

    PubMed  PubMed Central  Google Scholar 

  31. Goodman LA, Corcoran C, Turner K, Yuan N, Green BL. Assessing traumatic event exposure: general issues and preliminary findings for the stressful life events screening questionnaire. J Trauma Stress. 1998;11(3):521–42.

    PubMed  Google Scholar 

  32. Schäfer M, Korn S, Smith PK, Hunter SC, Mora-Merchán JA, Singer MM, et al. Lonely in the crowd: Recollections of bullying. Br J Dev Psychol. 2004;22(3):379–94.

    Google Scholar 

  33. Friborg O, Martinussen M, Rosenvinge JH. Likert-based vs. semantic differential-based scorings of positive psychological constructs A psychometric comparison of two versions of a scale measuring resilience. Personal Individ Differ. 2006;40(5):873–84.

    Google Scholar 

  34. Friborg O, Barlaug D, Martinussen M, Rosenvinge JH, Hjemdal O. Resilience in relation to personality and intelligence. Int J Methods Psychiatr Res. 2005;14(1):29–42.

    PubMed  Google Scholar 

  35. Hjemdal O, Friborg O, Stiles TC, Rosenvinge JH, Martinussen M. Resilience predicting psychiatric symptoms: a prospective study of protective factors and their role in adjustment to stressful life events. Clin Psychol Psychother. 2006;13(3):194–201.

    Google Scholar 

  36. Morote R, Hjemdal O, Martinez Uribe P, Corveleyn J. Psychometric properties of the Resilience Scale for Adults (RSA) and its relationship with life-stress, anxiety and depression in a Hispanic Latin-American community sample. PLoS ONE. 2017;12(11): e0187954.

    PubMed  PubMed Central  Google Scholar 

  37. Kaufman EA, Xia M, Fosco G, Yaptangco M, Skidmore CR, Crowell SE. The difficulties in emotion regulation scale short form (DERS-SF): validation and replication in adolescent and adult samples. J Psychopathol Behav Assess. 2016;38(3):443–55.

    Google Scholar 

  38. Dundas I, Vøllestad J, Binder P-E, Sivertsen B. The five factor mindfulness questionnaire in Norway. Scand J Psychol. 2013;54(3):250–60.

    PubMed  Google Scholar 

  39. Gross JJ, Muñoz RF. Emotion regulation and mental health. Clin Psychol Sci Pract. 1995;2(2):151–64.

    Google Scholar 

  40. Paulus FW, Ohmann S, Möhler E, Plener P, Popow C. Emotional dysregulation in children and adolescents with psychiatric disorders a narrative review. Front Psychiatry. 2021;12:628252.

    PubMed  PubMed Central  Google Scholar 

  41. Thornton LM, Munn-Chernoff MA, Baker JH, Juréus A, Parker R, Henders AK, et al. The anorexia nervosa genetics initiative (ANGI): Overview and methods. Contemp Clin Trials. 2018;74:61–9.

    PubMed  PubMed Central  Google Scholar 

  42. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 5th ed. Washington DC: American Psychiatric Press; 2013.

    Google Scholar 

  43. Fairburn CG, Beglin SJ. Eating disorder examination questionnaire. Cognitive behavior therapy and eating disorders. New York: Guilford Press; 2008. p. 309–13.

    Google Scholar 

  44. Rø Ø, Reas DL, Lask B. Norms for the eating disorder examination questionnaire among female university students in Norway. Nord J Psychiatry. 2010;64(6):428–32.

    PubMed  Google Scholar 

  45. Rø Ø, Reas DL, Stedal K. Eating disorder examination questionnaire (EDE-Q) in Norwegian adults: discrimination between female controls and eating disorder patients. Eur Eat Disord Rev. 2015;23(5):408–12.

    PubMed  Google Scholar 

  46. Spitzer RL, Kroenke K, Williams JBW, Löwe B. A brief measure for assessing generalized anxiety disorder: The GAD-7. Arch Intern Med. 2006;166(10):1092–7.

    PubMed  Google Scholar 

  47. Johnson SU, Ulvenes PG, Øktedalen T, Hoffart A. Psychometric properties of the general anxiety disorder 7-item (GAD-7) scale in a heterogeneous psychiatric sample. Front Psychol. 2019;10:1713.

    PubMed  PubMed Central  Google Scholar 

  48. Kroenke K, Spitzer RL, Williams JBW. The PHQ-9. J Gen Intern Med. 2001;16(9):606–13.

    PubMed  PubMed Central  Google Scholar 

  49. Langkaas TF, Rognan E, Johnson SU. An introduction to assessment of depression with PHQ-9. Tidsskrift for Norsk psykologforening. 2021;58(3):176–87.

    Google Scholar 

  50. R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2019.

  51. Muthén LK, Muthén BO. Mplus User’s Guide. Eighth Edition. Los Angeles, CA: Muthén & Muthén; 1998–2017.

  52. Schwarz G. Estimating the dimension of a model. Annals Stat. 1978;6(2):461–4.

    Google Scholar 

  53. Sclove SL. Application of model-selection criteria to some problems in multivariate analysis. Psychometrika. 1987;52(3):333–43.

    Google Scholar 

  54. McLachlan GJ, Peel D. Finite mixture models. John Wiley & Sons; 2004.

    Google Scholar 

  55. Nylund KL, Asparouhov T, Muthén BO. Deciding on the number of classes in latent class analysis and growth mixture modeling: a Monte Carlo simulation study. Struct Equ Modeling. 2007;14(4):535–69.

    Google Scholar 

  56. Asparouhov T, Muthén B. Auxiliary variables in mixture modeling: three-step approaches using Mplus. Struct Equ Modeling. 2014;21(3):329–41.

    Google Scholar 

  57. Satorra A, Bentler PM. Ensuring positiveness of the scaled difference chi-square test statistic. Psychometrika. 2010;75(2):243–8.

    PubMed  PubMed Central  Google Scholar 

  58. IBM Corp. IBM SPSS Statistics for Windows, Version 28.0. Armonk, NY: IBM Corp.; 2021.

  59. Sharpe D. Chi-square test is statistically significant: Now what? Pract Assess Res Eval. 2015;20(8):1–10.

    Google Scholar 

  60. Janousch C, Anyan F, Kassis W, Morote R, Hjemdal O, Sidler P, et al. Resilience profiles across context: A latent profile analysis in a German, Greek, and Swiss sample of adolescents. PLoS ONE. 2022;17(1): e0263089.

    PubMed  PubMed Central  Google Scholar 

  61. Eichler J, Schmidt R, Bartl C, Benecke C, Strauss B, Brähler E, et al. Self-regulation profiles reflecting distinct levels of eating disorder and comorbid psychopathology in the adult population: A latent profile analysis. International Journal of Eating Disorders. 2022;n/a(n/a).

  62. Anyan F, Hjemdal O, Bizumic B, Friborg O. Measuring resilience across Australia and Norway. Eur J Psychol Assess. 2020;36(2):280–8.

    Google Scholar 

  63. Jakobsen IS, Madsen LMR, Mau M, Hjemdal O, Friborg O. The relationship between resilience and loneliness elucidated by a Danish version of the resilience scale for adults. BMC Psychol. 2020;8(1):131.

    PubMed  PubMed Central  Google Scholar 

  64. Smith BW, Dalen J, Wiggins K, Tooley E, Christopher P, Bernard J. The brief resilience scale: assessing the ability to bounce back. Int J Behav Med. 2008;15(3):194–200.

    PubMed  Google Scholar 

  65. Denckla CA, Cicchetti D, Kubzansky LD, Seedat S, Teicher MH, Williams DR, et al. Psychological resilience: an update on definitions, a critical appraisal, and research recommendations. Eur J Psychotraumatol. 2020;11(1):1822064.

    PubMed  PubMed Central  Google Scholar 

  66. Windle G, Bennett KM, Noyes J. A methodological review of resilience measurement scales. Health Qual Life Outcomes. 2011;9(1):8.

    PubMed  PubMed Central  Google Scholar 

  67. Friborg O, Sørlie T, Hansen KL. Resilience to discrimination among indigenous Sami and non-Sami populations: The SAMINOR2 study. J Cross Cult Psychol. 2017;48(7):1009–27.

    Google Scholar 

  68. Yubero S, de las Heras M, Navarro R, Larrañaga E. Relations among chronic bullying victimization, subjective well-being and resilience in university students: a preliminary study. Current Psychology. 2021.

  69. Wingo AP, Wrenn G, Pelletier T, Gutman AR, Bradley B, Ressler KJ. Moderating effects of resilience on depression in individuals with a history of childhood abuse or trauma exposure. J Affect Disord. 2010;126(3):411–4.

    PubMed  PubMed Central  Google Scholar 

  70. Las Hayas C, Padierna JA, Muñoz P, Aguirre M, Gómez del Barrio A, Beato-Fernández L, et al. Resilience in eating disorders: A qualitative study. Women Health. 2016;56(5):576–94.

    PubMed  Google Scholar 

  71. Grogan K, O’Daly H, Bramham J, Scriven M, Maher C, Fitzgerald A. A qualitative study on the multi-level process of resilience development for adults recovering from eating disorders. J Eat Disord. 2021;9(1):66.

    PubMed  PubMed Central  Google Scholar 

  72. de Vos JA, LaMarre A, Radstaak M, Bijkerk CA, Bohlmeijer ET, Westerhof GJ. Identifying fundamental criteria for eating disorder recovery: a systematic review and qualitative meta-analysis. J Eat Disord. 2017;5(1):34.

    PubMed  PubMed Central  Google Scholar 

  73. Friborg O. An improved method for counting stressful life events (SLEs) when predicting mental health and wellness. Psychol Health. 2019;34(1):64–83.

    PubMed  Google Scholar 

Download references

Acknowledgements

We thank Dr. Lasse Bang for his contributions to the early stages of the EDGE project. The authors would also like to thank all the participants for their willingness to contribute to this study, and to the Norwegian user organizations ROS and SPISFO for their support and assistance.

Funding

Open access funding provided by University of Oslo (incl Oslo University Hospital) This study is funded by the South-Eastern Norway Regional Health Authority (#2017083).

Author information

Authors and Affiliations

Authors

Contributions

SØL contributed to the conceptualization of the study, data collection, data analysis, interpretation of results, and wrote and revised the original manuscript. LW contributed to the planning of the study, data collection, interpretation of results, and editing of the original draft. KS contributed to the interpretation of the results and review and editing of the original draft. ØR contributed to the conception and planning of the study, interpretation of results, and revising the original draft. OF contributed to the design of the analytic plan, data analysis, interpretation and visualization of results, and writing and revising parts of the original manuscript. All authors read, revised, and approved the final manuscript prior to submission.

Corresponding author

Correspondence to Selma Øverland Lie.

Ethics declarations

Ethics approval and consent to participate

The study and all procedures were approved by the Regional Ethics Committee in Norway, South-East region (REK Sør-Øst, project id# 2017/1606). All participants signed informed consent to participate. All procedures were performed in accordance with ethical guidelines and regulations.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and Permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Lie, S.Ø., Wisting, L., Stedal, K. et al. Stressful life events and resilience in individuals with and without a history of eating disorders: a latent class analysis. J Eat Disord 11, 184 (2023). https://doi.org/10.1186/s40337-023-00907-8

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s40337-023-00907-8

Keywords