Examining the placement of atypical anorexia nervosa in the eating disorder diagnostic hierarchy relative to bulimia nervosa and binge-eating disorder
Abstract
Objective
Some individuals meet the criteria for atypical anorexia nervosa and another eating disorder simultaneously. The current study evaluated whether allowing a diagnosis of atypical anorexia nervosa to supersede a diagnosis of bulimia nervosa (BN) or binge-eating disorder (BED) provided additional information on psychological functioning.
Methods
Archival data from 650 university students (87.7% female, 69.4% white) who met Eating Disorder Diagnostic Survey for DSM-5 eating disorder criteria and completed questionnaires assessing quality of life, eating disorder-related impairment, and/or eating pathology at a single time point. Separate regression models used diagnostic category to predict quality of life and impairment. Two diagnostic schemes were used: the DSM-5 diagnostic scheme and an alternative scheme where atypical anorexia nervosa superseded all diagnoses except anorexia nervosa. Model fit was compared using the Davidson–Mackinnon J test. Analyses were pre-registered (https://osf.io/2ejcd).
Results
Allowing an atypical anorexia nervosa diagnosis to supersede a BN or BED diagnosis provided better fit to the data for eating disorder-related impairment (p = .02; n = 271), but not physical, psychological, or social quality of life (p's ≥ .33; n = 306). Allowing an atypical anorexia nervosa diagnosis to supersede a BN or BED diagnosis provided a better fit in cross-sectional models predicting purging (p = .02; n = 638), but not body dissatisfaction, binge eating, restricting, or excessive exercise (p's ≥ .08; n's = 633–647).
Discussion
The current data support retaining the DSM-5 diagnostic scheme. More longitudinal work is needed to understand the predictive validity of the atypical anorexia nervosa diagnosis.
Public Significance
The current study examined how changes to the diagnostic categories for eating disorders may change how diagnoses are associated with quality of life and impairment. Overall, findings suggest that the diagnostic hierarchy should be maintained.
1 INTRODUCTION
Atypical anorexia nervosa (atypical AN) is an other specified feeding or eating disorder in which the cognitive and behavioral criteria for anorexia nervosa (AN) are met and the individual has lost significant weight; however, the individual is not underweight (e.g., body mass index [BMI] ≥ 18.5 kg/m2; American Psychiatric Association, 2013, 2022). In the current Diagnostic and Statistical Manual, 5th Edition (DSM-5) nosological scheme, an individual receives an atypical AN diagnosis only if they do not also meet the full criteria for bulimia nervosa (BN) or binge-eating disorder (BED). At the present time, it is unclear whether having a diagnosis of BN or BED supersede a diagnosis of atypical AN is the most appropriate diagnostic hierarchy (Thomas & Gydus, 2023; Walsh et al., 2023), given overlaps in clinical presentation. An individual with BN, by definition, has the undue influence of shape or weight on self-evaluation. If this individual is also weight suppressed (i.e., current weight is below their highest lifetime weight) and has a fear of gaining weight, they meet all other criteria for atypical AN. Individuals with BN are often weight suppressed (Gorrell et al., 2019; Keel et al., 2019) and fear of gaining weight is the most common eating disorder fear in this population (Brown & Levinson, 2022). As a consequence, approximately 50% of those with BN also meet the criteria for atypical AN (Forney et al., 2017). Given the psychological and behavioral correlates of weight suppression (Gorrell et al., 2019; Lowe et al., 2018), it is unknown if characterizing an individual as having atypical AN or BN has more utility (Walsh et al., 2023). Similarly, significant body image concerns and the overevaluation of shape/weight, a feature of AN, atypical AN, and BN, are associated with greater eating pathology within BED (Grilo, 2013). However, it is also unknown if those with BED with significant body image concerns and weight suppression might be better characterized as atypical AN. Consistent with calls to empirically inform nosological decision-making (Walsh et al., 2023), the current study seeks to evaluate whether prioritizing an atypical AN diagnosis over a BN or BED diagnosis provides additional clinical information.
Weight suppression is correlated with eating disorder severity and/or illness duration within eating disorder populations (Bodell & Keel, 2015; Gorrell et al., 2019; Hagan et al., 2017; Keel et al., 2017; Lowe et al., 2018). Greater weight suppression prospectively predicts a longer illness course and worse treatment outcomes within BN (Keel & Heatherton, 2010; Lowe et al., 2011), although these findings are not always consistent (Dawkins et al., 2013; Jenkins et al., 2018; Zunker et al., 2011). As such, it remains unclear if communicating weight suppression through an atypical AN diagnosis may provide additional clinical information not explicitly represented by the BN and BED diagnoses.
Conversely, should the atypical AN diagnosis supersede the BN and BED diagnosis, the clinical picture of atypical AN will exhibit more heterogeneous eating disorder behaviors. A high proportion of the resulting clinical population would likely exhibit binge eating. Historically, distinguishing between AN “restricting” and “binge-purge” subtypes has had clinical utility (Peat et al., 2009), and the current DMS-5 criteria, in effect, distinguishes between atypical AN “restricting” and atypical AN “binge-purge” through the BN diagnosis. This information would potentially be lost if atypical AN were made more heterogeneous by including individuals with and without recurrent binge eating into a single atypical AN category.
The current study seeks to empirically evaluate the utility of the current diagnostic scheme (DSM-5 scheme) compared to atypical AN superseding diagnoses of BN and BED (atypical AN scheme). Specifically, we compared the utility of the DSM-5 scheme to an alternative atypical AN scheme in explaining variance in physical quality of life, psychological quality of life, social quality of life, and eating disorder-related impairment. We focused on quality of life and impairment given their transdiagnostic associations with eating disorder severity (Jenkins et al., 2011; Mitchison et al., 2012) and because associations with eating disorder symptoms, such as binge eating, are inherently confounded with diagnostic criteria. The limited research on atypical AN has conflicting findings; atypical AN is associated with more severe eating pathology than AN in some domains (Walsh et al., 2023), and weight suppression is often associated with more severe eating pathology (Gorrell et al., 2019). However, in an unselected sample, atypical AN was associated with less severe eating pathology than other DSM-5 eating disorders (Forney et al., 2017). As a result of the conflicting literature, we did not have directional a priori hypotheses in our university sample. Instead, we a priori registered how we would interpret findings, taking a conservative approach (osf.io/2ejcd). If the atypical AN scheme explains more variance than the DSM-5 scheme for at least three of the four outcomes (i.e., greater than 50% of outcomes), then this would support that atypical AN superseding a BN or BED diagnosis provides more clinical information than the existing DSM-5 scheme. Despite potential confounds with diagnostic criteria, we also explored the utility of the DSM-5 scheme relative to the atypical AN scheme by examining eating pathology outcomes, including body dissatisfaction, binge eating, restricting, purging, and excessive exercise.
2 METHODS
2.1 Participants
Data in the present study were pooled from five archival data sets from a university in the Appalachian region of the United States (Forney, Burton Murray, et al., 2023; Forney, Horvath, et al., 2023; Harris, 2023), a university in the southern United States (Forney, Burton Murray, et al., 2023), and a university in Ontario, Canada (Biderman et al., 2023). The final pooled sample comprised 650 participants who met the criteria for an eating disorder (see “Eating Disorder Diagnostic Scale” under Measures) and passed sufficient effort responding criteria (Curran, 2016; Huang et al., 2015) or other specified attention checks. Analyses were pre-registered at https://osf.io/2ejcd.
All participants were over the age of 17 (Mage = 20.11, SD = 4.77; range: 17–54). Participants primarily identified as white (N = 451, 69.4%), with 127 (19.5%) identifying as Asian or Asian American, 22 (3.4%) identifying as Black or African American, 21 (3.2%) identifying as Hispanic or Latino, 5 (0.8%) identifying as an Indigenous Person of Canada, 3 (0.4%) identifying as American Indian or Alaska Native, and 1 (0.2%) identifying as Native Hawaiian or Other Pacific Islander. Fifty-three (8.2%) participants chose “other” or “not listed” and provided a description of their racial/ethnic identity. The percentages for racial and ethnic identity sum to more than 100 because participants were instructed to “select all that apply” for racial and ethnic identities. The sample primarily identified as female (N = 569; 87.7%), with smaller proportions identifying as male (N = 73; 11.2%), transgender (N = 5; 0.8%), and genderqueer (N = 2; 0.3%). Data on sexual orientation were available for 252 respondents (38.8% of the total sample). Participants primarily identified as heterosexual (N = 186, 73.8%), with the rest identifying as bisexual (N = 48, 19.0%) and gay or lesbian (N = 6, 2.4%). Twelve participants (4.8%) chose “Another term describes me.” Self-reported body mass index had a M (SD) of 25.45 (6.61) kg/m2 (range: 13.18–59.38 kg/m2).
2.2 Procedure
The institutional review boards at all three sites approved the study procedures and all participants provided informed consent. All data were collected online. The majority of participants (N = 539) received partial course credit for participation. A subset (N = 111) chose whether to receive partial course credit or to be entered into a raffle for one of two $25 gift cards.
2.3 Measures
2.3.1 Eating Disorder Diagnostic Scale for DSM-5 (EDDS; Stice, n.d.; Stice et al., 2000)
The scoring algorithm for the Eating Disorder Diagnostic Scale for DSM-5 (EDDS-5) was used with one modification for the DSM-5 scheme. Five percent, rather than 10%, weight loss defined atypical AN, given data that a 5% weight loss definition is associated with large differences in drive for thinness and eating-related distress when comparing women with atypical AN to those with comparable amounts of weight loss without an eating disorder (Forney et al., 2017). In the EDDS-5 algorithm, a diagnosis of atypical AN supersedes low-frequency BN, low-frequency BED, purging disorder, and night eating syndrome. The algorithm was re-applied with atypical AN superseding a BN or BED diagnosis for the atypical AN scheme. This variable was available for all participants (N = 650). We are unaware of data evaluating the psychometric properties of the EDDS for DSM-5; however, eating disorder diagnoses made with the DSM-IV version of EDDS show good concordance with interview-based diagnoses (Stice et al., 2004). Agreement between the EDDS for DSM-5 and Eating Disorder Assessment for DSM-5 ranges from poor (K = .27; no diagnosis on EDDS) to excellent (K = .77 for anorexia nervosa) in treatment-seeking eating disorder samples (Sysko et al., 2015). See Table 1 for diagnoses made with the EDDS.
| Diagnosis | DSM-5 scheme, N (%) | Atypical AN scheme, N (%) |
|---|---|---|
| AN | 28 (4.3) | 28 (4.3) |
| Bulimia nervosa | 228 (35.1) | 130 (20.0) |
| Binge-eating disorder | 11 (1.7) | 7 (1.1) |
| Atypical AN | 234 (36.0) | 336 (51.7) |
| Low frequency bulimia nervosa | 4 (0.6) | 4 (0.6) |
| Low-frequency binge-eating disorder | 1 (0.2) | 1 (0.2) |
| Purging disorder | 25 (3.8) | 25 (3.8) |
| Night eating syndrome | 119 (18.3) | 119 (18.3) |
- Note: Diagnoses were made using the Eating Disorder Diagnostic Scale for DSM-5.
- Abbreviation: DSM-5, Diagnostic and Statistical Manual, 5th Edition.
2.3.2 World Quality Organization Quality of Life Assessment (WHOQOL-BREF; The Whoqol Group, 1998)
The WHOQOL-BREF assesses quality of life in four domains (i.e., physical, psychological, environmental, and social) over the past 2 weeks on a 5-point Likert scale (The Whoqol Group, 1998). Higher scores indicate better quality of life. The Physical (seven items), Psychological (six items), and Social (three items) Quality of Life subscales were used in the present study and transformed to be on a 0 to 100 scale (The Whoqol Group, 1998). Lower quality of life in these domains has been associated with higher levels of eating pathology in community and clinical samples of women with eating disorders compared to women without eating disorders (Mond et al., 2012). McDonald's omega suggested acceptable reliability (Physical ω = .76, Psychological ω = .79, and Social ω = .76). These variables were present for 305 (Social Quality of Life) to 306 (Physical and Psychological Quality of Life) participants (Forney, Burton Murray, et al., 2023; Forney, Horvath, et al., 2023; Harris, 2023).
2.3.3 Clinical impairment assessment (Bohn et al., 2008)
The Clinical Impairment Assessment is a 16-item measure that assesses impairment due to eating disorder symptoms over the past month on a 4-point Likert scale. Higher scores indicate greater impairment. The total score is positively correlated with other self-report measures of eating disorder symptoms and clinician ratings of impairment (Bohn et al., 2008). The total score was used in the present study and reliability was excellent (McDonald's ω = .93). Data were available for 271 participants (Biderman et al., 2023; Forney, Horvath, et al., 2023).
2.3.4 Eating pathology symptoms inventory (EPSI; Forbush et al., 2013)
The Eating Pathology Symptoms Inventory (EPSI) assesses eight domains of eating pathology on a 5-point Likert scale over the past 4 weeks. The instructions of the EPSI were modified to be administered online with permission from the copyright holder. Higher scores indicate greater eating pathology. The Body Dissatisfaction (seven items), Binge Eating (eight items), Restricting (six items), Purging (six items), and Excessive Exercise (five items) subscales were used in exploratory analyses in the present study. Reliability of the subscales was acceptable (McDonald's ω = .77–.90). Data were available for 633–647 participants, depending upon subscale.
2.4 Data analytic plan
We used SPSS version 29.0 to compute descriptive statistics and reliability estimates. We used R version 3.6.0 for all other analyses (R Core Team, 2019). Eating disorder diagnoses were group coded with atypical AN as the referent. Four sets of analyses were run using the following dependent variables with the following sample sizes: Physical Quality of Life (n = 306), Psychological Quality of Life (n = 306), Social Quality of Life (n = 306), and Clinical Impairment Assessment total score (n = 271). We regressed dependent variable (e.g., Psychological Quality of Life) onto DSM-5 eating disorder diagnosis (see Table 1, middle column). A second linear regression was run using the atypical AN scheme (see Table 1, right column).
We compared the fit of the regression models using the Davidson-MacKinnon J test in the “lmtest” R package (Zeileis & Hothorn, 2002). The DSM-5 scheme model was tested as the “first” model and the atypical AN scheme was tested as the “second” model. In the Davidson–MacKinnon J test, the fitted values from the “second” model are added as a predictor in the first model. A significant Davidson–MacKinnon J test indicates that the fitted values from the atypical AN scheme regression model significantly improved the fit of the DSM-5 scheme regression model. We pre-registered our data interpretation, such that a significant Davidson–MacKinnon J test for three of four dependent variables indicated that the atypical AN scheme provides more clinical information than the current DSM-5 diagnostic scheme. An alpha of .05 was used for interpretation. Missing data were handled using listwise deletion. As an exploratory aim, we repeated these models using body dissatisfaction (n = 640), binge eating (n = 633), restricting (n = 638), purging (n = 647), and excessive exercise as dependent variables (n = 638).
3 RESULTS
3.1 Primary analyses
Table 1 displays diagnostic rates using the two diagnostic schemes. A large proportion (42.9%; n = 98) of those with BN simultaneously met the criteria for atypical AN and 36.4% (n = 4) of those with BED simultaneously met the criteria for atypical AN. Table 2 presents the linear regression models using the DSM-5 scheme (left panel) and atypical AN scheme (right panel) in explaining Physical Quality of Life scores. The Davidson–MacKinnon J test was non-significant (t = .98, p = .33), suggesting that adding the fitted values from the atypical AN model did not improve model fit in explaining Physical Quality of Life scores. Similarly, the Davidson–MacKinnon J test was non-significant when examining psychological quality of life (see Table 3 for regression models; t = .10, p = .92) and social quality of life (see Table 4 for regression models; t = .62, p = .54). However, the atypical AN scheme explained additional variance in eating disorder-related impairment above and beyond the DSM-5 scheme (see Table 5 for regression models; t = 2.31, p = .02).
| Diagnosis | DSM-5 scheme | Atypical anorexia nervosa (AN) scheme | ||||
|---|---|---|---|---|---|---|
| Estimate | S.E. | p | Estimate | S.E. | p | |
| Intercept | 67.09 | 1.54 | <.001 | 65.27 | 1.23 | <.001 |
| Anorexia nervosa (AN) | −6.43 | 5.50 | .24 | −4.60 | 5.44 | .40 |
| Bulimia nervosa (BN) | −2.83 | 2.07 | .17 | 1.13 | 2.22 | .61 |
| Binge-eating disorder (BED) | 6.91 | 6.65 | .30 | 9.73 | 9.26 | .29 |
| Low-frequency BN | – | – | – | – | – | – |
| Low-frequency BED | – | – | – | – | – | – |
| Purging disorder | −1.67 | 6.18 | .79 | .16 | 6.14 | .98 |
| Night eating syndrome | −.37 | 2.78 | .89 | 1.45 | 2.63 | .58 |
- Note: Atypical AN is the referent.
- Abbreviation: DSM-5, Diagnostic and Statistical Manual, 5th Edition.
| Diagnosis | DSM-5 scheme | Atypical anorexia nervosa (AN) scheme | ||||
|---|---|---|---|---|---|---|
| Estimate | S.E. | p | Estimate | S.E. | p | |
| Intercept | 44.08 | 1.61 | <.001 | 43.28 | 1.29 | <.001 |
| Anorexia nervosa (AN) | −9.96 | 5.76 | .09 | −9.17 | 5.70 | .11 |
| Bulimia nervosa (BN) | −3.27 | 2.17 | .13 | −2.65 | 2.33 | .26 |
| Binge-eating disorder (BED) | 6.92 | 6.96 | .32 | .38 | 9.70 | .97 |
| Low-frequency BN | – | – | – | – | – | – |
| Low-frequency BED | – | – | – | – | – | – |
| Purging disorder | −.22 | 6.47 | .97 | .57 | 6.43 | .93 |
| Night eating syndrome | 2.90 | 2.91 | .32 | 3.70 | 2.75 | .18 |
- Note: Atypical AN is the referent.
- Abbreviation: DSM-5, Diagnostic and Statistical Manual, 5th Edition.
| Diagnosis | DSM-5 scheme | Atypical anorexia nervosa (AN) scheme | ||||
|---|---|---|---|---|---|---|
| Estimate | S.E. | p | Estimate | S.E. | p | |
| Intercept | 53.51 | 2.07 | <.001 | 53.83 | 1.65 | <.001 |
| Anorexia nervosa (AN) | −1.51 | 7.40 | .84 | −1.83 | 7.29 | .80 |
| Bulimia nervosa (BN) | 2.28 | 2.78 | .41 | 2.95 | 2.98 | .32 |
| Binge-eating disorder (BED) | −2.34 | 8.94 | .79 | −3.83 | 12.40 | .76 |
| Low-frequency BN | – | – | – | – | – | – |
| Low-frequency BED | – | – | – | – | – | – |
| Purging disorder | 8.06 | 8.31 | .33 | 7.74 | 8.22 | .35 |
| Night eating syndrome | 6.95 | 3.76 | .07 | 6.63 | 3.55 | .06 |
- Note: Atypical AN is the referent.
- Abbreviation: DSM-5, Diagnostic and Statistical Manual, 5th Edition.
| Diagnosis | DSM-5 scheme | Atypical anorexia nervosa (AN) scheme | ||||
|---|---|---|---|---|---|---|
| Estimate | S.E. | p | Estimate | S.E. | p | |
| Intercept | 24.90 | 1.10 | <.001 | 25.45 | 0.93 | <.001 |
| Anorexia nerovsa (AN) | −1.30 | 3.52 | .71 | −1.85 | 3.45 | .59 |
| Bulimia nervosa (BN) | 6.02 | 1.58 | <.001 | 7.48 | 1.72 | <.001 |
| Binge-eating disorder (BED) | −3.57 | 4.45 | .42 | 1.30 | 5.33 | .81 |
| Low-frequency BN | −.40 | 7.55 | .96 | −.95 | 7.49 | .90 |
| Low-frequency BED | – | – | – | – | – | – |
| Purging disorder | 4.02 | 3.13 | .20 | 3.48 | 3.06 | .26 |
| Night eating syndrome | −4.64 | 1.74 | .01 | −5.18 | 1.64 | <.01 |
- Note: Atypical AN is the referent.
- Abbreviation: DSM-5, Diagnostic and Statistical Manual, 5th Edition.
Because the subsample who provided data on the quality of life measure (n = 305–306) had limited overlap (n = 54) with the subsample who completed the Clinical Impairment Assessment (n = 271), we post hoc compared the samples on eating pathology to see if there were differences in diagnosis prevalence and symptom severity. There were differences in the diagnostic distributions by measure (Likelihood ratio (12) = 23.0, p = .03; see Table S1). The sample that only completed the WHOQOL-Bref had greater binge-eating (d = .33, p < .001) and lower restricting (d = .45, p < .001) scores than the subsample that only completed the Clinical Impairment Assessment.
3.2 Pre-registered exploratory analyses
Tables S2–S6 present regression models comparing diagnostic schemes in explaining eating disorder symptom severity. A non-significant Davidson–MacKinnon J test suggested that the atypical AN model did not improve model fit in explaining body dissatisfaction (t = .11, p = .91), binge eating (t = 1.73, p = .08), restricting (t = −.30, p = .77), and excessive exercise (t = −.56, p = .57). However, the atypical AN scheme did improve model fit when evaluating purging (t = −2.40, p = .02).
3.3 Post hoc interpretation of primary regression models
Given that relatively little data exist comparing atypical AN to other DSM-5 eating disorders, we post hoc examined regression coefficients from the models in our primary analyses. Because atypical AN is coded as the referent, any statistically significant regression coefficients indicate a difference between diagnostic groups. Across physical, social, and psychological quality of life assessments, there were no statistically significant differences between atypical AN and any other eating disorder. In understanding eating disorder-related impairment, across both diagnostic schemes, individuals with atypical AN endorsed greater impairment than those with night eating syndrome and less impairment than those with BN.
4 DISCUSSION
The present pre-registered study sought to address nosological questions about whether individuals who meet criteria for both atypical AN and BN or BED should be categorized as BN and BED (the DSM-5 scheme) or atypical AN (atypical AN scheme). We replicated previous findings of a high overlap between BN and atypical AN (Forney et al., 2017) and provided preliminary data suggesting overlap between BED and atypical AN. The atypical AN scheme better explained eating disorder-related impairment than the DSM-5 scheme, but it did not provide additional information on physical, psychological, and social quality of life. In exploratory analyses examining eating disorder symptoms, the atypical AN scheme provided additional information over the DSM-5 scheme in understanding purging, and did not provide additional information on body dissatisfaction, binge eating, restricting, and excessive exercise.
The reason for the current findings is not clear. It may be that those who meet the criteria for both BN and atypical AN experience their eating disorder more ego-syntonically, and therefore report less impairment, than those who meet the criteria for only BN. Thus, moving those who meet criteria for both atypical AN and BN to the atypical AN diagnosis (i.e., atypical AN Scheme) increased group differences in eating disorder-related impairment between BN and atypical AN relative to the DSM-5 scheme (DSM-5 scheme group difference Cohen's d = .60; atypical AN scheme Cohen's d = .76). Indeed, prior work in a non-clinical sample suggested reduced distress and lower likelihood of endorsing having an eating disorder in atypical AN compared to a comparison group that largely comprised BN and BED (Forney et al., 2017).
The post hoc finding of reduced eating disorder-related distress in atypical AN relative to BN is inconsistent with a recent systematic review suggesting increased eating pathology in atypical AN relative to AN (Walsh et al., 2023). This may reflect the absence of treatment-seeking bias in our university sample or it may reflect limitations associated with questionnaire-based diagnosis. Given that individuals tend to endorse more eating-related pathology on questionnaires compared to interview (Berg et al., 2011), the current operationalization of atypical AN may have been overinclusive. The majority of the work comparing atypical AN and AN (Walsh et al., 2023) has occurred in specialty eating disorder treatment settings where severity is confounded with treatment seeking. Research is needed in both treatment and non-treatment settings, as only a small minority of individuals with eating disorders seeks treatment (Field et al., 2023). Given low rates of treatment seeking among individuals with eating disorders, the current university sample may be more representative of the general population of individuals with eating disorders than a tertiary care, treatment-seeking sample. More work is needed in non-treatment-seeking samples to fully understand differences in presentation to inform interventions that may be effective for all individuals when they choose to seek treatment.
In addressing these nosological questions, we purposefully took a conservative approach in interpreting data because empirically supported psychological and psychopharmacological treatments exist for BN and BED (Hagan & Walsh, 2021; Hilbert, 2023; Monteleone et al., 2022) and do not yet exist for atypical AN (Walsh et al., 2023). While transdiagnostic cognitive behavioral therapy for eating disorders was designed to address other specified feeding or eating disorders like atypical AN (Fairburn, 2008), this treatment is not empirically supported for atypical AN's most diagnostically similar eating disorder presentation, AN (Monteleone et al., 2022). Indeed, there is no empirically supported treatment for adults with AN (Monteleone et al., 2022). Thus, retaining BN and BED's placement in the diagnostic hierarchy should result in the implementation of more evidence-based treatment interventions. Importantly, our findings only speak to cross-sectional differences in quality of life and impairment. More work is needed to understand longitudinal course and treatment response. Finally, the current work only addresses eating disorder classification as it relates to current categorical schemes. Hybrid or dimensional approaches may have more utility (Forbush et al., 2023; Luo et al., 2016), particularly in communicating about weight suppression.
The current study had a number of strengths including pre-registration of data interpretation, a sample from multiple regions in North America, and a focus on clinical outcomes that are not specifically tied to behavioral symptoms. However, we are limited by the use of an undergraduate sample, questionnaire-based diagnoses, self-reported BMI, and a reduced, albeit still large, sample size for the primary dependent variables. In particular, the EDDS algorithm for atypical AN makes a diagnosis based on the presence of cognitive features in the absence of consideration of eating disorder behaviors or restriction. Importantly, the lack of behavioral assessment does not mean that eating disorder behaviors are absent. Only a minority of those with a DSM-5 atypical AN diagnosis (7.3%) denied engaging in compensatory behaviors. There remains debate about whether or how the recency of weight loss should be considered in atypical AN (Walsh et al., 2023); the current study relied on lifetime weight history consistent with the EDDS algorithm and cannot speak to the recency of weight loss. Further, the prevalence of BED was rather low compared to epidemiological data (American Psychiatric Association, 2013), suggesting that some individuals with “true” BED may have been miscategorized as BN. This does not change our interpretation, however, as all eating disorders were examined in models concurrently. Our post hoc interpretation of regression coefficients comparing diagnostic groups was underpowered for many diagnoses. Although our sample does not have treatment-seeking biases, it is limited in socioeconomic diversity, racial diversity, and developmental stage. Although the WHOQOL-Bref had good variability in our sample (lowest scores ranged 0–6, highest scores ranged 81–100; SDs ranged 15.83–21.28) and quality of life is an important metric of well-being, aspects of quality of life may not be as useful as impairment in differentiating between eating disorders. Our cross-sectional approach cannot speak to prognostic or treatment-approach information, which is also important when considering the utility of nosological schemes. A longitudinal approach is particularly important, given the prevalence of diagnostic cross-over in eating disorders (Schaumberg et al., 2019).
In sum, the current study provides one approach for addressing questions of how to categorize or diagnose individuals who meet the criteria for atypical AN and BN or BED. We found that having an atypical AN diagnosis supersede a BN or BED diagnosis was associated with a better understanding of eating disorder-related impairment, but not physical quality of life, psychological quality of life, or social quality of life. These findings appear to be driven by the fact that those with BN reported more impairment than those with atypical AN. Altogether, our findings do not support changing the DSM-5 diagnostic scheme. However, this study represents only one piece of information in a large puzzle. More longitudinal, epidemiological, and treatment research is needed to revise diagnostic schemes to best provide information about clinical presentation, prognosis, and appropriate treatment approach.
AUTHOR CONTRIBUTIONS
K. Jean Forney: Conceptualization; data curation; formal analysis; project administration; writing – original draft; writing – review and editing. Taylor Rezeppa: Writing – original draft; writing – review and editing. Naomi G. Hill: Writing – original draft; writing – review and editing. Lindsay P. Bodell: Conceptualization; data curation; project administration; writing – review and editing. Tiffany A. Brown: Data curation; project administration; writing – review and editing.
ACKNOWLEDGMENTS
Data collection was supported by internal funding from Western University (LPB).
CONFLICT OF INTEREST STATEMENT
The authors have no conflicts to disclose.
Open Research
OPEN RESEARCH BADGES
This article has earned a Preregistered Research Designs badge for having a preregistered research design, available at https://osf.io/2ejcd.
DATA AVAILABILITY STATEMENT
Data and code are available upon request.
