Eating disorders (EDs), particularly anorexia nervosa (AN), bulimia nervosa (BN) and binge eating disorder (BED), are complex psychiatric disorders associated with severe disturbances in eating behaviors. They have poor treatment outcomes and high relapse rates (Pehlivan et al., Reference Pehlivan, Miskovic-Wheatley, Le, Maloney, Touyz and Maguire2022). The Diagnostic and Statistical Manual of Mental Disorders (Third Edition; DSM-III) was published by the American Psychiatric Association (APA) in 1980 and laid the foundation for eating disorder classification, providing diagnostic criteria for AN and BN (Pope et al., Reference Pope, Lalonde, Pindyck, Walsh, Bulik, Crow, McElroy, Rosenthal and Hudson2006). Since then, there have been two additional versions, the most recent being the DSM-5 in 2013, each with adjustments in diagnostic categories and additional categories of EDs, including BED, pica, rumination disorder, and avoidant/restrictive food intake disorder (ARFID).
Accurate and clinically useful classification systems are critical to our understanding of the etiology of EDs and to our research methods and outcomes. How we classify EDs is fundamental to the conduct of research, as the meaningful grouping of cases is needed to accurately investigate illness features, causes and responses to treatment (Bulik et al., Reference Bulik, Thornton, Parker, Kennedy, Baker, MacDermod, Guintivano, Cleland, Miller, Harper and Larsen2021). ED diagnostic categories based on the DSM models have been shown to have limited predictability of clinical outcomes and stability over time (Eddy et al., Reference Eddy, Dorer, Franko, Tahilani, Thompson-Brenner and Herzog2008; Fichter & Quadflieg, Reference Fichter and Quadflieg2007; Schaumberg et al., Reference Schaumberg, Jangmo, Thornton, Birgegård, Almqvist, Norring, Larsson and Bulik2019; Thompson-Brenner et al., Reference Thompson-Brenner, Shingleton, Sauer-Zavala, Richards and Pratt2015).
One of the critical issues of the DSM-IV diagnostic criteria was the high percentage of individuals with clinically significant ED symptoms who did not meet the specific diagnostic criteria of the two formally recognized ED diagnoses (AN, BN) at the time and were thus categorized into a heterogenous residual category — Eating Disorder Otherwise Not Specified (EDNOS). Evidence showed that an EDNOS diagnosis represented 40−60% of individuals presenting with an ED (Fairburn et al., Reference Fairburn, Cooper, Bohn, O’Connor, Doll and Palmer2007; Stice et al., Reference Stice, Marti and Rohde2013). Partially in response to the large number of individuals falling into the EDNOS category in the DSM-IV, the diagnostic criteria for AN and BN were relaxed (by removing the amenorrhea criterion for AN and reducing the required frequency of binge-eating/inappropriate compensatory behaviors for BN) in DSM-5 (Fairweather-Schmidt & Wade, Reference Fairweather-Schmidt and Wade2014). Further, BED, previously classified as a research category within EDNOS, was formally recognized as an ED diagnosis in the DSM-5, giving those presenting with binge-eating episodes without inappropriate compensatory behaviors a specific ED diagnosis. In the DSM-5, EDNOS was also replaced with new two categories that differentiate between cases where full-threshold criteria for one of the EDs is not met for a specific reason — Other Specified Feeding and Eating Disorder (OSFED) — and where the clinician has insufficient information to determine an individual’s clinical characteristics and/or provide a definitive diagnosis, Unspecified Feeding or Eating Disorder (UFED). Both OSFED and UFED are considered subthreshold disorders, with OSFED allowing the clinician to specify the full-threshold ED diagnosis that the case most closely resembles; for example, using the specifier, low frequency BN (subthreshold BN) or limited duration BED (subthreshold BED). Other subthreshold disorders in OSFED include atypical anorexia nervosa (A-AN — AN without the weight criterion), purging disorder (PD), and night eating syndrome (NES). Research has suggested that these subthreshold disorders can be highly impactful, with outcomes as severe as in full-threshold disorders (e.g., AN, BN, and BED; Garber et al., Reference Garber, Cheng, Accurso, Adams, Buckelew, Kapphahn, Kreiter, Le Grange, Machen, Moscicki and Saffran2019). However, even with the broadening of criteria for full-threshold categories in the DSM-5, research continues to indicate that as many as 40% of cases fall into the OSFED category, with migration in and out of full-threshold disorders not uncommon (Fairweather-Schmidt & Wade, Reference Fairweather-Schmidt and Wade2014; Forbush et al., Reference Forbush, Chen, Hagan, Chapa, Gould, Eaton and Krueger2018; Vo et al., Reference Vo, Accurso, Goldschmidt and Le Grange2017). Further, individuals with subthreshold and full-threshold EDs cannot yet be differentiated based on factors such as psychiatric impairment, comorbid disorders, and genetic and environmental risk factors (Bulik et al., Reference Bulik, Thornton, Root, Pisetsky, Lichtenstein and Pedersen2010; Forbush et al., Reference Forbush, Hagan, Kite, Chapa, Bohrer and Gould2017). This ambiguity in diagnosis leads to questions about the diagnostic integrity of the current classification system and the implications for research in our field. This article will explore both our current nosological system for EDs and alternative dimensional models that can be utilized for classification and research.
Diagnostic Stability and DSM Classification Models
Longitudinal studies have demonstrated that individuals diagnosed with an ED using DSM criteria at baseline may be diagnosed with a different ED at follow-up (Forbush et al., Reference Forbush, Hagan, Kite, Chapa, Bohrer and Gould2017). Milos et al. (Reference Milos, Spindler, Schnyder and Fairburn2005) examined the course of a full range of EDs (including subthreshold and full-threshold disorders) in a prospective study using DSM-IV categories; their group found considerable diagnostic instability, with only 29% of the initial sample of 192 women retaining the same diagnosis across three assessment points over 30 months. Similarly, Tozzi et al. (Reference Tozzi, Thornton, Klump, Fichter, Halmi, Kaplan, Strober, Woodside, Crow, Mitchell and Rotondo2005) found 36% of patients initially diagnosed with AN crossed over to BN within the first 5 years of illness, and 27% of individuals initially diagnosed with BN crossed over to AN binge-purge subtype (AN-BP). In another prospective study of 294 treatment-seeking women, Eddy et al. (Reference Eddy, Dorer, Franko, Tahilani, Thompson-Brenner and Herzog2008) found high rates of diagnostic crossover between AN subtypes and from AN to BN but reported lower rates of crossover from BN and BED into AN.
It is difficult to ascertain whether the diagnostic crossover observed is a failure of initial classification or a true transition to a distinctly different illness category. For instance, a cross-over in diagnosis from AN binge/purge subtype to BN would only require a change in weight status. The current DSM-5 diagnostic criteria were predicted to decrease diagnostic crossover (Fairweather-Schmidt & Wade, Reference Fairweather-Schmidt and Wade2014). Yet studies using DSM-5 diagnostic criteria continue to show diagnostic migration, albeit mostly between subthreshold conditions rather than full-threshold disorders. Most commonly, diagnostic crossover occurs between full-threshold and subthreshold disorders, followed by crossover from AN to BN and then BN to AN (Breithaupt et al., Reference Breithaupt, Kahn, Slattery, Plessow, Mancuso, Izquierdo, Dreier, Becker, Franko, Thomas and Holsen2022; Fichter et al., Reference Fichter, Quadflieg, Crosby and Koch2017; Schaumberg et al., Reference Schaumberg, Jangmo, Thornton, Birgegård, Almqvist, Norring, Larsson and Bulik2019; Yao et al., Reference Yao, Larsson, Norring, Birgegård, Lichtenstein, D’Onofrio, Almqvist, Thornton, Bulik and Kuja-Halkola2021). These findings support the hypothesis that the nosological system is creating diagnostic ‘crossover’ as an artefact of its poor design. Such diagnostic migration may reflect the arbitrary boundaries of the DSM-5 rather than clinically meaningful illness factors and changes in pathology.
Individuals assigned a diagnosis of OSFED with a lifetime experience of AN or BN may meet the criteria for AN or BN at one point in time and narrowly miss meeting the full criteria for that disorder at another point in time (Breithaupt et al., Reference Breithaupt, Kahn, Slattery, Plessow, Mancuso, Izquierdo, Dreier, Becker, Franko, Thomas and Holsen2022; Eddy et al., Reference Eddy, Swanson, Crosby, Franko, Engel and Herzog2010; Fichter & Quadflieg, Reference Fichter and Quadflieg2007; Milos et al., Reference Milos, Spindler, Schnyder and Fairburn2005; Tozzi et al., Reference Tozzi, Thornton, Klump, Fichter, Halmi, Kaplan, Strober, Woodside, Crow, Mitchell and Rotondo2005). Thus, a minor symptom fluctuation can change a person’s diagnosis from one ED to another, reducing the validity of diagnosis and treatment options. Indeed, OSFED, as a subthreshold ED disorder, almost certainly acts also as a precursor to a full-threshold disorder or a diagnostic marker on a pathway to remission or recovery (Breithaupt et al., Reference Breithaupt, Kahn, Slattery, Plessow, Mancuso, Izquierdo, Dreier, Becker, Franko, Thomas and Holsen2022). Symptom fluctuation is part of the illness course for any ED and should not be a trigger for crossover to a diagnosis of another distinct disorder. A definition of diagnostic categories that allows for symptom fluctuation over time in line with the observed true course of illness would make for a more robust and meaningful diagnostic system. With evidence displaying poor longitudinal stability over time, diagnostic instability remains a problem using the current DSM-V categorical model for clinical diagnosis, treatment, and research on understanding these illnesses’ etiology.
Comorbidity and the DSM Diagnostic Model
Approximately 80% of individuals with an ED have a co-occurring comorbid psychiatric illness, most commonly mood disorders, generalized anxiety disorder, trauma-related or obsessive compulsive disorder (Forbush et al., Reference Forbush, Hagan, Kite, Chapa, Bohrer and Gould2017). There is evidence that the presence of a comorbid mood disorder may predict diagnostic instability (Castellini et al., Reference Castellini, Lo Sauro, Mannucci, Ravaldi, Rotella, Faravelli and Ricca2011). A 6-year longitudinal study evaluating diagnostic crossover in EDs indicated that mood disorder comorbidity explained some symptom changes between EDs, particularly with AN and BN diagnoses (Castellini et al., Reference Castellini, Lo Sauro, Mannucci, Ravaldi, Rotella, Faravelli and Ricca2011). Some scholars have even suggested that BN should be a variant of a mood disorder (Forbush et al., Reference Forbush, Hagan, Kite, Chapa, Bohrer and Gould2017). Considering the role of psychiatric comorbidity in symptom changes and diagnostic crossover, it may be that the presence or absence of comorbid symptoms determines illness trajectory more accurately than the single primary diagnostic category. Additionally, comorbidity utilizing the DSM-5 is typically looked at as two or more co-occurring discrete categorical illnesses, and therefore treatment plans are shaped around the two distinct disorders (Helle et al., Reference Helle, Trull, Watts, McDowell and Sher2020). However, studies have shown that comorbidity can be demonstrated more accurately by looking at symptoms across a hierarchical dimensional spectrum, allowing for a diagnosis and treatment plan encompassing all elements of psychopathology (Kotov et al., Reference Kotov, Jonas, Carpenter, Dretsch, Eaton, Forbes, Forbush, Hobbs, Reininghaus, Slade and South2020; Newson et al., Reference Newson, Pastukh and Thiagarajan2021).
Approaches to Reduce Diagnostic Instability
Several research groups proposed methods to inform the new classification of EDs in the DSM-5 to deal with heterogeneity and instability across diagnostic categories. Fairburn et al. (Reference Fairburn, Cooper and Shafran2003)’s Transdiagnostic Cognitive Behavioural model for EDs suggested that all EDs, despite their different presentations, have common core mechanisms and thus should be treated as a single illness category. Fairburn proposed a transdiagnostic model for the DSM-5, stating the diagnostic criteria of the DSM-IV categories were essentially arbitrary and that individual changes in the clinical presentation do not reflect recovery but rather the evolution of a single ED sharing a core psychopathology (Fairburn et al., Reference Fairburn, Cooper, Bohn, O’Connor, Doll and Palmer2007). Fairburn recognized that the psychopathology of EDNOS closely resembled that of AN and BN and suggested the high prevalence of EDNOS may indicate the need for a categorized ‘mixed’ category in DSM-5 for cases where clinical features of AN and BN combine in subtly different ways (Fairburn et al., Reference Fairburn, Cooper, Bohn, O’Connor, Doll and Palmer2007).
Other researchers have identified fundamental issues with a transdiagnostic model of EDs, debunking that AN and BN are a single disorder with a common cause using Hill’s Criteria of Causation (Birmingham et al., Reference Birmingham, Touyz and Harbottle2009). Analyses of the clustering of eating behavior symptom presentations suggest distinct symptom profiles and treatment needs, challenging the idea that all EDs share a common underlying mechanism that can be targeted with a transdiagnostic approach. Clinton et al. (Reference Clinton, Button, Norring and Palmer2004) explored natural groupings of patients who presented with a wide range of ED symptoms (N = 1103) and found three clusters: one that corresponds to AN characterized by low weight, amenorrhoea and absence of binge eating; another they named ‘overeating’ characterized by high weight and moderate levels of binge eating, and a third called ‘generalized eating disorder’, which was characterized by high levels of ED psychopathology on all variables except weight and menstrual functioning. Another recent study by Rossi et al. (Reference Rossi, Pietrabissa, Tagliagambe, Scuderi, Montecchiani, Castelnuovo, Mannarini and Dalla Ragione2023) on a study of patients with AN and BED (N = 421) found that while there were commonalities in risk factors for illness onset and progression, disease-specific pathways also contributed to AN and BED development. Also, using a transdiagnostic model for ED may lead to treatment selection and efficacy difficulties, as current evidence suggests ED treatments work differently for individuals with different ED diagnoses (Birmingham et al., Reference Birmingham, Touyz and Harbottle2009).
Dimensional Models of Classification
The DSM-5 model included several improvements from the DSM-IV, such as the lowered diagnostic thresholds for BN and the inclusion of BED as a threshold diagnostic category. However, the large percentage of individuals with OSFED diagnoses, even after these changes were made to the DSM-5, demonstrate problems with the validity and clinical utility. Many psychological theorists have utilized a dimensional model of psychopathology rather than a categorical classification model. Cluster analytic studies have shown ‘clusters’ of people with similar symptom profiles, but cluster profile studies do not test for the presence of dimension and fluidity in a diagnosis (Williamson et al., Reference Williamson, Gleaves and Stewart2005). Peterson et al. (Reference Peterson, Crow, Swanson, Crosby, Wonderlich, Mitchell, Agras and Halmi2011) in a study investigating the longitudinal stability of ED symptoms over 2 years, found three latent classes (binge eating and purging, binge eating, and low BMI), which had greater longitudinal stability and less crossover at the five examination points than the DSM-IV diagnostic categories. Thus, empirical classification approaches may better predict symptom stability and clinically relevant outcomes.
HiTOP Model and Implications for Nosology
In the past five years, a new bottom-up model called the Hierarchical Taxonomy of Psychopathology (HiTOP) was proposed that views mental health as a spectrum (Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark and Eaton2017). HiTOP classifies psychopathology dimensions at multiple levels of a hierarchy, allowing for more focus on behavioral phenotypes rather than diagnostic categories.
In the HiTOP model, the bottom levels are comprised of homogenous symptom groups, such as meal skipping behaviors or purging, that combine into dimensional syndromes, some of which roughly correspond to DSM diagnoses. Syndromes combine into subfactors, in the context of EDs, an ‘eating pathology’ subfactor (Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark and Eaton2017). Subfactors then combine to form spectra. Kotov et al. identified six spectra, including the internalizing spectrum, which is comprised of distress (unipolar mood, generalized anxiety, and PTSD), fear (social anxiety disorder, OCD, agoraphobia, and panic disorder), eating pathology (AN, BN, and BED) and sexual problems. The externalizing spectrum includes adult antisocial behaviors, drug and alcohol misuse, intermittent explosive disorder, gambling disorder, and ADHD. These spectra can be aggregated to form a general ‘p-factor’ that describes the association between them (Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark and Eaton2017). Twin studies in mental health research have supported common genetic vulnerabilities underlying the presence of a p-factor (Conway et al., Reference Conway, Forbes, Forbush, Fried, Hallquist, Kotov, Mullins-Sweatt, Shackman, Skodol, South and Sunderland2019). Conway et al. (Reference Conway, Forbes, Forbush, Fried, Hallquist, Kotov, Mullins-Sweatt, Shackman, Skodol, South and Sunderland2019) also found in their review that genetic influences have been identified at the lower levels of the hierarchy at the spectrum and syndrome levels. The narrow levels of the hierarchy can provide good insight for symptom-specific treatments in heterogeneous presentations of disease and account for comorbidity.
Several studies have evaluated and validated the utility of the HiTOP model in clinical settings and a few for classification purposes in EDs specifically (Forbush et al., Reference Forbush, Chen, Hagan, Chapa, Gould, Eaton and Krueger2018; Kotov et al., Reference Kotov, Jonas, Carpenter, Dretsch, Eaton, Forbes, Forbush, Hobbs, Reininghaus, Slade and South2020). Forbush et al. (Reference Forbush, Chen, Hagan, Chapa, Gould, Eaton and Krueger2018) developed the Hierarchical Taxonomy of Internalizing Dimensions for EDs (Hi-TIDE) model guided by the Hi-TOP model (Figure 1). Their HiTIDE model conceptualizes EDs as an internalizing disorder and identifies 15 dimensions characterizing individuals with eating disorders: distress, wellbeing, OCD and mania, restricting, negative attitudes towards obesity, excessive exercise, binge eating, body dissatisfaction, insomnia, lassitude, PTSD, social anxiety, mindlessness, purging and claustrophobia (Forbush et al., Reference Forbush, Chen, Hagan, Chapa, Gould, Eaton and Krueger2018). In a 3-year longitudinal study to test the validity of this model using a community sample of 243 individuals diagnosed with a DSM-V ED, they found that Hi-TIDE was a more robust predictor of 6-month outcomes of those diagnosed with an ED than the DSM-5 diagnostic categories and explained more variance in eating psychopathology severity scores (Forbush et al., Reference Forbush, Chen, Hagan, Chapa, Gould, Eaton and Krueger2018). Further, the dimensional model of EDs was more likely to stay the same (i.e., had greater stability) from initial classification to 6-month follow-up than the DSM-5 categorical modeling, the most stable dimensions being distress, restricting, negative attitudes towards obesity and OCD contamination, with coefficients of congruence showing high structural stability (>.80; Forbush et al., Reference Forbush, Chen, Hagan, Chapa, Gould, Eaton and Krueger2018). A more dimensional way of conceptualizing EDs, which focuses on behavioral phenotypes, may account for some factors not explained by categorical classification methods.
Figure 1. Hi-TIDE model developed by Forbush et al. (Reference Forbush, Chen, Hagan, Chapa, Gould, Eaton and Krueger2018).
Implications for Future Genetic Studies of EDs
Diagnostic stability poses a particular challenge to genetic research because genetic samples need to be collected from individuals who have met the criteria for a particular mental illness. Furthermore, the longitudinal stability of a trait sets an upper limit to its heritability and hence to the power to find causal genes by genomewide association studies (GWAS). If participants have met criteria for a number of related illnesses, arbitrary decisions, such as which diagnosis is the primary, need to be made. More research is required to validate whether using a dimensional model for genetic studies of EDs can increase the power to detect causal genes. Using a dimensional model may prevent individuals from being excluded, misclassified, or categorized differently. It is widely accepted that a continuous trait confers greater power than a binary categorisation of affected/unaffected.
To date, genetic studies have all been dependent on the DSM-5 diagnostic categories (Waszczuk et al., Reference Waszczuk, Eaton, Krueger, Shackman, Waldman, Zald, Lahey, Patrick, Conway, Ormel and Hyman2020), with large sample numbers needed due to the heterogenous presentation of EDs. Studies examining the genetic basis of full-threshold EDs, have sometimes extended their scope to include those with subthreshold EDs; including, for example, partial/broad AN or BN (Bulik et al., Reference Bulik, Thornton, Root, Pisetsky, Lichtenstein and Pedersen2010; Strober et al., Reference Strober, Freeman, Lampert, Diamond and Kaye2000). A genetic basis for these subthreshold EDs has been established (Lundgren et al., Reference Lundgren, Allison and Stunkard2006; Munn-Chernoff et al., Reference Munn-Chernoff, Keel, Klump, Grant, Bucholz, Madden, Heath and Duncan2015), albeit to a lesser extent than full-threshold cases, in which heritability estimates and familial aggregation have been higher (Bulik et al., Reference Bulik, Thornton, Root, Pisetsky, Lichtenstein and Pedersen2010; Strober et al., Reference Strober, Freeman, Lampert, Diamond and Kaye2000). Incorporating a Hi-TOP or Hi-TIDE like dimensional framework into genetic research so that all participants have a score on all behavioral phenotype dimensions (e.g., 10-point Likert scale for restriction, purging, bingeing behaviors) could increase statistical power by providing consistent phenotypic targets at levels of the hierarchy and more accurately characterize how genetic predispositions to EDs could present themselves (Waszczuk et al., Reference Waszczuk, Eaton, Krueger, Shackman, Waldman, Zald, Lahey, Patrick, Conway, Ormel and Hyman2020).
In the broader mental health field, there is mounting evidence that genetic risk factors do not adhere to the diagnostic categories that the DSM-V currently identifies (Waszczuk et al., Reference Waszczuk, Eaton, Krueger, Shackman, Waldman, Zald, Lahey, Patrick, Conway, Ormel and Hyman2020). When studying the genetic components of OCD, researchers have found that the five symptoms (checking, hoarding, obsessing, ordering, and washing) have independent genetic influences that account for 11−23% of the total phenotypic variance beyond what is accounted by a single, common genetic factor (Iervolino et al., Reference Iervolino, Rijsdijk, Cherkas, Fullana and Mataix-Cols2011). Given the similarities between OCD symptom profiles and EDs, genetic risk research may identify broad risk factors that cut across diagnoses of EDs, such as diuretics usage, self-induced vomiting, fasting, and meal-skipping behaviors and identify independent genetic influences on particular behaviors.
The HiTOP and Hi-TIDE models should be explored in future genetic studies of EDs as a data-driven classification system to identify the genetic markers more effectively to EDs rather than the traditional diagnostic categories currently utilized. HiTOP-compatible scales exist that allow for higher order and bifactor modeling (Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark and Eaton2017). Furthermore, some EDs are too rare to study in a typical G where large sample sizes are needed. The incorporation of Hi-TIDE and genotyping based on behavioral phenotype could increase statistical power by providing consistent phenotypic targets at every level of the hierarchy and more accurately characterize how genetic predispositions to EDs could present themselves (Waszczuk et al., Reference Waszczuk, Eaton, Krueger, Shackman, Waldman, Zald, Lahey, Patrick, Conway, Ormel and Hyman2020). With genetic research comes a more comprehensive understanding of EDs, not only as a psychological illness but as a metabolic one. A question arises as we accumulate data on genetics and EDs whether genetic data and cross-disorder heritability challenge the DSM paradigm. Disorders that the DSM defines as mutually exclusive are challenged by genomic studies, particularly twin studies that show evidence for a general psychopathology factor and internalizing and externalizing factors (Allegrini et al., Reference Allegrini, Cheesman, Rimfeld, Selzam, Pingault, Eley and Plomin2020; Liu et al., Reference Liu, Lichtenstein, Kotov, Larsson, D’Onofrio and Pettersson2024; Smoller et al., Reference Smoller, Andreassen, Edenberg, Faraone, Glatt and Kendler2019).
The genetics of complex traits (like EDs) does not follow traditional Mendelian inheritance patterns and all traits studied with sufficient power have proved to be highly polygenic, with hundreds or even thousands of genetics risk variants of individually small effect, but large cumulative effect (Breithaupt et al., Reference Breithaupt, Hubel and Bulik2018). Furthermore, many risk variants seem to be shared between related traits and this can be summarized as a genetic correlation; evidence has shown that individuals presenting with OSFED share a common genetic basis with other EDs (Fairweather-Schmidt & Wade, Reference Fairweather-Schmidt and Wade2014); twin studies have shown that 50% of genetic factors related to AN and BN contribute to both disorders (Bulik et al., Reference Bulik, Thornton, Root, Pisetsky, Lichtenstein and Pedersen2010). Further, cross-disorder genetic studies suggest that genetic influences operate across the DSM diagnostic boundaries. For example, a meta-analysis of GWAS showed that GAD, panic, agoraphobia, social anxiety, and specific phobia identified common variants associated with a higher order anxiety factor, consistent with the HiTOP fear subfactor (Conway et al., Reference Conway, Forbes, Forbush, Fried, Hallquist, Kotov, Mullins-Sweatt, Shackman, Skodol, South and Sunderland2019). In a study on general psychopathology, Allegrini et al. (Reference Allegrini, Cheesman, Rimfeld, Selzam, Pingault, Eley and Plomin2020) found that different behavioral phenotypes load on a common p-factor that is highly heritable and stable over time.
When applied to categorical phenotypes, polygenic risk scores (PRSs) appear to largely capture transdiagnostic genetic influences (Waszczuk et al., Reference Waszczuk, Eaton, Krueger, Shackman, Waldman, Zald, Lahey, Patrick, Conway, Ormel and Hyman2020). The cross-disorder genetic overlap, such as correlations between AN and bipolar disorder, may indicate the existence of a higher p-factor. Various advanced statistical genetic techniques are actively being used to explore these genetic relationships between traits, notably genomic structural equation modelling [gSEM, (Grotzinger et al., Reference Grotzinger, Rhemtulla, de Vlaming, Ritchie, Mallard, Hill, Ip, Marioni, McIntosh and Deary2019)].
Discussion and Conclusion
There is a narrow divide between subthreshold and full-threshold EDs. Diagnostic criteria and markers are essential, but the severity of the disease, comorbid psychiatric disorders, and outcomes are similar between sub- and full-threshold presentation of EDs. EDs are neither entirely independent nor overlapping with one another, and a dimensional model of classification could represent the heterogenous presentation of each ED in a more holistic manner. Considering the flux between OSFED, AN, BN, and BED in some cases, more attention should be given to the behaviors and symptomatology to treat someone with an ED. For example, someone with PD, a subthreshold variant of OSFED, engages in repeated harmful behaviors that could have a variety of consequences (Stice et al., Reference Stice, Marti, Shaw and Jaconis2009) but would be excluded in genetic studies that only examine those with full-threshold EDs (e.g., the Eating Disorder Genetics Initiative), even though the core psychopathology may be the same.
It is essential to consider how we classify participants in genetic research by disorder. It is vital to collect the phenotype of each patient. However, someone with AN might have a parent or relative who had BN, but their biological predispositions (genetics) combined with environmental factors led them to present as AN. It may be more important to look at the genetic predispositions to what behaviors develop rather than the illness itself. What distinguishes someone from AN-BP and BN is the weight of the patient or BMI, but the symptomatology is very similar. Therefore, a drug developed for AN-BP may be just as effective for BN. Questions arise as to why distinguishing AN-BP from BN is important in genetic research if we know that these genetic predispositions may present almost identical behavioral phenotypes.
Looking at a more dimensional model is critical to how we move forward in evaluating genetics, nosology, and general etiology of EDs. There is already a significant amount of evidence demonstrating the clinical utility of HiTOP models for psychiatric illness. There is evidence that GWAS studies of HiTOP dimensions can identify more genetic risk loci than studies of DSM-5 disorders and could produce more precise PRSs. When considering environmental risk factors versus genetics, several studies show that environmental risk factors influence HiTOP spectra rather than a specific disorder. EDs are complex illnesses and thus it is important to consult individuals with lived or living experience who may provide important insights into whether a label of AN, BN or BED or a more hierarchical model that identifies specific behaviors or syndromes may better encapsulate their experience.
Financial support
JL gratefully acknowledges financial support for this publication by the Fulbright U.S. Student Program.
Competing interests
None.
Open access
