Personality correlates of eating pathology severity and subtypes in the National Comorbidity Survey Adolescent Supplement (NCS-A)

Abstract

Objective:

We examined how personality traits are associated with eating pathology across a range of severities and symptom subtypes.

Method:

The National Comorbidity Survey Adolescent Supplement (N = 10,148) was used to compare personality across severities (i.e., clinical Eating Disorders [ED], subclinical disordered eating, pre-clinical weight concerns, or no weight/eating pathology [No W/EP]), subclinical subtypes (i.e., Binge Eating Only, Binge-Purge, Binge-Restrict, Restrict-Purge, or No W/EP), and clinical subtypes (i.e., Anorexia Nervosa [AN]), Bulimia Nervosa [BN], Binge Eating Disorder [BED], internalizing disorders, or no ED or internalizing disorder) of eating pathology.

Results:

More severe eating pathology was associated with more extreme personality trait endorsements. Impulsivity-related traits did not consistently distinguish binge eating/purging from restricting subtypes, although Behavioral Disinhibition differentiated adolescents with BN or BED from AN.

Conclusion:

Personality traits related to affectivity and impulsivity were more consistently associated with severity rather than subtype of eating pathology.

Introduction

Roughly one third of adolescent males and more than half of adolescent females report engaging in subclinical disordered eating behaviors such as binge eating, extreme dieting, and purging (Croll, Neumark-Sztainer, Story, & Ireland, 2002), making these behaviors an important concern in healthcare and educational settings that serve adolescents. Disordered eating behaviors are positively associated with depression, anxiety, thoughts of suicide, and self-harm (Lock, Reisel, & Steiner, 2001), and are robust risk factors for the later development of clinical Eating Disorders (EDs; Bryla, 2003; Houeto, Magnard, Dalley, Belin, & Carnicella, 2016), which, in turn, have high rates of mortality and physical and psychiatric morbidity (Arcelus, Mitchell, Wales, & Nielsen, 2011; Keel, Dorer, Franko, Jackson, & Herzog, 2005). Early- to mid-adolescence is a critical period of risk for the development of disordered eating behaviors and EDs (e.g., Jones, Bennett, Olmsted, Lawson, & Rodin, 2001; Swanson, Crow, Le Grange, Swendsen, & Merikangas, 2011; Volpe et al., 2016), suggesting that identifying risk factors for eating pathology during this period is a research and health priority.

Personality traits hold an important role in the etiology, symptomatology, and maintenance of both subclinical disordered eating behaviors and clinical EDs (see Cassin & von Ranson, 2005; Farstad, McGeown, & von Ranson, 2016; Lilenfeld, Wonderlich, Riso, Crosby, & Mitchell, 2006, for reviews). A sizable body of literature has demonstrated that certain personality traits distinguish adults with and without eating pathology (without controlling for other types of psychopathology; Farstad et al., 2016; Liley, Watson, Seah, Priddis, & Kane, 2013). Certain personality traits may also differentiate between clinical ED diagnostic groups and between subclinical disordered eating subtype groups (Atiye, Miettunen, & Raevuori-Helkamaa, 2015; Cassin & von Ranson, 2005; Claes, Vandereycken, & Vertommen, 2005). However, very little research has examined these associations in representative samples of adolescents. Given that personality traits typically become more stable in the transition from adolescence to adulthood (e.g., Borghuis et al., 2017; Shiner, Allen, & Masten, 2017), it is unclear if the associations between personality traits and eating pathology observed among adults would replicate in adolescents. Moreover, in line with an increasing movement in the field toward dimensional conceptualizations of psychopathology (Hudziak, Achenbach, Althoff, & Pine, 2007; Sanislow et al., 2010; Widiger, 2005), it would be informative to understand how personality traits relate to eating pathology across a gradient of severity, ranging from adolescents who are exhibiting pre-clinical weight concerns to those who meet full diagnostic criteria for an ED, as well as across groupings of symptom subtypes. Understanding how personality traits differentiate eating pathology severity and symptom expressions can inform etiological models of eating pathology, as well as early intervention and prevention efforts.

Robust Personality Correlates of Eating Pathology: Neuroticism, Negative Emotionality, and Impulsivity

Recent systematic and meta-analytic reviews of the literature suggest that neuroticism and impulsivity are positively and robustly associated with the presence of eating pathology (Cassin & von Ranson, 2005; Farstad et al., 2016). Among adult females (≥ 18 years old), those with Bulimia Nervosa (BN), Anorexia Nervosa-Bingeing/Purging subtype (AN-B/P), Anorexia Nervosa-Restrictive subtype (AN-R), and Binge Eating Disorder (BED) score higher on trait neuroticism measures relative to those without an ED diagnosis (without controlling for other psychopathology; Izydorczyk, 2012; Tasca et al., 2009). In adolescents and young adults aged 12–21, higher neuroticism was associated with increased odds of having a diagnosed clinical ED (Gual et al., 2002). Similarly, in community samples aged 13–19, neuroticism was higher among adolescents with a history of BED, subclinical binge eating (Lee-Winn, Townsend, Reinblatt, & Mendelson, 2016), and loss of control eating (Goldschmidt, Lavender, Hipwell, Stepp, & Keenan, 2017), relative to those without any eating pathology, but did not appear to be related to dietary restraint (Thamotharan, Hubbard, & Fields, 2015). These results suggest that neuroticism may differentiate between eating pathology that is primarily characterized by binge eating or purging from eating pathology that is characterized primarily by restricting. Conclusions about these relationships among adolescents should be considered tentative at this time. Moreover, it is important to interpret these findings in view of recent dimensional models asserting that neuroticism functions as, or contributes to, an underlying transdiagnostic vulnerability to many types of psychopathology including internalizing disorders (e.g., anxiety and depression), and externalizing problems (e.g., substance use and conduct problems; Caspi et al., 2014; Griffith et al., 2010). Thus, whether neuroticism confers unique risk for eating pathology above and beyond the risk it confers for general psychopathology remains an open question.

Negative emotionality is a trait that is closely related to neuroticism, and encompasses constructs such as stress reactivity, alienation, negative engagement, aggression, and negative affect (Peterson et al., 2010). A recent systematic literature review found that negative emotionality differentiates young adults and adults (i.e., ≥ 18 years old) with and without eating pathology, and further differentiates between clinical ED diagnostic groups (Farstad et al., 2016). Specifically, the review concluded that adults with BN or AN-B/P score higher on measures of negative emotionality relative to adults with BED or AN-R (Farstad et al., 2016). Only one known study has examined this association in adolescents and found that negative emotionality was positively associated with subclinical disordered eating engagement in adolescent males (Keel, Klump, Leon, & Fulkerson, 1998).

Impulsivity and sensation seeking (i.e., a facet of impulsivity characterized by a tendency to pursue novel and thrilling experiences) have also emerged as robust correlates of eating pathology and appear to differentiate between eating pathology symptom expressions. Both impulsivity and sensation seeking are elevated in adults and adolescents with eating pathology characterized by binge eating (i.e., subclinical binge eating behavior, BN or BED) and lower in those with eating pathology characterized by restricting (i.e., AN-R), relative to adults and adolescents without eating pathology (Lee-Winn et al., 2016; see Waxman, 2009 for a review). Moreover, evidence suggests that higher impulsivity further differentiates adolescents and adults with an ED with significant purging behavior (e.g., AN-B/P, BN-Purging subtype) from those with non-purging ED subtypes (e.g., AN-R, BN-Non-purging subtype; Claes, Nederkoorn, Vandereycken, Guerrieri, & Vertommen, 2006; Favaro et al., 2005; Waxman, 2009). Several literature reviews have now concluded that there is good evidence suggesting that eating pathology characterized by binge eating or purging is associated with a tendency to act on urges without premeditation and a tendency to seek out novelty (Cassin & von Ranson, 2005; Farstad et al., 2016; Waxman, 2009). Research regarding how neuroticism, negative emotionality and impulsivity relate to severity of eating pathology, particularly in adolescents, remains sparse.

Tentative Personality Correlates of Eating Pathology: Extraversion, Behavioral Disinhibition, and Aggression

While several additional personality traits have been linked to EDs and disordered eating behavior in young adult and adult samples, studies examining these links in adolescents are limited and, in some cases, diverge from the young adult and adult literature. For instance, two studies of young adults found that extraversion is lower among those with AN-R, AN-B/P, or BN, relative to young adults without eating pathology (without controlling for other psychopathology; Massoubre et al., 2005; Tasca et al., 2009), whereas a study of adolescents did not find an association between extraversion and disordered eating behaviors (Gual et al., 2002). One review (Farstad et al., 2016) noted that discrepancies in the literature pertaining to extraversion and eating pathology may be due to differences between studies that examine subclinical disordered eating symptoms (i.e., Gual et al., 2002) rather than clinical EDs, further highlighting the importance of examining personality across a range of eating pathology severities. Other research found that behavioral disinhibition (i.e., difficulties withholding a prepotent behavioral response) tends to be higher among young adults who display more binge eating or purging symptoms relative to young adults who display primarily restricting symptoms (Claes et al., 2006; Claes, Robinson, Muehlenkamp, Vandereycken, & Bijttebier, 2010). Antisocial aggression, which may be related to behavioral disinhibition, has been found to be higher in adult patients with BN compared to those without an ED (without controlling for other psychopathology; Bruce, Steiger, Koerner, Israel, & Young, 2004). To our knowledge, very few studies have examined how behavioral disinhibition or aggression relates to eating pathology in adolescents, nor have they examined how these traits are expressed across a range of eating pathology severities.

Research Gaps

Although there is a sizable body of literature examining personality correlates of eating pathology, the conclusions and clinical directions that can be drawn from these studies remain limited for several reasons. First, while many studies have examined associations between personality traits and eating pathology in community samples of adults, very few have examined these associations in community-based adolescents where a full range of eating pathology severities are likely to be observed, therefore circumventing the floor and ceiling effects on personality inventories that are often occur in clinical samples (Bach, Sellbom, & Simonsen, 2018; Gutierrez et al., 2017; Pedersen, Arnevik, Hummelen, Walderhaug, & Wilberg, 2017). Moreover, knowing whether the associations between personality and eating pathology in adults are also observed in adolescents can inform etiological models of eating pathology and, in combination with existing literature, may help to illuminate the role of personality in eating pathology across the lifespan. Research that compares how personality traits are associated with eating pathology across severity (e.g., pre-clinical weight concerns, versus subclinical disordered eating behaviors, versus clinical EDs) as well as distinct symptom expressions (e.g., binging/purging versus restricting) would be particularly useful in informing early detection and targeted intervention programs for adolescents (e.g., Hill, Peck, Wierenga, & Kaye, 2016; Turner et al., 2014). Second, while many studies compare samples with eating pathology against age-matched samples without any eating pathology, few studies adequately account for the association of personality traits with general psychopathology (i.e., the “p factor”; Caspi et al., 2014), leaving open the question of whether traits such as neuroticism and impulsivity confer unique risk for EDs, or whether they function as transdiagnostic vulnerability factors for a broader range of psychopathology. Comparing individuals with an ED against meaningful clinical comparison groups, such as individuals with internalizing disorders, would help to elucidate the unique or non-specific role of personality in eating pathology.

The Current Study

The current study examined associations between personality traits and various eating pathology severity and subtype groups in a major epidemiological study with a nationally-representative, community sample of adolescents (the National Comorbidity Survey Replication Adolescent Supplement [NCS-A] 2001–2004; Kessler et al., 2009; Merikangas, Avenevoli, Costello, Kortez, & Kessler, 2009). Past work using the NCS-A data to examine personality correlates of eating pathology found that higher neuroticism, impulsivity, and a combination of both traits, were associated with increased odds of experiencing lifetime BED or subclinical binge eating (Lee-Win et al., 2016). This study extends past work in three ways. First, we examined associations between personality traits and eating pathology severities by comparing adolescents who (a) met full diagnostic criteria for a clinical ED, (b) engaged in subclinical disordered eating behavior, (c) had pre-clinical weight concerns, or (d) had neither weight concerns nor eating pathology (No W/EP). Second, we examined whether personality traits were uniquely related to specific eating pathology symptom expressions by comparing adolescents who reported engagement in subclinical (a) binge eating (Binge-Only), (b) binge eating with purging (Binge-Purge), (c) binge eating with restricting (Binge-Restrict), restricting with purging (Restrict-Purge), or No W/EP. Third, we examined whether personality traits were associated with specific clinical EDs, and whether these associations were unique to EDs relative to other forms of psychopathology, by comparing adolescents who met full diagnostic criteria for (a) AN, (b) BN, (c) BED, (d) an internalizing disorder but no ED, or (e) neither an ED nor an internalizing disorder. Finally, we extend previous work by considering all seven personality traits that were included in the NCS-A protocol: Neuroticism-Anxiety, Impulsivity-Sensation Seeking, Emotionality, Aggression-Hostility, Sociability, Need for Activity, and Behavioral Disinhibition. Consistent with past research, we expected that adolescents with clinical EDs or subclinical disordered eating behavior would score significantly higher on measures of Neuroticism-Anxiety, Impulsivity-Sensation Seeking, and Emotionality, relative to adolescents with pre-clinical weight concerns or No W/EP. We further hypothesized that Impulsivity-Sensation Seeking and Behavioral Disinhibition would be higher among adolescents who experienced eating pathology that was characterized by binge eating (i.e., those in the BN, BED, Binge-Only and Binge-Purge groups), relative to those with eating pathology that did not include significant binge eating (i.e., those in the AN or Restrict-Purge groups). No hypotheses were made pertaining to Aggression-Hostility, Sociability (a purported facet of extraversion), or Need for Activity (related to the liveliness facet of extraversion), nor did we have any hypotheses regarding the differences between the ED and internalizing groups due to the paucity of studies investigating these objectives.

Method

Participants

The NCS-A was a nationally representative study of 10,148 adolescents (51% male), ages 13–18 years old, from the U.S. (Kessler et al., 2009; Merikangas et al., 2009). The NCS-A was administered between 2001–2004 in a dual-frame sample, including both a household (n = 904; 85.9% response rate) and school subsample (n = 9,244; 74.7% response rate) of adolescents (Kessler et al., 2009). Parents provided written informed consent and adolescents provided written informed assent of their participation in the NCS-A study. The NCS-A study was conducted in accordance with the Code of Ethics of the World Medical Association, and all procedures for the study were approved by the human research ethics boards of Harvard Medical School and the University of Michigan at Ann Arbor. The goals of the NCS-A were to estimate prevalence rates of mental health disorders (in accordance with Diagnostic and Statistical Manual of Mental Disorders 4th ed. [DSM-IV] criteria; American Psychiatric Association [APA], 2000), health-related behaviors, and mental health service utilization of U.S. adolescents. Details regarding the NCS-A study design, field procedures, measurement strategy, and sample weightings are described elsewhere (see Kessler et al., 2009; Merikangas et al., 2009). We received approval to access the restricted NCS-A data from the Inter-University Consortium for Political and Social Research at the University of Michigan (Kessler, 2017).

Measures

Personality traits.

Adolescents responded to 45 items that assessed a broad range of personality traits including: 1) Impulsivity-Sensation Seeking, i.e., lack of planning, acting without thinking, novelty- or excitement-seeking (11 items; ⍺ = .76); 2) Neuroticism-Anxiety, i.e., frequency of emotional upset, fearfulness, lack of self-confidence, and worry (7 items; ⍺ = .72); 3) Aggression-Hostility, i.e., expressiveness of aggression, rudeness, and temper (7 items; ⍺ = .80); 4) Need for Activity, i.e., restlessness, need to engage in activity when there is none, preference for challenging and high energy work, and busyness (7 items; ⍺ = .80); 5) Sociability, i.e., preference for social interactions, comfort with social events, and intolerance of isolation (7 items; ⍺ = .67); 6) Emotionality, i.e., emotional lability, perceptions of emotional intensity, and actions under strong emotion (4 items; ⍺ = .76); and 7) Behavioral Disinhibition, i.e., resistance to temptation and behavioral restraint (Lee-Win et al., 2016; Merikangas et al., 2009) (2 items; ⍺ = .63). Each item was rated on a 4-point Likert-type scale (1 = Very true; 4 = Not at all true), and items were reverse scored where necessary so that total scores reflect more extreme trait endorsement. Items for the NCS-A personality questionnaire were drawn from various versions of the Zuckerman-Kuhlman Personality Questionnaire (ZKPQ; Aluja et al., 2006; Zuckerman, Kuhlman, & Camac, 1988; Zuckerman, Kuhlman, Joireman, Teta, & Kraft, 1993; Zuckerman, Kuhlman, Thornquist, & Kiers, 1991; Zuckerman, 2002), and supplemented with additional items that assessed Emotionality (Eisenberg & Spinrad, 2004) and Behavioral Disinhibition (Rosenberg & Kagan, 1989). The short and long versions of the ZKPQ have demonstrated good cross-cultural adaptability, construct validity, and concurrent and discriminant validity, and adequate to good internal consistency and reliability (see Aluja, Garcia, & Garcia, 2003; Aluja, Garcia, & Garcia, 2004; Zuckerman & Cloninger, 1996; Zuckerman et al., 1993; Wang, Hu, Zheng, & Liu, 2017; Wu, Wang, Du, Li, Jiang, & Wang, 2000). Correlations between the personality subscales are reported in Table 1.

Table 1.

Pearson’s bivariate correlations between all personality scales included in NCS-A.

	N-Anx	Agg-Host	Act	Soc	Emo	Behav Dis
Imp-SS	.159**	.411**	.162**	.264**	.303**	.471**
N-Anx		.211**	-.025*	-.248**	.589**	.321**
Agg-Host			.003	.027**	.410**	.379**
Act				.226**	.074**	.129**
Soc					-.108**	.017
Emo						.460**

Note.

^*=correlation is significant at p <.05.

^**=correlation is significant a p <.01. Imp-SS = Impulsivity-Sensation Seeking, N-Anx = Neuroticism-Anxiety, Agg-Host = Aggression-Hostility, Act = Need for Activity, Soc = Sociability, Emo = Emotionality, Behav Dis = Behavioral Disinhibition.

Internalizing Disorders.

Adolescents completed the World Health Organization - Composite International Diagnostic Interview (CIDI) Version 3.0 (Kessler & Ustun, 2004) to assess lifetime presence of DSM-IV internalizing disorders (APA, 2000). The CIDI was modified for this study to ensure reasonable validity and applicability of the interview within an adolescent sample (see Kessler & Ustun, 2004; Merikangas et al., 2009). Reliability estimates for this version of the CIDI have not been reported. For the purposes of our analyses, we coded adolescents as having an internalizing disorder if they met full diagnostic criteria for any of the following disorders: Agoraphobia with or without Panic Disorder, Generalized Anxiety Disorder, Dysthymia, Major Depressive Disorder, Panic Disorder, Posttraumatic Stress Disorder, Separation Anxiety Disorder, Social Phobia, or Specific Phobia Disorder.

Eating Disorder Diagnoses.

The CIDI (Kessler & Ustun, 2004) was also used to assess DSM-IV diagnostic criteria for AN or BN. The CIDI provides information about BED with the following deviations from the DSM-IV criteria: first, the CIDI did not directly assess loss of control during binge eating (which is required by the DSM-IV criteria) and instead used the following item to approximate this criterion - “Did you often get upset both during and after the binges and think that your eating was out of your control?” (Kessler et al., 2009; Merikangas et al., 2009; Swanson et al., 2011); second, the CIDI asked if binge episodes occurred at least twice a week for three months or more, which is a briefer period than required by the DSM-IV criteria (i.e., 2 days a week for 6 months), but is more stringent than the DSM-5 criteria (i.e., 1 day a week for 3 months; APA, 2013) for BED. Diagnostic hierarchies were applied in the NCS-A protocol such that if an individual met criteria for more than one eating disorder at the same time, a diagnosis of AN superseded BN, and BN superseded BED (i.e., if an adolescent met diagnostic criteria for both AN and BN during the same year, only a diagnosis of AN was recorded). More than one eating disorder diagnosis was assigned only if there was evidence for a temporal distinction between the two disorders (Swanson et al., 2011).

Disordered Eating Behaviors.

The current study used various items from the CIDI to examine adolescents’ engagement in specific subclinical disordered eating behaviors (i.e., maladaptive eating behavior that did not meet full diagnostic criteria for an ED). Formation of the disordered eating groups was informed both by the literature on common presentations of eating concerns (Stice, Marti, Shaw, & Jaconis, 2009) as well as the skip logic of the CIDI. Two stem items were presented to all participants: the first inquired about the presence of weight preoccupations (“...was there ever a time in your life when you worried a great deal or strongly feared being too fat or overweight?”), while the second inquired about the presence of binge eating episodes (“...where a person eats a large amount of food during a short time like two hours. By ‘a large amount’ I mean eating so much food that it would be like eating two or more entire meals in one sitting, or eating so much of one particular food -- like candy or ice cream -- that it would make most people feel sick. With that definition in mind, did you ever have a time in your life when you went on eating binges at least twice a week for several months or longer?”). Participants who endorsed the first stem item were then asked whether these concerns persisted despite a low body weight, and, if so, were asked about their engagement in restricting and purging behaviors (“did you do things to keep your body weight low, such as dieting or exercising?”). Participants who endorsed the second stem item were asked whether these binge episodes met clinical specifiers (e.g., characterized by rapid eating, physical discomfort, negative affect following binge episode), and, if so, whether they engaged in restricting (i.e., inhibited eating for 8 or more hours following a binge episode in order to control weight) or purging (i.e., self-induced vomiting, laxative/enema use, over-exercise, or taking water pills, diuretics, or weight control medication use after a binge episode in order to control weight) in response to binge eating. Based on this item administration protocol, we were able to examine the following presentations of subclinical disordered eating: (1) Weight Concerns (i.e., endorsed the first stem item, but did not endorse engaging in restricting and/or purging behaviors, and did not endorse the second stem item), (2) Restrict-Purge (i.e., endorsed the first stem item, endorsed having low weight and engaging in restricting and/or purging behaviors, but did not endorse the second stem item), (3) Binge-Only (i.e., endorsed the second stem item and denied any restricting and purging in response to binge eating), (4) Binge-Purge (i.e., endorsed the second stem item, endorsed purging, but denied any restricting in response to binge eating), (5) Binge-Restrict (i.e., endorsed the second stem item, endorsed restricting, but denied any purging in response to binge eating), or (6) No Weight Concerns or Eating Pathology (No W/EP; participants said no to both stem items, and therefore did not receive any follow-up items). A seventh possible group, binge eating with both restricting and purging, was not included in our analyses due to low numbers. Please see Figure 1 for a visual representation of the NCS-A skip logic and how DE groups were formed.

Figure 1.

Visual representation of NCS-A skip logic and how various eating pathology-realated groups were formed.

Data Analysis

All analyses for the current study were conducted on weighted data to account for the stratified sampling of the NCS-A and to ensure that results can be assumed to reflect the demographics of adolescents according to 2000 U.S. census data. Consistent with the three study aims, we examined differences in seven personality traits in three sets of comparisons: 1) across levels of severity - comparing adolescents who met diagnostic criteria for a clinical ED, who engaged in subclinical disordered eating behavior but did not meet criteria for an ED, who had pre-clinical weight concerns but did not endorse any disordered eating behavior, or who had No W/EP; 2) across unique patterns of symptom expressions - comparing adolescents who engaged in subclinical Binge-Only, Binge-Purge, Binge-Restrict, Restrict-Purge, or No W/EP; and (3) across clinical diagnoses - comparing adolescents who met criteria for AN, BN, BED, an internalizing disorder but no ED, or adolescents with neither an internalizing disorder nor an ED. To examine these differences, we conducted a series of ANCOVAs using the Complex Samples General Linear Model module in IBM SPSS Statistics Version 25.0. In all analyses, the participant groups were entered as the independent variable and each personality trait was entered as the dependent variable. Given that prevalence rates of clinical and subclinical eating pathology are known to differ by gender, race, and age (Coffino, Udo, & Grilo, 2019; Croll et al., 2002; Pritchard, 2008; Stice, Marti, & Rohde, 2013), these three variables were included as covariates. Because the Complex Samples module did not allow for multivariate tests, we used the Holm’s Sequential Bonferroni correction procedure (Eichstaedt, Kovatch, & Maroof, 2013) to set thresholds for significance and reduce family-wise error. The traditional Bonferroni procedure has been criticized for being overly conservative and increasing the possibility of a Type II error, while the Holm’s-Bonferroni correction procedures are applied stepwise (i.e., becomes more conservative as the number of tests increases), therefore controlling for family-wise error while increasing statistical power (Eichstaedt et al., 2013). A total of 21 ANCOVAs were conducted: one family of seven tests (i.e., seven personality traits) was run for each of the three sets of comparisons; significance thresholds for each main effect ranged from p < 0.007 to p < 0.05. The same correction procedure was applied to obtain significance thresholds for planned post-hoc pairwise comparisons, which were conducted where significant main effects were detected. For Aim 1, there were six post-hoc pairwise comparisons per significant main effect; significance thresholds ranged from p < 0.008 to p < 0.05. For Aims 2 and 3, there were ten post-hoc pairwise comparisons per significant main effect; significance thresholds ranged from p < 0.005 to p < 0.05.

Results

The NCS-A sample included 282 adolescents who met diagnostic criteria for a clinical ED in their lifetime (M_age = 15.47 , SD = 1.47; 73% female, 48% White), 424 who engaged in subclinical disordered eating but did not meet criteria for an ED (M_age = 15.34, SD = 1.55 ; 55% female, 49% White), 2,382 adolescents who had pre-clinical weight concerns but did not engage in any disordered eating behaviors (M_age = 15.24, SD = 1.51; 69% female, 60% White), and 7,008 with No W/EP (M_age = 15.14, SD = 1.50; 44% female; 55% White). Table 2 presents contrasts on personality traits across eating pathology severity, revealing significant main effects for Neuroticism-Anxiety, Emotionality, Impulsivity-Sensation Seeking, Behavioral Disinhibition, and Aggression-Hostility. Corrected significant p-values from the post-hoc comparisons showed that adolescents with No W/EP scored significantly lower than all of the other eating pathology-related groups on all five of these personality traits. Moreover, adolescents who met criteria for a clinical ED scored higher on these five personality traits relative to adolescents with pre-clinical weight concerns, and scored higher on Neuroticism-Anxiety, Emotionality and Behavioral Disinhibition relative to adolescents with subclinical disordered eating, but did not differ from adolescents with subclinical disordered eating on Impulsivity-Sensation Seeking or Aggression-Hostility. Adolescents with subclinical disordered eating, in turn, scored higher than adolescents with pre-clinical weight concerns on measures of Impulsivity-Sensation Seeking and Behavioral Disinhibition, but not Neuroticism-Anxiety, Emotionality, or Aggression-Hostility. There were no significant differences between the groups (i.e., main effects) for Need for Activity (Wald F [3,40] = .82, p = .489) or Sociability (Wald F [3,40] = 2.38, p = .090).

Table 2.

GLM significant main effects and post-hoc contrast results for personality measures between adolescents with a clinical ED, subthreshold disordered eating, weight concerns, and no weight concerns or eating pathology.

					95% CI		p (Cohen’s d)
DV	Wald F(3,40), p	IV	Mean	Std. Error	Lower	Upper	WC	DE	ED
N-Anx	119.73, <.001*	No W/EP	14.00	.144	13.71	14.29	<.001* (.202)	<.001* (.226)	<.001* (.473)
		ED	18.33	.279	17.77	18.90	<.001* (.393)	<.001* (.328)
		DE	16.27	.367	15.53	17.01	.501 (.038)
		WC	16.01	.128	15.75	16.26
Emo	115.67, <.001*	No W/EP	8.85	.085	8.67	9.01	<.001* (.227)	<.001* (.212)	<.001* (.460)
		ED	11.40	.197	11.00	11.80	<.001* (.576)	<.002* (.248)
		DE	10.23	.282	9.66	10.80	.447 (.055)
		WC	10.00	.093	9.81	10.19
Imp-SS	26.53, <.001*	No W/EP	26.48	.166	26.15	26.81	<.001* (.094)	<.001* (.220)	<.001* (.266)
		ED	29.42	.426	28.56	30.27	.001* (.212)	.391 (.065)
		DE	28.95	.365	27.57	29.37	.001* (.153)
		WC	27.58	.188	27.20	27.96
Behav Dis	39.83, <.001*	No W/EP	3.90	.042	3.82	3.99	<.001* (.135)	<.001* (.322)	<.001* (.410)
		ED	5.08	.122	4.83	5.32	<.001* (.320)	.048* (.137)
		DE	4.81	.092	4.62	4.99	<.001* (.212)
		WC	4.37	.051	4.27	4.48
Agg-Host	32.45, <.001*	No W/EP	18.06	.151	17.75	18.36	<.001* (.166)	<.001* (.264)	<.001* (.290)
		ED	20.41	.402	19.60	21.22	.025* (.156)	.766 (.023)
		DE	20.24	.379	19.48	21.09	.018 (.125)
		WC	19.28	.163	18.95	19.60

Note.

^*=significant at respective threshold after Holm’s Sequential Bonferroni Procedure. No W/EP = adolescents who did not endorse any weight concerns nor eating pathology, ED = adolescents who met lifetime diagnostic criteria for a clinical Eating Disorder including Anorexia Nervosa, Bulimia Nervosa, and Binge Eating Disorder, DE = adolescents who did not meet criteria for an Eating Disorder but endorsed engaging in subclinical disordered eating including Binge-Only, Binge-Restrict, Binge-Purge, and Restrict-Purge behaviors, WC = adolescents who endorsed having weight concerns but did not endorse engaging in any eating pathology. Imp-SS = Impulsivity-Sensation Seeking, N-Anx = Neuroticism-Anxiety, Agg-Host = Aggression-Hostility, Emo = Emotionality, Behav Dis = Behavioral Disinhibition. ED group n = 513,001 (weighted) or 282 (unweighted), DE group n = 776,629 (weighted) or 424 (unweighted), WC group = 4,618,437 (weighted) or 2,382 (unweighted), No W/EP group n =13,262,228 (weighted) or 7,008 (unweighted). Cohen’s d effect sizes were calculated using weighted means and standard deviations and unweighted sample sizes.

With respect to unique patterns of eating pathology symptom expression, the NCS-A sample contained 269 adolescents who engaged in Binge-Only behaviors (M_age = 15.25, SD = 1.52; 43% female; 48% White), 98 with Restrict-Purge behaviors (M_age = 15.52, SD = 1.60 ; 90% female, 68% White), 36 with Binge-Restrict behaviors (M_age = 15.36, SD = 1.61; 56% female; 22% White), 21 with Binge-Purge behaviors (M_age = 15.71, SD = 1.45; 48% female; 33% White), and 7,008 had No W/EP (M_age = 15.14, SD = 1.50; 44% female; 55% White). Table 3 presents contrasts on personality traits between these subclinical disordered eating groups, revealing significant main effects for Neuroticism-Anxiety, Emotionality, Impulsivity-Sensation Seeking, Behavioral Disinhibition and Aggression-Hostility. Corrected significant p-values from the post-hoc analyses demonstrate that adolescents with No W/EP scored significantly lower than adolescents who engaged in Restrict-Purge, Binge-Restrict, and Binge-Only behaviors on all of these traits except Aggression-Hostility, and did not differ from adolescents who engaged in Binge-Purge behaviors on any of the traits. Adolescents who engaged in Binge-Only scored significantly lower on Neuroticism-Anxiety, Emotionality, and Behavioral Disinhibition relative to adolescents with Restrict-Purge behavior. Finally, adolescents with Binge-Purge behavior scored significantly lower on Emotionality relative to adolescents with Restrict-Purge behavior. There were no main effects for Sociability (Wald F [4,39] = .43, p = .786) or Need for Activity (Wald F (4,39) = 1.19, p = .327).

Table 3.

GLM significant main effects and post-hoc contrast results for personality measures between those who engaged in subclinical Binge-Only, Binge-Purge, Binge-Restrict and Restrict-Purge disordered eating behaviours, and no weight concerns or eating pathology.

					95% CI		p (Cohen’s d)
DV	Wald F(4, 39), p	IV	Mean	Std. Error	Lower	Upper	Restrict-Purge	Binge- Restrict	Binge- Purge	Binge- Only
N-Anx	9.23, <.001*	No W/EP	14.03	.166	13.69	14.36	<.001* (.358)	<.003* (.318)	.020 (.270)	<.001* (.138)
		Binge Only	15.50	.348	14.80	16.20	.004* (.374)	.061 (.320)	<.252 (.250)
		Binge-Purge	16.87	1.149	14.54	19.19	.381 (.181)	.652 (.104)
		Binge-Restrict	17.50	1.123	15.23	19.76	.657 (.080)
		Restrict-Purge	18.09	.805	16.46	19.71
Emo	7.29, <.001*	No W/EP	8.84	.095	8.64	9.02	<.001* (.390)	.001* (.371)	.261 (.088)	.002* (.143)
		Binge Only	9.75	.258	9.23	10.27	.004* (.354)	.038 (.346)	.437 (.122)
		Binge-Purge	9.35	.409	8.53	10.17	<.001* (.505)	.028 (.585)
		Binge-Restrict	11.17	.663	9.83	12.51	.649 (.081)
		Restrict-Purge	11.58	.602	10.37	12.80
Imp-SS	17.43, <.001*	No W/EP	26.42	.187	26.04	26.79	<.001* (.315)	<.002* (.322)	.978 (.005)	<.001* (.165)
		Binge Only	28.45	.462	27.51	29.38	.099 (.237)	.122 (.257)	.365 (.234)
		Binge-Purge	26.36	2.20	21.92	30.79	.084 (.418)	.118 (.457)
		Binge-Restrict	30.32	1.16	27.99	32.66	.890 (.026)
		Restrict-Purge	30.54	1.00	28.52	32.56
Behav Dis	26.40, <.001*	No W/EP	3.90	.044	3.81	3.99	<.001* (.517)	.002* (.389)	.098 (.270)	<.001* (.229)
		Binge Only	4.58	.123	4.33	4.83	<.001* (.432)	.202 (.237)	.791 (.066)
		Binge-Purge	4.72	.479	3.75	5.68	.153 (.338)	.588 (.164)
		Binge-Restrict	5.08	.366	4.34	5.82	.489 (.154)
		Restrict-Purge	5.38	.169	5.04	5.72
Agg-Host	14.91, <.001*	No W/EP	18.08	.138	17.80	18.36	<.001* (.305)	.011 (.333)	.527 (.106)	<.001* (.216)
		Binge Only	20.26	.511	19.24	21.30	.460 (.097)	.439 (.132)	.047 (.410)
		Binge-Purge	17.06	1.568	13.90	20.23	.029 (.549)	.017 (.587)
		Binge-Restrict	21.29	1.203	18.87	23.73	.844 (.037)
		Restrict-Purge	21.02	.731	19.55	22.50

Note.

^*=significant at respective threshold after Holm’s Sequential Bonferroni Procedure. No W/EP = adolescents who did not endorse weight concerns nor eating pathology, Binge-Only = adolescents who endorsed binge eating behaviors but no restricting or purging behaviours, Binge-Purge = adolescents who endorsed binge eating followed by purging behaviors, Binge-Restrict = adolescents who endorsed binge eating followed by restricting behaviors, Restrict-Purge = adolescents who endorsed restricting and/or purging, but no binge eating behaviors. Imp-SS = Impulsivity-Sensation Seeking, N-Anx = Neuroticism-Anxiety, Agg-Host = Aggression-Hostility, Emo = Emotionality, Behav Dis = Behavioral Disinhibition. No W/EP group n = 1,326,228 (weighted) or 7,008 (unweighted), Binge-Only group n = 495,573 (weighted) or 269 (unweighted), Binge-Purge group n = 52,925 (weighted) or 21 (unweighted), Binge-Restrict group n = 50,342 (weighted) or 36 (unweighted), Restrict-Purge group n = 177,787 (weighted) or 98 (unweighted). Cohen’s d effect sizes were calculated using weighted means and standard deviations and unweighted sample sizes.

Of those with a clinical ED, 75% had a comorbid internalizing disorder. The most common internalizing disorders were, in order, Specific Phobia Disorder, Social Phobia, Major Depressive Disorder, Separation Anxiety, Dysthymia, Posttraumatic Stress Disorder, Agoraphobia with/without Panic Disorder, Agoraphobia, Generalized Anxiety Disorder, and Panic Disorder. With respect to unique clinical disorders, the NCS-A sample contained 34 adolescents who met diagnostic criteria for AN in their lifetime (M_age = 15.06 , SD = 1.59; 59% female; 62% White; 56% had an internalizing disorder), 86 who met diagnostic criteria for BN (M_age = 15.48, SD = 1.41; 79% female; 38% White; 81% had an internalizing disorder), 162 who met diagnostic criteria BED (M_age = 15.55, SD = 1.48; 73% female; 50% White; 76% had an internalizing disorder), 3,661 who met criteria for an internalizing disorder but had never met criteria for an ED (M_age = 15.28, SD = 1.50; 59% female; 52% White), and 6,205 had never met criteria for either a ED or an internalizing disorder (M_age = 15.11, SD =1.51 ; 46% Female; 58% White). Table 4 presents contrasts on personality traits across these clinical groups, revealing significant main effects for Neuroticism-Anxiety, Emotionality, Impulsivity-Sensation Seeking, Behavioral Disinhibition, Aggression-Hostility, and Sociability. Corrected significant p-values from the post-hoc analyses demonstrated that adolescents with neither internalizing nor eating disorders scored significantly lower than adolescents with an internalizing disorder, BN or BED on Impulsivity-Sensation Seeking, Neuroticism-Anxiety, Aggression-Hostility, Emotionality, and Behavioral Disinhibition, and lower than adolescents with an internalizing disorder or BED on Sociability. Adolescents with BED scored significantly higher on Neuroticism-Anxiety, Emotionality and Behavioral Disinhibition relative to adolescents with an internalizing disorder. Adolescents with BN scored significantly higher on Neuroticism-Anxiety, Emotionality, Behavioral Disinhibition, and Aggression-Hostility, relative to adolescents with an internalizing disorder. Finally, relative to adolescents with AN, adolescents with BN scored significantly higher on Aggression-Hostility and Behavioral Disinhibition, while adolescents with BED scored significantly higher on Behavioral Disinhibition only. There were no significant main effects for Need for Activity (Wald F [4,39] = 1.05, p = .394).

Table 4.

GLM significant main effects and post-hoc contrast results for personality measures between those with BED, BN, AN, an internalizing disorder, and without an ED or an internalizing disorder.

					95% CI			p (Cohen’s d)
DV	Wald F(4, 39), p	IV	Mean	Std. Error	Lower	Upper	INT	AN	BN	BED
N-Anx	283.58, <.001*	Healthy	13.36	.114	13.13	13.60	<.001* (.371)	.022 (.350)	<.001* (.780)	<.001* (.623)
		BED	18.16	.484	17.18	19.14	.003* (.239)	.122 (.309)	.258 (.181)
		BN	19.24	.619	18.00	20.50	<.001* (.403)	.026 (.490)
		AN	16.15	1.17	13.78	18.52	.783 (.045)
		INT	16.48	.129	16.21	16.74
Emo	387.90, <.001*	Healthy	8.42	.069	8.28	8.56	<.001* (.335)	.031 (.385)	<.001* (.708)	<.001* (.648)
		BED	11.41	.283	10.84	11.98	.002* (.202)	.351 (.195)	.689 (.058)
		BN	11.61	.359	10.89	12.34	.003* (.246)	.296 (.244)
		AN	10.51	.931	8.63	12.39	.885 (.022)
		INT	10.38	.103	10.17	10.59
Imp-SS	26.79, <.001*	Healthy	26.12	.157	25.81	26.43	<.001* (.157)	.162 (.158)	<.001* (.371)	<.001* (.336)
		BED	29.43	.507	28.41	30.46	.011 (.151)	.103 (.281)	.617 (.071)
		BN	29.93	.825	28.27	31.60	.029 (.194)	.088 (.327)
		AN	27.66	1.054	25.53	29.79	.752 (.038)
		INT	28.01	.193	27.62	28.40
Behav Dis	84.85, <.001*	Healthy	3.74	.042	3.65	3.83	<.001* (.284)	.034 (.226)	<.001* (.564)	<.001* (.512)
		BED	5.14	.157	4.82	5.45	.001* (.253)	.006* (.464)	.731 (.043)
		BN	5.22	.182	4.85	5.59	.001* (.306)	.005* (.564)
		AN	4.32	.257	3.80	4.84	.350 (.122)
		INT	4.57	.041	4.49	4.65
Agg-Host	45.67, <.001*	Healthy	17.75	.148	17.45	18.05	<.001* (.161)	.983 (.003)	<.001* (.411)	<.001* (.268)
		BED	20.29	.519	19.25	21.34	.220 (.085)	.048 (.381)	.059 (.195)
		BN	21.41	.510	20.39	22.44	.001* (.228)	.005* (.635)
		AN	17.78	1.124	15.51	20.04	.131 (.200)
		INT	19.54	.175	19.18	19.89
Soc	19.24, <.001*	Healthy	19.61	.078	19.46	19.77	<.001* (.158)	.594 (.063)	.209 (.140)	<.001* (.278)
		BED	18.05	.396	17.25	18.85	.314 (.060)	.015 (.442)	.331 (.121)
		BN	18.74	.680	17.37	20.11	.699 (.038)	.094 (.235)
		AN	19.92	.555	18.80	21.04	.014 (.232)
		INT	18.46	.137	18.19	18.74

Note.

^*=significant at respective threshold after Holm’s Sequential Bonferroni Procedure. AN = adolescents who met diagnostic criteria for Anorexia Nervosa, BN = adolescents who met diagnostic criteria for Bulimia Nervosa, BED = adolescents who met diagnostic criteria for Binge Eating Disorder, INT = adolescents who met diagnostic criteria for an internalizing disorder but did not meet criteria for an Eating Disorder, Healthy = adolescents who did not meet diagnostic criteria for an internalizing disorder nor an Eating Disorder. Imp-SS = Impulsivity-Sensation Seeking, N-Anx = Neuroticism-Anxiety, Agg-Host = Aggression-Hostility, Soc = Sociability, Emo = Emotionality, Behav Dis = Behavioral Disinhibition. Healthy group n = 11,689,121 (weighted) or 6,205 (unweighted) , BED group n = 296,227 (weighted) or 162 (unweighted), BN group n = 161,339 (weighted) or 86 (unweighted), AN group n = 55,434 (weighted) or 34 (unweighted) and INT group = 7,971,529 (weighted) 3,661 (unweighted). Cohen’s d effect sizes were calculated using weighted means and standard deviations and unweighted sample sizes.

Discussion

Past research suggests that some personality traits, such as neuroticism and impulsivity, reliably differentiate adults with and without eating pathology, and may further differentiate between eating pathology subtypes (Cassin & von Ranson, 2005; Farstad et al., 2016; Lilenfeld et al., 2006). This study builds on previous literature by comparing personality traits in a nationally-representative community sample of adolescents across a range of eating pathology severities, symptom expression subtypes, and distinct clinical ED groups. Our results illuminate three major findings. First, personality traits related to both affectivity (e.g., Neuroticism-Anxiety, Emotionality) and impulsivity/externalizing behaviors (e.g., Impulsivity-Sensation Seeking, Behavioral Disinhibition, Aggression-Hostility) differentiated adolescents with no eating pathology or pre-clinical weight concerns from those with any eating or pre-clinical weight concerns. Moreover, Neuroticism-Anxiety, Emotionality, and Behavioral Disinhibition distinguished adolescents with clinical EDs from those with less severe eating concerns (i.e., subclinical disordered eating or pre-clinical weight concerns), while traits related to impulsivity/externalizing behaviors (i.e., Impulsivity-Sensation Seeking and Behavioral Disinhibition) distinguished adolescents with subclinical disordered eating from those with pre-clinical weight concerns. Second, with respect to specific symptom expressions, our results suggest that adolescents who engage in restrictive eating and/or purging in the absence of significant binge eating (Restrict-Purge) had the most extreme scores across a range of personality traits, and were differentiated from adolescents who engage binge eating in the absence of restricting (Binge-Only, Binge-Purge) on Emotionality. Contrary to our expectations, however, we did not find consistent evidence that impulsivity-related traits differentiated specific symptom expressions. Third, our results show that while many personality traits can be conceptualized as non-specific correlates of psychopathology that are elevated in adolescents with either internalizing disorders or EDs, three traits (i.e., Neuroticism-Anxiety, Emotionality and Behavioral Disinhibition) may be specific features that differentiate adolescents with EDs from those with internalizing psychopathology, and one trait (Behavioral Disinhibition) further distinguished ED subtypes. We discuss the theoretical and clinical implications of each of these findings in turn.

With respect to personality traits that differentiate adolescents across a gradient of symptom severity, our results suggest that traits related to both affectivity (i.e., Neuroticism-Anxiety, Emotionality) and impulsivity (i.e., Behavioral Disinhibition) differentiated adolescents with clinical EDs from those with subclinical disordered eating, pre-clinical weight concerns, or without any weight or eating concerns. The distinction between clinical ED and subclinical disordered eating was mainly in traits related to affectivity (i.e., Neuroticism-Anxiety, Emotionality, as well as Behavioral Disinhibition), while the distinction between adolescents with subclinical disordered eating versus pre-clinical weight concerns was mainly related to impulsivity and externalizing traits (i.e., Impulsivity-Sensation Seeking and Behavioral Disinhibition). These findings extend previous results that neuroticism and impulsivity reliably differentiate individuals with eating pathology from healthy controls (Cassin & von Ranson, 2005; Farstad et al., 2016) by showing where in the severity continuum specific traits may be relevant. Together, these findings align with dimensional models of psychopathology (Sanislow et al., 2010) in suggesting that more extreme personality trait endorsements are useful in differentiating severity of eating difficulties along a gradient from mild to severe (e.g., Wolz et al., 2015; see also Hawkins & Clement, 1984; Blodgett Salafia & Lemer, 2012), and demonstrate how categorical groups may be used to complement a dimensional approach. Expanding these dimensional conceptualizations of eating pathology and its risk factors may facilitate a deeper understanding of risk trajectories from pre-clinical weight concerns to clinical EDs, diagnostic crossover between ED diagnoses (Stice et al., 2009), and co-occurrence of EDs with other forms of psychopathology (e.g., substance use disorders; anxiety disorders; various Personality Disorders; Carbaugh & Sias, 2010; Levinson et al., 2018; Farstad et al., 2016; Serpell, Hirani, Willoughby, Neiderman, & Lask, 2006). Although our results are consistent with a dimensional approach, it is important to note that the cross-sectional nature of the NCS-A did not allow us to disentangle whether extreme personality traits precede or follow the development of weight concerns and eating pathology.

With respect to personality differences across disordered eating subtypes, we found that few personality traits consistently differentiated between adolescents who engaged in disordered eating characterized primarily by binge eating, binge eating with compensatory behaviors (either restricting or purging), or purely restricting/purging. Indeed, the differences we did identify (i.e., higher Neuroticism-Anxiety, Emotionality and Behavioral Disinhibition in participants with Restrict-Purge behavior compared to Binge-Only, and higher Emotionality in Restrict-Purge relative to Binge-Purge) were inconsistent with our expectation that impulsivity-related traits would be key for differentiating over- and under-controlled eating presentations (Cassin & von Ranson, 2005). On the other hand, and consistent with past studies (Claes et al., 2006; Claes et al., 2010), we did find that Behavioral Disinhibition differentiated adolescents with AN from those with BN or BED, suggesting that ability to inhibit prepotent behavioral responses may play a role in differentiating symptom profiles at the clinical level. These results suggest that personality traits related to affectivity and impulsivity may be more important for differentiating the severity, rather than the symptom presentation, of eating pathology. Two notable limitations that may temper this interpretation, however, are the fact that the NCS-A skip logic did not permit examination of a purely restricting group, nor were we able to examine AN subtypes (e.g., AN-restricting versus AN-binge/purging subtype). Nonetheless, these results point to the importance of examining personality traits within both clinical EDs and subclinical disordered eating groups as results may diverge depending on the severity of the eating pathology.

Finally, with respect to examining the associations of personality traits with distinct clinical presentations, our results suggested that while several personality traits, namely Neuroticism-Anxiety, Emotionality, Impulsivity-Sensation Seeking, Behavioral Disinhibition, and Aggression-Hostility, differentiated adolescents with an ED from those without any eating or internalizing disorders, these traits also differentiated adolescents with an internalizing disorder from those who were healthy, suggesting that many trait elevations may be indicative of the presence of general vulnerability to psychopathology rather than specific to EDs (see Caspi et al., 2014). A more limited set of traits, namely Neuroticism-Anxiety, Emotionality, and Behavioral Disinhibition, differentiated adolescents with BN or BED from those with an Internalizing disorder. No traits differentiated adolescents with AN from those with an internalizing disorder, perhaps because AN is often characterized by a need for control and compulsivity (Waxman, 2009), which also characterizes many anxiety disorders (e.g. Obsessive Compulsive Disorder, Generalized Anxiety Disorder, Panic Disorder; Moulding, & Kyrios, 2006). Together, these results suggest that EDs may be differentiated from other forms of psychopathology by the degree rather than the type of personality elevations, with BN and BED marked by more extreme personality elevations relative to other types of internalizing problems. Similar to our first set of results, these findings seem to support dimensional conceptualizations of psychopathology that emphasize continuities across vulnerability factors and symptom expression (Forbush et al., 2017; Hudziak et al., 2007; Sanislow et al., 2010; Widiger, 2005).

In terms of implications for clinical practice, the burgeoning literature linking personality traits to EDs has prompted calls to for clinicians to adapt ED treatments to more explicitly target relevant traits, to use personality assessment to facilitate treatment matching, and to develop transdiagnostic treatments that could reduce not only ED symptoms but also psychological problems that commonly co-occur with EDs (e.g., Accurso et al., 2016; Kaye, Wierenga, Knatz, Jian, Boutelle, Hill & Eisler, 2014; Martinez & Craighead, 2015). Consistent with these calls, our results underscore that personality trait elevations indicate risk for eating pathology across a broad range of severity, suggesting the utility of personality assessment for informing targeted prevention efforts. For instance, if an adolescent is identified as having pre-clinical weight concerns, prevention efforts might seek to reduce impulsivity (for instance, by teaching distress tolerance or healthy coping behaviors) in an effort to mitigate onset of subclinical disordered eating. Adolescents with subclinical disordered eating, in turn, may benefit from interventions that focus on reducing vulnerability to both emotion dysregulation and impulsivity. With respect to assessment, our results suggest that including neuroticism, emotionality and behavioral disinhibition in a pre-treatment assessment may be particularly valuable for identifying possible clinical profiles. However, prospective research disentangling whether trait elevations predict, co-occur or follow symptomatic escalations is needed to inform more specific interpretation of assessment results. Emerging research showing that personality traits such as extraversion, sensation seeking, and affectivity predict treatment response in people with EDs (Accurso et al., 2016; Deumens, Noorthoorn, & Verbraak, 2012) underscores the importance of assessing these traits in future clinical trials.

A number of limitations warrant consideration. First and foremost, the NCS-A is a cross-sectional study, precluding any examination of the temporal ordering of personality trait elevations relative to eating pathology. This is an especially critical limitation in this field, as previous research shows that eating problems, especially long periods of restriction and starvation, are associated with changes in the expression of emotions and personality trait endorsement on standardized measures (Halmi, Schork, & Marcus, 1989; Keys, Brozek, Henschel, Mickelsen, & Taylor, 1950). Second, the rules for assigning ED diagnoses made it difficult to examine diagnostic cross-over or comorbidity of ED diagnoses in this sample (Tozzi et al., 2005), and the small numbers of ED diagnoses captured in this community sample precluded a detailed examination of how personality traits differ according to psychiatric problems that co-occur with EDs (e.g., internalizing disorders, substance use, etc.). Extending the current results in large samples where a broad range of psychopathology is assessed would help elucidate specific and non-specific personality features that relate to eating pathology presentations. Third, given previous studies suggesting high rates of distortion and denial among ED patients (Vitousek & Manke, 1994), the interviews used in the NCS-A may underestimate the prevalence of some eating concerns. Fourth, weight and height information were not released in this restricted access dataset, so we were not able to examine how BMI might moderate the associations between personality traits and eating concerns. This is important for future work, given established associations between chronic low weight and personality (Keys et al., 1950; Vitousek et al., 2004). Fifth, although the NCS-A study aimed to consider a broad range of personality traits that are relevant to psychopathology (Kessler et al., 2009; Merikangas et al., 2009), several traits that correlate with eating pathology in previous studies (e.g., openness, conscientiousness, perfectionism; see Cassin & von Ranson, 2005; Farstad et al., 2016) were not included in the protocol. Moreover, several personality trait scales were comprised of very few items and had mediocre internal consistency (i.e., Behavioral Disinhibition, Sociability). Although the internal consistencies observed in this study are consistent with other personality studies, particularly for very short subscales (i.e., 0.5 and upwards; Aluja et al., 2003; Aluja et al., 2004; Gutierrez et al., 2017; Pedersen et al., 2017), the limited reliability of these scales is important to consider. As well, our univariate analyses did not account for correlations between personality traits, and future research should aim to employ a multivariate approach. Finally, while the NCS-A remains one of the few epidemiological surveys that has focused specifically on psychopathology in American adolescents, the data were collected in 2001 to 2004. Given some research suggests that the presence of eating pathology behaviors may be increasing over time (e.g., White, Reynolds-Malear, & Cordero, 2005), replication of these findings in a recent sample would enhance confidence in these findings.

These limitations notwithstanding, the current study contributes valuable information to our understanding of the links between personality and eating pathology by focusing on how these traits relate to both severity and expression of eating pathology. Overall, our results suggest that personality traits related to affectivity and impulsivity may be more strongly connected to the severity of eating pathology rather than specific symptom experiences. Some traits, particularly Neuroticism-Anxiety, Emotionality, and Behavioral Disinhibition, appear to be useful in discriminating adolescents who experience clinical EDs, subclinical disordered eating, pre-clinical weight concerns, or no eating or weight concerns, as well as adolescents who engage in subclinical disordered eating characterized by binge eating, with or without purging, versus subclinical disordered eating without binge eating. Prospective research is a crucial next step in understanding how the relationships between eating pathology and personality trait elevations unfold over time.