Executive function in eating disorders: The role of state anxiety




We examined the influence of depression and anxiety on executive function in individuals with a DSM-IV diagnosis of anorexia nervosa-restricting type, anorexia nervosa-binge-eating/purging type, bulimia nervosa, or eating disorder not otherwise specified.


We assessed 106 women after their inpatient treatment in an eating disorders program. All participants were nutritionally stable at the time of testing.


Thirty percent of the total sample showed impaired performance on one or more tests of executive function. No differences in executive function were observed among diagnostic groups. Anxiety scores accounted for significant variance in performance for all groups.


Executive function deficits were found in a minority of our sample, with significant variance in performance accounted for by self-reported anxiety. State anxiety appears to contribute to diminished executive function in women with eating disorders. © 2013 by Wiley Periodicals, Inc. (Int J Eat Disord 2013)


A variety of cognitive deficits have been reported in individuals with eating disorders.1–3 General intellectual ability appears to be preserved and is often in the high average range, but deficits involving attention, working memory, executive function, motor, and visual–spatial skills have been observed.2 Some discrepancies have been found across eating disorder subtypes, including anorexia nervosa-restricting type (ANR), anorexia nervosa-binge-eating/purging type (ANB), bulimia nervosa (BN), and eating disorder not otherwise specified (EDNOS). Although the existing literature shows relative consistency with respect to executive function problems in individuals with anorexia nervosa (both subtypes), a recent study of BN and EDNOS showed no differences between these groups and healthy controls on tests of attention, inhibitory control, and decision-making.4 Still others have found executive function deficits in individuals with BN.5, 6 Given a preponderance of evidence across studies, Tchanturia and colleagues have asserted that poor executive function, as defined by performance on tasks requiring mental flexibility and cognitive set-shifting, is a trait feature and candidate endophenotype of anorexia nervosa and bulimia.7–9

Notwithstanding Tchanturia and colleagues' assertion, only some patients display deficits, implying that other factors contribute to compromised executive function. Concurrent psychiatric symptoms may influence executive function in individuals with eating disorders, as more than 90% of inpatients have a co-morbid mood disorder, and approximately 67% have a history of anxiety disorder.10 Multiple studies have shown significantly greater anxiety and depression in participants with eating disorders relative to healthy controls, but few have examined the predictive value of depression and anxiety symptoms relative to executive function. Depression and anxiety have been shown to affect executive function performance in other psychiatric populations.11–14 One report suggests that at least one type of executive function, cognitive set-shifting, is diminished in anxious and depressed women with eating disorders.9

We examined executive function in women being discharged from an inpatient eating disorders program at the Laureate Psychiatric Clinic and Hospital. We expected to find executive function deficits in our sample of participants. Given the prevalence of anxiety and mood disorders in the eating disorders population and the influence of anxiety and depression on executive function in other psychiatric populations,9, 11–15 we hypothesized that depression and anxiety scores would predict performance on executive function tasks.



One hundred six women in an inpatient eating disorders program at the Laureate Psychiatric Clinic and Hospital in Tulsa, OK, were tested on a battery of standardized neuropsychological tests near the time of their discharge from the program, between 2005 and 2009. Participant characteristics are listed in Table 1.

Table 1. Patient characteristics in the total sample and by eating disorder subtype
 Total Sample (n = 106)AN-R (n = 23)AN-B (n = 58)BN (n = 13)ED-NOS (n = 12)
  • Note: Means are depicted for the first ten measures with standard deviations listed in parentheses. AN-R, anorexia nervosa, restricting type; AN-B, anorexia nervosa, binge-purge type; BN, bulimia nervosa; ED-NOS, eating disorder, not otherwise specified; n, sample size; EDI, Eating Disorders Inventory; CGI, clinical global impression with a range of 1–7 with increasing degree of severity; BDI, Beck Depression Inventory; STAI, State Trait Anxiety Inventory; Hx, history; Dx, diagnosis; SD, standard deviation.

  • a

    Full-Scale Intelligence Quotient (IQ) is based on the Wechsler Test of Adult Reading (WTAR).16

  • b

    Pairwise comparisons showed the following difference in BMI: AN-R < ED-NOS (p < .05)

  • c

    Pairwise comparisons showed the following differences on the EDI Risk Composite: AN-R > AN-B (p < .05), AN-R>BN (p < .001), ED-NOS > BN (p < .05).

  • d

    Pairwise comparisons showed a greater number of inpatient days for AN-R compared with AN-B (p < .05).

  • e

    State anxiety scores differed between BN and ED-NOS subtypes (p < .07).

  • f

    Other Axis I Disorder, Axis I disorders, not including eating disorders, mood, or anxiety disorders.

  • g

    A greater proportion of participants with AN-B had other Axis I Disorders than participants with AN-R (p < .05).

  • h

    Admission defined here as inpatient hospital admission due to an eating disorder. Differences between groups are not significant unless indicated.

Age24.2 (7.3)23.6 (7.3)24.5 (6.8)23.2 (7.3)24.9 (9.9)
Education14.2 (2.0)14.4 (2.4)14.1 (2.0)13.8 (1.8)14.3 (1.8)
Duration of illness (years)9.0 (7.8)8.3 (8.0)9.7 (7.4)7.9 (9.6)7.7 (7.9)
Full-scale IQ estimatea110.6 (6.2)110.8 (7.3)110.7 (6.2)110.0 (4.6)110.4 (5.7)
Body mass indexb21.5 (1.3)21.0 (0.4)21.5 (1.4)22.0 (1.2)22.2 (1.6)
EDI risk composite at testingc43.6 (10.9)49.5 (9.8)42.4 (9.9)34.5 (10.9)46.8 (10.5)
CGI at testing3.6 (1.0)3.8 (1.0)3.6 (1.1)3.1 (0.8)3.4 (0.8)
Number of days as inpatientsd94.7 (31.1)111.6 (28.6)90.8 (33.1)87.1 (25.1)90.6 (21.0)
BDI score15.5 (11.5)16.6 (12.9)13.7 (10.7)14.6 (11.4)21.7 (10.9)
STAI state anxiety51.9 (14.0)55.5 (11.7)49.9 (14.7)45.8 (12.9)e59.8 (10.8)e
Hx of mood disorder Dx79/10613/2346/589/1311/12
Hx of anxiety disorder Dx75/10616/2342/589/138/12
Hx of other axis I Dxf44/1065/23g28/58g8/133/12
Hx of axis II Dx36/1065/2322/584/135/12
Hx of prior admissionh19/1066/2311/582/130/12


All participants gave informed consent to participate in the IRB-approved study. All participants were discharged and tested after nutritional stabilization, as indicated by blood and liver function tests and a BMI within a range of 19–25. Blood and liver function tests included glucose, chloride, sodium, potassium, serum albumin protein, and total protein. Exclusion criteria included a BMI under 19, a history of head injury or neurological disorder, or an estimated full-scale IQ of less than 80.

Most participants (87/106) did not have a prior history of inpatient treatment for an eating disorder, and neither age of onset nor illness duration differed significantly by group (see also Table 1). Participants were medicated according to individual needs, with only one patient not taking any psychiatric medication. As part of the inpatient eating disorders program, participants were stabilized on medication prior to discharge and testing. The percentage of individuals in the total sample taking psychiatric medication was as follows: selective serotonin reuptake inhibitors (SSRIs), 61.3%; atypical antipsychotics, 43.4%; anticonvulsants, 34.9%; selective norepinephrine reuptake inhibitors, 30.2%; benzodiazepines, 34.0%; hypnotics, 22.6%; tricyclic/tetracyclic antidepressants, 10.4%; stimulants, 8.5%; buspirone, 8.5%; metoclopramide, 6.6%; bupropion, 5.7%; lithium, 4.7%; tegaserod, 4.7%; ramelteon, 3.8%; ropinirole, 1.9%; dexamethasone, 1.9%; acamprosate, 1.9%.

Test Battery

The clinical neuropsychological test battery included the following tests that measure different aspects of executive function. The Digit Span subtest from the Wechsler Adult Intelligence Scale, Third Edition (WAIS-III)17 assesses auditory working memory and attention span. The Digit Symbol-Coding test from the WAIS-III measures visual working memory and speed of information processing. The Ruff Figural Fluency Test18 assesses nonverbal ideational fluency and mental flexibility. The Controlled Oral Word Association (COWA) test19 assesses verbal fluency and mental flexibility. The Trail Making Test (Parts A and B)20 requires set-shifting in a visual–spatial array of letters and numbers, and it assesses speed of information processing. A full-scale intelligence quotient (IQ) was estimated from performance on the Wechsler Test of Adult Reading (WTAR).16

Psychiatric measures administered were the Structured Clinical Interviews for DSM-IV Axis I and Axis II Disorders (SCID-I and SCID-II),21, 22 the Beck Depression Inventory, 2nd Edition (BDI-II),23 and the State Trait Anxiety Inventory (STAI).24 Severity of eating disorder was estimated for participants based on the research interviewer's clinical global impression (CGI; with symptom severity ranging from 1 to 7) and the eating disorders risk composite score from the Eating Disorders Inventory, 3rd Edition (EDI-3).25 Table 1 lists patient characteristics.

Statistical Methods

We used stepwise linear regression for its predictive value to determine if anxiety and depression ratings predicted executive function performance.26 We included age, education level, estimated full-scale IQ, and the eating disorders risk composite score (from the EDI-3) in each regression analysis to determine if depression or anxiety scores predicted executive function performance beyond variance explained by these other factors. Logarithmic transformations were applied to test scores, where appropriate, based on skewness and kurtosis, to approximate normal distributions.27 Selection of the best regression model for each test measure was based on adjusted R-square and Schwarz's Bayesian Information Criterion.28


Neuropsychological Test Scores

Mean standardized test scores are shown in Table2. Means for each measure fell within the average to above average range of standardized scores for every test (based on the participant's age, and education, depending on normative values), and within each eating disorder subtype. The number of participants with scores falling below 1.5 standard deviations of the standardized mean of the normative sample (below the 6th percentile) was calculated. In the entire sample, 30.2% (32 of 106) of participants scored lower than 1.5 standard deviations below the standardized mean on at least one test. Across individual tests performance was poorest on the Ruff Figural Fluency Test, with 22.6% of the sample showing a deficit. The proportion of individuals performing below 1.5 standard deviations of the standardized mean did not differ by eating disorder subtype for any test (see Table 2).

Table 2. Mean (+SD) scores and deficit counts on executive function and working memory measures for the total sample and by eating disorder subtype
Total Sample (n = 106)AN-R (n = 23)AN-B (n = 58)BN (n = 13)ED-NOS (n = 12)
  • Note: Deficit count (in italics) is expressed as the number of participants in each category with performance values falling within the clinically abnormal range based upon normative thresholds (1.5 standard deviations or more below the standardized mean). SS, scaled score with a standardized mean of 10 and a standard deviation of 3; T, T-score with a standardized mean of 50 and a standard deviation of 10; z, z-score with a standardized mean of 0 and a standard deviation of 1; n, sample size.

  • a

    The adjusted score for verbal fluency is based on the Controlled Oral Word Association normative values from the Multilingual Aphasia Examination,29 where adjusted scores between 31 and 44 represent the 25th to 75th percentile (adjusted scores below 25 equate to less than 1.5 standard deviations below the standardized mean, or below the 6th percentile). Note: no differences in deficit count frequency were observed across eating disorder subtypes.

Deficit count     
Digit span (SS)10.8 (2.4)10.7 (2.6)10.6 (2.3)12.2 (3.1)10.2 (1.9)
  0/104 0/23 0/56 0/13 0/12
Ruff figural43.3 (9.1)41.5 (10.1)44.7 (9.7)42.3 (5.4)40.9 (6.2)
Fluency test (T) 24/106 7/23 13/58 1/13 3/12
COWA verbal45.2 (11.1)45.1 (9.9)45.4 (11.9)44.4 (9.7)45.6 (12.1)
Fluency testa 3/105 0/23 2/57 0/13 1/12
Symbol digit11.8 (2.8)11.6 (3.0)12.1 (2.8)12.2 (2.3)10.5 (2.7)
Coding test (SS) 1/105 0/23 1/57 0/13 0/12
Trails A (z)0.08 (1.1)−0.05 (0.9)0.08 (1.2)0.07 (1.0)0.33 (0.8)
  7/105 2/23 4/57 0/13 1/12
Trails B (z)−0.06 (1.4)−0.08 (1.7)−0.16 (1.3)0.14 (1.0)0.21 (0.9)
  12/105 2/23 8/57 1/13 1/12

Regression Prediction of Executive Function

Stepwise linear regression procedures were first performed for the entire eating disorders sample, given the lack of significant differences in test performance by eating disorder subtype. Factors entered into the stepwise procedures as predictors were age, education, full-scale IQ, eating disorders risk composite score, depression score, and state anxiety score (see also Methods section). Results showed that, along with estimated full-scale IQ, state anxiety predicted executive function performance on the figural fluency test (Ruff Figural Fluency Test, R2 = .124, p < .01; partial correlations: anxiety, r = −.27, p < .01; full-scale IQ, r = .24, p < .02) and the verbal fluency test (Controlled Oral Word Association Test, R2 = .225, p < .001; partial correlations: anxiety, r = −.24, p < .01; full-scale IQ, r = .42, p < .001). A similar trend was observed for visual–spatial coding (Digit Symbol-Coding subtest from the WAIS-III, R2 = .049, p < .10; partial correlations: anxiety, r = −.18, p = .07; full-scale IQ, r = .14, p < .17). In each case, higher state anxiety scores predicted poorer executive function, beyond variance accounted for by full-scale IQ (see Table 3 and Fig. 1).

Figure 1.

Fitted regression models. Panels indicate performance on the (a) Ruff figural fluency test, (b) Controlled Oral Word Association (COWA) verbal fluency test, and (c) digit symbol-coding test. Note that scales for the state anxiety, Ruff figural fluency, COWA verbal fluency, and digit symbol-coding test scores reflect logarithmic transformations (see Statistical Methods section). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Table 3. Linear regression analyses predicting executive function performance in the entire eating disorders sample
Variable BSE (B)β tSig. (p)
  • Notes: COWA, Controlled Oral Word Association; B, unstandardized coefficients; β, standardized coefficient; SE, standard error; t, t-statistic; Sig., significance; p, probability.

  • a

    Full-Scale IQ is not based on an estimate derived from the Digit Symbol-Coding subtest of the Wechsler Scale of Intelligence, 3rd Edition (WAIS-III).17 The full-scale IQ estimate is based on the Wechsler Test of Adult Reading (WTAR).16

Ruff figural fluency test
Full-scale IQ0.0090.0030.2432.529.013
State anxiety−0.1920.069−0.269−2.794.006
R2 = .124
COWA verbal fluency test
Full-scale IQ0.0170.0040.4214.649<.000
State anxiety−0.2000.075−0.242−2.677.009
R2 = .225
Digit symbol-coding testa
Full-scale IQ0.0060.0040.1391.388.168
State anxiety−0.1510.083−0.181−1.807.074
R2 = .049

For each measure of executive function, education, age, eating disorder risk composite scores, and depression scores failed to account for significant variance. A correlation was observed between eating disorder risk composite and anxiety scores (r (96) = .59, p < .001), but only anxiety accounted for significant variance in executive function performance in the regression models. Given the diagnostic heterogeneity of our total sample, we ran the regression procedures a second time by including only those diagnosed with ANR and ANB (n = 81). The same significant regression results were found; however, the trend observed for the spatial coding test was no longer present.


Tchanturia and coworkers have suggested that deficits of mental flexibility and set-shifting are trait features of eating disorders.7–9 We administered standardized clinical tests of executive function that require mental flexibility and set-shifting to a sample of 106 women with eating disorders and found that thirty percent of our participants were impaired on one or more tests. Thus, poor executive function performance, relative to standardized normative values, was restricted to a minority of our sample. Previous studies of performance by individuals with eating disorders on executive function tests have shown significant differences compared with healthy controls. Our study suggests that, at least for the tests included here, clinically significant executive function problems in this population are restricted to a minority of women with eating disorders. Moreover, for three measures of executive function, women who performed poorly were likely to have reported higher levels of anxiety. Roberts et al. report of increased anxiety in women with eating disorders and poor set-shifting ability,9 along with the results of this investigation, suggest that anxiety symptoms in individuals with eating disorders may impact executive function, particularly when mental flexibility is required by the task.

Executive function is significantly influenced by anxiety in other psychiatric populations. Executive function deficits have been reported in individuals with post-traumatic stress disorder30 and other anxiety disorders31; co-morbid anxiety also has been shown to exacerbate executive function deficits in individuals with depression11, 15 and bipolar II disorder.13 A full range of anxiety disorder diagnoses was found within our sample, with social phobia, obsessive-compulsive disorder, and general anxiety disorder being most common across groups. One surprising finding here, however, given associations previously reported between executive function and depression in other populations,12, 14, 32 is that depressive symptoms did not significantly account for variance in executive function performance in our sample. One reason may be that the tests we used did not include a reward or affective feedback component, to which individuals with depression appear particularly sensitive.33–35 Since depression severity also may impact executive function performance,32 future studies should include a larger sample of participants with a greater range of depression severity than found in our sample.

The heterogeneity of psychotropic medication taken by our participants represents a limitation to the generalizability of our findings. Given the high prevalence of co-morbid psychiatric diagnoses in our sample, and self-reported emotional distress, it was not feasible to take participants off medication prior to neuropsychological testing. Regardless, antipsychotics and benzodiazepines have been shown to alter performance across multiple cognitive domains,36–39 including executive function. Both positive and negative effects on performance have been reported.36–39 The extent to which these medications and antidepressants affect executive function specifically in women with eating disorders is not known and should be a topic of future research.

Another limitation of this study is a small sample for two eating disorder subtypes: bulimia nervosa and eating disorder-not otherwise specified. A recent study of these two subtypes failed to find executive function deficits relative to a control sample4; our analyses, on the other hand, did not suggest any differences between these subtypes and those with anorexia nervosa. Neuropsychological and psychiatric function in these subtypes remains understudied. Our findings—even in light of these limitations—may have implications for the therapy and management of women with eating disorders. Executive function deficits have been cited as relevant to treatment resistance in this population,40 and to the extent that such deficits are related to anxiety symptoms, the latter may be particularly targeted when treating women with eating disorders.

Earn CE credit for this article!

Visit: http://www.ce-credit.com for additional information. There may be a delay in the posting of the article, so continue to check back and look for the section on Eating Disorders. Additional information about the program is available at www.aedweb.org