Objective

This study aimed to identify discrete neuropsychological profiles and their relationship to clinical symptoms in 253 female children and adolescents with anorexia nervosa (AN) and 170 healthy controls (HCs) using a standardised neuropsychological assessment battery.

Method

Hierarchical cluster analysis was used to identify the optimum number of clusters, and participants were assigned using K-means cluster analysis. Confirmatory discriminant function analysis determined which combination of neuropsychological variables best distinguished the clusters

Results

Three distinct clusters in the AN sample emerged- AN cluster 1 (19%) - "neuropsychologically low average to average"; AN cluster 2 (33%) - "verbal/visuo-spatial discrepancy"; and AN cluster 3 (48%) - "verbally strong and neuropsychologically average to high average". Two distinct clusters in HCs were identified. HC cluster 1 (48%) demonstrated poor visuo-spatial memory scores and high verbal fluency scores, whilst HC cluster 2 (52%) scored within the average range on all neuropsychological tasks. Neuropsychological performance was associated with clinical symptoms of body mass index centile, Eating Disorder Examination subscale and global score, anxiety, depression and obsessions, and compulsions between the AN and HC groups. However, niether significant differences emerged between AN clusters only nor HC clusters only at the post-hoc level.

Discussion

An underlying neuropsychological heterogeneity may exist in AN. We encourage future studies to investigate whether the identified profiles and their association with clinical characteristics are replicable. We cautiously suggest that neuropsychological profiling may have potential to both inform future research and have possible clinical benefits through individually tailored treatment strategies.

Introduction

Anorexia nervosa (AN) is an eating disorder characterized by an intense fear of weight gain, severe anxiety around eating and food, a relentless pursuit of thinness which can lead to life-threatening weight loss, as well as a distorted body image. The disorder affects between 0.5% and 1% of 12–23-year-old females (Machado, Machado, Goncalves, & Hoek, 2007) are more common in this age range than later adult life and occurs much more rarely in men. Sufferers often continue attempts to lose weight despite being perilously unwell, to the extent where their life is gravely at risk.

There is convincing evidence for an interaction between social, psychological and neurobiological factors in the pathogenesis of AN (Lask & Frampton, 2011). However, the precise nature of the neurobiological contribution remains unclear. Studies in the neurobiology of AN have revealed anomalies in neuropeptide and neuroendocrine functioning (Bailer & Kaye, 2003), genetic associations (Trace, Baker, Peñas-Lledó, & Bulik, 2013) and changes in brain structure and function (Friederich, Wu, Simon, & Herzog, 2013). The behavioral and psychological aspects that are commonly associated with AN include obsessionality, rigidity and perfectionism. A particularly challenging concern is the distorted thoughts and beliefs around shape, weight and food, and the “thin ideal.” When the disorder is viewed as ego-syntonic, then these may act as perpetuating features that potentially underlie the high levels of treatment resistance and relapse rates found in AN (Abbate-Daga, Amianto, Delsedime, De-Bacco, & Fassino, 2013).

An approach taken to better understand this complex disorder has been the conceptualization of the cognitive features seen in AN, in particular neuropsychological functioning. Neuropsychology is conceptualized as the mediator between mental health symptoms including emotional, behavioral and cognitive functioning, and neurobiology (Frampton & Rose, 2013). Hence, in order to gain a greater understanding of the distorted thinking in AN and its role in motivation for change and treatment compliance, the neuropsychological performance of patients with AN has been investigated.

Neuropsychological Functioning in Adults With AN

According to both reviews and meta-analyses of the neuropsychological functioning of adult patients with AN, there has been some consistency in findings of weaker performance in visuo-spatial memory and central coherence (Duchesne et al., 2004; Jáuregui-Lobera, 2013; Lang, Lopez, Stahl, Tchanturia, & Treasure, 2014a; Lena, Fiocco, & Leyenaar, 2004; Lopez, Tchanturia, Stahl, & Treasure, 2008; Roberts, Tchanturia, Stahl, Southgate, & Treasure, 2007; Zakzanis, Campbell, & Polsinelli, 2010). Findings indicate that patients with AN tend to have difficulty in replicating and recall fewer details from complex visual designs, as well as identifying simple shapes embedded within a background. Findings from reviews suggest that patients with AN show a preference for local processing at the expense of global processing. This has been linked to the clinical observation of a cognitive bias of focusing on finer details and difficulties in the integrative processing of information. Reviews and a meta-analysis suggest attentional bias in AN, as found using the Stroop paradigm (Stroop, 1935) that incorporated illness-related stimuli (i.e., the color words were replaced with food, shape, or body image words or images; Dobson & Dozois, 2004; Johansson, Ghaderi, & Andersson, 2005). Selective attention has been implicated in body image processing.

There is generally consistent evidence of weaknesses in executive functioning in those with AN compared to healthy controls (HCs). Specifically, set-shifting is often characterized by either excessive or preservative errors which are often seen in patients with AN (Westwood, Stahl, Mandy, & Tchanturia, 2016; Wu et al., 2014). This has been linked to clinical observations of persistent dieting behaviors and rigidity of thinking which may underpin difficulties in developing alternative healthy strategies even when individuals have expressed a desire for change during treatment. There is also general support for altered reward-related decision-making in adults with AN (Wu et al., 2016). Findings suggest that in decision-making tasks, patients with AN have difficulty in learning the optimal response to maximize a long-term reward over less valuable but more immediate short-term rewards. Such findings have been linked to either an inability in incorporating reward based feedback, or that those with AN have a different qualitative experience in response to rewards compared to HCs. This may manifest as either failures in learning and proneness to making disadvantageous decisions, or that biologically driven responses to food can be overridden in preference for rewards derives from a sense of pride or achievement via behaviors such as dietary restriction or excessive exercise.

A profiling study that utilized a cluster analysis by Renwick et al. (2015) analyzed neuropsychological and social performance from 100 adult participants with AN. Three clusters were identified with 45, 38, and 17 participants in each cluster. The latter cluster showed more substantial weaknesses in that scores on the Wisconsin Card Sorting Task (to assess executive functioning) were greater than 2 SDs below the norm as well as poorer performance on the Reading the Mind in Films Task (assessing emotion recognition). All three clusters had Rey Complex Figure Test (Rey) central coherence scores (assessing visuo-spatial constructional ability) near 1 SD below the norm indicating a mild to poor weakness. Performance across all clusters were within 1 to −2.5 SDs of the norm; hence, no clusters demonstrated substantial strengths nor weaknesses (i.e., greater than ±2.5 SDs). Interestingly, the three clusters did not differ in clinical characteristics, that is, primary diagnosis, anti-depressant medication, number of previous inpatient admissions nor mean number of treatment sessions as part of a larger randomized controlled trial.

Neuropsychological Functioning in Children and Adolescents With AN

Compared to the adult literature, the child and adolescent neuropsychological literature in AN has been traditionally lacking. For example, only 4 out of 14 (Lena et al., 2004) and 18 out of 71 (Reville, O'Connor, & Frampton, 2016) reviewed studies focused on younger people with AN. Systematic reviews based on younger people with AN have been conducted in three domains: set-shifting, central coherence, and decision-making. In terms of set-shifting, Lang, Stahl, Espie, Treasure, and Tchanturia (2014b) suggest performance weaknesses to be less pronounced in children compared to adults. Similarly, Westwood et al. (2016) concluded from a meta-analyses of four studies that overall there were no significant differences in set-shifting between child and adolescents with AN and HCs. However, Wu et al. (2014) reported no differences between adult and adolescent samples in that set-shifting inefficiencies were found in both age groups. Regarding central coherence, Lang and Tchanturia (2014c) report evidence of inefficient global processing in children. In terms of decision-making, Wu et al. (2016) reported a lack of supportive evidence for disadvantageous reward-related decision-making in children and adolescents with AN. Neuropsychological profiling studies in young people with AN compared to HCs have mostly identified only subtle inefficiencies. Andrés-Perpiña et al. (2011) reported the only significant difference to emerge from a battery assessing intelligence, memory, visuo-spatial ability, cognitive flexibility, inhibition, and verbal fluency was longer Rey copy times in patients with AN. Calderoni et al. (2013) reported finding only subtle inefficiencies in automatic response inhibition and cognitive flexibility in audio-motor responses, and impaired cognitive performance in visuo-spatial tasks in those with AN on the Developmental Neuropsychological Assessment (NEPSY-II; Korkman, Kirk, & Kemp, 2007). The examination of the neurocognitive and intelligence profile by Kjaersdam Telléus et al. (2015) revealed only significantly poorer performance in verbal memory and motor speed in patients with AN.

Association between Neuropsychological Functioning and Clinical Factors

There is debate regarding the association between neuropsychological performance and clinical factors. For example, neuropsychological weaknesses have been shown to be independent of AN symptom severity (e.g., Mikos et al., 2008), weight status (e.g., Moser et al., 2003), or recovery (e.g., Tchanturia et al., 2012) suggesting that they may reflect underlying vulnerability factors. Furthermore, the role of psychiatric comorbidity on neuropsychological performance in AN is not fully understood (Zakzanis et al., 2010) with the suggestion that some cognitive weaknesses may be secondary to comorbid conditions (Duchesne et al., 2004). Anxiety, depression, and obsessive–compulsive (OC) disorder have all been shown to be highly comorbid with AN, for example, 12 out of 17 (Lopez et al., 2008) and 11 out of 15 studies (Roberts et al., 2007) reported comorbid conditions, as well as confounding neuropsychological performance, for example, depression and obsessionality onset-shifting (Wilsdon & Wade, 2006) and depression on decision-making (Guillaume et al., 2010).

A need for Further Understanding of the Neuropsychological Functioning in Children and Adolescents With AN

In other fields such as depression, research focusing on younger individuals has demonstrated within-group differences in neuropsychological functioning with potential to inform early detection and targeted intervention strategies (Hermens et al., 2011). It is important that this is explored as findings may indicate different risk factors, pathophysiologies, developmental trajectories, clinical presentation, treatment responsive, and outcomes. Deriving distinct profiles in children and adolescents may validate this construct in a developmental context and imply that profiles measure underlying traits, in-line with the theoretical understanding of AN as a “final common pathway” outcome of genetic, epigenetic, and environmental risk factors (Lask & Frampton, 2011).

A number of factors possibly hinder our understanding of the neuropsychology of AN, for instance different samples of patients along the lifespan and different tasks used which examine different functions (Bühren et al., 2012). Furthermore, the relative rarity of the disorder generally leads to small sample sizes, longitudinal studies including long-term follow-up have many logistical problems, and it is possible that neuropsychological heterogeneity may mask profiles. In an attempt to challenge this incomparability between studies, Stedal, Rose, Frampton, Landrø, and Lask (2012) reported on a neuropsychological battery assessing visuo-spatial functioning and executive functioning (Ravello Profile; Rose, Davis, Frampton, & Lask, 2011) administered to 155 patients aged 9–27 with AN. Mixed performance was observed for executive functioning, in that patients scored significantly above the normative mean on verbal fluency and inhibition, whereas set-shifting was significantly below the normative mean. Measures of visuo-spatial memory and central coherence were significantly below the normative mean. In addition, neuropsychological performance was found to be independent of weight status. Stedal and colleagues concluded that a common neuropsychological profile existed in AN that comprised of a relative strength in verbal fluency and relative weakness in visuo-spatial skills.

However, a number of limitations in this study prevent conclusions being drawn that are specific to children and adolescents with AN. For example, firstly adults were included in the sample. Inclusion of adults may mean these particular participants met diagnostic criteria for an eating disorder years prior to the study; hence, duration of illness and sustained starvation may have influenced findings. Secondly, performance was assessed by comparison to normative data for all but two tasks. For central coherence and the Brixton Spatial Anticipation Task, a HC group was used for comparison purposes. Since the neuropsychological performance of those with AN was not compared to an age- and gender-matched control group for all tasks, it is not clear whether the profile was (a) specific to an AN population nor (b) present in healthy females. Thirdly, comorbid anxiety, depression, and OC symptoms were reported as descriptive data only. Thus, it is unknown whether comorbidity may have influenced neuropsychological performance. Furthermore, patient heterogeneity was not explored and may have masked particular neuropsychological strengths and weaknesses.

Compared to adults, relatively little is still known about neuropsychological performance in younger individuals with AN. Hence, further investigation is required on the neuropsychological performance of younger individuals with AN. This study aims to assess if distinct neuropsychological profiles indicating different cognitive strengths and weaknesses in female children and adolescents with AN exist in comparison to an age- and gender-matched HC group using the Ravello Profile neuropsychological battery. Thus, two hypotheses will be tested. Firstly, neuropsychological strengths and weaknesses will cluster together to form one or more distinct neuropsychological profiles in patients with AN (Hypothesis 1). Secondly, neuropsychological profiles observed in patients with AN will be significantly different from the neuropsychological profiles found in HCs (Hypothesis 2). The secondary aim is to explore the relationship between neuropsychological performance and clinical symptoms. It is hypothesized that eating disorder symptoms and anxiety, depression, and OC symptom severities will not be associated with AN neuropsychological profiles (Hypothesis 3).

Materials and Methods

Participants

Two hundred and fifty-three female patients with a clinical diagnosis of AN based on either DSM IV criteria (Association, 1994) or Great Ormond Street diagnostic checklist criteria (Bryant‐Waugh, 2000) were recruited from a multicenter collaboration that included 17 specialist eating disorders units from the United Kingdom, Norway, Germany, and Switzerland (see Table A1). In addition, data from 170 female HCs were recruited from two secondary schools in the United Kingdom and via flyer and Internet announcements in Germany.

All participants met the following inclusion criteria: aged between 9 and 18 (inclusive; mean age = 15.6, SD = 1.8, range 10.3–18.9); all were women; all completed the battery of neuropsychological tests and had a derived IQ ≥ 85. Intelligence was measured using the Wechsler scales (in Switzerland, Norway and the UK) or Culture Fair Intelligence Test—20— Revision (in Germany; Weiß, 2006). IQ scores were estimated from the Vocabulary and Matrix Reasoning subtests from either the Wechsler Abbreviated Scales of Intelligence (WASI; Wechsler, 1999) or the Wechsler Adult Intelligence Scale-Third Edition (WAIS-III; Wechsler, 1997). Since Vocabulary and Matrix Reasoning are significantly highly correlated, 0.88 and 0.66, respectively, these scores were used to produce a prorated two subtest full-scale IQ score (Wechsler, 1999).

Neuropsychological Assessment

This study used a standardized neuropsychological assessment battery for AN, the Ravello Profile (Rose et al., 2011), which was informed by review studies at the time of development (Duchesne et al., 2004; Lena et al., 2004; Lopez et al., 2008; Tchanturia, Campbell, Morris, & Treasure, 2005). The tests were chosen on the basis of: (a) published normative data; (b) covering a broad age range; (c) brief and easy to administer; (d) relatively low cost; (e) easy to acquire; (f) established psychometric properties; and (g) empirical evidence of performance differences between patients with AN and HCs. Full details of each test can be found in our previous papers (Rose et al., 2011; Stedal, Frampton, Landrø, & Lask, 2011; Stedal et al., 2012; van Noort, Pfeiffer, Lehmkuhl, & Kappel, 2013). The neuropsychological test battery consisted of the following tests: Rey Complex Figure Test (Rey; Meyers & Meyers, 1995; Rey, 1941) to assess visual memory (Immediate Recall, Delayed Recall, and Recognition Trial) and three central coherence measures (Booth, 2006) applied to the Copy Trial (Central Coherence Index—CCI, Order of Construction Index—OCI, and Style Index-SI); the Brixton Spatial Anticipation Test (Brixton; Burgess & Shallice, 1997) to assess cognitive flexibility; four subtests from the Delis–Kaplin Executive Function System (Delis, Kaplan, & Kramer, 2001): the Verbal Fluency Test to assess word productivity (Verbal Fluency—Letters; Category; Flexibility Total) and cognitive flexibility (Verbal Fluency—Flexibility Accuracy), Trail Making Test (TMT) for assessing cognitive flexibility; Color–Word Interference Test to assess inhibition (CWI3) and combined inhibition/cognitive flexibility (CWI4), and the Tower Test (TT) to assess planning. While the Hayling Sentence Completion Test (Burgess & Shallice, 1997) has been previously included in the battery (Rose et al., 2011), it is English language specific and was not administered in Germany, Switzerland, nor Norway and thus not included in the analysis for this study.

Clinical Symptom Measures

Full details of each questionnaire can be found in our previous papers (Rose et al., 2011; Stedal et al., 2011; Stedal et al., 2012). AN psychopathology was assessed using either the Eating Disorder Examination (EDE; Fairburn & Cooper, 1993) for those aged 15 or older, or the Child Eating Disorder Examination (ChEDE; Watkins, Frampton, Lask, & Bryant-Waugh, 2005) for those aged 14 or under. The EDE-Questionnaire (EDE-Q; Luce & Crowther, 1999) was administered if data from an interview version were unavailable. The EDE-Q has been shown to have a moderate to strong correspondence with the interview version (Binford, Le Grange, & Jellar, 2005). For depressive symptoms, the Beck Depression Inventory II (BDI-II; A Beck, Steer, & Brown, 1996) was used for those aged 17 plus or the Children's Depression Inventory (CDI; Kovacs, 1992) for those aged 16 years or under. To measure anxiety the Beck Anxiety Inventory (BAI; A. Beck & Steer, 1993) or the State/Trait Anxiety Inventory (STAI; Spielberger, Goursuch, & Lushene, 1983) for 14 year olds and over and the State/Trait Anxiety Inventory for Children (STAIC; Spielberger, 1973) for 13 year olds and under. To explore OC symptoms, the Yale-Brown OC Scale-Short Form or the Clark-Beck OC Inventory (Clark, Antony, Beck, Swinson, & Steer, 2005) was used for those aged 17 plus and the Children's Obsessional-Compulsive Inventory (ChOCI; Shafran et al., 2003) for those aged 16 and under.

Procedure

The neuropsychological assessments were conducted by a researcher or clinician with specialist training in administration and scoring of the battery under the supervision of a clinical psychologist. Quarterly quality control workshops were held for researchers and clinicians to ensure uniformity and inter-rater reliability. Time taken to complete the neuropsychological battery was approximately 1.5–2.5 hr over one to three sessions. All neuropsychological tests were administered according to the original manuals. Attention was taken to ensure that Rey trials were administered according to time frames indicated in the manual, with no visuo-spatial tasks administered between them to reduce interference with visuo-spatial memory.

Statistical Analysis

Data preparation

Statistical analyses were conducted using IBM SPSS Statistics for Windows, Version 22.0. All tests were two tailed. Data were screened for missing values, normality, and outliers. Missing values were explored via the MCAR test to investigate whether data were missing randomly. The MCAR test was carried out on predictor variables and was significant for AN data (χ2 = 672.325, df = 551, p < .001) and non-significant for control data (χ2 = 83.433, df = 80, p = .374). For AN data, the percentage of missing data for each predictor variable were as follows: Rey-IR 1.7%, Rey-DR 1.7%, Verbal Fluency—Letters 3.6%, CWI3 5.5%, Trails 7.4%, Towers 8.1%, Brixton 8.4%, Style Index 9.2%, Order of Construction Index 9.5%, Central Coherence Index 9.7%, Verbal Fluency—Flexibility Total 19.7%, Verbal Fluency—Category 24.3%, CWI4 22.4%, Rey-RT 23.9%, and Verbal Fluency—Flexibility Accuracy 34.0%. Thus, it was assumed that missing AN data were “not missing at random,” that is, there appeared to be a systematic reason why data were missing. Missing data were found to be associated with clinical site, which may reflect that the neuropsychological battery was sometimes administered over several sessions. An expectation–maximization algorithm was applied to the data to calculate expected values for missing observations (Tabachnick & Fidell, 2001).

Raw scores obtained on each of the neuropsychological tests were standardized (scaled or t-scores) using the published age-based norms for Germany, Scandinavia, or the United Kingdom/USA. All standardized scores were Z scale transformed (to mean = 0, SD = 1) to allow direct comparison between subtests. Where published age appropriate norms were unavailable, a normal sample database available at the point of AN data analysis was used with the following mean and standard deviations: Brixton under 18 years (mean =  15.4, SD = 3.6), OCI (mean = 1.91, SD =  0.58), SI (mean = 1.01, SD = 0.45), and CCI (mean = 1.93, SD = 0.45). The Brixton Z scores were reversed so that negative values reflect poor performance. Prior to the analyses, outliers beyond ±3 Z scores were curtailed to ±3, to prevent extreme scores from influencing the emergent clusters.

Due to the different questionnaires used to assess comorbidity, a summary Z score was created for depression, anxiety, and OC symptoms. Means and standard deviations used in the Z transformation were taken from either the standardized scores (CDI, STAI, and STAIC) or a normal sample database with the following mean and standard deviations: ChOCI (mean = 7.64, SD = 7.18), Y-BOCS (mean = 2.05, SD = 3.26), BDI-II (mean = 4.75, SD = 5.49).

Cluster analysis

A two-step cluster analysis procedure was used to group the AN and HC samples separately, to maximize similarity of scores within clusters while minimizing the similarity of scores between clusters. For the first step, a hierarchical cluster analysis using Ward's method of minimum variance with a squared Euclidean distance measure was conducted with the 15 neuropsychological variables. Inspection of the dendrogram and agglomeration schedule coefficient informed the optimum number of clusters. The second step used a k-means cluster analysis to assign membership to the specified number of clusters identified in the previous step.

Discriminant function analysis

A confirmatory discriminant function analysis was conducted to determine which combination of the neuropsychological variables best distinguished the cluster groups and whether these combinations could reliably predict cluster-group membership. Leave-one-out cross-validation was used to assess the predictive reliability of clusters.

Between-group differences

Demographic and clinical group differences were tested using either t or Mann–Whitney U tests, depending on whether parametric assumptions were met. Multivariate analysis of variance (MANOVA) models were fitted to explore the effect of (a) group (independent variables: HC vs. AN) on neuropsychological performance (dependent variables were the 15 neuropsychological tests), (b) clusters (independent variables: AN and HC clusters) on neuropsychological performance (dependent variables were the 15 neuropsychological tests), and (c) clusters (independent variables: AN and HC clusters) on clinical symptoms (dependent variables were depression, anxiety, and OC symptom Z scores, and EDE subscale and global scores). Pillai's Trace and Type III method for computing sum of squares were used for MANOVAs as these were more conservative procedures if assumptions were violated (Warner, 2013). Between-subjects analyses (one-way between-subjects analysis of variance; ANOVAs) were also conducted for each of the MANOVAs. A post-hoc Bonferroni test for multiple comparisons explored the mean score differences on each task between clusters.

Due to the number and variety of statistical procedures, three types of effect size calculations were used. To evaluate pair-wise group comparisons, Cohen's d (d = mean difference/mean SD) was used if data met parametric assumptions; r (r = z/square root of N) for Mann–Whitney U tests. Partial eta squared $ηp2$ was used for MANOVAs. Effect size strengths were established as follows: Cohen's d : small 0.2, medium 0.5 and large 0.8; r: small 0.1, medium 0.3 and large 0.5; and $ηp2$ small 0.01, medium 0.06 and large 0.14 (Cohen, 1988).

Results

Descriptive Statistics for AN and HC Groups

Table 1 presents demographic and clinical characteristics of the AN and HC groups. Groups differed significantly in terms of age (U = 27,752.00, z = 5.068, p < .001, r = 0.25), weight status (BMI: U 3,811.00, z = −13.769, p < .001, r = −0.69; BMI age corrected centile: U = 1,756.50, z = −15.367, p < .001, r = −0.78), eating disorder symptomatology (Restraint: U = 26,017.50, z = 11.354, p  < .001, r = 0.60; Eating Concern: U = 26,657.5, z = 12.063, p < .001, r = 0.64; Weight Concern: U = 25,200, z = 10.476, p < .001, r = 0.55; Shape Concern: U = 25,802, = 11.096, p < .001, r = 0.59; and global score: U = 26,420.50, z = 11.738, p < .001, r = 0.62), depression Z score (U = 39,100, z = 14.332, p < .001, r = 0.70), anxiety Z score (= −20.840, p < .001, $ηp2$ = 0.51) and OC Z score (U = 35,804.50, z = 11.693, p < .001, r = 0.57). Effect sizes for these differences were medium to large. There was no significant difference in IQ between AN and HC groups (t = −0.093, p = .926, $ηp2$ = 0.00).

Table 1.

Demographic and clinical characteristics of AN and HC groups

AN (N = 253) HC (N = 170) Test statistica p Effect sizeb
n Mean SD/25–75th percentiles n Mean SD/25–75th percentiles
Age 253 15.67 1.77 170 14.50 2.28 27, 752.00Δ .000 0.25
IQ 253 107.34 12.57 168 107.24 10.70 −0.093◊ .926 0.00
Vocabulary 226 55.57 9.25 75 55.35 8.25 −0.183◊ .855 0.02
Matrix reasoning 226 53.01 8.55 75 49.91 7.84 −2.78◊ .006 0.37
BMI 230 15.98 14.77–17.28 170 20.80 18.28–22.75 3, 811.00Δ .000 −0.69
BMI percentile 224 2.04 0.17–7.53 168 63.27 36.19–79.43 1,756.50Δ .000 −0.78
EDE global 219 3.75 2.00–4.78 139 0.49 0.16–1.16 26,420.50Δ .000 0.62
EDE Restraint 219 3.60 1.60–4.80 139 0.20 0.00–0.80 26,017.50Δ .000 0.60
EDE Eating Concern 219 3.20 1.20–4.40 139 0.20 0.00–0.40 26,657.50Δ .000 0.64
EDE Weight Concern 219 3.80 1.60–5.20 139 0.60 0.20–1.40 25,200.00Δ .000 0.55
EDE Shape Concern 219 4.20 2.38–5.60 139 0.75 0.25–1.75 25,802.00Δ .000 0.59
Depression Z score 253 2.18 1.20–3.00 170 −0.60 −1.10 to −0.10 39,100.00Δ .000 0.70
Anxiety Z score 253 1.32 1.12 170 −0.53 0.72 −20.840◊ .000 0.51
OC Z score 253 2.55 1.05–3.00 170 −0.09 −1.06–0.65 35,804.50Δ .000 0.57
AN (N = 253) HC (N = 170) Test statistica p Effect sizeb
n Mean SD/25–75th percentiles n Mean SD/25–75th percentiles
Age 253 15.67 1.77 170 14.50 2.28 27, 752.00Δ .000 0.25
IQ 253 107.34 12.57 168 107.24 10.70 −0.093◊ .926 0.00
Vocabulary 226 55.57 9.25 75 55.35 8.25 −0.183◊ .855 0.02
Matrix reasoning 226 53.01 8.55 75 49.91 7.84 −2.78◊ .006 0.37
BMI 230 15.98 14.77–17.28 170 20.80 18.28–22.75 3, 811.00Δ .000 −0.69
BMI percentile 224 2.04 0.17–7.53 168 63.27 36.19–79.43 1,756.50Δ .000 −0.78
EDE global 219 3.75 2.00–4.78 139 0.49 0.16–1.16 26,420.50Δ .000 0.62
EDE Restraint 219 3.60 1.60–4.80 139 0.20 0.00–0.80 26,017.50Δ .000 0.60
EDE Eating Concern 219 3.20 1.20–4.40 139 0.20 0.00–0.40 26,657.50Δ .000 0.64
EDE Weight Concern 219 3.80 1.60–5.20 139 0.60 0.20–1.40 25,200.00Δ .000 0.55
EDE Shape Concern 219 4.20 2.38–5.60 139 0.75 0.25–1.75 25,802.00Δ .000 0.59
Depression Z score 253 2.18 1.20–3.00 170 −0.60 −1.10 to −0.10 39,100.00Δ .000 0.70
Anxiety Z score 253 1.32 1.12 170 −0.53 0.72 −20.840◊ .000 0.51
OC Z score 253 2.55 1.05–3.00 170 −0.09 −1.06–0.65 35,804.50Δ .000 0.57

Notes: AN = anorexia nervosa; HC = healthy control; N = number of participants in full sample; n = number of participants in subsample; SD = standard deviation; BMI = body mass index; EDE = Eating Disorder Examination; OC = obsessive–compulsive.

aTest statistics: Δ = Mann–Whitney U, median values displayed with 25th and 75th percentiles; ◊ = t test, mean values displayed with standard deviations.

bEffect sizes for pair-wise group comparisons: Cohen's d (d = mean difference/mean standard deviation) for parametric tests; r (r = z/square root of N) for nonparametric tests. Strengths of effect were as follows: Cohen's d: small 0.2, medium 0.5 and large 0.8; r: small 0.1, medium 0.3 and large 0.5 (Cohen, 1988).

Table 2 shows a MANOVA between the independent variable of group (HC vs. AN) and the dependent variables of neuropsychological tests. Overall, a main effect of group (AN vs. HC) on neuropsychological tasks was found with a large effect size, F (15, 407) = 9.738, Pillai's trace = 0.264, p < .001, =  0.264. Tables A2 and A3 show raw, scaled, or t-scores of neuropsychological performance reported by AN and HC groups as well as by clusters.

Table 2.

Neuropsychological performance between AN and HC groups in Z scores

AN (N = 253) HC (N = 170) Test statistica Effect sizeb
n Mean SD n Mean SD F p
Immediate recall 253 −0.65 1.11 170 −0.58 1.18 0.461 .498 0.001
Delayed recall 253 −0.69 1.16 170 −0.53 1.19 1.942 .164 0.005
Recognition trial 253 −0.46 1.10 170 −0.20 1.23 5.235 .023 0.012
Order of construction index 253 0.28 1.14 170 −0.10 0.96 12.750 .000 0.029
Style index 253 0.14 0.95 170 −0.08 1.00 5.576 .019 0.013
Central coherence index 253 0.43 0.89 170 −0.02 1.01 23.312 .000 0.052
VF—Letters 253 0.92 1.16 170 0.07 0.96 63.353 .000 0.131
VP—Categories 253 1.16 1.02 170 0.89 1.08 6.759 .010 0.016
VF—Flexibility total 253 1.23 1.13 170 0.80 1.01 16.525 .000 0.038
VF—Flexibility accuracy 253 1.23 1.07 170 0.89 0.90 11.327 .001 0.026
CWI Condition 3 253 0.22 1.11 170 0.40 0.57 3.719 .054 0.003
CWI Condition 4 253 0.38 0.90 170 0.48 0.73 1.473 .226 0.009
TT 253 0.02 0.70 170 0.19 0.78 5.007 .026 0.012
TMT 253 −0.15 0.95 170 0.14 0.80 11.160 .001 0.026
Brixton 253 −0.25 1.15 170 −0.05 0.86 3.783 .052 0.009
AN (N = 253) HC (N = 170) Test statistica Effect sizeb
n Mean SD n Mean SD F p
Immediate recall 253 −0.65 1.11 170 −0.58 1.18 0.461 .498 0.001
Delayed recall 253 −0.69 1.16 170 −0.53 1.19 1.942 .164 0.005
Recognition trial 253 −0.46 1.10 170 −0.20 1.23 5.235 .023 0.012
Order of construction index 253 0.28 1.14 170 −0.10 0.96 12.750 .000 0.029
Style index 253 0.14 0.95 170 −0.08 1.00 5.576 .019 0.013
Central coherence index 253 0.43 0.89 170 −0.02 1.01 23.312 .000 0.052
VF—Letters 253 0.92 1.16 170 0.07 0.96 63.353 .000 0.131
VP—Categories 253 1.16 1.02 170 0.89 1.08 6.759 .010 0.016
VF—Flexibility total 253 1.23 1.13 170 0.80 1.01 16.525 .000 0.038
VF—Flexibility accuracy 253 1.23 1.07 170 0.89 0.90 11.327 .001 0.026
CWI Condition 3 253 0.22 1.11 170 0.40 0.57 3.719 .054 0.003
CWI Condition 4 253 0.38 0.90 170 0.48 0.73 1.473 .226 0.009
TT 253 0.02 0.70 170 0.19 0.78 5.007 .026 0.012
TMT 253 −0.15 0.95 170 0.14 0.80 11.160 .001 0.026
Brixton 253 −0.25 1.15 170 −0.05 0.86 3.783 .052 0.009

Notes: All scores are Z scaled. AN = anorexia nervosa; HC = healthy control; N = number of participants in full sample; n = number of participants in subsample; SD = standard deviation; VF = verbal fluency; TMT = Trail Making Test; CWI = Color–Word Interference; TT = Tower Test.

aAnalysis of variance.

bEffect size is partial eta squared ($ηp2$). Strength of effect size as follows: small 0.01, medium 0.06 and large 0.14 (Cohen, 1988).

Cluster Analysis

AN sample cluster analysis

Agglomeration coefficients generated by the hierarchical cluster analysis revealed a demarcation point between three- and four-cluster solutions, suggesting that a three-cluster solution best distinguished the cases; this was confirmed by inspection of the dendrogram. K-means cluster analysis was used to generate the three-cluster solution. The size of the clusters were as follows: Cluster 1: = 49 (19%); Cluster 2: n = 83 (33%); Cluster 3: n = 121 (48%).

Table 3 shows the cluster-group mean Z score and standard deviation for each of the 15 neuropsychological variables. ANOVA determined main effects of cluster for each neuropsychological variable. There were significant differences between clusters on all neuropsychological tests except the Tower Test. The neuropsychological profiles of the three AN clusters are represented by a line graph in Fig. 1. The cluster profiles can be described as follows: AN Cluster 1 represented the smallest proportion (19%) with scores across functions in the average range. However, scores on seven tasks (all three memory trials and inhibition/flexibility) were below a Z score of −0.50. This cluster did not show any strengths with all scores below 0.58. Hence, this group was termed “neuropsychologically low average to average.” AN Cluster 2 (33%) showed weakness in visuo-spatial memory (all three memory trials were in the average to low average range Z = −0.69 to −1.57) and relative strengths in verbal fluency and verbal flexibility (Z scores ranged 0.92–1.07) All other functions were in the average range. Hence, this group was labeled “visuo-spatial/verbal discrepancy.” In contrast, AN Cluster 3 represented the largest proportion of the full sample (48%) and showed strengths in verbal fluency and verbal flexibility. All other scores were in the average range. This group was termed “verbally strong and neuropsychologically average to high average.”.

Fig. 1.

Canonical discriminant functions of participant cluster membership in (a) the AN group and (b) the HC group. Notes: AN clusters: Function 1 = “Visuo-spatial processing,” Function 2 = “Verbal fluency, verbal flexibility and inhibition”; HC clusters: Function = “Immediate Recall and Delayed Recall.”

Fig. 1.

Canonical discriminant functions of participant cluster membership in (a) the AN group and (b) the HC group. Notes: AN clusters: Function 1 = “Visuo-spatial processing,” Function 2 = “Verbal fluency, verbal flexibility and inhibition”; HC clusters: Function = “Immediate Recall and Delayed Recall.”

Table 3.

Neuropsychological performance of AN group shown by three clusters

AN clusters mean scores (N = 253) Test statistica Effect sizeb
1 (n = 49, 19%) 2 (n = 83, 33%) 3 (n = 121, 48%)
Mean SD Mean SD Mean SD F p
Immediate recall −1.07 0.88 −1.48 0.87 0.08 0.81 90.125 .000 0.419
Delayed recall −1.07 1.02 −1.57 0.92 0.07 0.80 89.277 .000 0.417
Recognition trial −0.72 1.39 −0.69 1.00 −0.20 0.98 6.741 .001 0.051
Order of construction index 0.58 0.87 −0.66 1.07 0.79 0.85 63.611 .000 0.337
Style index 0.02 0.92 −0.51 0.75 0.64 0.79 51.077 .000 0.290
Central coherence index 0.38 0.75 −0.30 0.80 0.95 0.61 77.413 .000 0.382
VF—Letters −0.26 0.87 0.91 1.07 1.41 0.96 50.802 .000 0.289
VP—Categories 0.18 0.85 1.16 0.96 1.55 0.84 42.163 .000 0.252
VF—Flexibility total −0.08 0.90 1.44 0.99 1.62 0.89 62.713 .000 0.334
VF—Flexibility accuracy 0.01 0.91 1.46 0.92 1.57 0.85 57.975 .000 0.317
CWI Condition 3 −0.51 1.22 0.61 0.61 0.58 0.68 38.161 .000 0.234
CWI Condition 4 −0.73 1.44 0.27 1.03 0.58 0.72 29.729 .000 0.192
TT −0.12 0.72 0.00 0.71 0.10 0.69 1.729 .180 0.014
TMT −0.94 1.06 −0.05 0.68 0.10 0.89 26.260 .000 0.174
Brixton −0.53 1.20 −0.49 1.21 0.03 1.03 7.042 .001 0.053
AN clusters mean scores (N = 253) Test statistica Effect sizeb
1 (n = 49, 19%) 2 (n = 83, 33%) 3 (n = 121, 48%)
Mean SD Mean SD Mean SD F p
Immediate recall −1.07 0.88 −1.48 0.87 0.08 0.81 90.125 .000 0.419
Delayed recall −1.07 1.02 −1.57 0.92 0.07 0.80 89.277 .000 0.417
Recognition trial −0.72 1.39 −0.69 1.00 −0.20 0.98 6.741 .001 0.051
Order of construction index 0.58 0.87 −0.66 1.07 0.79 0.85 63.611 .000 0.337
Style index 0.02 0.92 −0.51 0.75 0.64 0.79 51.077 .000 0.290
Central coherence index 0.38 0.75 −0.30 0.80 0.95 0.61 77.413 .000 0.382
VF—Letters −0.26 0.87 0.91 1.07 1.41 0.96 50.802 .000 0.289
VP—Categories 0.18 0.85 1.16 0.96 1.55 0.84 42.163 .000 0.252
VF—Flexibility total −0.08 0.90 1.44 0.99 1.62 0.89 62.713 .000 0.334
VF—Flexibility accuracy 0.01 0.91 1.46 0.92 1.57 0.85 57.975 .000 0.317
CWI Condition 3 −0.51 1.22 0.61 0.61 0.58 0.68 38.161 .000 0.234
CWI Condition 4 −0.73 1.44 0.27 1.03 0.58 0.72 29.729 .000 0.192
TT −0.12 0.72 0.00 0.71 0.10 0.69 1.729 .180 0.014
TMT −0.94 1.06 −0.05 0.68 0.10 0.89 26.260 .000 0.174
Brixton −0.53 1.20 −0.49 1.21 0.03 1.03 7.042 .001 0.053

Notes: All scores are Z scaled. AN = anorexia nervosa; N = number of participants in full sample; n = number of participants in subsample; SD = standard deviation; VF = verbal fluency; TMT = Trial Making Test; CWI = Color–Word Interference; TT = Tower Test.

aAnalysis of variance.

bEffect size is partial eta squared ($ηp2$). Strength of effect size as follows: small 0.01, medium 0.06 and large 0.14 (Cohen, 1988).

AN sample discriminant function analysis

Significant mean differences between the three clusters were observed for all the predictors except Towers Task. Box's M indicated that the assumption of equality of covariance matrices was violated (Box's M = 439.451, F (240, 71659.284) = 1.65, p < .001). However, given the large sample, this problem was not regarded as serious. Two functions were generated to separate the three clusters. The first function accounted for 62% of the variance between clusters (Wilk's λ =  0.165, p < .001). The second function accounted for 38% of between cluster variance (Wilk's λ =  0.467, p < .001). The structure matrix revealed predictors with the following discriminant loadings on each function. Function 1 loads on “visuo-spatial processing”: Rey-IR (0.619), Rey-DR (0.612), Central Coherence Index (0.533), and Order of Construction Index (0.446). Function 2 loads on “verbal fluency, verbal flexibility and inhibition”: Verbal Fluency—Flexibility Accuracy (0.556), Verbal Fluency—Flexibility Total (0.549), CWI3 (0.473), and Verbal Fluency—Letters (0.407).

The overall correct classification rate was 81.6%, 94%, and 99.2% for Clusters 1, 2, and 3, respectively. The cross-validation showed that overall 90% were correctly classified; specifically 71.4%, 86.7%, and 97.5% for Clusters 1–3, respectively. The hit ratio was greater than 25% above what would be achieved by chance alone (three clusters would give a 33% chance of correct classification) suggesting that the two functions have good predictive accuracy. Figure 2a shows a scatterplot of the distribution of discriminant scores for Function 1 “Visuo-spatial processing” and 2 “Verbal Fluency, verbal flexibility and inhibition” across the three clusters. These findings support the cluster analysis, the discriminant function analysis providing additional support that the three clusters in children and adolescents were maximally separated by verbal fluency, verbal flexibility, inhibition, and visuo-spatial processing.

Fig. 2.

Profile plot of AN and HC cluster mean scores (with 95% confidence interval) across neuropsychological tasks. Notes: AN = anorexia nervosa; HC = healthy control; VF = verbal fluency; TMT = Trail Making Test; CWI = Color–Word Interference; TT = Tower Test. All scores are Z scaled.

Fig. 2.

Profile plot of AN and HC cluster mean scores (with 95% confidence interval) across neuropsychological tasks. Notes: AN = anorexia nervosa; HC = healthy control; VF = verbal fluency; TMT = Trail Making Test; CWI = Color–Word Interference; TT = Tower Test. All scores are Z scaled.

HC cluster analysis

Agglomeration coefficients generated by the hierarchical cluster analysis revealed a demarcation point between two- and three-cluster solutions, suggesting that a two-cluster solution best distinguished the control participants; this was confirmed by inspection of the dendrogram. K-means cluster analysis was used to generate the two-cluster solution. HC Cluster 1 had 82 participants (48%), whereas HC Cluster 2 had 88 participants (52%). ANOVAs revealed that there was a significant difference between clusters based on all tasks except Verbal Fluency—Letters, CWI3 and 4, Tower Test and Brixton (see Table 4). Participants in HC Cluster 1 had poorer visuo-spatial memory scores (Z scores ranged from −0.46 to −1.37) and relatively higher verbal fluency scores (Z scores ranged from 0.19 to 1.14). HC Cluster 2, however, had scores within the average range on all neuropsychological tasks. The neuropsychological profiles of the HC clusters are also represented by the line graph shown in Figure 1.

Table 4.

Neuropsychological performance of HC group shown by two clusters

HC cluster scores (N = 170) Test statistica Effect sizeb
1 (n = 82, 48%) 2 (n = 88, 52%)
Mean SD Mean SD F p
Immediate recall −1.37 0.91 0.17 0.89 124.711 .000 0.426
Delayed recall −1.32 0.88 0.21 0.94 120.444 .000 0.418
Recognition trial −0.46 1.35 0.04 1.06 7.267 .008 0.041
Order of construction index −0.55 0.89 0.32 0.81 44.497 .000 0.209
Style index −0.57 0.93 0.37 0.84 47.170 .000 0.219
Central coherence index −0.51 0.71 0.44 1.03 48.442 .000 0.224
VF—Letters 0.19 0.81 −0.05 1.07 2.680 .104 0.016
VP—Categories 1.14 1.10 0.65 1.01 9.237 .003 0.052
VF—Flexibility total 1.02 1.10 0.65 1.01 7.769 .006 0.044
VF—Flexibility accuracy 1.13 0.92 0.67 0.82 11.594 .001 0.065
CWI Condition 3 0.54 0.78 0.43 0.68 0.937 .334 0.006
CWI Condition 4 0.42 0.59 0.38 0.55 0.261 .610 0.002
TT 0.12 0.70 0.25 0.86 1.027 .312 0.006
TMT 0.28 0.69 0.02 0.86 4.825 .029 0.028
Brixton −0.17 0.96 0.07 0.75 3.271 .072 0.019
HC cluster scores (N = 170) Test statistica Effect sizeb
1 (n = 82, 48%) 2 (n = 88, 52%)
Mean SD Mean SD F p
Immediate recall −1.37 0.91 0.17 0.89 124.711 .000 0.426
Delayed recall −1.32 0.88 0.21 0.94 120.444 .000 0.418
Recognition trial −0.46 1.35 0.04 1.06 7.267 .008 0.041
Order of construction index −0.55 0.89 0.32 0.81 44.497 .000 0.209
Style index −0.57 0.93 0.37 0.84 47.170 .000 0.219
Central coherence index −0.51 0.71 0.44 1.03 48.442 .000 0.224
VF—Letters 0.19 0.81 −0.05 1.07 2.680 .104 0.016
VP—Categories 1.14 1.10 0.65 1.01 9.237 .003 0.052
VF—Flexibility total 1.02 1.10 0.65 1.01 7.769 .006 0.044
VF—Flexibility accuracy 1.13 0.92 0.67 0.82 11.594 .001 0.065
CWI Condition 3 0.54 0.78 0.43 0.68 0.937 .334 0.006
CWI Condition 4 0.42 0.59 0.38 0.55 0.261 .610 0.002
TT 0.12 0.70 0.25 0.86 1.027 .312 0.006
TMT 0.28 0.69 0.02 0.86 4.825 .029 0.028
Brixton −0.17 0.96 0.07 0.75 3.271 .072 0.019

Notes: All scores are Z scaled. HC = healthy control; N = number of participants in full sample; n = number of participants in subsample; SD = standard deviation; VF = verbal fluency; TMT = Trail Making Test; CWI = Color–Word Interference; TT = Tower Test.

aAnalysis of variance.

bEffect size is partial eta squared ($ηp2$). Strength of effect size as follows: small 0.01, medium 0.06 and large 0.14 (Cohen, 1988).

HC discriminant function analysis

Box's M showed that the assumption of quality of covariance matrices were violated (Box's M = 604.039, F (120, 86614.132) = 4.56, p < .001), though the large sample size limits the seriousness of this violation. One function was generated which accounted for 100% of variance between clusters (Wilk's λ = 0.293, p < .001). The structure matrix shows that the most important predictors were Immediate Recall (0.621) and Delayed Recall (0.373). The overall correct classification rate was 98.8% and 98.9% for Clusters 1 and 2, respectively. Cross-validation showed that overall 95.3% were correctly classified; specifically 93.9% and 96.6% for HC Clusters 1 and 2, respectively. The hit ratio was greater than 25% above what would be achieved by chance alone (two clusters would give a 50% chance of correct classification), suggesting the function had good predictive accuracy. Figure 2b shows histograms of the discriminant function score for the function.

Exploration of cluster analysis results

A MANOVA explored the independent variables of AN and HC clusters on the dependent variables of the 15 neuropsychological tasks. There was a main effect of clusters on neuropsychological tasks with a large effect size, F (60, 1628) = 14.361, p < .001, Pillai's Trace = 1.384, $ηp2$ = 0.346. Box's M test was significant (Box's M = 1353.743, F (480,178284.372) = 2.59, p < .001) which suggests that there were significant differences among the regions in the covariance matrices; this may indicate an increased possibility of Type 1 error. However, the observed power was generally strong (power to detect effects ranged from 0.62 to 1) thus limiting this possibility. Table 5 details post-hoc Bonferroni analyses of significant differences between neuropsychological tasks and clusters. AN versus HC clusters generally differed on verbal fluency, verbal flexibility, and central coherence. AN Cluster 1 was significantly different to the HC clusters on 9 out of the 15 neuropsychological tests, specifically central coherence, verbal abilities, and flexibility as measured by Trails. In addition, AN Cluster 2 versus HC Cluster 2 differed on 11 out of 15 tasks, specifically visuo-spatial processing (memory and central coherence), verbal abilities and flexibility as measured by the Brixton.

Table 5.

Post-hoc Bonferroni analysis showing significant differences between neuropsychological tasks and AN–HC clusters

HC clusters Number of significant differences out of six clusters
AN clusters
Immediate recall — — ** (HC↓) ** (AN↓) ** (AN↓) —
Delayed recall — — ** (HC↓) ** (AN↓) ** (AN↓) —
Recognition trial — — — ** (AN↓) ** (AN↓) —
Order of construction index ** (HC↓) — ** (HC↓) — ** (AN↓) ** (HC↓)
Style index ** (HC↓) — ** (HC↓) — ** (AN↓) —
Central coherence index ** (HC↓) — ** (HC↓) — ** (AN↓) ** (HC↓)
VF—Letters — ** (HC↓) ** (HC↓) — ** (HC↓) ** (HC↓)
VP—Categories ** (AN↓) — * (HC↓) — * (HC↓) ** (HC↓)
VF—Flexibility total ** (AN↓) * (HC↓) ** (HC↓) ** (AN↓) ** (HC↓) ** (HC↓)
VF—Flexibility accuracy ** (AN↓) — ** (HC↓) ** (AN↓) ** (HC↓) ** (HC↓)
CWI Condition 3 ** (AN↓) — — ** (AN↓) — —
CWI Condition 4 ** (AN↓) — — ** (AN↓) — —
TT — — — — — —
TMT ** (AN↓) — — ** (AN↓) — —
Brixton — — — * (AN↓) ** (AN↓) —
Number of significant differences out of 15 tasks 11
HC clusters Number of significant differences out of six clusters
AN clusters
Immediate recall — — ** (HC↓) ** (AN↓) ** (AN↓) —
Delayed recall — — ** (HC↓) ** (AN↓) ** (AN↓) —
Recognition trial — — — ** (AN↓) ** (AN↓) —
Order of construction index ** (HC↓) — ** (HC↓) — ** (AN↓) ** (HC↓)
Style index ** (HC↓) — ** (HC↓) — ** (AN↓) —
Central coherence index ** (HC↓) — ** (HC↓) — ** (AN↓) ** (HC↓)
VF—Letters — ** (HC↓) ** (HC↓) — ** (HC↓) ** (HC↓)
VP—Categories ** (AN↓) — * (HC↓) — * (HC↓) ** (HC↓)
VF—Flexibility total ** (AN↓) * (HC↓) ** (HC↓) ** (AN↓) ** (HC↓) ** (HC↓)
VF—Flexibility accuracy ** (AN↓) — ** (HC↓) ** (AN↓) ** (HC↓) ** (HC↓)
CWI Condition 3 ** (AN↓) — — ** (AN↓) — —
CWI Condition 4 ** (AN↓) — — ** (AN↓) — —
TT — — — — — —
TMT ** (AN↓) — — ** (AN↓) — —
Brixton — — — * (AN↓) ** (AN↓) —
Number of significant differences out of 15 tasks 11

Notes: HC = healthy control; AN = anorexia nervosa; VF = verbal fluency; TMT = Trail Making Test; CWI = Color–Word Interference; TT = Tower Test; ↓, poorer performance. *p ≤ .05, **p ≤ .001.

Clinical symptoms across clusters

As shown in Table 6, a MANOVA explored the independent variables of AN and HC clusters on the dependent variables of BMI centile, EDE subscale and global scores, and depression, anxiety, and OC symptoms. There was a significant main effect of cluster on clinical symptoms with a large effect size, F (36, 1304) = 9.544, p < .001, Pillai's Trace = 0.834,  $ηp2$ = 0.209. Box's M test was significant (Box's M = 804.055, F (180, 138653.392) = 4.192, p < .001) which suggests there were significant differences among the regions in the covariance matrices and thus increased possibility of Type 1 error. However, the observed power was strong limiting the risk (power to detect the effect was 1 for all variables). Post-hoc Bonferroni analyses showed that the AN clusters differed to the HC clusters on BMI centile, EDE subscale and global scores, and depression, anxiety, and OC symptoms. No post-hoc differences were found between AN clusters only nor between HC clusters only.

Table 6.

Clinical characteristics of AN and HC clusters

AN (N = 253) HC (N = 170) Test statistica Effect sizeb
1 (n = 49, 19%) 2 (n = 82, 33%) 3 (n = 121, 48%) 1 (n = 82, 48%) 2 (n = 88, 52%) F p
BMI centile 6.75 (12.08) 6.66 (11.08) 8.55 (13.67) 63.35 (28.01) 55.75 (25.81) 141.427 .000 0.631
EDE Restraint 3.24 (2.01) 3.50 (1.92) 2.99 (1.81) 0.78 (1.08) 0.54 (0.81) 54.658 .000 0.398
EDE Eating Concern 2.91 (1.98) 3.07 (1.75) 2.70 (1.73) 0.40 (0.68) 0.40 (0.73) 59.395 .000 0.418
EDE Weight Concern 3.41 (2.03) 3.68 (1.99) 3.24 (1.93) 1.22 (1.32) 0.89 (1.12) 40.476 .000 0.328
EDE Shape Concern 3.71 (1.87) 4.08 (1.92) 3.61 (1.91) 1.30 (1.28) 1.12 (1.18) 49.933 .000 0.376
EDE global score 3.31 (1.71) 3.57 (1.74) 3.14 (1.67) 0.92 (0.98) 0.76 (0.89) 61.838 .000 0.428
Depression Z score 1.85 (1.48) 1.77 (1.37) 1.78 (1.37) −0.48 (0.77) −0.53 (0.84) 75.723 .000 0.478
Anxiety Z score 1.13 (1.14) 1.37 (1.19) 1.25 (1.16) −0.55 (0.68) −0.50 (0.67) 65.331 .000 0.441
OC Z score 2.15 (1.16) 1.71 (1.54) 1.79 (1.55) 0.09 (1.01) 0.05 (0.97) 39.097 .000 0.321
AN (N = 253) HC (N = 170) Test statistica Effect sizeb
1 (n = 49, 19%) 2 (n = 82, 33%) 3 (n = 121, 48%) 1 (n = 82, 48%) 2 (n = 88, 52%) F p
BMI centile 6.75 (12.08) 6.66 (11.08) 8.55 (13.67) 63.35 (28.01) 55.75 (25.81) 141.427 .000 0.631
EDE Restraint 3.24 (2.01) 3.50 (1.92) 2.99 (1.81) 0.78 (1.08) 0.54 (0.81) 54.658 .000 0.398
EDE Eating Concern 2.91 (1.98) 3.07 (1.75) 2.70 (1.73) 0.40 (0.68) 0.40 (0.73) 59.395 .000 0.418
EDE Weight Concern 3.41 (2.03) 3.68 (1.99) 3.24 (1.93) 1.22 (1.32) 0.89 (1.12) 40.476 .000 0.328
EDE Shape Concern 3.71 (1.87) 4.08 (1.92) 3.61 (1.91) 1.30 (1.28) 1.12 (1.18) 49.933 .000 0.376
EDE global score 3.31 (1.71) 3.57 (1.74) 3.14 (1.67) 0.92 (0.98) 0.76 (0.89) 61.838 .000 0.428
Depression Z score 1.85 (1.48) 1.77 (1.37) 1.78 (1.37) −0.48 (0.77) −0.53 (0.84) 75.723 .000 0.478
Anxiety Z score 1.13 (1.14) 1.37 (1.19) 1.25 (1.16) −0.55 (0.68) −0.50 (0.67) 65.331 .000 0.441
OC Z score 2.15 (1.16) 1.71 (1.54) 1.79 (1.55) 0.09 (1.01) 0.05 (0.97) 39.097 .000 0.321

Notes: All scores are Z scaled. AN = anorexia nervosa; HC = healthy control; N = number of participants in full sample; n = number of participants in subsample; SD = standard deviation; BMI = Body Mass Index; EDE = Eating Disorder Examination; OC = obsessive–compulsive.

aAnalysis of variance.

bEffect size is partial eta squared ($ηp2$). Strength of effect size as follows: small 0.01, medium 0.06 and large 0.14 (Cohen, 1988).

Discussion

The primary aim of this study was to identify whether discrete neuropsychological profiles could be identified in female children and adolescents with AN, and whether these profiles were similar to those in a healthy female child and adolescent comparison group. Three distinct clusters were identified in the AN sample and two distinct clusters in the HC sample. In the AN group, AN Cluster 1 (19%, neuropsychologically low average to average) was characterized by low scores in verbal fluency and executive functioning with significantly lower scores on all but one (TT) of these tasks. This group had the poorest verbal skills compared with the other AN and HC clusters which were significantly different on all four verbal tasks. This cluster stood out as being significantly different to the HC clusters on more of the neuropsychological tests than the other AN clusters. AN Cluster 2 (33%, visuo-spatial/verbal discrepancy) was characterized by the largest visual/verbal skills performance discrepancy (2.02 SDs) of the AN and HC groups. AN Cluster 3 represented the largest proportion and was characterized by average to high average performance (48%, verbally strong and neuropsychologically average to high average). Discriminant function analysis determined the strongest distinguishing characteristic was visuo-spatial processing (both recall trials and central coherence). Strong verbal fluency, verbal flexibility, and inhibition were also distinguishing characteristics, as seen from the discriminant functions. While the between-subjects ANOVAs showed significant differences between AN clusters in planning and three measures of flexibility, the discriminant function analyses showed the TT, TMT, and Brixton to not significantly contribute to the AN clusters. Two clusters emerged in the HC sample. HC Cluster 1 (48%) was characterized by poorer visuo-spatial memory scores and higher verbal fluency scores. HC Cluster 2 (52%) demonstrated neuropsychological performance within the average range on all neuropsychological tasks. Discriminant function analysis suggested the strongest predictor of HC clusters was visuo-spatial memory.

There was a main effect of neuropsychological performance between AN and HC groups. Comparison of overall group means indicate that visuo-spatial memory recall and inhibition did not significantly differ. Those with AN showed poorer performance on recognition memory and planning compared to controls, whereas central coherence and verbal fluency performance were significantly greater than controls. Cognitive flexibility performance was mixed, in that verbal flexibility was significantly greater in those with AN, whereas the Trail Making Test revealed significantly poorer performance in those with AN. Performance on the Brixton Task did not significantly differ between groups. However, it should be noted though that while significant differences emerged, differences in absolute scores were all within ±1.23 SD of the norm (i.e., Z score of zero), thus these data do not indicate “impaired” functioning in the assessed neuropsychological domains. Arguably, this may call for a change in perspective in the field of eating disorders as it appears that children and young people who were ill with AN (over 84% requiring inpatient care) and the effects of starvation were able to complete over 1½ hr of neuropsychological testing with results demonstrating attainment scores within 2 SDs of the norm.

Inspection of the post-hoc differences between clusters showed strongest evidence for those with AN to show significantly greater verbal flexibility and fluency than controls. Central coherence performance was generally weaker in controls, in that controls generally started the copy trial with less global elements, copied key global elements in a fragmented fashion, and demonstrated a more local processing copy style compared to those with AN. With regard to visuo-spatial memory, inhibition, and cognitive flexibility, AN Clusters 1 and 2 generally showed significantly poorer performance than controls that were masked by comparison of the overall group means. Interestingly, AN Cluster 2 and HC Cluster 1 demonstrated a similar response pattern of weaker visuo-spatial skills and stronger verbal skills. In addition, AN Cluster 3 and HC Cluster 2 also showed similar performance in that these two clusters showed average or higher neuropsychological performance. These similarities highlight the importance of a HC sample to identify the clinically meaningful patterns of responding. When bearing in mind differences between AN and HC clusters, it appears that AN Cluster 1 is of most interest given the consistent significant differences on visuo-spatial skills (both central coherence and memory), verbal fluency and flexibility as measured by the TMT.

The secondary aim of this study explored the relationship between neuropsychological performance and clinical symptoms. It appears that there was a main effect of neuropsychological performance (cluster membership) on the clinical symptoms of BMI centile, EDE subscale and global score, anxiety, depression, and OC symptoms between the AN and HC groups. Thus, when the groups were compared overall, those with AN demonstrated significantly higher scores of AN symptom severity (both higher symptom ratings and lower weight status) and comorbid depression, anxiety, and OC symptoms compared to controls. For the AN group, it is also worth noting the high level of AN symptom severity relative to community norms (Carter, Stewart, & Fairburn, 2001 found global mean score = 1.6, standard deviation = 1.4; median global score in the current study was 3.75). However, when neuropsychological performance was further explored at a post-hoc level between clusters, neither significant differences emerged between AN clusters only nor HC clusters only.

To the authors’ knowledge, this study is the first to conduct neuropsychological profiling using clustering techniques in a large cohort of female children and adolescents with AN. The distinct profiles provide evidence of neuropsychological heterogeneity in AN, that is, there is not one single pattern of weakness nor specific profile that characterizes AN. Furthermore, given (a) all neuropsychological performance was within ±1.23 SDs of the norm, (b) the majority of effect sizes between AN and HC groups were in the small range, and (c) the greatest discrepancy between groups was found between only two domains—visuo-spatial memory and verbal fluency, these suggest the neuropsychological differences between those with AN and HCs are relatively subtle. This is in-line with other profiling studies that have focused on children and adolescents with AN (Andrés-Perpiña et al., 2011; Calderoni et al., 2013; Kjaersdam Telléus et al., 2015).

The neurobiological correlates of the neuropsychological variability are not well understood, but it is possible that our findings may represent different underlying pathophysiologies of AN. The characteristic behavioral, cognitive and emotional features of AN could arise as a “consequence” of underlying neuropsychological function. Alternatively, eating disorder symptoms, such as maintained low weight, could themselves be the “cause” of disruption in neural mechanisms. Subsequently, this might lead to interference with cognitive functioning, and hence to secondary neuropsychological dysfunction.

Our study supports the general finding in the AN literature of weak to borderline performance in visuo-spatial memory, as shown by AN Cluster 2. However, executive functions, namely inhibition, cognitive flexibility, and planning, did not emerge as weaknesses relative to normative data, as evidenced by the average performance seen across all AN clusters. Some reviews (Lena et al., 2004; Zakzanis et al., 2010) have suggested poor verbal fluency in those with AN. However, in-line with a meta-analysis of verbal fluency in AN (Stedal et al., 2011), those in AN Cluster 3 (48% of the AN group) were distinguished by strong verbal skills. Verbal fluency in the remaining clusters was in the average range relative to population norms. Similar to Renwick et al. (2015), three distinct clusters emerged from a sample of AN patients based on neuropsychological performance. Two measures used in their study were also used in this study—the Brixton task and central coherence based on the Rey. Interestingly, the lowest scores on both of these tasks across clusters in the Renwick study (−1.74 and −1.26, respectively) were much lower than those within this study (−0.53 and −0.30, respectively). Furthermore, the Central Coherence Index score in this study was significantly higher in those with AN compared to controls suggesting overall those with AN showed a greater degree of efficient global processing than age-matched healthy peers, a finding that goes against the literature (Lang & Tchanturia, 2014c). This may suggest that age (or more specifically, illness chronicity) influenced neuropsychological functioning, either as a state or trait phenomenon, that is, limited its development. Supporting evidence for this comes from Shott et al. (2012), who report that duration of illness affects onset-shifting ability. Regarding specific clinical characteristics, findings were generally consistent with those in the current literature. Several studies have shown a high degree of comorbidity with AN similarly identified within this study, for example, Godart et al. (2007). Our findings support the independence of neuropsychological weaknesses from symptom severity (Mikos et al., 2008) or weight status (Moser et al., 2003). Independence of clinical symptoms on neuropsychological performance between AN clusters was also found by Renwick et al. (2015). In terms of IQ, a systematic review suggested that individuals with AN performed in the high average range (Lopez, Stahl, & Tchanturia, 2010). It is apparent from our findings that within a large cohort of patients there was a range of IQ performance from average to high average.

The finding that planning and two measures of flexibility, TMT and Brixton, did not significantly predict AN cluster-group membership, despite all three showing significant post-hoc differences between AN and HC clusters, is particularly intriguing given the general consensus in the adult AN literature of weaknesses in these domains. This finding may in part be due to this study focusing on children and adolescents. For example, Reville et al. (2016) highlighted 15 studies that assessed the cognitive construct of decision-making and planning and found evidence of generally poorer performance in those with AN compared to controls. Zakzanis et al. (2010) found a medium effect size in their meta-analysis of problem solving studies (Cohen's d = 0.43, 95% confidence interval 0.07–0.79). In contrast, studies that have assessed planning ability in younger individuals with AN have shown TT scores to be within 1 SD of healthy population norms at pretreatment (Dahlgren, Lask, Landrø, & Ro, 2013), negative findings of gross impairments when compared to HCs (Dmitrzak-Weglarz et al., 2013) and that inefficient decision-making is independent of a diagnosis of AN (Fornasari et al., 2014). Regarding cognitive flexibility in adults with AN, a meta-analysis of 15 studies (Roberts et al., 2007; 2 included studies used adolescents) report generally poorer performance in those with AN compared to HCs (small-to-large effect sizes). However, a review by Zakzanis et al. (2010) report flexibility differences to be in the small-to-medium effect size range and suggest that previous research “…may have overemphasized the reliability of set-shifting impairments in patients with AN…” (p. 101). Similarly, Wu et al. (2014) reported a medium effect size for restricting type AN. This contrasts to reviews of set-shifting in younger populations which appear to find a lack of evidence of significant set-shifting differences between those with AN and HCs (Lang et al., 2014b; Westwood et al., 2016).

It has been recommended by Zakzanis (2001) that interpretation of effect sizes in the field of neuropsychological research should be done with caution. Effect size cut offs are essentially based on Cohen's review of treatment efficacy literature in psychology and their relative typicality in relation to a broad range of social sciences literature (Cohen, 1988). In the study of relationships between brain and behavior, Zakzanis argues such cut offs as used to broadly define social sciences research may be inappropriate. For example, if the difference between two groups on a test score produced an effect size of d = 8 which is considered large, this still equates to an overlap in the distribution of test scores of 52.6%. In other words, only 47.4% of test scores were not obtained by both groups. While one may conclude the magnitude of the effect was “large,” clearly such a test would not be reliably sensitive enough to distinguish between these two groups. Zakzanis suggests an effect size d > 3.0 be used as an appropriate cut-off to indicate markers in neuropsychological research which equates to an overlap of 7.2%, that is, 92.8% of scores would be unique. With regard to this study, the use of ANOVA's meant reporting the partial eta squared effect size was which represents the proportion of the variance of a particular dependent variable that can be explained by the independent variable while controlling for the other dependent variables. Inspection of overall group means on the neuropsychological tasks show that only the group difference in verbal fluency was of a large effect size (0.131), yet this difference only accounts for 13% of the variance between AN and HC groups. The difference in CCI approached a medium effect size (0.052), yet this only accounted for 5% of variance between AN and HC groups. Converting F to Cohen's d using the formula $d=2Fdf(error)$ allows an overlap percentage to be obtained. Verbal fluency d = 0.72 and CCI d =  0.50 correspond to a 57/66.6% (respectively) overlap in the distribution of scores. Thus, 43% of verbal fluency scores and 33.4% of CCI scores were not obtained by both groups.

There are several important limitations to this study. (a) Diagnostic subtypes were not recorded. Eating Disorder Examination data were collected at the subscale level only, not at the item level, thus subtype could not be ascertained from the individual's EDE. One review noted that the majority of studies did not separate their findings by restrictive and purging subtypes which may distort or mask results (Duchesne et al., 2004). However, the reliability of the binge/purge versus restrictive distinction has been deemed questionable as demonstrated by the high rate of diagnostic drift within eating disorders (Eddy et al., 2008). Arguably, subtypes defined by neuropsychological status may be a potential alternative compared to those defined by “surface” symptoms. However, this would require investigation and validation by other clinicians and researchers. (b) The length of illness was unknown. This is commonly very difficult to measure accurately because some patients are ill for many years before coming to the attention of services, illness onset may be gradual and self-report potentially unreliable to assess retrospectively the exact date of onset. (c) Medication use was not recorded. Psychoactive drugs can affect neuropsychological functioning, and thus may confound the findings of this study. (d) Participants completed a battery of several tasks, thus effort and fatigue may have a bearing on the findings of this study. Every effort was made to ensure fatigue was minimized by administering tests over two or more sessions. However, we did not directly assess this. (e) Due to the nature of recruitment at clinical sites, administration of neuropsychological tests and clinical symptom measures may have been at different time points. For a subset of the patients, it is unknown whether the patient's mood at the time of completion of the psychological measures was similar to the mood on the day of neuropsychological testing. Related to this is the combination of standard scores from diverse instruments to create composite depression, anxiety, and OC symptom scores. This procedure may lack robustness. (f) Assessment of eating disorder psychopathology may have benefited from a more consistent approach. Both EDE interviews and questionnaires were used. Unfortunately, EDE type (interview or questionnaire) was not recorded, thus these data could not be analyzed separately. (g) Missing data were shown to be associated with clinical site, thus these may have influenced the clusters. While there is no clear answer to the optimum management of missing data, we have tried to reduce the impact of these by using Maximum Likelihood imputation. (h) The Ravello Profile battery does not contain direct measures of decision-making, non-visuo-spatial memory, or processing speed. Hence, this study was unable to shed light on whether there were group or cluster differences across these neuropsychological functions, which have previously be identified as weaknesses for AN patients (Jáuregui-Lobera, 2013; Zakzanis et al., 2010). (i) Finally, the cross-sectional survey design of this study has some limitations including sampling bias. The majority of clinical sites in this study were specialist inpatient units, thus individuals with AN treated in community outpatient clinics were under-represented. Nor can any direction of causality be inferred. It is unknown whether the identified neuropsychological profiles represent a consequence of AN or a predisposing risk factor.

In contrast, the study has some strengths. (a) To the best of our knowledge, this is the first study to undertake neuropsychological profiling using a clustering technique in a large cohort of children and adolescents with AN. (b) The benefit of focusing on younger individuals is that results from such studies have the potential to inform early detection and targeted intervention strategies, as already developed in the field of depression (Hermens et al., 2011). (c) The large sample size offers an advantage of statistical power. We reviewed the existing neuropsychological literature at the time the battery was developed. Thirty-four studies incorporated one or more measures used in the neuropsychological battery. Of these, nine met criteria for case-control studies with significant findings. Based on these studies, and adopting 80% power at p < .05, sample sizes for individual tests ranged from n = 12 to n = 125, compared with our AN sample size of 253. (d) The large sample size allows for a degree of heterogeneity. Thus, it could be argued that the sample is more representative of the AN population as a whole than studies based on smaller numbers from just one or two clinics. (e) This study uses both normative data as well as an age- and gender-matched control group. Interpretation of results relative to normative data can offer insight into the actual prevalence of neuropsychological strengths and weaknesses in this clinical population. (f) This study is consistent with the Research Domain Criteria project (RDoC) developed by the US National Institute of Mental Health (NIMH) to define neuroscientific dimensions of functioning, a framework that has been applied to eating disorders (Wildes & Marcus, 2015).

As noted in reviews focusing on young people with AN (Lang et al., 2014b; Lang & Tchanturia, 2014c), there are relatively few studies which limit firm conclusions being drawn about neuropsychological profiles. In-line with the need for further research, this study adds to the growing field of neuropsychological research in children and adolescents with AN. While the findings of this study are preliminary, we cautiously suggest that neuropsychological markers may have the potential to inform the testing of neuroscience models as well as interventions. For example, Marsh, Maia, and Peterson (2009) propose the cortico-striato-thalamo-cortical circuits that are believed to mediate self-regulatory control may be dysfunctional in childhood and adolescence and as such, may be a vulnerability factor to the development of AN. Based on this model, one may expect poorer performance of tasks that assess cognitive and inhibitory control. Thus, research that can shed light on the presence of neuropsychological markers may help to inform hypotheses about development of precise cognitive probes of underlying neural systems. Regarding interventions, the current literature is mixed on whether weaknesses in neuropsychological functioning improve with treatment (Jáuregui-Lobera, 2013), yet there are encouraging findings from interventions which directly target neuropsychological dysfunction through the use of cognitive remediation therapy (see Tchanturia, Lloyd, & Lang, 2013 for an overview). In any event, it may now be possible to investigate the association between neuropsychological clusters with factors such as outcome.

There are several avenues of potential research based on the findings of the current study. The primary focus of this study was to identify and explore whether distinct neuropsychological profiles existed in children and adolescents with AN. We encourage the research community to investigate whether the identified profiles can be replicated in other studies. Future studies may investigate (a) the specificity of the profiles to AN, in terms of the potential clinical utility of the profiles; (b) the association of profiles to factors such as outcome, risk factors, and illness trajectories; (c) to consider the weaknesses of this study as these may influence the expression of neuropsychological profiles; and (d) biological maturation, cognitive developmental trajectory as well as puberty may influence neuropsychological performance, thus investigating age by distinct profile may be of interest.

Conclusion

In conclusion, this study suggests that distinct neuropsychological profiles exist in young people with AN and secondly, the neuropsychological differences between individuals with AN and HCs are relatively subtle. This study also highlights the importance of comparing with an HC sample to identify the clinically meaningful neuropsychological patterns of responding. While our findings require replication, we are intrigued by the prospect that future neuropsychological profiling studies may have the potential to offer an insight into the underlying neurobiology of AN, to guide treatment strategies, and to inform future research.

Table A1.

Characteristics of collaborating sites and number of AN patients recruited

Country Site type n
United Kingdom Independent specialist inpatient unit 48
Independent specialist inpatient unit 26
Independent specialist inpatient unit 29
NHS inpatient and community specialist service 11
NHS outpatient and day care specialist service 10
NHS inpatient and community specialist service 15
Independent specialist inpatient unit 33
Independent specialist inpatient unit 13
Independent specialist outpatient service
NHS inpatient and community specialist service
Switzerland Specialist inpatient unit
Germany Specialist inpatient unit 23
Norway Specialist inpatient unit 28
Specialist inpatient unit
Specialist inpatient unit
Total  253
Country Site type n
United Kingdom Independent specialist inpatient unit 48
Independent specialist inpatient unit 26
Independent specialist inpatient unit 29
NHS inpatient and community specialist service 11
NHS outpatient and day care specialist service 10
NHS inpatient and community specialist service 15
Independent specialist inpatient unit 33
Independent specialist inpatient unit 13
Independent specialist outpatient service
NHS inpatient and community specialist service
Switzerland Specialist inpatient unit
Germany Specialist inpatient unit 23
Norway Specialist inpatient unit 28
Specialist inpatient unit
Specialist inpatient unit
Total  253

Notes: AN = anorexia nervosa; n = number of participants; NHS = UK National Health Service (state-delivered universal healthcare).

Table A2.

Raw/scaled/t-scores of neuropsychological performance of AN group shown by three clusters

Overall (N = 253) AN clusters mean scores
1 (n = 49, 19%) 2 (n = 83, 33%) 3 (n = 121, 48%)
Mean SD Mean SD Mean SD Mean SD
Immediate recall 43.46 11.05 39.35 8.79 35.24 8.72 50.77 8.13
Delayed recall 43.10 11.55 39.31 10.24 34.32 9.23 50.66 7.96
Recognition trial 45.39 11.04 42.84 13.92 43.11 10.00 47.99 9.84
Order of construction index 2.07 0.66 2.25 0.50 1.53 0.62 2.37 0.49
Style index 1.07 0.43 1.02 0.42 0.78 0.34 1.30 0.35
Central coherence index 2.12 0.40 2.10 0.34 1.80 0.36 2.36 0.27
VF—Letters 12.77 3.48 9.21 2.60 12.72 3.22 14.24 2.88
VP—Categories 13.47 3.06 10.53 2.55 13.48 2.87 14.66 2.53
VF—Flexibility total 13.70 3.38 9.75 2.71 14.33 2.96 14.87 2.66
VF—Flexibility accuracy 13.69 3.20 10.03 2.74 14.37 2.75 14.70 2.56
CWI Condition 3 11.14 2.71 8.47 3.65 11.84 1.83 11.73 2.05
CWI Condition 4 10.67 3.33 7.82 4.33 10.80 3.10 11.73 2.17
TT 10.07 2.11 9.64 2.15 10.00 2.13 10.29 2.08
TMT 9.55 2.84 7.17 3.18 9.84 2.05 10.31 2.66
Brixton 16.30 4.16 17.30 4.31 17.17 4.37 15.30 3.73
Overall (N = 253) AN clusters mean scores
1 (n = 49, 19%) 2 (n = 83, 33%) 3 (n = 121, 48%)
Mean SD Mean SD Mean SD Mean SD
Immediate recall 43.46 11.05 39.35 8.79 35.24 8.72 50.77 8.13
Delayed recall 43.10 11.55 39.31 10.24 34.32 9.23 50.66 7.96
Recognition trial 45.39 11.04 42.84 13.92 43.11 10.00 47.99 9.84
Order of construction index 2.07 0.66 2.25 0.50 1.53 0.62 2.37 0.49
Style index 1.07 0.43 1.02 0.42 0.78 0.34 1.30 0.35
Central coherence index 2.12 0.40 2.10 0.34 1.80 0.36 2.36 0.27
VF—Letters 12.77 3.48 9.21 2.60 12.72 3.22 14.24 2.88
VP—Categories 13.47 3.06 10.53 2.55 13.48 2.87 14.66 2.53
VF—Flexibility total 13.70 3.38 9.75 2.71 14.33 2.96 14.87 2.66
VF—Flexibility accuracy 13.69 3.20 10.03 2.74 14.37 2.75 14.70 2.56
CWI Condition 3 11.14 2.71 8.47 3.65 11.84 1.83 11.73 2.05
CWI Condition 4 10.67 3.33 7.82 4.33 10.80 3.10 11.73 2.17
TT 10.07 2.11 9.64 2.15 10.00 2.13 10.29 2.08
TMT 9.55 2.84 7.17 3.18 9.84 2.05 10.31 2.66
Brixton 16.30 4.16 17.30 4.31 17.17 4.37 15.30 3.73

Notes: AN = anorexia nervosa; N = number of participants in full sample; n = number of participants in subsample; SD = standard deviation; VF = verbal fluency; TMT = Trail Making Test; CWI = Color–Word Interference; TT = Tower Test. Score type is as follows: Rey Complex Figure Test = t-score, central coherence measures = raw score, Delis–Kaplin Executive Function System = scaled scores, Brixton Spatial Anticipation Test = raw score.

Table A3.

Raw/scaled/t-scores of the neuropsychological performance of HC group shown by two clusters

Overall (N = 170) HC cluster scores
1 (n = 82, 48%) 2 (n = 88, 52%)
Mean SD Mean SD Mean SD
Immediate recall 44.23 11.83 36.25 9.12 51.66 8.86
Delayed recall 44.72 11.93 36.76 8.81 52.14 9.42
Recognition trial 48.01 12.29 45.43 13.46 50.42 10.61
Order of construction index 1.85 0.55 1.59 0.52 2.09 0.47
Style index 0.97 0.45 0.76 0.42 1.17 0.38
Central coherence index 1.92 0.45 1.70 0.32 2.13 0.47
VF—Letters 10.20 2.87 10.57 2.44 9.85 3.20
VP—Categories 12.67 3.23 13.43 3.30 11.95 3.02
VF—Flexibility total 12.39 3.04 13.05 3.31 11.77 2.64
VF—Flexibility accuracy 12.68 2.70 13.39 2.77 12.02 2.47
CWI Condition 3 11.44 2.20 11.61 2.35 11.28 2.05
CWI Condition 4 11.20 1.71 11.27 1.78 11.14 1.65
TT 10.56 2.35 10.37 2.09 10.74 2.57
TMT 10.43 2.39 10.84 2.08 10.05 2.59
Brixton 15.57 3.10 16.02 3.45 15.16 2.68
Overall (N = 170) HC cluster scores
1 (n = 82, 48%) 2 (n = 88, 52%)
Mean SD Mean SD Mean SD
Immediate recall 44.23 11.83 36.25 9.12 51.66 8.86
Delayed recall 44.72 11.93 36.76 8.81 52.14 9.42
Recognition trial 48.01 12.29 45.43 13.46 50.42 10.61
Order of construction index 1.85 0.55 1.59 0.52 2.09 0.47
Style index 0.97 0.45 0.76 0.42 1.17 0.38
Central coherence index 1.92 0.45 1.70 0.32 2.13 0.47
VF—Letters 10.20 2.87 10.57 2.44 9.85 3.20
VP—Categories 12.67 3.23 13.43 3.30 11.95 3.02
VF—Flexibility total 12.39 3.04 13.05 3.31 11.77 2.64
VF—Flexibility accuracy 12.68 2.70 13.39 2.77 12.02 2.47
CWI Condition 3 11.44 2.20 11.61 2.35 11.28 2.05
CWI Condition 4 11.20 1.71 11.27 1.78 11.14 1.65
TT 10.56 2.35 10.37 2.09 10.74 2.57
TMT 10.43 2.39 10.84 2.08 10.05 2.59
Brixton 15.57 3.10 16.02 3.45 15.16 2.68

Notes: HC = healthy control; N = number of participants in full sample; n = number of participants in subsample; SD = standard deviation; VF = verbal fluency; TMT = Trail Making Test; CWI = Color–Word Interference; TT = Tower Test.Score type is as follows: Rey Complex Figure Test = t-score, central coherence measures = raw score, Delis–Kaplin Executive Function System = scaled scores, Brixton Spatial Anticipation Test = raw score.

Funding

This project has been funded by Care UK, UK; Forest Healthcare, UK; Goldsmiths University of London, UK; Norwegian Government Health Board, Health South East Norway; Regional Avdeling vor Spiseforstyrrelser (RASP), Oslo Universitetssykehus Ullevål, Norway; The Rosetrees Trust, UK; and The Huntercombe Group, UK.

None declared.

Acknowledgements

The authors wish to thank the following collaborators for contributing data to the Ravello Profile study: Barnesenteret, Oslo Universitetssykehus Aker, Gaustad, Norway; Cheshire & Merseyside Eating Disorder Service for Adolescents, Cheshire & Wirral Partnership NHS Foundation Trust, UK; Child and Adolescent Eating Disorders Service, Springfield University Hospital, South West London and St George's Mental Health NHS Trust, London, UK; Child and Adolescent Mental Health Service, Weston General Hospital, Avon and Wiltshire Mental Health Partnership NHS Trust, UK; Child & Family Practice, London; Ellern Mede Eating Disorders Service, Forest Healthcare, London, UK; Feeding and Eating Disorders Service, Great Ormond Street Hospital for Children NHS Foundation Trust, London, UK; Follo BUP, Akershus Universitetssykehus, Røde kors klinikken, Norway; Frimley Childrens Centre, Surrey and Borders Partnership NHS Foundation Trust, UK; Helse Bergen, Haukeland Universitetssykehus, Norway; Huntercombe Hospitals-Maidenhead, Stafford & Edinburgh, The Huntercombe Group, Maidenhead, UK; Institut fur Psychologie, Universitat Basel, Switzerland; Klinik für Psychiatrie, Psychosomatik und Psychotherapie des Kindes-und Jugendalters, Charite, Universitatsmedizin Berlin, Germany; Psychiatry Institute, Hospital das Clínicas da Universidade de São Paulo, Brazil; and Regional Avdeling vor Spiseforstyrrelser (RASP), Oslo Universitetssykehus Ullevål, Oslo, Norway. The authors would like to thank Nick Lange for generating statistical graphics.

References

Abbate-Daga
,
G.
,
Amianto
,
F.
,
Delsedime
,
N.
,
De-Bacco
,
C.
, &
Fassino
,
S.
(
2013
).
Resistance to treatment and change in anorexia nervosa [corrected]: A clinical overview
.
BMC Psychiatry
,
13
,
294
. .
Andrés-Perpiña
,
S.
,
Lozano-Serra
,
E.
,
Puig
,
O.
,
Lera-Miguel
,
S.
,
Lázaro
,
L.
, &
Castro-Fornieles
,
J.
(
2011
).
Clinical and biological correlates of adolescent anorexia nervosa with impaired cognitive profile
.
,
20
(11-12)
,
541
549
. .
Association
,
A. P.
(
1994
).
American Psychiatric Association-Diagnostic and Statistical Manual of Mental Disorders (DSM−IV)
.
Washington DC
:
American Psychiatric Association Press
.
Bailer
,
U.
, &
Kaye
,
W.
(
2003
).
A review of neuropeptide and neuroendocrine dysregulation in anorexia and bulimia nervosa
.
Current Drug Targets: CNS and Neurological Disorders
,
2
(1)
,
53
59
.
Beck
,
A.
,
Steer
,
R.
, &
Brown
,
G.
(
1996
).
Manual for Beck Depression Inventory-II
.
San Antonio, TX
:
Psychological Corporation
.
Beck
,
A.
, &
Steer
,
R. A.
(
1993
).
Beck anxiety inventory manual
.
San Antonio
:
Harcourt Brace and Company
.
Binford
,
R.
,
Le Grange
,
D.
, &
Jellar
,
C.
(
2005
).
Eating Disorders Examination versus Eating Disorders Examination-Questionnaire in adolescents with full and partial-syndrome bulimia nervosa and anorexia nervosa
.
International Journal of Eating Disorders
,
37
(1)
,
44
49
.
Booth
,
R.
(
2006
).
Local-global processing and cognitive style in autism spectrum disorders and typical development (PhD )
.
London
:
King's College, University of London
.
Bryant‐Waugh
,
R.
(
2000
). Overview of the eating disorders. In
,
B.
, &
Bryant‐Waugh
,
R.
(Eds.)
Anorexia nervosa and related eating disorders in childhood and adolescence
.
Hove
:
Psychology Press
.
Bühren
,
K.
,
Mainz
,
V.
,
Herpertz-Dahlmann
,
B.
,
Schäfer
,
K.
,
Kahraman-Lanzerath
,
B.
,
Lente
,
C.
, et al.
(
2012
).
Cognitive flexibility in juvenile anorexia nervosa patients before and after weight recovery
.
Journal of Neural Transmission
,
119
(9)
,
1047
1057
.
Burgess
,
P.
, &
Shallice
,
T.
(
1997
).
The Hayling and Brixton Tests Manual
.
Bury st Edmunds
:
Thames Valley Test Company
.
Calderoni
,
S.
,
Muratori
,
F.
,
Leggero
,
C.
,
Narzisi
,
A.
,
Apicella
,
F.
,
Balottin
,
U.
, et al.
(
2013
).
Neuropsychological functioning in children and adolescents with restrictive-type anorexia nervosa: an in-depth investigation with NEPSY-II
.
Journal of Clinical and Experimental Neuropsychology
,
35
(2)
,
167
179
. .
Carter
,
J.
,
Stewart
,
D.
, &
Fairburn
,
C.
(
2001
).
Eating disorder examination questionnaire: Norms for young adolescent girls
.
Behaviour Research and Therapy
,
39
,
625
632
.
Clark
,
D.
,
Antony
,
M.
,
Beck
,
A.
,
Swinson
,
R.
, &
Steer
,
R.
(
2005
).
Screening for obsessive and compulsive symptoms: Validation of the clark-beck obsessive-compulsive inventory
.
Psychological Assessment
,
17
(2)
,
132
143
.
Cohen
,
J.
(
1988
).
Statistical power analysis for the behavioral sciences
(
2nd ed
).
Hillsdale, NJ
:
Lawrence Erlbaum Associates
.
Dahlgren
,
C. L.
,
,
B.
,
Landrø
,
N. I.
, &
Ro
,
Ø.
(
2013
).
Neuropsychological functioning in adolescents with anorexia nervosa before and after cognitive remediation therapy: A feasibility trial
.
International Journal of Eating Disorders
,
46
(6)
,
576
581
. .
Delis
,
D.
,
Kaplan
,
E.
, &
Kramer
,
J.
(
2001
).
Delis kaplin executive function system
.
San Antonio
:
The Psychological Corporation
.
Dmitrzak-Weglarz
,
M.
,
Skibinska
,
M.
,
Slopien
,
A.
,
Tyszkiewicz
,
M.
,
Pawlak
,
J.
,
Maciukiewicz
,
M.
, et al.
(
2013
).
Serum neurotrophin concentrations in polish adolescent girls with anorexia nervosa
.
Neuropsychobiology
,
67
(1)
,
25
32
. .
Dobson
,
K. S.
, &
Dozois
,
D. J.
(
2004
).
Attentional biases in eating disorders: A meta-analytic review of Stroop performance
.
Clinical Psychology Review
,
23
(8)
,
1001
1022
.
Duchesne
,
M.
,
Mattos
,
P.
,
Fontenelle
,
L.
,
Veiga
,
H.
,
Rizo
,
L.
, &
Appolinario
,
J.
(
2004
).
Neuropsychology of eating disorders: A systematic review of the literature
.
Revista Brasileira de Psiquiatria
,
26
(2)
,
107
117
.
Eddy
,
K. T.
,
Dorer
,
D. J.
,
Franko
,
D. L.
,
Tahilani
,
K.
,
Thompson-Brenner
,
H.
, &
Herzog
,
D. B.
(
2008
).
Diagnostic crossover in anorexia nervosa and bulimia nervosa: Implications for DSM-V
.
American Journal of Psychiatry
,
165
,
245
250
.
Fairburn
,
C.
, &
Cooper
,
Z.
(
1993
).
The eating disorder examination
(
12th ed
).
New York
:
Guilford Press
.
Fornasari
,
L.
,
Gregoraci
,
G.
,
Isola
,
M.
,
Laura Negri
,
G. A.
,
Rambaldelli
,
G.
,
Cremaschi
,
S.
, et al.
(
2014
).
Psychopathological and personality traits underlie decision making in recent onset medication naive anorexia nervosa: A pilot study
.
Psychiatry Research
,
216
(1)
,
89
96
. .
Frampton
,
I.
, &
Rose
,
M.
(
2013
). Eating disorders and the brain. In
,
B.
, &
Bryant-Waugh
,
R.
(Eds.)
Eating Disorders in Childhood and Adolescence
.
London
:
Wiley
.
Friederich
,
H.
,
Wu
,
M.
,
Simon
,
J.
, &
Herzog
,
W.
(
2013
).
Neurocircuit function in eating disorders
.
International Journal of Eating Disorders
,
46
(5)
,
425
432
.
Godart
,
N. T.
,
Perdereau
,
F.
,
Rein
,
Z.
,
Berthoz
,
S.
,
Wallier
,
J.
,
Jeammet
,
P.
, et al.
(
2007
).
Comorbidity studies of eating disorders and mood disorders. Critical review of the literature
.
Journal of Affective Disorders
,
97
(1-3)
,
37
49
. .
Guillaume
,
S.
,
Sang
,
C. N. T.
,
Jaussent
,
I.
,
Raingeard
,
I.
,
Bringer
,
J.
,
Jollant
,
F.
, et al.
(
2010
).
Is decision making really impaired in eating disorders
.
Neuropsychology
,
24
(6)
,
808
812
.
Hermens
,
D.
,
Hodge
,
M.
,
Naismith
,
S.
,
Kaur
,
M.
,
Scott
,
E.
, &
Hickie
,
I.
(
2011
).
Neuropsychological clustering highlights cognitive differences in young people presenting with depressive symptoms
.
Journal of the International Neuropsychological Society
,
17
,
267
276
.
Jáuregui-Lobera
,
I.
(
2013
).
Neuropsychology of eating disorders: 1995-2012
.
Neuropsychiatric Disease and Treatment
,
9
,
415
430
.
Johansson
,
L.
,
,
A.
, &
,
G.
(
2005
).
Stroop interference for food- and body-related words: A meta-analysis
.
Eating Behaviors
,
6
(3)
,
271
281
. .
Kjaersdam Telléus
,
G.
,
Jepsen
,
J. R.
,
Bentz
,
M.
,
Christiansen
,
E.
,
Jensen
,
S. O.
,
Fagerlund
,
B.
, et al.
(
2015
).
Cognitive profile of children and adolescents with anorexia nervosa
.
European Eating Disorders Review
,
23
(1)
,
34
42
. .
Korkman
,
M.
,
Kirk
,
U.
, &
Kemp
,
S.
(
2007
).
NEPSY-II. A developmental neuropsychology assessment
.
San Antonio, TX
:
The Psychological Corporation
.
Kovacs
,
M.
(
1992
).
Children’s Depression Inventory (CDI)
.
New York
:
Multi-health Systems, Inc
.
Lang
,
K.
,
Lopez
,
C.
,
Stahl
,
D.
,
Tchanturia
,
K.
, &
Treasure
,
J.
(
2014
a).
Central coherence in eating disorders: An updated systematic review and meta-analysis
.
The World Journal of Biological Psychiatry
,
15
(8)
,
586
598
. .
Lang
,
K.
,
Stahl
,
D.
,
Espie
,
J.
,
Treasure
,
J.
, &
Tchanturia
,
K.
(
2014
b).
Set shifting in children and adolescents with anorexia nervosa: An exploratory systematic review and meta-analysis
.
International Journal of Eating Disorders
,
47
(4)
,
394
399
. .
Lang
,
K.
, &
Tchanturia
,
K.
(
2014
c).
A systematic review of central coherence in children and adolescents with anorexia nervosa
.
Journal of Child and Adolescent Behavior
,
2
(140)
,
B.
, &
Frampton
,
I.
(
2011
).
Eating disorders and the brain
.
Chichester
:
Wiley-Blackwell
.
Lena
,
S. M.
,
Fiocco
,
A. M.
, &
Leyenaar
,
J. K.
(
2004
).
The role of cognitive deficits in the development of eating disorders
.
Neuropsychological Review
,
14
(2)
,
99
113
.
Lopez
,
C.
,
Stahl
,
D.
, &
Tchanturia
,
K.
(
2010
).
Estimated intelligence quotient in anorexia nervosa: A systematic review and meta-analysis of the literature
.
Annals of General Psychiatry
,
9
(40)
.
Lopez
,
C.
,
Tchanturia
,
K.
,
Stahl
,
D.
, &
Treasure
,
J.
(
2008
).
Central coherence in eating disorders: A systematic review
.
Psychological Medicine
,
38
(10)
,
1393
1404
.
Luce
,
K. H.
, &
Crowther
,
J. H.
(
1999
).
The reliability of the eating disorder examination-self‐report questionnaire version (EDE‐Q)
.
International Journal of Eating Disorders
,
25
(3)
,
349
351
.
,
P. P.
,
,
B. C.
,
Goncalves
,
S.
, &
Hoek
,
H. W.
(
2007
).
The prevalence of eating disorders not otherwise specified
.
International Journal of Eating Disorders
,
40
(3)
,
212
217
. .
Marsh
,
R.
,
Maia
,
T.
, &
Peterson
,
B.
(
2009
).
Functional disturbances within frontostriatal circuits across multiple childhood psychopathologies
.
American Journal of Psychiatry
,
166
(6)
,
664
674
.
Meyers
,
J.
, &
Meyers
,
K.
(
1995
).
Rey complex figure test and recognition trial
.
Odessa
:
Psychological Assessment Resources, Inc
.
Mikos
,
A. E.
,
McDowell
,
B. D.
,
Moser
,
D. J.
,
Bayless
,
J. D.
,
Bowers
,
W. A.
,
Anderson
,
A. E.
, et al.
(
2008
).
Stability of neuropsychological performance in AN
.
Annals of Clinical Psychiatry
,
20
(1)
,
9
13
.
Moser
,
D. J.
,
Benjamin
,
M. L.
,
Bayless
,
J. D.
,
McDowell
,
B. D.
,
Paulsen
,
J. S.
,
Bowers
,
W. A.
, et al.
(
2003
).
Neuropsychological functioning pretreatment and posttreatment in an inpatient eating disorders program
.
International Journal of Eating Disorders
,
33
(1)
,
64
70
. .
Renwick
,
B.
,
Musiat
,
P.
,
Lose
,
A.
,
DeJong
,
H.
,
,
H.
,
Kenyon
,
M.
, et al.
(
2015
).
Neuro- and social-cognitive clustering highlights distinct profiles in adults with anorexia nervosa
.
International Journal of Eating Disorders
,
48
(1)
,
26
34
.
Reville
,
M.
,
O'Connor
,
L.
, &
Frampton
,
I.
(
2016
).
Literature review of cognitive neuroscience and Anorexia Nervosa
.
Current Psychiatric Reports
,
18
(18)
. .
Rey
,
A.
(
1941
).
L'examin psychologique dans les cas d'encepalopathie traumatique (psychological examination of traumatic encephalopathy)
.
Archives de Psychologie
,
28
,
286
340
.
Roberts
,
M. E.
,
Tchanturia
,
K.
,
Stahl
,
D.
,
Southgate
,
L.
, &
Treasure
,
J.
(
2007
).
A systematic review and meta-analysis of set-shifting ability in eating disorders
.
Psychological Medicine
,
37
,
1075
1084
.
Rose
,
M.
,
Davis
,
J.
,
Frampton
,
I.
, &
,
B.
(
2011
).
The Ravello Profile—Development of a global standard neuropsychological assessment for young people with anorexia nervosa
.
Clinical Psychology and Psychiatry
,
16
(2)
,
195
202
.
Shafran
,
R.
,
Frampton
,
I.
,
Heyman
,
I.
,
Reynolds
,
M.
,
Teachman
,
B.
, &
Rachman
,
S.
(
2003
).
The preliminary development of a newself-report measure for OCD in young people
.
,
26
,
137
142
.
Shott
,
M. E.
,
Filoteo
,
J. V.
,
Bhatnagar
,
K. A.
,
Peak
,
N. J.
,
Hagman
,
J. O.
,
Rockwell
,
R.
, et al.
(
2012
).
Cognitive set-shifting in anorexia nervosa
.
European Eating Disorders Review
,
20
(5)
,
343
349
. .
Spielberger
,
C.
(
1973
).
Preliminary Manuel for the State-Trait Anxiety Inventory for children
.
Palo Alto
:
Consulting Psychologists Press
.
Spielberger
,
C.
,
Goursuch
,
R.
, &
Lushene
,
R.
(
1983
). Manual for the state trait anxiety inventory. In
Consulting Psychologist Press
.
Mountain View, CA
.
Stedal
,
K.
,
Frampton
,
I.
,
Landrø
,
N.
, &
,
B.
(
2011
).
An examination of the Ravello Profile—A neuropsychological test battery for anorexia nervosa
.
European Eating Disorders Review
,
20
(3)
,
175
181
. .
Stedal
,
K.
,
Rose
,
M.
,
Frampton
,
I.
,
Landrø
,
N.
, &
,
B.
(
2012
).
The neuropsychological profile of children, adolescents and young adults with anorexia nervosa
.
Archives of Clinical Neuropsychology
,
27
(3)
,
329
337
. .
Stroop
,
J.
(
1935
).
Studies of interference in serial verbal reactions
.
Journal of Experimental Psychology
,
18
,
643
662
.
Tabachnick
,
B. G.
, &
Fidell
,
L. S.
(
2001
).
Using multivariate statistics
(
4th ed.
).
Boston
:
Allyn & Bacon
.
Tchanturia
,
K.
,
Campbell
,
I. C.
,
Morris
,
R.
, &
Treasure
,
J.
(
2005
).
Neuropsychological studies in anorexia nervosa
.
International Journal of Eating Disorders
,
37
(Supp. S72–76)
,
S87
79
. discussion.
Tchanturia
,
K.
,
Davies
,
H.
,
Roberts
,
M.
,
Harrison
,
A.
,
Nakazato
,
M.
,
Schmidt
,
U.
, et al.
(
2012
).
Poor cognitive flexibility in eating disorders: Examining the evidence using the Wisconsin Card Sorting Task
.
PloS ONE
,
7
(1)
,
e28331
. .
Tchanturia
,
K.
,
Lloyd
,
S.
, &
Lang
,
K.
(
2013
).
Cognitive remediation therapy for anorexia nervosa: Current evidence and future research directions
.
International Journal of Eating Disorders
,
46
(5)
,
492
495
. .
Trace
,
S.
,
Baker
,
J.
,
Peñas-Lledó
,
E.
, &
Bulik
,
C.
(
2013
).
The genetics of eating disorders
.
Annual Review of Clinical Psychology
,
9
,
589
620
.
van Noort
,
B. M.
,
Pfeiffer
,
E.
,
Lehmkuhl
,
U.
, &
Kappel
,
V.
(
2013
).
Eine internationale neuropsychologische testbatterie für kinder, jugendliche und erwachsene mit anorexia nervosa—Die deutsche Adaptation des Ravello Profils
.
Zeitschrift für Kinder-und Jugendpsychiatrie und Psychotherapie
,
41
(6)
,
371
381
.
Warner
,
R.
(
2013
).
Applied statistics: From bivariate through multivariate techniques
(
2nd ed.
).
Thousand Oaks, CA
:
Sage
.
Watkins
,
B.
,
Frampton
,
I.
,
,
B.
, &
Bryant-Waugh
,
R.
(
2005
).
Reliability and validity of the child version of the eating disorder examination: A preliminary investigation
.
International Journal of Eating Disorders
,
38
(2)
,
183
187
.
Wechsler
,
D.
(
1997
).
(
3rd ed.
).
San Antonio
:
The Psychological Corporation
.
Wechsler
,
D.
(
1999
).
Wechsler Abbreviated Scale of Intelligence (WASI) Manual
.
San Antonio
:
The Psychological Corporation
.
Weiß
,
R.
(
2006
).
Grundintelligenztest Skala 2—Revision (CFT 20-R) mit Wortschatztest und Zahlenfolgentest—Revision (WS/ZF-R)
.
Göttingen
:
Hogrefe
.
Westwood
,
H.
,
Stahl
,
D.
,
Mandy
,
W.
, &
Tchanturia
,
K.
(
2016
).
The set-shifting profiles of anorexia nervosa and autism spectrum disorder using the Wisconsin Card Sorting Test: A systematic review and meta-analysis
.
Psychological Medicine
, (pp.
1
19
). .
Wildes
,
J. E.
, &
Marcus
,
M. D.
(
2015
).
Application of the Research Domain Criteria (RDoC) framework to eating disorders: Emerging concepts and research
.
Current Psychiatry Reports
,
17
(5)
,
30
.
Wilsdon
,
A.
, &
,
T. D.
(
2006
).
Executive functioning in anorexia nervosa: Exploration of the role of obsessionality, depression and starvation
.
Journal of Psychiatric Research
,
40
(8)
,
746
754
.
Wu
,
M.
,
Brockmeyer
,
T.
,
Hartmann
,
M.
,
Skunde
,
M.
,
Herzog
,
W.
, &
Friederich
,
H. C.
(
2014
).
Set-shifting ability across the spectrum of eating disorders and in overweight and obesity: A systematic review and meta-analysis
.
Psychological Medicine
,
44
(16)
,
3365
3385
.
Wu
,
M.
,
Brockmeyer
,
T.
,
Hartmann
,
M.
,
Skunde
,
M.
,
Herzog
,
W.
, &
Friederich
,
H. C.
(
2016
).
Reward-related decision making in eating and weight disorders: A systematic review and meta-analysis of the evidence from neuropsychological studies
.
Neuroscience Biobehavioral Reviews
,
61
,
177
196
. .
Zakzanis
,
K.
(
2001
).
Statistics to tell the truth, the whole truth, and nothing but the truth: Formulae, illustrative numerical examples, and heuristic interpretation of effect size analyses for neuropsychological researchers
.
Archives of Clinical Neuropsychology
,
16
(7)
,
653
667
.
Zakzanis
,
K.
,
Campbell
,
Z.
, &
Polsinelli
,
A.
(
2010
).
Quantitative evidence for distinct cognitive impairment in anorexia nervosa and bulimia nervosa
.
Journal of Neuropsychology
,
4
(1)
,
89
106
.