RESEARCH ARTICLE
Set-Shifting and its Relation to Clinical and Personality Variables in Full Recovery of Anorexia Nervosa Susanne E. Lindner1, Manfred M. Fichter1,2* & Norbert Quadflieg1 1 2
Department of Psychiatry, University of Munich (LMU), Munich, Germany Schoen Hospital Roseneck for Behavioral Medicine, Prien, Germany
Abstract Objective: First, this study aimed to explore whether set-shifting is inefficient after full recovery of anorexia nervosa (recAN). Second, this study wanted to explore the relation of set-shifting to clinical and personality variables. Method: A total of 100 recAN women were compared with 100 healthy women. Set-shifting was assessed with Berg’s Card Sorting Test. Expert interviews yielded assessments for the inclusion/exclusion criteria, self-ratings for clinical and personality variables. Results: Compared with the healthy control group, the recAN participants achieved fewer categories, showed more perseverations and spent less time for shifting set. Perfectionism is correlated with set-shifting but in converse directions in the two groups. Discussion: Our study supports the findings of inefficiencies in set-shifting after full recovery from AN. Higher perfectionism in the recAN group is associated with better set-shifting ability, whereas higher perfectionism in the healthy control group is related to worse set-shifting ability. Copyright © 2014 John Wiley & Sons, Ltd and Eating Disorders Association. Keywords set-shifting; cognitive flexibility; anorexia nervosa; Berg’s Card Sorting Test; perfectionism *Correspondence Manfred M. Fichter, MD, Dipl.-Psych., Professor of Psychiatry, Schoen Hospital Roseneck for Behavioral Medicine, University of Munich (LMU), Am Roseneck 6, 83209 Prien, Germany. Tel: +49 8051 68-3001; Fax: +49 8051 68-3003. Email: [email protected] Published online 7 May 2014 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/erv.2293
Introduction Set-shifting is an important executive function. It refers to the ability to move back and forth between multiple tasks, operations or mental sets (Miyake, Friedman, Emerson, Witzki, & Howerter, 2000), for example, the ability to display flexibility in case of changing rules. Individuals with problems in set-shifting show rigid approaches to problem solving (cognitive inflexibility) or perseverative behaviours (response inflexibility) (Roberts, Tchanturia, Stahl, Southgate, & Treasure, 2007). Anorexia nervosa (AN) patients are often characterised by rigidity, perfectionism, obsessive–compulsive traits and high persistence (e.g. Cassin & von Ranson, 2005; Halmi et al., 2000). In their eating behaviour (e.g. eating routines) and their perpetual concerns about weight and shape, these personality traits are obvious. Inefficiencies in set-shifting can be associated with these phenomenological traits of inflexibility, rigidity and persistence of AN patients (Roberts, Tchanturia, & Treasure, 2010; Tchanturia et al., 2012; Treasure, Lopez, & Roberts, 2007). Thus, some theoretical papers discuss inefficient set-shifting as a risk factor for developing an eating disorder (Southgate, Tchanturia, & Treasure, 2005; Steinglass & Walsh, 2006; Tchanturia, Campbell, Morris, & Treasure, 2005). Recent studies tried to broaden the endophenotype of AN to ensure a better understanding of the aetiology and maintenance of the disorder and to enhance treatment approaches
252
(Roberts et al., 2007; Tenconi et al., 2010; Treasure et al., 2007). Set-shifting represents a promising endophenotype candidate for AN. An endophenotype has to be heritable, measurable, associated with the illness under investigation but independent of the actual state of this illness and more frequent in close relatives than in the normal population (Gottesman & Gould, 2003). Inefficiencies in set-shifting seem to be closely related to the diagnosis of AN. Roberts et al. (2007) conducted a systematic review and meta-analysis regarding set-shifting in eating disorders. Across different measures, samples and studies, this meta-analysis revealed inefficient set-shifting ability in eating disorder groups compared with healthy controls. Most of the more recent studies support the finding of the meta-analysis by Roberts et al. (2007) that especially AN patients showed inefficiencies in set-shifting compared with a healthy control group (Abbate-Daga et al., 2011; Danner et al., 2012; Fagundo et al., 2012; Friederich et al., 2012; Galimberti et al., 2013; Kim, Kim, & Kim, 2010; Lounes, Khan, & Tchanturia, 2011; Nakazato et al., 2009; Roberts et al., 2010; Stedal, Rose, Frampton, Landrø, & Lask, 2012; Tchanturia et al., 2011; Tchanturia et al., 2012; Tenconi et al., 2010; Zastrow et al., 2009). There are only few studies that did not find inefficient set-shifting in AN samples when compared with healthy controls (Cavedini et al., 2004; Galimberti, Martoni, Cavallini, Erzegovesi, & Bellodi, 2012; Gillberg, Råstam, Wentz, & Gillberg, 2007; Tokley & Kemps, 2007).
Eur. Eat. Disorders Rev. 22 (2014) 252–259 © 2014 John Wiley & Sons, Ltd and Eating Disorders Association.
S. E. Lindner et al.
Inefficiencies in set-shifting were also found in unaffected firstdegree relatives and sisters of AN patients (Galimberti et al., 2013; Holliday, Tchanturia, Landau, Collier, & Treasure, 2005; Roberts et al., 2010; Tenconi et al., 2010). Kanakam, Raoult, Collier, and Treasure (2013) found set-shifting difficulties in unaffected twin siblings. Therefore, inefficiencies in set-shifting seem to be more frequent in unaffected close relatives of AN patients than in healthy individuals. It is not clear yet if set-shifting remains inefficient after recovery from AN. Quite a few studies have examined the question whether inefficient set-shifting is only a temporary state caused by malnutrition or whether it is a trait factor rather than a state factor. These studies also assessed recovered AN patients (recAN) in addition to currently ill AN patients in comparison with a healthy control group. Some studies showed that the inefficiency in set-shifting persists after recovery of AN (Danner et al., 2012; Holliday et al., 2005; Tchanturia, Morris, Surguladze, & Treasure, 2002; Tchanturia et al., 2012; Tenconi et al., 2010). In other studies, the performance in set-shifting of recAN patients lay between the performances of the AN group and the healthy control group. However, both differences were not statistically significant (Friederich et al., 2012; Nakazato et al., 2009; Roberts et al., 2010; Tchanturia et al., 2004; Tchanturia et al., 2011). In most of these studies, the sample size was rather small (13-30 recAN participants; Danner et al., 2012; Friederich et al., 2012; Holliday et al., 2005; Nakazato et al., 2009; Roberts et al., 2010; Tchanturia et al., 2004; Tchanturia et al., 2002). Tenconi et al. (2010) studied 63 weight recovered and 29 fully recovered AN patients and found no difference from the acutely ill AN patients, that is, the recAN patients showed inefficiencies in set-shifting compared with the healthy control group. Tchanturia et al. (2011, 2012) studied 72 and 90 recAN patients, respectively, with two different measures of set-shifting. They found an intermediate performance of the recAN participants, which was not significantly different from the performances of the AN group nor of the healthy control group regarding the Brixton Spatial Anticipation Test (Tchanturia et al., 2011). Regarding the Wisconsin Card Sorting Task, there was a significant difference between the recAN participants and the healthy control group (Tchanturia et al., 2012). Set-shifting inefficiencies are measurable, associated with AN and more frequent in healthy close relatives compared with a control group. It still needs to be resolved whether these inefficiencies are unrelated to the acute state of AN and therefore still come out inefficient after symptomatic recovery from AN. The present study aimed to investigate whether set-shifting ability is inefficient after full recovery of AN. Some studies have already examined set-shifting in a sample of recAN, but either the sample was rather small, or the definition of recovery was not sufficiently clear and rigorous. This study assessed set-shifting in a very large sample of truly recovered AN patients. The definition of recovery in this study considered physiological, behavioural and psychological variables. Furthermore, we took into account relevant personality, clinical (frequent comorbid disorders: depression, anxiety and obsessive-compulsive traits) and eating disorder variables for a better understanding of the relation between set-shifting and these variables. As mentioned earlier, relevant personality variables are perfectionism and impulsiveness as the opposite of rigidity. Especially perfectionism was of interest. AN patients show a consistent
Set-Shifting in Recovered AN
pattern of elevated perfectionism scores compared with healthy controls, which seems to be specific for AN compared with other psychiatric disorders, particularly one aspect of perfectionism ‘concerns over mistakes’ (Bardone-Cone et al., 2007; Bulik et al., 2003; Egan, Wade, & Shafran, 2011). Even after weight restoration, AN patients show still higher levels of perfectionism (Bardone-Cone et al., 2007). Regarding full recovery of AN, there are inconsistent findings (Bardone-Cone, Sturm, Lawson, Robinson & Smith, 2010; Srinivasagam et al., 1995). As Friederich and Herzog (2011) stated, perfectionism and obsessional traits are associated with impaired cognitive-behavioural flexibility. Also, Buehren et al. (2012) found a relation between perfectionism and cognitive inflexibility in an adolescent sample.
Methods After providing detailed information, all participants gave their informed consent to this study (case-control design with one cross-sectional assessment), previously approved by the ethical review committee of the University of Munich.
Participants The sample has been described already by Lindner, Fichter, and Quadflieg (2012, 2013). One hundred female participants with a lifetime diagnosis of AN who had fully recovered from AN (recAN) were recruited in the context of a large longitudinal study with former inpatients of the Schoen Hospital Roseneck for Behavioral Medicine in Prien (Germany). Full recovery of AN was defined as follows: no eating disorder according to DSM-IV-TR (American Psychiatric Association, 1994), body mass index (BMI = kg/m2) between 18.5 and 26, a regular menstrual cycle and absence of significant eating disorderspecific cognitions (fear of gaining weight, body image distortion, self-esteem dependent on body and shape with no or only slight manifestations) as well as absence of bingeing and compensatory behaviours for at least 1 year. Thus, the definition of recovery considered physical, behavioural and psychological components (Bachner-Melman, Zohar, & Ebstein, 2006; Bardone-Cone, Harney et al., 2010). One hundred healthy participants were recruited by means of notices in hospitals, universities and nurseries. Inclusion criteria for the healthy controls were the following: BMI above 18.5 since adolescence, no absent menstrual cycles for more than three months together with a low body weight, no pronounced eating disorder-specific cognitions and no close relative with an eating disorder. Exclusion criteria for both groups were severe manifestations of depression, anxiety and obsessive-compulsive traits; past or current diseases that could potentially affect brain functioning; past or current substance dependence, or current substance abuse; past or current schizophrenia; any other mental or physical condition that would have compromised the participation; and insufficient knowledge of the German language. The recAN and healthy participants were matched one to one for gender (all female), age (±2 years) and educational level (9, 10 and 12 years of school education and university graduates).
Eur. Eat. Disorders Rev. 22 (2014) 252–259 © 2014 John Wiley & Sons, Ltd and Eating Disorders Association.
253
S. E. Lindner et al.
Set-Shifting in Recovered AN
Measures The measures, especially the ones for the inclusion/exclusion criteria, are described in more detail by Lindner et al. (2012, 2013). For assessing eating disorder, we used the expert interview of the Structured Inventory for Anorexic and Bulimic Syndromes (Fichter & Quadflieg, 1999) with regard to the year prior to assessment (only recAN group) and lifetime (both groups). The Eating Attitudes Test (EAT-26; Garner, Olmsted, Bohr, & Garfinkel, 1982) complemented the control group assessment. Participants with a score higher than 20 in the EAT-26 would have been excluded; this was not the case in our study. The German short version of the Patient Health Questionnaire (Löwe, Spitzer, Zipfel, & Herzog, 2002) was used in both groups. A verbal intelligence measurement (Mehrfachwahl-WortschatzIntelligenztest; Lehrl, 2005) examined the matching based on educational level. The participant has to detect an existing word between other non-existing words in ascending difficulty level. Self-rating instruments assessed relevant clinical and personality variables: depression (Beck Depression Inventory I, BDI-I; Hautzinger, Bailer, Worall, & Keller, 1994), anxiety (Beck Anxiety Inventory, BAI; Margraf & Ehlers, 2007), obsessive-compulsive traits (Maudsley Obsessive-Compulsive Inventory, MOCI; Hodgson & Rachman, 1977; German: Kulick, 1979), impulsiveness (Barratt Impulsiveness Scale, BIS-11; Patton, Stanford, & Barratt, 1995), and perfectionism (Frost Multidimensional Perfectionism Scale German version, FMPS-D; Stöber, 1995). Apart from the FMPS-D total score, we also calculated the scores for the four subscales: concerns over mistakes and doubts about actions, parental expectations and criticism, personal standards and organisation (Stöber, 1998). We used these four subscales instead of the original six ones because of their better factorial stability (Stöber, 1998). Grant and Berg (1948) developed a card sorting test (Berg’s Card Sorting Test, BCST) for assessing set-shifting. The BCST consists of 64 cards with four different symbols (triangle, star, cross and circle) in four different numbers (1-4) and four different colours (red, green, yellow and blue). Four stimulus cards are presented to each participant. Other cards are consecutively shown, and the participant has to assign them to one of the four stimulus cards. The rule for this assignment can be symbol, number or colour, which is unknown to the participant; the participant just receives feedback whether the assignment is right or wrong. If the assignment is right, the participant has to stick to this rule until the feedback switches to wrong, which means the rule had changed. This occurs after five successful trials (category). The change is not predictable for the participant. The test ends when nine categories (three runs for each rule) are completed or each card has been shown twice (128 cards). We used a computer version of the BCST, which corresponds to the original test of Grant and Berg (1948), but it differed from the widely used Wisconsin Card Sorting Test (WCST) in the number of correct trials before the rule changes. In our test, the rule changed after five successful trials; in the WCST, the rule changes after ten correct trials. We calculated three outcome variables. First is the number of categories achieved. A category is achieved if the participant sorted five cards correctly. There were nine categories to be achieved, for each rule 254
three runs. The number of categories achieved measures not only set-shifting; it is a more global measure for building concepts and categories. Second is the number of perseverations. A perseveration occurs when the participant clings to the last rule and do not change to a new rule. Higher number of perseverations is a sign for inefficiency in set-shifting. Third is the time spent for a shift (shift cost). This indicates the difference between the reaction time of the first trial with the feedback ‘wrong’, indicating a rule change, and the mean reaction time of the three trials before. Shift cost measures cognitive flexibility on the basis of reaction time and not of errors (Deák, 2003). Statistical analyses We used t-tests to compare the recAN group with the healthy control group on demographic and clinical variables. Analyses of variance addressed the results in set-shifting. The Bonferroni procedure allowed for corrections of multiple comparisons (significance level at p = .017 for the BCST). Pearson correlations permitted to express the relationship between set-shifting variables and clinical characteristics. For BDI-I, BAI and FMPS-D, there were significant differences between the groups. Therefore, we executed the correlation analyses separately for each group. For BIS-11 and MOCI, the total sample was considered. Subsequently, an analysis of covariance took into account relevant control variables. The recAN group was further divided into subgroups: women who suffered only from the restricting type of AN lifetime (recAN rest) and women who suffered from both the restricting and the binge eating/purging subtype of AN or any other eating disorder lifetime (recAN other). For detecting group differences, Scheffé post-hoc tests were executed. Because of the explorative character of the further analyses, the p-values were not corrected. In order to estimate the relevance of significant results, we calculated Cohen’s effect sizes (Cohen, 1988). Cohen’s effect sizes (d) are defined as small (d = 0.2), medium (d = 0.5) and large (d = 0.8).
Results Sample Table 1 presents the demographic and clinical variables. The sample is described in more detail by Lindner et al. (2012, 2013). The scores of depression (BDI-I), anxiety (BAI) and perfectionism (FMPS-D) all came out higher in the recAN group. Regarding obsessive-compulsive traits (MOCI) and impulsiveness (BIS-11), the two groups were not significantly different. Set-shifting The recAN group achieved significantly fewer categories as the healthy control group (F(1, 199) = 7.47, p = .007, Cohen’s d = .39). The recAN group showed significantly more perseverations than the healthy control group (F(1, 199) = 6.07, p = .015, Cohen’s d = .35). The reaction time for shifts was shorter in the recAN group compared with the healthy control group (F(1, 196) = 6.53, p = .011, Cohen’s d = .36). The effect sizes were small. The results are shown in Table 2.
Eur. Eat. Disorders Rev. 22 (2014) 252–259 © 2014 John Wiley & Sons, Ltd and Eating Disorders Association.
S. E. Lindner et al.
Set-Shifting in Recovered AN
Table 1 Demographic and clinical variables RecAN group
Age (years: range 22-51) Intelligence estimate (MWT-B: range 0-37) Current BMI Minimal BMI Age of onset of AN (years)* Duration of AN (years)** Duration of recovery from AN (years)* Depression (BDI-I: range 0-63)*** Anxiety (BAI: range 0-63) Perfectionism (FMPS-D: range 29-145)*** CMD PEC PS O Obsessive-compulsive traits (MOCI: range 0-30) Impulsiveness (BIS-11: range 30-120)
Healthy control group
t-test
M
SD
M
SD
t
p
34.49 30.43 20.86 13.92 18.95 3.88 6.37 3.35 4.40 71.78 30.41 22.31 19.05 23.13 3.81 57.49
7.13 3.32 1.31 1.81 4.19 3.17 3.93 4.17 5.07 20.78 10.31 9.03 6.40 3.83 2.92 6.75
34.53 30.88 21.80 20.09
7.26 3.29 1.37 .93
0.034 0.963 4.983 30.288
.973 .337
Set-Shifting and its Relation to Clinical and Personality Variables in Full Recovery of Anorexia Nervosa Susanne E. Lindner1, Manfred M. Fichter1,2* & Norbert Quadflieg1 1 2
Department of Psychiatry, University of Munich (LMU), Munich, Germany Schoen Hospital Roseneck for Behavioral Medicine, Prien, Germany
Abstract Objective: First, this study aimed to explore whether set-shifting is inefficient after full recovery of anorexia nervosa (recAN). Second, this study wanted to explore the relation of set-shifting to clinical and personality variables. Method: A total of 100 recAN women were compared with 100 healthy women. Set-shifting was assessed with Berg’s Card Sorting Test. Expert interviews yielded assessments for the inclusion/exclusion criteria, self-ratings for clinical and personality variables. Results: Compared with the healthy control group, the recAN participants achieved fewer categories, showed more perseverations and spent less time for shifting set. Perfectionism is correlated with set-shifting but in converse directions in the two groups. Discussion: Our study supports the findings of inefficiencies in set-shifting after full recovery from AN. Higher perfectionism in the recAN group is associated with better set-shifting ability, whereas higher perfectionism in the healthy control group is related to worse set-shifting ability. Copyright © 2014 John Wiley & Sons, Ltd and Eating Disorders Association. Keywords set-shifting; cognitive flexibility; anorexia nervosa; Berg’s Card Sorting Test; perfectionism *Correspondence Manfred M. Fichter, MD, Dipl.-Psych., Professor of Psychiatry, Schoen Hospital Roseneck for Behavioral Medicine, University of Munich (LMU), Am Roseneck 6, 83209 Prien, Germany. Tel: +49 8051 68-3001; Fax: +49 8051 68-3003. Email: [email protected] Published online 7 May 2014 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/erv.2293
Introduction Set-shifting is an important executive function. It refers to the ability to move back and forth between multiple tasks, operations or mental sets (Miyake, Friedman, Emerson, Witzki, & Howerter, 2000), for example, the ability to display flexibility in case of changing rules. Individuals with problems in set-shifting show rigid approaches to problem solving (cognitive inflexibility) or perseverative behaviours (response inflexibility) (Roberts, Tchanturia, Stahl, Southgate, & Treasure, 2007). Anorexia nervosa (AN) patients are often characterised by rigidity, perfectionism, obsessive–compulsive traits and high persistence (e.g. Cassin & von Ranson, 2005; Halmi et al., 2000). In their eating behaviour (e.g. eating routines) and their perpetual concerns about weight and shape, these personality traits are obvious. Inefficiencies in set-shifting can be associated with these phenomenological traits of inflexibility, rigidity and persistence of AN patients (Roberts, Tchanturia, & Treasure, 2010; Tchanturia et al., 2012; Treasure, Lopez, & Roberts, 2007). Thus, some theoretical papers discuss inefficient set-shifting as a risk factor for developing an eating disorder (Southgate, Tchanturia, & Treasure, 2005; Steinglass & Walsh, 2006; Tchanturia, Campbell, Morris, & Treasure, 2005). Recent studies tried to broaden the endophenotype of AN to ensure a better understanding of the aetiology and maintenance of the disorder and to enhance treatment approaches
252
(Roberts et al., 2007; Tenconi et al., 2010; Treasure et al., 2007). Set-shifting represents a promising endophenotype candidate for AN. An endophenotype has to be heritable, measurable, associated with the illness under investigation but independent of the actual state of this illness and more frequent in close relatives than in the normal population (Gottesman & Gould, 2003). Inefficiencies in set-shifting seem to be closely related to the diagnosis of AN. Roberts et al. (2007) conducted a systematic review and meta-analysis regarding set-shifting in eating disorders. Across different measures, samples and studies, this meta-analysis revealed inefficient set-shifting ability in eating disorder groups compared with healthy controls. Most of the more recent studies support the finding of the meta-analysis by Roberts et al. (2007) that especially AN patients showed inefficiencies in set-shifting compared with a healthy control group (Abbate-Daga et al., 2011; Danner et al., 2012; Fagundo et al., 2012; Friederich et al., 2012; Galimberti et al., 2013; Kim, Kim, & Kim, 2010; Lounes, Khan, & Tchanturia, 2011; Nakazato et al., 2009; Roberts et al., 2010; Stedal, Rose, Frampton, Landrø, & Lask, 2012; Tchanturia et al., 2011; Tchanturia et al., 2012; Tenconi et al., 2010; Zastrow et al., 2009). There are only few studies that did not find inefficient set-shifting in AN samples when compared with healthy controls (Cavedini et al., 2004; Galimberti, Martoni, Cavallini, Erzegovesi, & Bellodi, 2012; Gillberg, Råstam, Wentz, & Gillberg, 2007; Tokley & Kemps, 2007).
Eur. Eat. Disorders Rev. 22 (2014) 252–259 © 2014 John Wiley & Sons, Ltd and Eating Disorders Association.
S. E. Lindner et al.
Inefficiencies in set-shifting were also found in unaffected firstdegree relatives and sisters of AN patients (Galimberti et al., 2013; Holliday, Tchanturia, Landau, Collier, & Treasure, 2005; Roberts et al., 2010; Tenconi et al., 2010). Kanakam, Raoult, Collier, and Treasure (2013) found set-shifting difficulties in unaffected twin siblings. Therefore, inefficiencies in set-shifting seem to be more frequent in unaffected close relatives of AN patients than in healthy individuals. It is not clear yet if set-shifting remains inefficient after recovery from AN. Quite a few studies have examined the question whether inefficient set-shifting is only a temporary state caused by malnutrition or whether it is a trait factor rather than a state factor. These studies also assessed recovered AN patients (recAN) in addition to currently ill AN patients in comparison with a healthy control group. Some studies showed that the inefficiency in set-shifting persists after recovery of AN (Danner et al., 2012; Holliday et al., 2005; Tchanturia, Morris, Surguladze, & Treasure, 2002; Tchanturia et al., 2012; Tenconi et al., 2010). In other studies, the performance in set-shifting of recAN patients lay between the performances of the AN group and the healthy control group. However, both differences were not statistically significant (Friederich et al., 2012; Nakazato et al., 2009; Roberts et al., 2010; Tchanturia et al., 2004; Tchanturia et al., 2011). In most of these studies, the sample size was rather small (13-30 recAN participants; Danner et al., 2012; Friederich et al., 2012; Holliday et al., 2005; Nakazato et al., 2009; Roberts et al., 2010; Tchanturia et al., 2004; Tchanturia et al., 2002). Tenconi et al. (2010) studied 63 weight recovered and 29 fully recovered AN patients and found no difference from the acutely ill AN patients, that is, the recAN patients showed inefficiencies in set-shifting compared with the healthy control group. Tchanturia et al. (2011, 2012) studied 72 and 90 recAN patients, respectively, with two different measures of set-shifting. They found an intermediate performance of the recAN participants, which was not significantly different from the performances of the AN group nor of the healthy control group regarding the Brixton Spatial Anticipation Test (Tchanturia et al., 2011). Regarding the Wisconsin Card Sorting Task, there was a significant difference between the recAN participants and the healthy control group (Tchanturia et al., 2012). Set-shifting inefficiencies are measurable, associated with AN and more frequent in healthy close relatives compared with a control group. It still needs to be resolved whether these inefficiencies are unrelated to the acute state of AN and therefore still come out inefficient after symptomatic recovery from AN. The present study aimed to investigate whether set-shifting ability is inefficient after full recovery of AN. Some studies have already examined set-shifting in a sample of recAN, but either the sample was rather small, or the definition of recovery was not sufficiently clear and rigorous. This study assessed set-shifting in a very large sample of truly recovered AN patients. The definition of recovery in this study considered physiological, behavioural and psychological variables. Furthermore, we took into account relevant personality, clinical (frequent comorbid disorders: depression, anxiety and obsessive-compulsive traits) and eating disorder variables for a better understanding of the relation between set-shifting and these variables. As mentioned earlier, relevant personality variables are perfectionism and impulsiveness as the opposite of rigidity. Especially perfectionism was of interest. AN patients show a consistent
Set-Shifting in Recovered AN
pattern of elevated perfectionism scores compared with healthy controls, which seems to be specific for AN compared with other psychiatric disorders, particularly one aspect of perfectionism ‘concerns over mistakes’ (Bardone-Cone et al., 2007; Bulik et al., 2003; Egan, Wade, & Shafran, 2011). Even after weight restoration, AN patients show still higher levels of perfectionism (Bardone-Cone et al., 2007). Regarding full recovery of AN, there are inconsistent findings (Bardone-Cone, Sturm, Lawson, Robinson & Smith, 2010; Srinivasagam et al., 1995). As Friederich and Herzog (2011) stated, perfectionism and obsessional traits are associated with impaired cognitive-behavioural flexibility. Also, Buehren et al. (2012) found a relation between perfectionism and cognitive inflexibility in an adolescent sample.
Methods After providing detailed information, all participants gave their informed consent to this study (case-control design with one cross-sectional assessment), previously approved by the ethical review committee of the University of Munich.
Participants The sample has been described already by Lindner, Fichter, and Quadflieg (2012, 2013). One hundred female participants with a lifetime diagnosis of AN who had fully recovered from AN (recAN) were recruited in the context of a large longitudinal study with former inpatients of the Schoen Hospital Roseneck for Behavioral Medicine in Prien (Germany). Full recovery of AN was defined as follows: no eating disorder according to DSM-IV-TR (American Psychiatric Association, 1994), body mass index (BMI = kg/m2) between 18.5 and 26, a regular menstrual cycle and absence of significant eating disorderspecific cognitions (fear of gaining weight, body image distortion, self-esteem dependent on body and shape with no or only slight manifestations) as well as absence of bingeing and compensatory behaviours for at least 1 year. Thus, the definition of recovery considered physical, behavioural and psychological components (Bachner-Melman, Zohar, & Ebstein, 2006; Bardone-Cone, Harney et al., 2010). One hundred healthy participants were recruited by means of notices in hospitals, universities and nurseries. Inclusion criteria for the healthy controls were the following: BMI above 18.5 since adolescence, no absent menstrual cycles for more than three months together with a low body weight, no pronounced eating disorder-specific cognitions and no close relative with an eating disorder. Exclusion criteria for both groups were severe manifestations of depression, anxiety and obsessive-compulsive traits; past or current diseases that could potentially affect brain functioning; past or current substance dependence, or current substance abuse; past or current schizophrenia; any other mental or physical condition that would have compromised the participation; and insufficient knowledge of the German language. The recAN and healthy participants were matched one to one for gender (all female), age (±2 years) and educational level (9, 10 and 12 years of school education and university graduates).
Eur. Eat. Disorders Rev. 22 (2014) 252–259 © 2014 John Wiley & Sons, Ltd and Eating Disorders Association.
253
S. E. Lindner et al.
Set-Shifting in Recovered AN
Measures The measures, especially the ones for the inclusion/exclusion criteria, are described in more detail by Lindner et al. (2012, 2013). For assessing eating disorder, we used the expert interview of the Structured Inventory for Anorexic and Bulimic Syndromes (Fichter & Quadflieg, 1999) with regard to the year prior to assessment (only recAN group) and lifetime (both groups). The Eating Attitudes Test (EAT-26; Garner, Olmsted, Bohr, & Garfinkel, 1982) complemented the control group assessment. Participants with a score higher than 20 in the EAT-26 would have been excluded; this was not the case in our study. The German short version of the Patient Health Questionnaire (Löwe, Spitzer, Zipfel, & Herzog, 2002) was used in both groups. A verbal intelligence measurement (Mehrfachwahl-WortschatzIntelligenztest; Lehrl, 2005) examined the matching based on educational level. The participant has to detect an existing word between other non-existing words in ascending difficulty level. Self-rating instruments assessed relevant clinical and personality variables: depression (Beck Depression Inventory I, BDI-I; Hautzinger, Bailer, Worall, & Keller, 1994), anxiety (Beck Anxiety Inventory, BAI; Margraf & Ehlers, 2007), obsessive-compulsive traits (Maudsley Obsessive-Compulsive Inventory, MOCI; Hodgson & Rachman, 1977; German: Kulick, 1979), impulsiveness (Barratt Impulsiveness Scale, BIS-11; Patton, Stanford, & Barratt, 1995), and perfectionism (Frost Multidimensional Perfectionism Scale German version, FMPS-D; Stöber, 1995). Apart from the FMPS-D total score, we also calculated the scores for the four subscales: concerns over mistakes and doubts about actions, parental expectations and criticism, personal standards and organisation (Stöber, 1998). We used these four subscales instead of the original six ones because of their better factorial stability (Stöber, 1998). Grant and Berg (1948) developed a card sorting test (Berg’s Card Sorting Test, BCST) for assessing set-shifting. The BCST consists of 64 cards with four different symbols (triangle, star, cross and circle) in four different numbers (1-4) and four different colours (red, green, yellow and blue). Four stimulus cards are presented to each participant. Other cards are consecutively shown, and the participant has to assign them to one of the four stimulus cards. The rule for this assignment can be symbol, number or colour, which is unknown to the participant; the participant just receives feedback whether the assignment is right or wrong. If the assignment is right, the participant has to stick to this rule until the feedback switches to wrong, which means the rule had changed. This occurs after five successful trials (category). The change is not predictable for the participant. The test ends when nine categories (three runs for each rule) are completed or each card has been shown twice (128 cards). We used a computer version of the BCST, which corresponds to the original test of Grant and Berg (1948), but it differed from the widely used Wisconsin Card Sorting Test (WCST) in the number of correct trials before the rule changes. In our test, the rule changed after five successful trials; in the WCST, the rule changes after ten correct trials. We calculated three outcome variables. First is the number of categories achieved. A category is achieved if the participant sorted five cards correctly. There were nine categories to be achieved, for each rule 254
three runs. The number of categories achieved measures not only set-shifting; it is a more global measure for building concepts and categories. Second is the number of perseverations. A perseveration occurs when the participant clings to the last rule and do not change to a new rule. Higher number of perseverations is a sign for inefficiency in set-shifting. Third is the time spent for a shift (shift cost). This indicates the difference between the reaction time of the first trial with the feedback ‘wrong’, indicating a rule change, and the mean reaction time of the three trials before. Shift cost measures cognitive flexibility on the basis of reaction time and not of errors (Deák, 2003). Statistical analyses We used t-tests to compare the recAN group with the healthy control group on demographic and clinical variables. Analyses of variance addressed the results in set-shifting. The Bonferroni procedure allowed for corrections of multiple comparisons (significance level at p = .017 for the BCST). Pearson correlations permitted to express the relationship between set-shifting variables and clinical characteristics. For BDI-I, BAI and FMPS-D, there were significant differences between the groups. Therefore, we executed the correlation analyses separately for each group. For BIS-11 and MOCI, the total sample was considered. Subsequently, an analysis of covariance took into account relevant control variables. The recAN group was further divided into subgroups: women who suffered only from the restricting type of AN lifetime (recAN rest) and women who suffered from both the restricting and the binge eating/purging subtype of AN or any other eating disorder lifetime (recAN other). For detecting group differences, Scheffé post-hoc tests were executed. Because of the explorative character of the further analyses, the p-values were not corrected. In order to estimate the relevance of significant results, we calculated Cohen’s effect sizes (Cohen, 1988). Cohen’s effect sizes (d) are defined as small (d = 0.2), medium (d = 0.5) and large (d = 0.8).
Results Sample Table 1 presents the demographic and clinical variables. The sample is described in more detail by Lindner et al. (2012, 2013). The scores of depression (BDI-I), anxiety (BAI) and perfectionism (FMPS-D) all came out higher in the recAN group. Regarding obsessive-compulsive traits (MOCI) and impulsiveness (BIS-11), the two groups were not significantly different. Set-shifting The recAN group achieved significantly fewer categories as the healthy control group (F(1, 199) = 7.47, p = .007, Cohen’s d = .39). The recAN group showed significantly more perseverations than the healthy control group (F(1, 199) = 6.07, p = .015, Cohen’s d = .35). The reaction time for shifts was shorter in the recAN group compared with the healthy control group (F(1, 196) = 6.53, p = .011, Cohen’s d = .36). The effect sizes were small. The results are shown in Table 2.
Eur. Eat. Disorders Rev. 22 (2014) 252–259 © 2014 John Wiley & Sons, Ltd and Eating Disorders Association.
S. E. Lindner et al.
Set-Shifting in Recovered AN
Table 1 Demographic and clinical variables RecAN group
Age (years: range 22-51) Intelligence estimate (MWT-B: range 0-37) Current BMI Minimal BMI Age of onset of AN (years)* Duration of AN (years)** Duration of recovery from AN (years)* Depression (BDI-I: range 0-63)*** Anxiety (BAI: range 0-63) Perfectionism (FMPS-D: range 29-145)*** CMD PEC PS O Obsessive-compulsive traits (MOCI: range 0-30) Impulsiveness (BIS-11: range 30-120)
Healthy control group
t-test
M
SD
M
SD
t
p
34.49 30.43 20.86 13.92 18.95 3.88 6.37 3.35 4.40 71.78 30.41 22.31 19.05 23.13 3.81 57.49
7.13 3.32 1.31 1.81 4.19 3.17 3.93 4.17 5.07 20.78 10.31 9.03 6.40 3.83 2.92 6.75
34.53 30.88 21.80 20.09
7.26 3.29 1.37 .93
0.034 0.963 4.983 30.288
.973 .337
