Investigating differences in cognitive flexibility, clinical perfectionism, and eating disorder-specific rumination across anorexia nervosa illness states

ABSTRACT

Introduction: Cognitive inflexibility, clinical perfectionism, and eating disorder (ED)-specific rumination are common characteristics reported in anorexia nervosa (AN) and may contribute to the maintenance of the illness. It is suggested that clinical perfectionism and rumination may mediate the relationship between cognitive flexibility and AN pathology; however, research to date has not investigated all these factors together. The aim of the current study was to explore the relationships between these factors and how they may relate to ED symptoms in AN. Methods: Participants included 15 women with a current diagnosis of AN, 12 women who had a past diagnosis of AN and were currently weight-restored, and 15 healthy controls (HCs). Results: The results revealed that participants with both acute and weight-restored AN self-reported poorer cognitive flexibility than HCs, but the groups did not differ in performance on objective assessments of cognitive flexibility. Participants with AN also reported significantly greater clinical perfectionism and ED-specific rumination than HC. A parallel mediation analysis found that ED-specific rumination mediated the relationship between subjective cognitive flexibility and ED symptoms. Further, subjective cognitive flexibility directly influenced ED symptoms. However, the mediation model was not significant for objective cognitive flexibility. Conclusion: The findings of this study have implications for potential treatment barriers and factors which might contribute to the risk of relapse.

Clinical implications

• Perfectionism and rumination were present in in women with anorexia nervosa (AN).

• Self-reported cognitive flexibility was poor in women with AN, suggesting perceived rigidity.

• Women with AN performed well on a cognitive flexibility task, indicating adaptability.

• Perceived cognitive flexibility directly and indirectly impacted AN symptoms.

• Perceived cognitive flexibility, perfectionism and rumination may impact treatment.

Introduction

Anorexia nervosa (AN) is an eating disorder (ED) characterised by significantly low body weight and disturbances in how one’s body weight or shape is experienced (American Psychiatric Association, 2013). Existing AN treatments are inadequate in that many people do not experience long-term recovery from their illness (Eddy et al., 2017). Thus, further research is needed to better understand the causes and maintaining factors of AN so that they can be appropriately targeted in treatment. Cognitive flexibility (the ability to effectively adapt to changes in the environment and/or changing task demands; Deák, 2003) may be involved in AN symptom maintenance and treatment resistance (Tchanturia et al., 2013; Treasure & Schmidt, 2013). Research indicates that adults with acute AN self-report poor cognitive flexibility (Dell’osso et al., 2018; Herbrich et al., 2018; Lounes et al., 2011; Miles et al., 2020), and often perform worse than healthy controls (HCs) on neurocognitive tasks of cognitive flexibility (Abbate Daga et al., 2011; Aloi et al., 2015; Steward et al., 2019). Cognitive inflexibility is often observed in AN in real-world situations such as intense exercise routines, rules for meal times, and other food rituals (Rößner et al., 2017; Tchanturia et al., 2013). Targeting cognitive inflexibility may enable meaningful behaviour change and thus has potential implications for treatment (Baldock & Tchanturia, 2007; Pitt et al., 2010). Although cognitive flexibility difficulties in AN may be relevant to treatment, cognitive rigidity alone cannot explain all eating disorder symptomology. Building our understanding of how cognitive flexibility may relate to other relevant cognitive and psychological factors may provide new insights that further accelerate and improve treatments for AN.

Perfectionism is an established, common characteristic of people with AN (Fairburn et al., 2003), and research has found that people with acute AN score significantly higher than HCs in assessments of perfectionism (Dahlenburg et al., 2019; Halmi et al., 2000; Lindner et al., 2014; Moor et al., 2004; Sutandar-Pinnock et al., 2003; Vall & Wade, 2015; Wade et al., 2008). Very few papers have investigated perfectionism in AN after weight-restoration or full recovery, and overall, the results have been inconsistent regarding group differences (Bachner-Melman et al., 2006; Gárriz et al., 2020; Lindner et al., 2014; Srinivasagam et al., 1995; Sutandar-Pinnock et al., 2003). Overall, research to-date suggests that perfectionism in EDs is “not a stable trait but is malleable and sensitive to clinical interventions” (Wade et al., 2016, p. 217). Existing research has typically used multidimensional assessments of perfectionism which define perfectionism according to the ‘dimensions’ or ‘aspects’ that comprise perfectionism (Shafran et al., 2002). Multidimensional assessments of perfectionism are broad (e.g., can include an evaluation of parental expectations) and capture both the adaptive and maladaptive aspects of perfectionism, yet only the maladaptive aspects of perfectionism may be clinically relevant to the study and treatment of mental illnesses (Rhéaume et al., 2000). Clinical perfectionism refers to a system of self-evaluation in which self-worth is primarily judged on one’s capacity to successfully meet personally demanding goals despite any negative consequences (Fairburn et al., 2003; Shafran et al., 2002). Clinical perfectionism is considered dysfunctional and is associated with adverse effects such as anxiety, poor concentration, negative health effects, and/or social isolation (Shafran et al., 2002). Despite the abundance of research on perfectionism in AN, to our knowledge, no study has focused on clinical perfectionism in AN.

Another construct which may relate to cognitive flexibility is rumination. Rumination refers to a series of related thoughts that are repetitious, unintended, and difficult to reduce or stop (Martin & Tesser, 1996). People with AN often engage in rumination which is specifically related to their eating, shape, and weight (Dondzilo et al., 2016), and research suggests that ED-specific rumination may be associated with ED symptomology (Cowdrey & Park, 2012; Dondzilo et al., 2016). Cowdrey and Park (2011) found that participants with lifetime AN reported significantly greater ED-specific rumination than HCs. However, this study had statistical limitations and participants at various stages of AN were combined into a single group (i.e., acute, weight-restored, and fully recovered participants were not separated). The combination of these participants into a single group is potentially problematic, as there is currently insufficient research to determine if there are significant differences in ED-specific rumination between people at different AN illness stages. Understanding potential differences in ED-specific rumination between people at different stages of AN could inform treatment goals and contribute to our knowledge of rumination as a state or trait factor.

Although cognitive flexibility, perfectionism, and rumination may be common factors in AN, these factors should not be considered to be entirely independent. These factors may relate to one another and may play a combined role in AN symptomology. One study found that HCs who ruminate perform poorer on a cognitive flexibility test compared to people who do not ruminate (Davis & Nolen-Hoeksema, 2000). The only study to consider cognitive flexibility and rumination in people with AN, however, found that general rumination did not significantly correlate with performance on neurocognitive assessments of cognitive flexibility (Startup et al., 2013). One study supported links between cognitive flexibility and perfectionism in AN (Bühren et al., 2012); however, other research has been inconsistent or has not supported a link between these factors (Buzzichelli et al., 2018; Dell’osso et al., 2018; Lindner et al., 2014; Talbot et al., 2015; Vall & Wade, 2015).

Perfectionism and rumination may mediate the relationship between cognitive flexibility and AN pathology (Buzzichelli et al., 2018; Startup et al., 2013). Poor cognitive flexibility may contribute to the development of rigid rules, high expectations for oneself, and an unwillingness to change or act flexibly even in response to negative consequences, hence reinforcing clinical perfectionism traits. This clinical perfectionism may then be applied to the context of body image and eating, thus encouraging disordered eating and the development of ED symptoms. Further, if someone experiences inflexibility or perceives themselves to be cognitively rigid, they might struggle to disengage from a thought and find themselves engaging in repetitive and cyclical thought processes. Therefore, they may ruminate on ED-related content and have difficulties in focusing on non-ED content. This increased ED-specific rumination may then lead to increases in ED symptoms such as binge eating, food restriction, and poor body image.

Despite the considerable research which has considered cognitive flexibility, perfectionism, and rumination in AN, no study has jointly considered the relationships between cognitive flexibility, clinical perfectionism, ED-specific rumination, and ED symptoms. Therefore, the aim of the current study was to explore the relationships between these variables in AN. It was hypothesised that compared to HCs, cognitive flexibility (as assessed by a self-report questionnaire and a neurocognitive task) would be poorer in people with acute AN and those weight-restored from AN (AN-WR). It was also predicted that clinical perfectionism and ED-specific rumination would be significantly higher in acute AN and AN-WR groups compared to HCs. Finally, it was hypothesised that clinical perfectionism and ED-specific rumination would mediate the relationship between cognitive flexibility (as assessed by a self-report questionnaire and a neurocognitive task) and ED symptoms across the entire sample.

Methods

Participants

Participants in this study comprised 15 women with a current diagnosis of AN (M age = 26.72, SD = 7.74), 12 women who had a past diagnosis of AN and were currently weight-restored (AN-WR; M age = 21.27, SD = 2.66), and 15 HCs (M age = 25.37, SD = 4.28). Table 1 and Supplementary Material provide further details of the samples.

Table 1. Sample characteristics.

	Group			Test statistics
	AN	AN-WR	HC	F	p	Significant group differences	Effect size
Age	26.72 (7.74)	21.27 (2.66)	25.37 (4.28)	3.49	.040	AN > AN-WR; AN = HC; AN-WR = HC	.15^^
Current BMI	16.88 (2.93)	22.38 (2.27)	23.52 (2.15)	30.01	>.001	AN < AN-WR, HC; AN-WR = HC	.61^^
WTAR	111.07 (9.48)	111.82 (10.25)	104.87 (6.84)	2.61	.086	-	.12^^
				t	p
Age of AN onset	15.87 (4.94)	15.18 (4.12)	-	.37	.712	-	.15^#
Illness duration	10.41 (8.47)	3.72 (2.63)	-	2.52	.019	AN > AN-WR	1.00^#
Duration of weight restoration	-	1.95 (1.74)	-	-	-	-

Note:age, age of AN onset, duration of illness, and duration of weight restoration are reported in years; WTAR standard scores are reported; AN = anorexia nervosa; AN-WR = anorexia nervosa, weight-restored; BMI = body mass index, kg/m²; HC = healthy control; WTAR = Wechsler Test of Adult Reading.

^Eta-squared effect size.

#Cohen’s d effect size.

Inclusion criteria for all participants were female sex, over the age of 18 years old, fluent in English, no history of a head injury or a neurological condition (e.g., autism spectrum disorder), no history of psychosis or psychotic symptoms, and no family history of psychotic disorders. Participants in the acute AN group were required to have a current primary diagnosis of AN as determined by a psychologist or psychiatrist. Participants in the AN-WR group were required to have past diagnosis of AN and to have maintained a body mass index (BMI) over 18 for a minimum of 3 months. Participants self-reported their lowest weight over the past 3 months and this information was used to calculate lowest BMI and ensure that participants met inclusion criteria. Due to the poor consensus in the literature regarding the effects of psychiatric medication or comorbid diagnoses on cognitive performance (Lopez et al., 2008), participants in the AN and AN-WR groups were not excluded if they were taking psychiatric medication or had a diagnosis of another mental illness (other than a psychotic disorder). HCs were required to have no personal or family history of an ED and were excluded if they were currently taking psychiatric medication or had a current formal diagnosis of a mental illness as determined by a psychologist or psychiatrist.

Procedure and measures

The study was approved by the Swinburne University Human Research Ethics committee (Ref: 20211127–8734). Participants were recruited from an established participant registry of people with a lifetime diagnosis of an ED, social media posts, posters placed in public places, and advertisements with Australian ED organisations including Eating Disorders Victoria and the National Eating Disorders Collaboration. After an initial screening interview via email or phone call, participants attended a face-to-face testing session that took approximately an hour and a half. The clinical interviews, self-report questionnaires, physical measurements, and neurocognitive testing all took place in this session. Participation was voluntary, and all participants provided written informed consent at the beginning of the study. Participants were compensated for their time/travel expenses.

The MINI International Neuropsychiatric Interview 7.0.2 (Sheehan et al., 1998) was used to screen participants for psychiatric disorders according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), and to confirm acute AN diagnoses. A detailed clinical demographics questionnaire which included questions on AN age of onset, illness duration, and duration of weight-restoration was completed, and participants’ height and weight were measured to calculate BMI (see Table 1). AN age of onset and illness duration were calculated from time of significant symptom onset rather than time of AN diagnosis. The Wechsler Test of Adult Reading (WTAR; Wechsler, 2001) was used as an estimate of premorbid intelligence (Green et al., 2008). The WTAR comprises 50 words of increasing difficulty which participants are asked to read aloud. The total number of correctly pronounced words is summed and converted to an age-adjusted standardised score.

Self-report questionnaires

Eating Disorders Examination-Questionnaire version 6.0 (EDE-Q; Fairburn & Beglin, 2008). The 28-item EDE-Q was used to assess the severity and frequency of ED symptomatology over the past 28 days. The EDE-Q includes a mix of short answer questions and items which participants respond to using a seven-point Likert-type scale (e.g. Have you had a definite desire to have an empty stomach with the aim of influencing your shape or weight?). Higher scores on the EDE-Q represent greater severity/frequency of ED symptoms. The EDE-Q comprises four subscales, Restraint, Eating Concern, Weight Concern, and Shape Concern, in addition to the global score. In the current study, internal consistency in the AN group was poor for the Eating Concern and Weight Concern subscales (Cronbach’s alphas =.65); however, internal consistency was good to excellent for the other subscales across the three groups (Cronbach’s alpha =.77–.95).

Cognitive Flexibility Inventory (CFI; Dennis & Vander Wal, 2010). The 20-item CFI was included as a self-report assessment of cognitive flexibility. The CFI evaluates cognitive flexibility across three domains: (a) the tendency to perceive challenging situations as controllable; (b) the ability to consider several alternative explanations for events and human behaviour; and (c) the ability to generate multiple alternative solutions to difficult situations. Items on the CFI (e.g. I like to look at difficult situations from many different angles) are responded to using a seven-point Likert-type scale ranging from strongly disagree (1) to strongly agree (7). Although the CFI includes two subscales (Alternatives and Control), only the total sum score was used in the current study. Lower scores on the CFI indicate poorer cognitive flexibility. In the current study, internal consistency for the CFI total score was good in all participant groups (Cronbach’s alphas ≥.76).

Clinical Perfectionism Questionnaire (CPQ; Fairburn et al., unpublished, cited in Riley et al., 2007). The CPQ is a 12-item questionnaire used to assess clinical perfectionism in domains other than eating, weight, or appearance. Participants used a four-point Likert-type scale to rate how well the items described them over the past month (e.g. Have you felt a failure as a person because you have not succeeded in meeting your goals?). Higher scores on the CPQ represent higher levels of clinical perfectionism. In the current study, internal consistency for the CPQ was good in all participant groups (Cronbach’s alphas ≥.75).

Ruminative Response Scale for Eating Disorders (RRS-ED; Cowdrey & Park, 2011). The RRS-ED comprises nine questions designed to assess ED-specific rumination (e.g. How frequently do you think ‘why do I have such issues with my eating, weight and/or shape?’). Each question is responded to on a four-point Likert-type scale ranging from ‘almost never’ (1) to ‘almost always’ (4). Higher scores on the RRS-ED are indicative of more frequent ED-related rumination. The RRS-ED includes two subscales, Brooding and Reflection; however, only the total score was used in this study. In the current study, internal consistency for the RRS-ED was good in all participant groups (Cronbach’s alphas ≥.74).

Neurocognitive task

A computerised version of the Wisconsin Card Sorting Test (WCST; Heaton et al., 1993) was used as a neurocognitive assessment of cognitive flexibility. The WCST was administered using the Inquisit Web software and was hosted online by Millisecond (Millisecond Software, 2016).

The WCST is a card-based test in which participants sort a response card with one of four multidimensional stimulus cards that vary in colour, form, and number of stimuli. Participants used a computer mouse to select a stimulus card by clicking on it. No explicit instructions were given to the participants on how to sort the response cards, and they were required to use the feedback provided after each trial (correct or wrong) to learn the sorting rule and correctly sort the response cards. After correctly sorting ten response cards in a row, the sorting rule changed without warning and participants were required to use trial and error again to establish the new sorting rule. The test ended after the completion of all six categories (colour, form, number, colour, form, number) or 128 trials (Heaton et al., 1993). Following the recommendations of Miles et al. (2021), the test was scored in accordance with the standardised manual (Heaton et al., 1993) using custom scripts that were written in MATLAB (2018). Two outcomes were used to assess cognitive flexibility: the percentage of perseverative responses (i.e., persistent responses made on the basis of a stimulus dimension that the participant has been previously informed is incorrect) and the percentage perseverative errors (i.e., perseverative responses that are incorrect) made. A higher percentage of perseverative responses and perseverative errors indicates poorer cognitive flexibility. To test the mediation hypothesis, only the percentage of perseverative responses variable was included in the model to avoid repetitive and potentially redundant analyses. This variable was selected as it includes perseverative responses that are correct and those that are incorrect (i.e., perseverative errors).

Statistical analysis

Analyses were completed using SPSS version 27, and the alpha level was set at .05 for all analyses. To reduce the number of variables and the need to correct for multiple comparisons, the CFI and RRS-ED total scores were used in all analyses. Across the variables of interest, the data were determined to be normally distributed and there were no univariate outliers within the participant groups. t-tests and one-way between groups analyses of variance (ANOVAs) with a-priori contrasts (AN and AN-WR, AN and HC, AN-WR and HC) were conducted, as appropriate. Where the assumption of homogeneity of variances was violated, Welch’s F is reported. Cohen’s d effect sizes are reported for the t-tests (where d = 0.2 represents a small effect, d = 0.5 indicates a medium effect, and d = .8 is a large effect; Cohen, 1988), and eta-squared (where η² = 0.01 represents a small effect, η² = 0.06 indicates a medium effect, and η² = .14 is a large effect; Cohen, 1988) is reported for the ANOVAs.

Two parallel mediation analyses using ordinarily least squares path analysis were conducted using the PROCESS macro for SPSS (Hayes, 2017). In both analyses, clinical perfectionism (CPQ Total) and ED-specific rumination (RRS-ED Total) were included as potential mediators and overall eating disorder symptoms (EDE-Q Global score) was the dependent variable. In the first parallel mediation analysis, self-reported cognitive flexibility (CFI Total) was the independent variable, and in the second parallel mediation analysis WCST percentage of perseverative responses was the independent variable. As recommended by Hayes (2017), 10000 bootstrap samples and percentile bootstrap confidence intervals were calculated for the specific indirect effects. The absence of the value of zero in the 95% percentile bootstrap confidence interval would indicate a statistically significant indirect effect.

Results

Group comparisons

With large effect sizes, significant group differences were found for the EDE-Q. AN and AN-WR participants scored significantly higher on the EDE-Q compared to HCs, and acute AN participants scored significantly higher on these assessments than AN-WR participants (see Table 2). Significant group differences were found for the CFI, CPQ, and RRS-ED, with large effect sizes (see Table 2). Compared to HCs, the AN and AN-WR groups scored significantly higher on the CPQ and RRS-ED and significantly lower on the CFI. There were no significant differences between AN and AN-WR participants on the RRS-ED and CFI; however, on the CPQ, AN participants scored significantly higher than AN-WR participants. There were no significant differences between the groups on the cognitive flexibility outcomes of the WCST.

Table 2. Group comparisons for eating disorder symptoms, cognitive flexibility, clinical perfectionism, and eating disorder-specific rumination.

		Group			Test statistics
		AN	AN-WR	HC	F	p	Significant group differences	η^2
EDE-Q
	Restraint	3.87 (1.64)	1.90 (1.84)	.84 (1.34)	13.70	>.001	AN > AN-WR, HC; AN-WR = HC	.41
	Eating concern	3.11 (1.25)	1.85 (1.43)	.39 (.67)	28.70^^	>.001	AN > AN-WR, HC; AN-WR > HC	.52
	Weight concern	4.69 (1.07)	2.83 (1.56)	1.00 (1.18)	32.11	>.001	AN > AN-WR, HC; AN-WR > HC	.62
	Shape concern	4.96 (1.23)	3.18 (1.77)	1.16 (1.28)	26.94	>.001	AN > AN-WR, HC; AN-WR > HC	.58
	Global score	4.16 (1.02)	2.44 (1.53)	.85 (1.09)	28.03	>.001	AN > AN-WR, HC; AN-WR > HC	.59
CFI Total		97.13 (14.54)	105.75 (12.76)	117.80 (9.12)	10.64	>.001	AN < HC; AN-WR < HC; AN = AN-WR	.35
WCST
	% of perseverative responses	12.64 (7.82)	15.36 (9.28)	13.12 (9.06)	.36	.701	-	.02
	% of perseverative errors	11.53 (6.39)	13.54 (7.24)	12.05 (7.75)	.28	.760	-	.01
CPQ Total		36.07 (5.30)	30.17 (6.01)	25.07 (5.85)	13.96	>.001	AN > AN-WR, HC; AN-WR > HC	.42
RRS-ED Total		25.13 (6.29)	21.83 (5.56)	11.60 (3.81)	26.25	>.001	AN, AN-WR > HC; AN = AN-WR	.57

Note: AN = anorexia nervosa; AN-WR = anorexia nervosa, weight-restored; CFI = Cognitive Flexibility Inventory; CPQ = Clinical Perfectionism Questionnaire; EDE-Q = Eating Disorders Examination-Questionnaire; HC = healthy control; RRS-ED = Ruminative Response Scale for Eating Disorders; WCST = Wisconsin Card Sorting Test; lower CFI scores = poorer cognitive flexibility, whereas higher scores on all other measures = increased symptomatology.

^{^}Welch’s F is reported as the assumption of homogeneity of variances was violated.

Parallel mediation analyses

Self-reported cognitive flexibility

From a parallel mediation analysis, CFI score was directly and indirectly related to EDE-Q Global score. As can be seen in Figure 1, CFI score did not directly related to CPQ score (a¹ = −.08). However, CPQ score was related to EDE-Q Global score (b¹ = .09) such that participants who experienced greater clinical perfectionism reported greater ED symptoms. The percentile bootstrap confidence interval for the specific indirect effect (ab¹ = −.007) included zero (−.02 to .003) indicating that the effect was not significant.

Figure 1. Mediation of the effect of self-reported cognitive flexibility on eating disorder symptoms through clinical perfectionism and eating disorder-specific rumination.

Note: Unstandardised regression coefficients presented on top and standard errors presented in parentheses with *p < .01 and **p < .001; CFI = Cognitive Flexibility Inventory; CPQ = Clinical Perfectionism Questionnaire; EDE-Q = Eating Disorders Examination-Questionnaire; RRS-ED = Ruminative Response Scale for Eating Disorders

CFI score was directly related to RRS-ED score (a² = −.21) such that participants who self-reported poorer cognitive flexibility experienced greater ED-specific ruminations. Further, RRS-ED score was significantly related to EDE-Q Global score (b² = .11) such that participants who experienced greater ED-specific ruminations reported greater ED symptoms. The percentile bootstrap confidence interval for the specific indirect effect (ab² = −.02) did not include zero (−.04 to −.009) indicating that the indirect effect was significant.

CFI score was directly and independently related to EDE-Q Global score (c’ = −.03, p = .008).

Neurocognitive assessment of cognitive flexibility

From a parallel mediation analysis, the percentage of perseverative responses on the WCST did not directly nor indirectly relate to EDE-Q Global score. As can be seen in Figure 2, WCST percentage of perseverative responses did not directly relate to CPQ score (a¹ = −.20). However, CPQ score was related to EDE-Q Global score (b¹ = .08) such that participants who experienced greater clinical perfectionism reported greater ED symptoms. The percentile bootstrap confidence interval for the specific indirect effect (ab¹ = −.02) included zero (−.04 to .004) indicating that the effect was not significant.

Figure 2. Mediation of the effect of a neurocognitive assessment of cognitive flexibility on eating disorder symptoms through clinical perfectionism and eating disorder-specific rumination.

Note: Unstandardised regression coefficients are presented on top and standard errors are presented in parentheses with *p < .05 and **p < .001; CPQ = Clinical Perfectionism Questionnaire; EDE-Q = Eating Disorders Examination-Questionnaire; RRS-ED = Ruminative Response Scale for Eating Disorders; WCST = Wisconsin Card Sorting Test.

WCST percentage of perseverative responses did not directly relate to RRS-ED score (a² = .07). However, RRS-ED score was significantly related to EDE-Q Global score (b² = .14) such that participants who experienced greater ED specific ruminations reported greater ED symptoms. The percentile bootstrap confidence interval for the specific indirect effect (ab² = .01) included zero (−.03 to .06) indicating that the indirect effect was not significant.

WCST percentage of perseverative responses did not directly and independently relate to EDE-Q Global score (c’ = −.01, p = .565).

Discussion

The aim of this study was to investigate cognitive flexibility, clinical perfectionism, and ED-specific rumination in women with acute AN, people weight-restored from AN, and HCs. The first hypothesis that cognitive flexibility would be poorer in women with AN than HCs was supported for subjective, but not objective, cognitive flexibility. It was found that both the acute AN and AN-WR groups self-reported significantly poorer cognitive flexibility than HCs, but the two AN groups did not differ from one another. These findings support past research that has found poorer self-reported cognitive flexibility in people with acute AN compared to HCs (Dell’osso et al., 2018; Herbrich et al., 2018; Lang et al., 2015; Lao-Kaim et al., 2015; Lounes et al., 2011; McAnarney et al., 2011). The absence of a significant difference between acute and weight-restored AN groups suggests that self-perceptions of inflexibility may be a trait feature of AN that is present irrespective of illness state. Alternatively, ongoing perceptions of poor flexibility after weight-restoration could suggest a scar effect. That is, perceptions of poor cognitive flexibility may be a consequence of AN which does not ameliorate with recovery. Whether these findings extend to individuals who are fully recovered from AN requires further research.

Significant group differences in cognitive flexibility were not found for the neurocognitive task, suggesting that objective cognitive flexibility may not be impaired in people with AN. Several other studies have also not found differences between people with past or current AN and HCs in cognitive flexibility as assessed by the WCST (Lao-Kaim et al., 2015; Paslakis et al., 2019; Perpiñá et al., 2017; Pignatti & Bernasconi, 2013; Roberts et al., 2010; Sato et al., 2013; Stedal et al., 2019; Talbot et al., 2015; Wollenhaupt et al., 2019). However, past research is inconsistent, and a number of studies have indeed reported poorer performance on the WCST in AN compared to HCs (Abbate Daga et al., 2011, 2014; Aloi et al., 2015; Buzzichelli et al., 2018; Fagundo et al., 2012; Galimberti et al., 2013; Steward et al., 2019; Talbot et al., 2015; Tenconi et al., 2010). It has been argued that the WCST lacks ecological validity and does not predict or relate to everyday executive functioning (Burgess et al., 2006; Chaytor et al., 2006). Indeed, the abstract stimuli in the WCST may be irrelevant to everyday life and unrelated to the experiences of people with AN. People with AN may perform differently to the current findings if illness-relevant stimuli (e.g., food pictures) were used in neurocognitive tasks. As limited research to date has investigated cognitive flexibility in EDs using illness-relevant stimuli, it is unclear if and/or how different stimuli may have changed findings.

The current findings indicate that neurotypical women with AN perceive themselves as inflexible and resistant to change but perform within the normal range on neurocognitive assessments of flexibility. Past research has demonstrated that self-report and neurocognitive assessments of cognitive flexibility are poorly related (Gonzalez et al., 2013; Howlett et al., 2022; Johnco et al., 2014; Lounes et al., 2011; Miles et al., 2022; Tchanturia et al., 2004). The results of this study add to the growing body of research which suggests a discrepancy between self-report and neurocognitive assessments of cognitive flexibility in people with AN (Lounes et al., 2011; Miles et al., 2022; Tchanturia et al., 2004).

Hypothesis two, which sought to investigate group differences in clinical perfectionism and ED-specific rumination, was supported. Regarding clinical perfectionism, women with acute AN reported the highest levels of clinical perfectionism, followed by the participants who were weight-restored, and then HCs. The results support existing research that, whilst not explicitly examining clinical perfectionism, has found high levels of perfectionism in acute AN (Dahlenburg et al., 2019; Halmi et al., 2000; Lindner et al., 2014; Moor et al., 2004; Sutandar-Pinnock et al., 2003; Vall & Wade, 2015; Wade et al., 2008) and indicates that people with acute and weight-restored AN experience a dysfunctional form of perfectionism that may have adverse effects and impact their self-evaluation. Encouragingly, the results suggest that clinical perfectionism may improve with weight-restoration and recovery. An alternate explanation for our findings is that clinical perfectionism may not change with illness state, but women who initially have lower levels of clinical perfectionism may experience a less severe illness course and may be more likely to recover. Indeed, despite the similar ages of AN onset, the participants with acute AN had a significantly longer illness duration than the participants who were weight-restored. Further research is required to investigate this possibility, and longitudinal research will be necessary to examine changes in clinical perfectionism from acute AN to weight-restoration to full recovery.

After comparing the groups in ED-specific rumination, it was found that the acute AN and AN-WR groups experienced significantly more ED ruminations than HCs, yet there were no significant differences between the AN groups. These results are in line with Cowdrey and Park (2011) and Cowdrey et al. (2014) who found that people with lifetime AN and people who were fully recovered from AN reported significantly greater ED-specific rumination than HCs; as well as other research that has demonstrated increased general rumination in acute AN compared to HCs (Dell’osso et al., 2018; Rawal et al., 2010; Startup et al., 2013). This study is the first to separate women with acute AN and women who are weight-restored and compare their experiences of ED-specific rumination. The findings indicate that at a weight-restored state, women experience significant ED-specific ruminations that may be a symptom of AN rather than a side effect of starvation alone. It is possible that ED-specific ruminations reduce further in frequency after full recovery from AN, however, further research is required to investigate this possibility. The findings of this study highlight the importance of considering psychological aspects of recovery from AN rather than uniquely considering weight improvements and behavioural symptoms.

Past research has demonstrated relationships between cognitive flexibility, perfectionism, rumination, and ED symptoms (Bühren et al., 2012; Cowdrey & Park, 2012; Davis & Nolen-Hoeksema, 2000; Dondzilo et al., 2016; Lindner et al., 2014; Whitmer & Gotlib, 2012), and this study aimed to test a proposed mediation model. Hypothesis three and the proposed mediation models were partially supported. Although clinical perfectionism was associated with ED symptoms, it was not found to mediate the relationship between subjective cognitive flexibility and ED symptoms. Importantly, however, ED-specific rumination mediated the relationship between subjective cognitive flexibility and ED symptoms. Furthermore, variations in subjective cognitive flexibility had a direct effect on ED symptoms. The current findings suggest that poor cognitive flexibility may contribute to increased ED-related rumination, which in turn may contribute to an increase in disordered eating symptoms.

Contrastingly, the same mediation model for the neurocognitive assessment of cognitive flexibility was not supported. Clinical perfectionism and ED-specific rumination did not mediate the relationship between objective cognitive flexibility and ED symptoms. Furthermore, variations in objective cognitive flexibility did not have a direct effect on ED symptoms. The results demonstrate that ED symptoms are not directly nor indirectly influenced by objective cognitive flexibility, clinical perfectionism, or ED-specific ruminations. These findings support the work of Buzzichelli et al. (2018) who found that perfectionism did not significantly mediate the relationship between the WCST and a drive for thinness in a sample of HCs and people with acute AN.

Limitations

There were several limitations to this study. Firstly, the sample size was small and thus the statistical power of the analyses was reduced. The lowered statistical power may have diminished the extent to which the findings can be meaningfully interpreted and hence the current study requires replication. In addition, internal consistency for the Eating Concern and Weight Concern subscales of the EDE-Q was poor in the acute AN group. Secondly, the sample of participants who were weight-restored was heterogeneous in terms of length of weight-restoration, and recovery was not defined with consideration of behavioural and psychological symptoms. It is possible that there are differences in cognitive flexibility, clinical perfectionism, and/or ED-specific rumination between people who are recently weight-restored and those who have been weight-restored for a longer time and no longer experience any ED symptoms (i.e. fully recovered). Thirdly, many participants with AN were taking psychiatric medication or had comorbid diagnoses and these factors may have impacted performance on the cognitive flexibility task. Due to the variability across participants in the combinations of medications which they were taking and comorbid diagnoses which they had received, medication and comorbid diagnoses could not be controlled for in the analyses. Lastly, this study only considered neurotypical women, and the results may not generalise to people who are neurodiverse.

Implications and future research directions

The ongoing self-perception of poor cognitive flexibility after women have attained weight-restoration has implications as a potential barrier for recovery from AN. If someone considers themself as rigid and unable to change, they may have ongoing difficulties engaging in treatment, attempting to change their behaviours, and achieving full recovery. Perceptions of cognitive inflexibility may need to be directly addressed throughout treatment and considered when challenging AN symptomology. As this study is one of the first to investigate self-reported cognitive flexibility in people who are weight-restored from AN, additional research is required to further elucidate the role of perceptions of cognitive inflexibility in recovery from AN. In addition, the findings of the current study also suggest that frequent ED ruminations remain even after weight-restoration. These ruminations are important to consider as they may slow recovery and/or increase the risk of relapse. ED ruminations (which are by nature repetitive and difficult to prevent or stop) may incite someone to engage in ED behaviours and may also increase cognitive load, thus potentially hampering attempts to ignore these thoughts and focus on recovery. As this is the first study to investigate ED-specific rumination in people who have achieved weight-restoration (but may not yet be fully recovered from AN), future research is required to explore the impacts of ED-specific rumination on recovery from AN.

The results from the mediation analyses indicate that although objective cognitive flexibility may not contribute to ED symptoms, subjective self-perceptions of poor cognitive flexibility may directly and indirectly (through ED-specific ruminations) influence ED symptoms. The results highlight the distinction between subjective and objective cognitive flexibility and provide potential avenues for the treatment of AN. It may be beneficial for therapists to initiate discussions with patients on flexible thinking and behaviours and explore differences between objective abilities and subjective perceptions of cognitive flexibility. Challenging patients to question their own perceptions of their ability to change may help to facilitate improvements in self-efficacy and modifications to habits. Existing treatments such as cognitive behaviour therapy and cognitive remediation therapy include aspects which may target cognitive flexibility. In cognitive behaviour therapy, patients are encouraged to think more adaptively and flexibly (Beck & Dozois, 2011), and cognitive remediation therapy is a cognitive training program that aims to improve cognitive and behavioural domains through the completion of cognitive exercises and self-reflective practices (Baldock & Tchanturia, 2007; Tchanturia et al., 2013). Cognitive remediation therapy is typically completed before the commencement of traditional AN therapies, however, the current findings indicate that people continue to experience poor subjective cognitive flexibility after weight-restoration. Hence, targeted subjective cognitive flexibility interventions may be necessary throughout the recovery process. It would be valuable for future research to examine if subjective cognitive flexibility improvements during therapy and recovery contribute to a reduction in AN symptoms.

Conclusion

To conclude, this study found that there are significant differences between people with AN and HCs in perceived cognitive flexibility, clinical perfectionism, and ED-specific rumination. The findings suggest that perceived cognitive inflexibility and clinical perfectionism may be state-specific factors that improve following weight-restoration; however, further longitudinal research is required to support this claim and investigate further potential improvements following full recovery. Further, subjective cognitive flexibility was found to have a direct and indirect impact on ED symptoms, suggesting that poor cognitive flexibility may contribute to AN symptomology.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

Data is not shared due to ethical restrictions.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

S.M. was the recipient of an Australian Government Research Training Program Scholarship. A.P. is supported by the National Health and Medical Research Council Project Grant (CIA; GNT1159953). These funding bodies played no role in the study design, analysis or interpretation of the data, writing of the manuscript, or the decision to submit the paper for publication.