Exploring autistic traits in anorexia: a clinical study

Anorexia nervosa (AN) is a potentially life-threatening illness associated with high levels of functional and social impairment [1], often with a chronic course [2, 3]. Autism spectrum conditions (ASC) and autistic traits are overrepresented in individuals with chronic AN [4, 5]; therefore, autistic traits may increase treatment resistance. Similarities between AN and ASC were first observed in the 1980s by Christopher Gillberg. He subsequently proposed that AN should be conceptualized as an empathy disorder on the same spectrum as autism [6]. Thirty years later, this hypothesis is gathering evidence, with reports of shared endophenotypes in AN and ASC [5, 7‐9], the suggestion that AN may be a female presentation of ASC [7] and work identifying atypical eating behaviours in ASC [10]. Gillberg’s hypothesis was informed by longitudinal studies indicating that ASC is overrepresented in the AN population [11]. In other samples, 16% of teenagers with AN had a premorbid diagnosis of ASC [12] and 23% of adults with AN met the clinical criteria for ASC [13].

Despite differences in presentation, such as a later age of onset in AN [14], a biased sex ratio towards females in AN vs. a biased sex ratio towards males in ASC [15, 16], and higher than average IQ in AN [17] vs. more heterogenous IQ levels in ASC [18], empirical studies from the last decade support an overlap in behavioural and cognitive features in AN and ASC. A brief review of relevant work is outlined next.

Deficits in social reciprocity represent part of the diagnostic criteria for ASC [19]. In AN, patients have difficulty recognizing emotional stimuli [14, 20] and expressing emotions [21]. In addition, theory of mind inefficiencies have been reported in AN [22] and bulimia nervosa [23]. Self-report measures reveal elevated levels of social anhedonia [24], reduced empathy [25] and impaired relationships and social leisure activities [1, 26] in AN. Considering these converging lines of evidence, findings in AN are likely to reflect the difficulties with empathy observed in ASC [27, 28], although it should be noted that nearly all psychopathologies are associated with some form of social difficulty and this is not unique to AN.

Similarities have also been noted in the non-social cognitive profiles of AN and ASC, including inefficiencies in set-shifting and global processing. In the autism literature, poor set-shifting, perseveration and repetitive responses in the context of changing environmental demands have been consistently documented [29, 30]. Studies investigating set-shifting with large samples of patients with AN have also reported problems in this domain [31, 32]; this is supported by a systematic review of cognitive flexibility studies in eating disorders [33]. The authors concluded that people with AN are less able both to adapt to changing rules and to employ strategies flexibly, relative to controls.

With regard to global processing, extensive research reported in the autism literature suggests that individuals with ASC have a superior eye for detail [34, 35]. Experimental research describing the same cognitive style in AN is summarized in a systematic review of the literature of AN [36] and in subsequent studies [37‐39]. It is important to note that despite similar behavioural outcomes for investigations of central coherence in the ASC and AN literature, recent fMRI studies demonstrated activation in different brain areas when comparing AN and ASC participants during an embedded figure task [40]. This is the first study of its kind and therefore requires further exploration.

Another paradigm employed to investigate commonalities in AN and ASC is the use of self-report questionnaires assessing characteristics associated with ASC, including superior ‘systemizing’ and below average ‘empathizing’ [41, 42]. Patients with AN scored significantly higher than controls on the autism spectrum quotient (AQ-10) [43], a measure of autistic traits in adults of average intelligence or above [44]. However, differences on measures of empathizing and systemizing were not observed. When replicated in a larger sample, the scores of participants with AN were significantly different to those of typical controls on all three measures [4].

In summation, the experimental and self-report evidence shows overlapping characteristics in the socio-emotional and non-social cognitive domains in people with AN or ASC. In both conditions, treatment success is limited and it may be that people with comorbid AN and ASC respond particularly poorly to standard treatment programmes for AN. Focusing on the similarities between the two may lead to better tailored interventions for both conditions. In this study we investigated the prevalence of autistic traits in a clinical population receiving treatment for AN in a large national eating disorder service in London. The main aims were to implement a brief screening measure (the AQ-10 [45]) for the first time in a population with AN and to explore associations between autistic traits and eating disorder symptoms, motivation to change, mood and work and social adjustment in a sample of AN and typical women. Based on the outlined literature, we predicted that individuals with AN would score higher on the AQ-10 than controls. No further predictions were made regarding associations between AQ-10 scores and other self-report measures, as this analysis was undertaken on an exploratory basis.

This study set out to investigate autistic traits in individuals with AN. Individuals with AN scored significantly higher on the AQ-10 than typical control participants. Furthermore, analysis of responses to individual items of this scale revealed significant differences for the majority of questions. This supports the hypothesis that women with anorexia possess a greater number of autistic traits than typical women and replicates the findings of previous studies [4, 44]. In addition, this study goes beyond the existing literature as the first study using the AQ-10 in a population with AN. Using the clinical cut-off recommended by the scale’s authors, more than a quarter of the AN group obtained a score high enough to warrant a referral to an ASC assessment service (referrals were made in these instances), closely reflecting the frequency of patients meeting full clinical criteria in an adult AN population [13]. In contrast, less than 2% of the control participants scored over the cut-off point.

Following on from these results, an exploratory analysis of how these elevated levels of autistic traits were related with other self-reported symptoms in the AN group revealed a strong correlation with mood, including both anxiety and depression, and the ability to maintain close relationships. Regression analyses showed a predictive role of the AQ-10 total score, indicating that higher levels of autistic traits are associated with feeling anxious and depressed and that this might explain difficulties in interpersonal interactions for people with AN [1]. It is interesting to note that significant correlations were not observed between autistic traits and eating disorder symptoms. It is important to consider whether the AQ-10 is in fact measuring mood symptoms rather than autistic traits, a situation that would threaten the validity of this instrument. Examining the individual items reveals that only certain questions may directly relate to emotional problems (for example, difficulty multitasking and coping with interruptions relates to poor concentration) with a greater number tapping into other cognitive or socio-emotional domains. This would suggest that symptoms of anxiety and depression might increase scores slightly; however, the range of questions appears to reflect typical autistic traits and therefore have good face validity. How can the significantly higher AQ-10 scores in individuals with AN be reconciled with a failure to demonstrate a direct link between autistic traits and disordered eating? It could be speculated that autistic traits exacerbate ‘maintaining factors’ for eating disorder, such as cognitive rigidity, low mood, low motivation or a lack of social skill, with the latter two factors relating to the significant correlations reported here, rather than directly causing the eating disorder behaviours. In a recent study, eating disorders (AN and bulimia nervosa), anxiety disorders and depression accounted for a large portion of difficulties with social functioning [53]. Although the outcome measure was somewhat different in this study, it suggests that eating symptoms, as well as mood, are expected to be associated with autistic traits falling within the social domain.

Nevertheless, the current study along with previous research indicates a significant trend: people with AN are more likely to show autistic traits. Several factors might play a part in this. For example, starvation has been shown to reduce the quality of interpersonal relationships, interest in other people, and libido [55]. Secondly, a detailed assessment of the items on the AQ-10 revealed that seven out of the ten items discriminated between patients with AN and control participants. These items tap into problems with ‘bigger picture’ (global) thinking (item 1, 2 and 7); inflexibility of thinking (item 3 and 4) and problems with social interactions (item 5 and 10).

In the adult literature, the presence of difficulties in these three domains in the AN population has received empirical and experimental support. For example, problems with abstract thinking are reported in a systematic review conducted by Lopez et al. [36] and have been shown to be experienced by AN patients in experimental studies [38, 39]. In addition, problems with flexibility of thinking are reported in the systematic review [33] and large data-based studies of Tchanturia et al. [31, 32]. It is also notable that social anhedonia [1] and problems with social cognition [5, 20] have been widely reported in the literature.

The items that did not significantly discriminate between patients with AN and controls appear to tap into theory of mind abilities, including difficulty reading faces and working out characters’ intentions. Despite the literature described (for example, [14, 20‐22]), this is in accordance with a psychometric study reporting that both AN and ASC groups demonstrated a need for sameness; however, other core features of autism, such as difficulty empathizing, were not seen in the AN group [56].

Individuals with AN scored higher than controls on the short version AQ-10, a measure that has not previously been used with this population. Levels of autistic traits correlated with self-reported mood and an ability to maintain close relationships; however, eating disorder symptoms were not associated with autistic traits. Further investigation of these findings would benefit from a comparison group of participants with an ASC diagnosis. The role of autistic traits in the presentation of eating disordered behaviour requires further investigation and has the potential to inform new intervention approaches based on joint working between ASC and eating disorder services.