Levels of autistic traits in anorexia nervosa: a comparative psychometric study

There is a growing consensus that a better characterization of social and cognitive impairments in anorexia nervosa (AN), which overlap with those found in Autism Spectrum Disorders (ASD), may help to develop better approaches tailored to enhance specific areas of social functioning [1, 2].

From a clinical perspective, a well-known Swedish prospective study on teenage-onset AN suggests that autistic symptomatology is over-represented in AN, that it is associated with poor outcome, and persists long-term after recovery [3, 4]. However, using well validated clinical assessment tools to assess autistic symptoms, a more recent study comparing children and adolescents with an early-onset eating disorder (EOED), others with ASD and typically developing controls, failed to find an increased prevalence of ASD among the EOED, but reported clinically significant levels of autistic traits (in particular repetitive and stereotyped behaviours) [5]. Regarding cognitive functioning, AN patients, often display lack of flexibility (poor set-shifting) and extreme attention to details (weak central coherence), two features commonly found in ASD. Interestingly, in AN patients, these cognitive difficulties, extend beyond the sole effects of starvation, and are often shared with family members [6‐10].

While atypical emotional and interpersonal functioning of individuals with ASD has been extensively studied, it is an emerging area of research in AN. Several recent studies showed that AN patients have higher scores on the Autism-spectrum Quotient (AQ) than healthy controls [4, 11] and that they have difficulties processing their own as well as others’ emotions [11‐18]. Yet the co-existence of autistic traits in AN is still little explored and the question of whether inflexibility and impairments in social cognitions in AN are ‘truly autistic’ remains unanswered [5], p.589. In particular, no previous study has directly compared self-reported social cognitive skills in AN and ASD adults.

Here, we report findings collected as part of an ongoing study comparing AN and ASD socio-affective functioning (DETENDOEMO; RGB:2007-A01068-45). As recommended [5], we used a dimensional approach to explore the extent to which AN and ASD patients present dissimilar scores on questionnaires classically used to measure ASD impairments. In addition, the level of symptoms and concerns characteristic of eating disorders and the level of current dysphoric affects were measured. We expected to replicate previous studies [4, 11, 17] suggesting that, relative to healthy controls, patients with AN have higher scores on questionnaires measuring autism-spectrum traits and alexithymia. Moreover, given the aforementioned socio-cognitive difficulties observed in AN, we predicted that the AN group would have lower scores on the empathy quotient and higher scores on the systemizing quotient questionnaires than their control group (despite previous negative results [11]). Conversely, relative to the ASD group, we expected the AN group to have similar scores on the autism-spectrum quotient and the alexithymia questionnaires, but higher scores on the empathy quotient and lower scores on the systemizing quotient questionnaires.

All participants gave their written informed consent. The study was approved by the Ethics Committee, Paris Ile de France VI.

Mean scores (SD, range) on each test measure (all complete cases) are presented in Table 1. The distributions of individual scores for each group according to gender are displayed in Figure 1. Between-group comparisons for all total scale scores are shown in Table 2.

This study explored the level of cognitive and affective autistic traits in AN relative to both healthy controls and ASD patients. We found few arguments in favour of the suggestion that AN might be considered as a female variant of Asperger’s syndrome [32]. We observed between-group differences on some but not all the dimensions.

Regarding AQ categories, we found that a third of our AN sample shared certain endophenotypical markers of a broader/medium autism phenotype. In line with the conceptualization that autistic traits exist along a continuum [33], we used a dimensional approach and observed that both clinical groups reported a marked need for sameness: for the AQ Attention switching scores, the AN scored mid-way between the ASD and the controls. High scores on this AQ scale correspond to the neuropsychological performances that have been described in AN [9, 10] and are in favour of considering cognitive remediation therapy for AN [34] as a promising mode of treatment. With respect to the other AQ scores, we observed clear difference between the two clinical groups. This pattern suggests no significant co-occurrence of any of the core features of ASD in AN.

In the study by Hambrook et al. on 22 AN and 45 controls, the AN had significantly higher AQ total and subscale scores for the AQ Social skills, Attention switching and Imagination scores [11]. The authors concluded this pattern of results suggests that patients with AN experience difficulties similar to those with ASD. In the present study, participants with AN were also found to have significantly higher AQ total and Attention switching scores, but the effect was smaller for AQ Communication and Social skills scores, and no differences were found for the AQ Attention to details and Imagination scores. Such discrepancies may be due to differences in sample size, and/or to variations in patients’ characteristics (i.e., greater age of illness onset and shorter illness duration in our AN sample). Here, by directly comparing AN and ASD patients, we were able to highlight that the difficulties are not as severe in both groups, and that the AN score midway between the ASD and the controls on some autistic traits.

Moreover, the AN and ASD groups differed on two important psychological dimensions within the Empathizing-Systemizing imbalance theory of autism (E-S theory: [35]). According to this model, individuals with ASD display poor Empathizing skills (inability to identify another person’s emotions and thoughts, and respond to these with an appropriate emotion) but high Systemizing skills (strong propensity to analyze, explore and construct a system, which provides a way of understanding and predicting non agent behaviours). Baron-Cohen et al. [36] argued that strong Systemizing and excellent attention to detail are related, and stated that the former relies on the latter. They further explained that strong systemizing might account for some of the non-social deficits found in autism, including repetitive behaviour, narrow repertoire of interests or resistance to change. Indeed, in order to systematically investigate the workings of a given phenomenon, it is best to only change its constituent features one at a time while keeping all else constant, and once a possible causal chain is identified, to verify the rule by repeating the sequence and observing the outcome [36], p.1378. Therefore, on the basis of this conceptualization and of the literature on the AN patients’ cognitive profile, we expected AN and ASD to have similar levels of Systemizing (SQ-short scores) in addition to attention to detail and need for sameness (AQ Details and Switching scores). However, our data seem to rule out the hypothesis that AN and ASD have the same Empathizing-Systemizing pattern.

Contrary to our prediction, but in line with the Hambrook et al. study [11], we found that adults with AN had similar Empathizing and Systemizing scores to their control group. In our study, the ASD group displayed low EQ-short but high SQ-short scores (the expected e-S pattern, typical of the ‘male brain type’), whereas the AN group reported a marked propensity to care for others and an ability to feel concerned in interpersonal situations (high EQ-short scores) but low Systemizing skills (an E-s pattern), which is considered the typical ‘female brain type’ in this model E-s pattern [27, 35, 37].

In the present psychometric study, in line with the idea that empathy is multidimensional, we included the IRI in addition to the EQ-short. The IRI includes two subscales that explore cognitive processes (Perspective Taking, Fantasy) and two measuring affective components (Personal Distress, Empathic Concern). Empathic Concern scores, which reflect the ability to experience appropriate emotions like compassion, tenderness or sadness for others, were significantly higher among the AN group than the ASD group. Nevertheless, we also observed that the AN group resembled the ASD group more than the controls on two other dimensions that are relevant for empathizing: the tendency to feel anxious when confronted to negative situations (example of an IRI Personal Distress item : “Being in a tense emotional situation scares me”) and the ability to identify with fictional characters (example of an IRI Fantasy item : “When I watch a good movie, I can very easily put myself in the place of a leading character”, negatively keyed). This is an important issue, as previous studies using experimental paradigms (e.g. facial or vocal expressions of emotions, pictures demonstrating basic emotions, film clips of social scenes, emotional melodies) revealed that individuals with AN have deficits in emotional awareness and recognition, and in understanding others’ emotional states [13]. As recently discussed [38‐40], future studies are needed for better characterizing the pathways (e.g. maladaptive attachment, anhedonia, lack of embodiment) that may explain the relationships between interpersonal or social adjustment difficulties and disordered eating.

As early as the 1960s, Hilde Bruch highlighted the difficulties experienced by anorexic patients in perceiving or controlling their bodily sensations, and confusion in the identification of their mental states [41, 42]. These deficits in interoception and emotional introspection in AN have since been explored further via studies of alexithymia, a multifaceted personality construct that has emerged in line with the idea that deficits in the ability to experience and symbolise emotions can have adverse effects on well-being [30]. Bagby and Taylor stated that alexithymic characteristics “reflect deficits both in the cognitive-experiential domain of emotion response system and at the level of interpersonal regulation of emotion. [. . .] Lacking knowledge of their own emotional experiences, alexithymic individuals cannot readily imagine themselves in another person’s situation and are consequently unempathetic and ineffective in modulating the emotional states of others”[43], p.30. Whereas alexithymia has been well documented in AN (see [13]), this is a current area of research in the autism literature. Interestingly, several recent experimental studies with ASD participants provided arguments for considering alexithymia as a potential part of the Broader Autism Phenotype. For instance, alexithymia has been identified in ASD family members [44, 45], and in ASD alexithymia is associated with higher AQ scores, as well as with lower empathy scores, poorer facial expression decoding performances and decreased attention resource allocation to relevant facial areas [46‐48]. Interestingly, the present study revealed no difference between the AN group and the ASD group on alexithymia scores (whereas the two clinical groups were significantly different from their respective control group), which might suggest that alexithymia accounts partly for impaired social skills in these two conditions. Yet, it appears that high alexithymia scores are not always associated with low empathy scores, as exemplified in the present sample of AN participants. This idea is further supported by the distribution of group scores across the first dimension of the PCA, where controls (AN-C and ASD-C) are located at one end of the axis and ASD at the other.

In order to provide a clearer picture of the potential overlap between socio-affective profiles in AN and ASD, and especially in order to determine whether there is a continuum between ASD, AN and healthy individuals, larger samples should be investigated. In particular, in order to account for discrepancies between our results and Hambrook et al. [11], future studies using larger cohorts of patients might be able to carefully take into account differences in AN patients’ characteristics (i.e., younger age of illness onset and longer illness duration in their AN sample) or AN subtype (2/3 of binge/purging subtype in our study but not specified in Hambrook et al. sample; see [38] for a recent discussion on this issue). Another potential limitation of the present study stems from the absence of IQ data collection in the AN group [49]. There is a need to address this issue in future studies. Moreover, our study relied on self-report measures only, so comparing AN and ASD behavioral performances on well-validated tasks as well as in terms of their neural activity while performing these tasks would be a useful next step (e.g. see [50]). In addition, the present study leaves an important issue unanswered: what could account for the fact AN and ASD are gender-related? Indeed, the female-to-male incidence ratio is opposite in AN (i.e. large majority of females) and ASD (i.e. large majority of males), as was the case in this data set. Gender differences in socio-affective skills have been well documented, and this argument underpins Baron-Cohen and colleagues’ ‘extreme male brain’ theory of autism. Emerging models for gender-biased neuropsychiatric disorders pointing to complex interactions between prenatal stress and gonadal hormones early in gestation [51], or the implication of testosterone and oxcytocin in human social and emotional behaviours [52, 53], might prove to be very informative in furthering our understanding of the biological underpinnings of these gender differences. Although these issues were beyond the scope of the present project, it is interesting to note that in the present study we observed that for some measures, the AN resembled the male participants more (e.g. the ASD and the ASD-C alexithymia and IRI Fantasy scores), but also that participants with ASD resembled the female participants more on other dimensions (e.g. the AN and AN-C AQ Attention to details and IRI Personal Distress scores). Future studies, with a more balanced gender ratio across groups should be conducted to clarify this question.

Overall, our study adds to the literature pointing to specific links between disturbed eating, emotional regulation and problems in social interactions. Nevertheless, as in the recent study by Pooni et al. [5], our results suggest that difficulties in AN are only superficially similar to those found in ASD. Our mixed results encourage the exploration of transdiagnostic similarities between these disorders, in particular in longitudinal studies which could help to determine whether or not the pattern observed here in AN is a mere effect of the acute phase of AN.