Background
Anorexia nervosa (AN) is a severe eating disorder (ED) characterised by low body weight, restriction of energy intake relative to requirements and undue influence of weight and shape on self-evaluation [
1]. The disorder tends to be diagnosed during adolescence [
2,
3] and affects significantly more females than males [
4,
5]. In contrast, autism spectrum disorder (ASD) is a developmental disorder defined by difficulties with social interaction, communication and repetitive and restricted behaviours (RRBs) [
1]. ASD affects approximately 1% of the population, 0.2% of whom are female [
6]. It is considered a heterogenic, dimensional disorder with the identification of subgroups, based on the presence of co-occurring psychiatric, medical and/or genetic risk of continuing importance to research and clinical practice [
7].
While both ASD and AN are rare disorders in the general population, ASD appears to be over-represented within ED populations, with a systematic review reporting a mean prevalence of 22.9% [
8]. Much of the research on ASD in AN has come from a Swedish cohort who was assessed for ASD at regular intervals over a 16-year period [
9]. In each of the four assessments, the diagnostic tools used to assess ASD differed, yielding differing rates of ASD diagnosis. This underlines the difficulty of assessing for ASD in AN, particularly in light of evolving diagnostic criteria and the presence of a distinct female autism phenotype [
10], which makes accurately assessing the disorder in females challenging [
11]. There is also a possibility that the acute, starved state associated with AN may exacerbate the presence of symptoms characteristic of ASD [
12]. This state versus trait argument [
12] has led to calls for further research to disentangle the complex relationship between AN and ASD.
One of the criticisms of research examining the presence of ASD in AN is the use of inconsistent screening tools and varying diagnostic criteria [
8,
13], instead of ‘gold-standard’ clinical assessment. More recently, studies in younger people, utilising parental-report methods [
14,
15], have reported the prevalence rate of ASD to be substantially lower than in adult studies. While this is consistent with the idea that AN may exacerbate the presence of ASD symptoms, diagnostic tools which rely on parental report rather than direct observations of autistic behaviour may result in under-reporting of ASD, particularly in a predominantly female sample, in which ASD may not be recognised.
Females with ASD display less RRBs than their male counterparts and may be able to mask their social difficulties, leading to under-recognition or missed diagnoses [
16,
17]. The high likelihood that ASD will remain undetected in females [
18] can cause stress and anxiety and may lead to the high proportion of females with ASD whom experience co-morbid mental health difficulties [
17,
19]. This applies not only to ASD but also to elevated symptoms of autism, which extend throughout the population [
20]. Thus, the presence of ASD symptoms may act as a risk factor for a range of emotional and behaviour issues [
21], including eating problems. It stands to reason that females with elevated ASD symptoms may be vulnerable to the development of secondary mental health problems and are thus likely to be over-represented in ED populations. Alternatively, the gender bias in diagnosis may result in females with ASD being mislabelled as having an ED, particularly as features of AN such as extreme rigidity in eating or obsession with exercise may mask the presence of ASD [
13,
22].
Despite ASD presenting differently in females, gender-specific diagnostic tools and criteria are yet to be widely implemented [
23]. As such, the diagnosis of adults with ASD relies on standardised assessment tools including the Autism Diagnostic Observation Schedule, 2nd Edition (ADOS-2; [
24]) which is yet to be tested with women with AN. In a small-scale feasibility study [
13], the ADOS-2 was used with a pre-selected sample of women with AN who presented with social and flexibility difficulties. Of ten women assessed, half received an ADOS-2 ASD classification. This study demonstrated the feasibility of using the ADOS-2 with women with AN. However, as the study used a small sample, pre-selected for having a suspected ASD, the high prevalence of observed autistic difficulties cannot be generalised to the wider AN population.
The ADOS-2 has been criticised for not showing adequate specificity and sensitivity to discriminate adults with comorbidities and no previous diagnosis of ASD [
16,
25], leading to the use of alternative measures such as the Ritvo Autism Asperger Diagnostic Scale, Revised (RAADS-R; [
26]), with people with EDs [
27]. Using the RAADS-R, 33% of participants were classified as having elevated ASD traits. However, the RAADS-R relies on self-report whereas the recommended diagnosis of ASD in adults should incorporate both self-report and observational measures [
28]. While observational measures should not be used in isolation and may not be necessary for a diagnosis of ASD to be given [
29], the ADOS-2 is nevertheless a widely used measure of ASD symptoms. Studies utilising this tool in AN populations will therefore provide a useful comparison to other studies, such as those utilising self-report measures to assess for ASD traits [
30].
This challenge of diagnosing ASD in AN is exacerbated by the high level of comorbidities in EDs, which may mediate the relationship between ASD and AN. Up to 97% of individuals hospitalised for AN are thought to have at least one psychiatric comorbidity [
31], with particularly high rates of depression [
32] and anxiety disorders including obsessive-compulsive disorder (OCD) [
33,
34]. The presence of such comorbidities along with the possibility that starvation exacerbates ASD symptoms [
35] confounds the ability to differentiate between these disorders.
Aims
The primary aim of the current study was to examine the presence of autistic symptoms, as measured via direct observation using the ADOS-2, in women hospitalised for AN. In addition, the study aimed to explore group differences in symptoms of eating pathology as well as other psychiatric symptoms of depression, anxiety, alexithymia and OCD between individuals with high, sub-clinical and no symptoms of ASD.
Discussion
This study is the first to examine the presence of observable ASD symptoms in a cross-sectional sample of adult women hospitalised for AN, using the ADOS-2. Of the 60 women assessed, 14 (23.3%) scored above clinical cutoff on the revised algorithm, indicating the presence of clinical symptoms associated with ASD. The finding is in line with previous prevalence estimates in adults with EDs, in particular from Huke et al.’s [
8] systematic review, which reported an estimated average prevalence rate of 22.9%. However, the rate is higher than in studies with young people, which have suggested only a slightly increased prevalence of 4% [
14,
15]. As these studies did not use observational measures of ASD symptoms, the results are not directly comparable.
Using the revised diagnostic algorithm [
36], all 14 women who scored above clinical cutoff displayed difficulties with social affect, but notably, only eight displayed RRBs. While RRBs are not included in the ADOS-2 diagnostic algorithm, they form part of the diagnostic criteria for ASD [
1]. This level of RRBs is also lower than in adults with ASD without AN, reported in one previously published study using the revised algorithm [
37]. The most common RRB displayed in the HAS group was speech abnormalities associated with ASD which includes unusual intonation, volume, rhythm or rate. There was no evidence of sensory interests or hand mannerisms in any participant.
Females with ASD have been found to display less RRBs than males [
10], and RRBs are less common in adults compared with younger individuals [
47]. Therefore, it may be that the women who scored above cutoff represent the typical, under-recognised profile of ASD often seen in high-functioning females without intellectual disability [
48], particularly as AN is characterised by above average intelligence (for systematic review: [
49]). In addition, it could be that the ADOS-2 was not sensitive enough to detect RRBs in females with AN. RRBs are notoriously hard to measure using direct observation, which is why they are not included in the ADOS-2 diagnostic algorithm [
24]. Therefore, they may have been present in the participants in this study but were not detected during assessment.
There is also a possibility that the RRBs displayed in participants in this study are not manifestations of ASD but are instead associated with other psychiatric symptoms including depression or anxiety. This could account for the flat intonation displayed by several participants and explain why other stereotypies associated with ASD, including unusual sensory interests or complex hand mannerisms, were not observed in any participants. In fact, the observed difficulties with social affect could also occur within the context of anxiety and depressive symptoms and are thus not necessarily indicative of ASD. Delineating differential diagnoses, overlapping symptoms and true comorbidities remains a key clinical challenge [
22], not only in the field of AN but also in the accurate diagnosis of ASD in all clinical populations.
Despite these findings supporting the idea of an overlap between AN and ASD, the aetiology of these symptoms remains unclear. Observational measures such as the ADOS-2 cannot determine the extent to which ASD is truly over-represented in AN or whether the observed symptoms, at least in some cases, represent epiphenomena, secondary to the ED. In this study, elevated ASD symptoms were not related to ED psychopathology, BMI or illness duration. These findings appear to be contrary to the suggestion of ASD symptoms arising from the starved state of AN [
12] as if this was the case, one would predict a linear association between ASD symptoms and ED symptoms. However, other symptoms of AN including lowest recorded BMI or the presence of amenorrhea were not examined in this study, and future research would benefit from examining the association between ED severity and ASD more closely, using a longitudinal design.
This study also suggests that in AN, the presence of elevated ASD symptoms is associated with the presence of alexithymia and obsessive-compulsive symptoms. While it is possible that symptoms such as these may mediate the relationship between AN and ASD, causing individuals to appear autistic as a secondary effect of other symptoms, it is also possible that the presence of ASD symptoms in females may indirectly cause additional mental health problems for a range of reasons, including the stress and anxiety caused by attempting to mask social difficulties [
22].
While disorders such as OCD are common in both AN and ASD separately [
31,
50], it may also be that the presence of both AN and elevated ASD symptoms increases the likelihood of other co-occurring psychiatric symptoms. Individuals with both diagnosed ASD and OCD score higher on the OCI-R than those with ASD alone [
51], and as these disorders are not mutually exclusive, it may be that this pattern is the same in individuals with AN. Additionally, both AN and ASD are associated with increased alexithymia [
52], with its presence contributing to poorer treatment outcome in AN [
53] and therefore the association between ASD symptoms and alexithymia in AN is worthy of consideration.
Psychiatric symptoms such as alexithymia may also compound the social difficulties and isolation of individuals with AN [
54], thus causing them to score highly on the ADOS-2, without having a ‘true’ ASD. While research on the use of the ADOS-2 in complex psychiatric groups is limited, overlap in symptoms between disorders may lead to false-positive results using its current diagnostic algorithm. A study using the ADOS and the Autism Diagnostic Interview, Revised (ADI-R; [
55]), in children with psychosis found that participants met criteria for ASD on this measure despite not receiving a clinical diagnosis from clinicians [
56]. Standard diagnostic assessment packages, such as the ADOS/ADI-R, may therefore have limited use, particularly for clinically complex groups, highlighting the need for robust clinical assessment.
The cross-sectional design of the current study and the small number of participants who scored above clinical cutoff for ASD limit analysis to the statistical associations between variables, preventing examination of predictor variables or causal relationships to be made. This is the main limitation of this study, and the underlying aetiology of ASD symptoms in AN remains inconclusive. In addition, the assessment of ASD was limited to a single observation of symptoms. While the ADOS-2 is a highly specific and sensitive assessment tool, clinical diagnosis of ASD is multi-model, utilising direct observation, self-report, developmental history and occupational information. Future longitudinal studies using both the ADOS-2 and a standardised parental report measure are needed to confirm the precise nature of the relationship between ASD and AN. As all participants in this study were recruited from inpatient or day-patient services, they are likely to have more severe or complex eating disorder pathology. Therefore, the findings may not translate to other patient groups, such as individuals receiving less intensive treatment. It is possible that this specific patient group experience higher levels of comorbidity, and thus, future studies examining ASD symptoms across the spectrum of EDs are warranted.
Acknowledgements
The authors would like to thank the IoPPN/MRC Excellence Studentship, the Psychiatry Research Trust and the Swiss Anorexia Foundation for their support.