Autism Spectrum Disorder and Anorexia Nervosa. Is There a Link? An Italian Experience


 Background: Potential overlaps exist between psychopathological features of Anorexia Nervosa (AN) and Autism Spectrum Disorder (ASD). The impact of malnutrition on autistic traits in patients with AN should be considered. This study aims to investigate possible associations between Eating Disorders (ED) psychopathology, ASD traits and BMI in a group of young patients with AN, using the EDI-3 (Eating Disorder Inventory-3) test and gold-standard measures for ASD.Methods: Prospective study involving 23 inpatients admitted to an Italian Centre for paediatric ED. ASD traits and ED psychopathology were assessed administering ADOS-2 (Autism Diagnostic Observation Schedule-2), AQ (Autism Quotient) and EDI-3 tests. Both present and past autistic traits were investigated using different versions of AQ. Possible effect of BMI, Obsessive Compulsive Disorder (OCD) comorbidity and psychopharmacological treatments on results were evaluated with multiple linear regressions.Results: An ASD diagnosis was possible in 22% of patients. EDI-3 Interpersonal Problems (IPC), Affective Problems (APC), and Global Psychological Maladjustment (GPMC) subscales showed correlations with different ASD measures on ADOS-2 and AQ. Scores for ASD and their correlations with EDI-3 subscales were independent of BMI, age, OCD and psychopharmacological treatments.Conclusions: ASD traits presented high prevalence in a group of young inpatients with AN. These traits were significantly correlated to 3 specific EDI-3 subscales and independent of BMI. This is the first study to investigate the relationship between ASD traits as measured with gold-standard measures, EDI-3 scores, and BMI.


Introduction and aims
Eating Disorders (EDs) are pathologic conditions characterized by a persistent disturbance of eating or eating-related behaviour, resulting in altered consumption of food and signi cantly psychosocial functioning (APA, 2013). Anorexia nervosa (AN) is a severe ED, characterized by signi cantly low body weight, intense fear of gaining weight, and undue in uence of weight and shape on self-evaluation. AN has one of the highest prevalence among EDs: 1.4% for women and 0.2% for men (Galmiche et al., 2019). The incidence of AN requiring inpatient treatment in Italy for the age group 10-19 years is 22.8 per 100.000 women and 2.0 per 100.000 men (Gigantesco et al., 2010). The lifetime prevalence of psychiatric comorbidities is very high among individuals affected by AN, from 45-97% (Herpertz-Dahlmann, 2015). Among comorbid psychiatric conditions, Autism Spectrum Disorder (ASD) or autistic traits are frequently documented in patients with AN (Jagielska et al., 2018).
ASD is a pervasive neurodevelopmental disorder starting in infancy, de ned by di culties with reciprocal social interaction and repetitive, restricted behaviours and interests (APA, 2013). Ledford and Gast described prevalence rates of problematic feeding behaviours in ASD ranging from 46-89% (Ledford and Gast, 2006). Food selectivity represents the most predominant feeding problem, affecting approximately 70% of ASD patients. Feeding disorders as selective or scarce feeding may represent early warning signs of ASD and are frequently associated with delay or stagnation of development (Parmeggiani et al., 2019). Feeding di culties should be investigated and considered together with the analysis of anthropometric parameters in children with ASD (Parmeggiani and Pruccoli, in press). In 2013, Huke and colleagues conducted the rst systematic review and meta-analysis investigating the prevalence of ASD among patients with AN, reporting an average prevalence of 22.9% (Huke et al., 2013). More recent reviews have con rmed these ndings (Westwood et al., 2017;Nickel et al., 2019). A second meta-analysis (Westwood et al., 2016) reported signi cantly greater autistic traits in AN patients than in healthy controls, although the latter did not meet the cut-off criteria for a diagnosis of ASD. The impact of BMI values on scores for ASD was not reported. All the reviews taken into consideration seem to indicate a higher prevalence of autistic traits and ASD in AN patients than in the general population.
Many common aspects between the neuropsychological pro les and the emotional and social functioning in AN and ASD make the potential link between the two disorders a subject of interest (Westwood and Tchanturia, 2017). Regarding empathy, fantasy and imagination, a review reports very similar pro les between patients with ED and subjects with ASD (Kerr-Gaffney et al., 2019). Theory of mind, the ability to infer the mental states of others, has been shown to present comparable impairment in ASD and AN (Leppanen et al., 2018). On the other hand, different researchers stated that malnutrition in patients with AN can determine mental rigidity, obsessive behaviours, and ritualism very similar to ASD symptoms and that these characteristics may then improve with normal nutrition (Mandy et al., 2015). Clinicians and researchers should investigate whether autistic traits in AN patients are an epiphenomenon of acute metabolic modi cations or speci c neurodevelopmental characteristics.
Psychopathological traits are more compromised when EDs overlap with ASD, especially regarding disturbances of thinking, cognitive in exibility, ritualistic behaviour, mood disorders, phobic aspects, and social impairment. These aspects do not seem to be correlated with the severity of the malnutrition or the BMI ( The aim of the present study was to investigate the relationship between ASD traits, ED psychopathology and BMI in a population of young inpatients with AN at an Italian Regional Centre for Feeding and Eating Disorders.

Materials And Methods
Twenty-three inpatients were enrolled in the study. Inclusion criteria were: a primary diagnosis of AN according to DSM-5 criteria; an age between 11 and 17 years at the moment of hospitalisation; the presence of one or more symptoms suggestive for ASD such as: obsessive-compulsive traits, social withdrawal, scarce emotional reciprocity, shared attention de cit, narrow interests, ritualistic and stereotyped behaviour, verbal communication impairment. Patients unable to attend a thorough clinical evaluation including all selected standardised measures were excluded from the study. Diagnostic procedures for AN and autistic traits were carried out by clinical psychologists and child neuropsychiatrists belonging to the same Centre where the study was conducted. and imagination. This study employed two different versions of the AQ for each patient: one was used to assess present symptoms and behaviour (AQ 12-15 years, which is a parent-report test or AQ 16 + years, which is a self-report test); the other was used to investigate ASD symptoms and behaviour in childhood (AQ 4-11 years, a parent-report test). This choice enables us to detect the possible presence of ASD features immediately after their typical onset age and before the appearance of an overlaying AN symptomatology.

Statistical analysis
All statistical analyses were conducted using JASP, version 0.13.1 for Windows. Adopted alpha error rate was 0.05 (two-tailed), with conservative statistical power of 95%. Descriptive statistics for demographic and clinical variables were calculated for the whole sample, considering means and standard deviations. Descriptive analyses and comparisons for total scores and subscales of ADOS-2, AQ and EDI-3 were performed. Pearson's or Spearman's rho correlations were calculated for selected clinical variables (BMI, EDI-3 scores, AQ-scores, ADOS-2 scores). Shapiro-Wilk and Levene tests were used to assess normality of data distribution and homogeneity of variance. Multiple linear regressions controlling for age, BMI and OCD comorbidity were conducted to investigate signi cant correlations between EDI-3 scores and ASD measures.

Ethical considerations
The study was approved by the local Ethical Committee (protocol code: ASD-AN-18). Written informed consent was obtained.
The AQ 4-11 years test was administered to the parents of the patients in order to collect data on the symptoms and behaviours that occurred in the patients' childhood. The mean total score in our sample was 54, 8

ADOS-2 and AQ
Five patients out of 23 (22%) had a diagnosis of "Autism" or "Autism Spectrum" at the ADOS-2. Only one of these patients had corresponding high scores at both AQ questionnaires. Two patients scored above threshold at ADOS-2 and AQ past.

Discussion
This is the rst study comparing ASD traits, which were evaluated with gold-standard measures (ADOS-2 and AQ tests), to EDI-3 scores and BMI, in a sample of young patients hospitalized for AN. The investigation of ASD traits in childhood with the employment of the AQ "past version" represents a second new addition to the literature on the subject. Along with the quantitative measurement, multidimensional clinical evaluation of the patient was always taken into account since our Centre specializes in the diagnosis and treatment of both ASD and AN.
Five patients, of the 23 assessed (22%), obtained scores in the ADOS-2 compatible with a diagnosis of ASD. On the whole, 12 patients out of 23 (52%) scored above clinical thresholds for ASD in at least one test (ADOS-2, AQ childhood and adolescence). Notably, the patients, whom we analysed for this study, were considered on account of the presence of autistic features. The rationale for this choice was to explore and con rm possible criteria to select AN patients to be screened for ASD in clinical practice.
Our ndings on the prevalence of ASD symptoms investigated with ADOS-2 in patients with EDs are consistent with most of the previous research. Particularly, a recent systematic review found a 26.5% prevalence of ASD in patients with ED (Nickel et al., 2019). A meta-analysis addressing the prevalence of ASD symptomatology among patients with AN found a wide spectrum of results, ranging from 4 to 52% (Westwood and Tchanturia, 2017).
Our nding of 26% patients scoring above thresholds for ASD in the AQ adolescence version is also consistent with previous research. Speci cally, the mean score obtained in our population (21.9) con rms the results reported by a previous systematic review (Westwood et al., 2016). We documented strong correlations between the two versions of the AQ test, namely, total scores, as well as 4 out of 6 pairs of subscales, showed signi cantly correlated results between the childhood and adolescence version of the test. This could indicate that ASD traits pre-existed the onset of the ED. In this regard, ADOS-2 and AQadolescence total scores and subtests showed no direct correlation with patients' BMI. These ndings suggest that autistic traits documented in adolescent patients with AN may not be due solely to the severity of their condition during clinical evaluation, which con rms and expands on existing evidence (Sedgewick et al., 2019).
Correlation measures between tests for ASD and indicators of EDs showed that scores obtained at speci c EDI-3 subscales, particularly IPC, APC and GMPC, were signi cantly correlated with AQ and ADOS-2 scores: this could suggest an association between ED psychopathology as assessed with EDI-3 and ASD traits. Multiple linear regressions adjusted for age, BMI and OCD comorbidity con rmed these ndings. These results corroborate and widen previous data (Dell 'Osso et al., 2018) for they show correlations between subthreshold ASD symptoms and selected EDI-2 subscales. In particular, the authors found Interpersonal Distrust EDI-2 subtest to be signi cantly correlated with total scores obtained at the Adult Autism Subthreshold Spectrum and with the Non-verbal communication and In exibility and adherence to routine domains. However, the ADOS-2 schedule was not adopted in this study. More generally, these results corroborate several previous studies, showing di culties in social interactions and low levels of social skills in patients with AN (Klump et al., 2000;Anckarsäter et al., 2011;Postorino et al., 2017).
The present study has a limited and non-homogeneous sample of patients with regard to gender, BMI, psychiatric drugs and interval of time passed from discharge to evaluation. The self-report and parentreport questionnaires employed in this study allowed us to include various perspectives in the evaluation of the patients; on clinical observation, however, some biased attitudes towards the tests were noted, both coming from patients and parents. Nonetheless, most patients and parents were observed to be objective and compliant, and the clinical perspective given by the ADOS-2 permitted a well-rounded investigation.
In conclusion our results, although consistent with previous studies, are unprecedented in literature and should be investigated in larger samples, given their likely importance for both clinical practice and research. They indicate the possibility that the EDI-3 --speci cally its subscales IPC, APC, and GMPC --be employed as a preliminary measure to identify and examine patients with AN for ASD traits with goldstandard measures.