Neurobiological underpinning of eating disorders: integrative biopsychosocial longitudinal analyses in adolescents

The eating disorders (EDs) [anorexia nervosa (AN), bulimia nervosa (BN), binge eating disorder (BED) and related syndromes] are common psychiatric disorders that affect up to 15% of young women and up to 4% of young men in developed countries. The cause of these disorders is complex and their development influenced by environmental, psychological and biological factors that may interact to create a risk for different clinical presentations of the disorder. While known risk factors include sociocultural influences (e.g. media exposure, idealisation of thinness) and personality factors, a range of other factors are either unknown or can currently only be considered as correlates. There is also the increasingly broad acceptance of EDs as being brain-based disorders sharing neurobiological overlaps with anxiety disorders and addictions. Identifying true risk factors for EDs and understanding how they contribute to the development of the specific aspects of the disorders, including behaviours related to reward and punishment, cognitive control and emotional processes, will be crucial for improving prevention and treatment of EDs.

To investigate this and understand how behaviours of dysfunctional eating develop, we will use a rich database collected a large population-based, longitudinal cohort of adolescents of the IMAGEN study. With over n=2000 participants recruited from four European countries, including the UK, Ireland, Germany and France, and followed-up at ages 14, 16, 19 and 23, IMAGEN is the largest and most comprehensively characterized longitudinal gene x neuroimaging cohort. The adolescents of this study underwent neuroimaging assessments and extensive assessments to monitor health and wellbeing, including personality measures, social determinants, and pre- and post-natal life-events questionnaires, neuropsychological and biological measures. There is also an extensive characterisation of behavioural measures related to ED symptoms, internalising and externalising behaviours and substance use. To demonstrate that any differences observed in the IMAGEN population can also apply to a clinical population, we will complement the IMAGEN sample with a sample of emerging adults, of comparable age to the most recent IMAGEN follow-up (i.e. ~ 23 years-old), diagnosed with a first episode ED (AN or BN/BED). Before undergoing treatment for their ED, those individuals, unmedicated at the time of recruitment, will be assessed in a similar way as the IMAGEN participants. By combining these two samples we will be in a position to identify multiple correlates of dysfunctional eating across the full spectrum of drinking patterns. The analyses of the repeated measures of the longitudinal IMAGEN dataset, will allow us to better establish which correlates have a causal role in the development of EDs (those occurring before appearance of eating disorder symptoms). They will also enable us to characterising possible protective and/or additional risk factors, both environmental and biological.

For this research, we have brought together unique resources, a leading expert in EDs and scientific leaders in neuroimaging, genetics and developmental psychopathology to form a multi-disciplinary team, which is excellently positioned to deliver the research necessary to understand the underlying mechanisms and heterogeneity of EDs and inform future prevention and treatment strategies.

This project aims to identify early biomarkers of eating disorders (EDs) by applying Big Data methods to the rich database derived from a representative population of adolescents and a clinical sample of emerging adults with an ED diagnosis.
First, we will characterise trajectories of ED symptoms in IMAGEN, the largest and most comprehensively characterised longitudinal gene x neuroimaging cohort n=2000 male and female adolescents followed-up at ages 14, 16, 19 and 23 years. We will classify individuals based on their ED symptoms trajectories, using growth mixture models or generalised estimating equations to model the longitudinal course of disordered eating behaviours, including core ED symptoms, a continuous global index of eating pathology and BMI changes. Next, we will use machine learning procedures based on cross-validated regularised logistic regression combining neuroimaging, genomic and psychometric data modalities to identify correlates of EDs and select the features that best classify individuals who endorse ED symptoms along with those who endorse no ED symptoms at any time point. To identify predictors of ED symptoms, we will use age 14 data to compare individuals who developed ED symptoms over time (symptoms absent at age 14 but present at later ages), or those who recovered (symptoms present at age 14, not at later ages) to those who never developed symptoms. Finally, we will select the most predictive and easily applicable components of our comprehensive profile for validation in a clinical sample. This will be patients, aged 18-25 with a 1st episode of DSM-5 AN or BN (n = 50 /diagnosis), assessed through standardised protocols in a way identical to IMAGEN.
This paradigmatic multimodal approach, in which neuroimaging may provide added value compared with the existing standard assessments, may yield potential application for early and differential ED diagnoses.

Adolescents who engage in persistently disturbed eating behaviours [from severe undereating/self-starvation to overeating/binge eating] and who initiate these behaviours early in life are at elevated risk for a range of negative outcomes that span emotional (anxiety and mood disorders), behavioural (self-harm, and substance use), social (poor peer, family and personal relationships), physical (death through malnutrition, obesity and related complications), educational (poor academic performance) and economic (reduced employability) impacts. Even sub-clinical syndromes are associated with negative health outcomes. Early diagnosis and prognosis of eating disorders (EDs) would be key to improving current treatments. The aetiology of EDs is however complex, with evidence for multiple overlapping and distinct biopsychosocial factors implicated in risk for different clinical EDs and/or subclinical disordered eating.
In this project we will for the first time comprehensively explore biopsychosocial risk factors and biomarkers for EDs. Our paradigmatic approach that uses a wide range of data modalities, in which neuroimaging may provide added value, compared with the existing standard assessments, may have application for early and differential ED diagnoses, making a significant contribution to prediction and re-classification of EDs. The identification of biomarkers for early diagnosis and prognosis based on brain networks will allow implementation of pre-existing intervention programmes at an earlier stage. It will also facilitate development of novel early intervention programmes, which target specific neurobehavioural phenotypes. Our longitudinal assessments of ED symptoms trajectories will establish what measures of brain or other features are associated with persistent disordered eating and provide information inform about reversibility of symptoms. The results of the project are thus preparing for future work identifying novel targets, as well as for the establishment of neurobehaviourally informed endpoints for interventions.