ESTRA-BED
Research type
Research Study
Full title
Multimodal longitudinal and machine learning-based predictive modelling to understand the development of eating disorders
IRAS ID
326571
Contact name
Zuo Zhang
Contact email
Sponsor organisation
King's College London
Duration of Study in the UK
1 years, 0 months, 1 days
Research summary
Eating disorders (EDs) are severe mental illnesses affecting up to 15% of young women and 3% of young men in high income countries. The causes of EDs are complex and involve many biological, psychological, and social factors. We are interested in understanding the connections between many factors (such as difficulties with attention) and disordered eating.
We will investigate causes and mechanisms of eating disorders. This will help us understand the basis of diagnostic classifications, which will promote early intervention and the identification of new areas to target in treatments.
We will analyse the data already collected in the STRATIFY study, including patients with Anorexia Nervosa (N=60), Bulimia Nervosa (N=49), Binge eating disorder (N=27) and healthy controls (N=69). We will also recruit 30 new participants with binge eating disorder using the original STRATIFY protocol to enlarge the binge eating disorder group, so that its sample size is comparable to the other groups.
We will use neuroimaging, cognitive, psychological and environmental data to assess if behavioural/neural processes are what differentiate one eating disorder from another and if there are similar processes across ED diagnoses.
We will use advanced statistical methods such as machine learning based models. We will use these to find the behavioural and neurological processed involved in EDs and will validate these findings using the IMAGEN dataset to test if we can predict future disease risk.
IMAGEN is a longitudinal population-based genetic and neuroimaging study of over 2,000 participants from adolescence to early adulthood.
Participants will fill online questionnaires, take an online clinical interview, and undergo a research visit, including one brain scan, blood and urine samples, and a range of social and behavioural measures. The study will end on 30/06/2024.
Results Summary
1. Machine learning-based models for predicting the diagnosis and risk of eating disordersWe built machine-learning-based models to predict the diagnosis of anorexia nervosa (AN) and bulimia nervosa (BN) by incorporating broad domains of data, including cognition, environment, personality, psychopathology, substance use. The models could accurately predict the diagnosis of AN and BN, even without using BMI information. These results demonstrate the capability of machine learning methods to accurately predict mental health diagnoses by using multi-domain psychosocial data. The classification model for AN may help eliminate the reliance of healthcare professionals on BMI for AN diagnosis, which has been decried for delaying diagnosis and getting in the way of early intervention.
The models demonstrated high transdiagnostic potential, as those trained for AN and BN were also accurate in classifying alcohol use disorder (AUD) and major depressive disorder (MDD) from healthy people, and vice versa. Shared predictors, such as neuroticism, hopelessness, and symptoms of attention-deficit/hyperactivity disorder, were identified as reliable classifiers for these four diagnoses. These results suggest shared mechanisms underlying these conditions and support the adoption of an integrative approach to their treatment, taking co-occurring conditions into account.
Furthermore, we built models utilising data collected at age 14 years to predict the development of future eating disorder (ED), MDD, and AUD symptoms in the coming 2–5 years. The models exhibited moderate performance in predicting the development of future symptoms of EDs, MDD and AUD. The most reliable predictors for future ED symptoms were being female, having a higher BMI, more advanced pubertal status, symptoms of depression, specific phobia, obsessive compulsive disorder, emotional symptoms, harmful drinking, and impulsivity. Particularly, impulsivity was a common reliable predictor of future symptom onset for all three disorder groups. Being female, more advanced pubertal status, and specific phobia symptoms were shared predictors of ED symptoms and depression.
Clinicians face significant challenges in treating youth psychopathology, as symptoms often manifest and progress differently in young people compared to adults. A lack of confidence, knowledge, and training are commonly cited by primary care practitioners as key barriers to effectively identifying and managing mental health issues in this population. Digital tools developed from the youth-focused models represent a promising avenue for addressing gaps in care. Such models can better capture the nuances of psychopathology in young people and potentially assist primary care practitioners in delivering early detection and intervention.
2. Multivariate personality profiles as precursors and correlates of eating disorders and comorbid mental health conditions
Eating disorders do not occur in isolation; they often co-occur with other mental health conditions, including depression, anxiety, and self-harm. Moving beyond previous studies that investigated each mental health condition individually, we adopted a comprehensive approach to examine how personality characteristics are associated with combinations of mental health symptoms. We identified two multivariate personality profiles assessed at age 14 years that were associated with combinations of mental health symptoms over the coming 2–5 years. The first profile combined neuroticism, impulsivity, and conscientiousness, which were associated with future anxiety symptoms and dieting behaviours. The second profile combined lower agreeableness and lower anxiety sensitivity, associated with future deliberate self-harm and purging behaviours. In the young-adult clinical sample, personality profiles comprising hopelessness and neuroticism were associated with concurrent depression, generalised anxiety, suicide risk in both AN and BN diagnoses. For BN, this profile also included impulsivity. Additionally, extraversion was found to be a protective factor that was linked to lower depressive risk in BN.
These findings deepen our understanding on the relationship between personalities and eating disorders. Neuroticism and impulsivity are both risk and diagnostic markers, while neuroticism and hopelessness are shared diagnostic markers for AN and BN. Specific to BN, impulsivity is a diagnostic marker, while extroversion serves as a protective factor. These findings may inform the design of more personalised prevention and intervention strategies. Prevention programmes targeting neuroticism and impulsivity may reduce the future risk of multiple mental health problems. Interventions targeting hopelessness and introversion, for example, by teaching healthy coping strategies and improving social skills, may improve clinical outcomes.
3. The interplay between neurobiology, genetics, and ED psychopathology
We found that disordered eating behaviours are associated with distinct developmental brain trajectories and identified genetic contributions underlying this relationship. Across ages 14 to 23 years, restrictive eaters and emotional/uncontrolled eaters exhibited smaller reductions in grey matter volume in the left cerebellum compared with the healthy eaters. The emotional/uncontrolled eaters also showed smaller reductions in several fronto-temporal regions and the cerebellum. Given that the grey matter volumes typically decrease during this age period, our results indicate that brain maturation may be delayed in unhealthy eaters. Delayed brain maturation also helped explain the association between genetic risk for high BMI and unhealthy eating behaviours at age 23. Our findings reveal a critical role of adolescent brain development in shaping eating habits in young adulthood, and suggest potential benefits of education targeting early dietary habits and maladaptive coping behaviours such as emotional eating, in order to prevent eating disorders and promote brain health.
To better understand how relationship between brain structure, brain function, and clinical symptoms in EDs, we analysed structural and functional neuroimaging data in a group of AN and BN patients. Reduced grey matter volumes were found in patients with AN compared to healthy volunteers, in the left lateral orbitofrontal cortex, which were linked to higher impulsivity. Lower cortical thickness was found in the left rostral middle frontal gyrus, which was associated with cognitive restraint in eating, suggesting that these regions may play key roles in ED psychopathology. Functional neuroimaging also revealed notable differences. During reward anticipation, patients with EDs exhibited deactivations in the cerebellum and right superior frontal gyrus, alongside reduced activation in the left lingual gyrus. These functional changes were associated with heightened neuroticism. Lower brain activations for reward anticipation and higher neuroticism helped explain the association between grey matter reductions and cognitive restraint in eating. The findings provide a more detailed picture of neural mechanisms of AN and BN, linking brain’s structural and functional differences to clinical symptoms and related personalities. The key brain regions we identified suggest potential brain-based targets for developing interventions.
4. New data collected
We have recruited 50 participants with a current binge-eating disorder, and collected multimodal data including questionnaires, psychological assessments, behavioural data under cognitive tasks. 41 out of 50 participants provided MRI data including structural and functional neuroimaging. 31 participants provided blood and urine samples.
The collected data have been pseudonymised and stored on a data server hosted by the STRATIFY and IMAGEN consortia (https://gbr01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fimagen2.cea.fr%2F&data=05%7C02%7Cgmsouth.rec%40hra.nhs.uk%7Cbce02d67c68a4de7d1c808ddcf8ded84%7C8e1f0acad87d4f20939e36243d574267%7C0%7C0%7C638894930694069204%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=K0pVUuqTDU1C%2BddlrqNGYn8R2nKUTQpl1BMGlzbjF2M%3D&reserved=0), with data sharing agreements in place. The blood and urine samples are stored in -80 degree freezers in the laboratory of the Social, Genetic & Developmental Psychiatry Centre, King’s College London.
The protocol for data collection matched those for the STRATIFY and IMAGEN studies. This ensured that the data collected for the participants with binge-eating disorders are comparable to the existing data in the STRATIFY and IMAGEN studies, including participants with anorexia nervosa, bulimia nervosa, and healthy controls. The study protocol has been drafted as a paper and submitted to the European Journal of Epidemiology for peer review.
The collected data are undergoing quality checks and pre-processing. I will use them to carry out cross-disorder comparisons to identify shared and distinct neural and psychological mechanisms underlying binge-eating disorder, anorexia nervosa, and bulimia nervosa.
Publications:
Zhang, Z., Robinson, L., Whelan, R., et al. (2024). Machine learning models for diagnosis and risk prediction in eating disorders, depression, and alcohol use disorder. Journal of Affective Disorders, 379, 889-899. https://gbr01.safelinks.protection.outlook.com/?url=https%3A%2F%2Ftrack.pstmrk.it%2F3ts%2Fdoi.org%252F10.1016%252Fj.jad.2024.12.053%2FNBTI%2FFQe-AQ%2FAQ%2F93c5e214-08b4-4de8-9a78-8fcd4a8e6ab0%2F2%2Ff9RWYXAOg2&data=05%7C02%7Cgmsouth.rec%40hra.nhs.uk%7Cbce02d67c68a4de7d1c808ddcf8ded84%7C8e1f0acad87d4f20939e36243d574267%7C0%7C0%7C638894930694085355%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=sOrpTgzdLgkgaw%2BzZo%2BarZ23d76Sfk6dw%2F7ivDv86FM%3D&reserved=0Zhang, Z., Robinson, L., Campbell, I., et al. (2024). Distinct personality profiles associated with disease risk and diagnostic status in eating disorders. Journal of Affective Disorders, 360, 146–155. https://gbr01.safelinks.protection.outlook.com/?url=https%3A%2F%2Ftrack.pstmrk.it%2F3ts%2Fdoi.org%252F10.1016%252Fj.jad.2024.05.132%2FNBTI%2FFQe-AQ%2FAQ%2F93c5e214-08b4-4de8-9a78-8fcd4a8e6ab0%2F3%2Fg_EzA5Dq7M&data=05%7C02%7Cgmsouth.rec%40hra.nhs.uk%7Cbce02d67c68a4de7d1c808ddcf8ded84%7C8e1f0acad87d4f20939e36243d574267%7C0%7C0%7C638894930694098994%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=xpp9tn%2FBo0mcfo51ucswoT42X9woJ4tLcdoNPiEym7g%3D&reserved=0
Yu, X., Zhang, Z., Herle, M., et al. (2025). Relationships of eating behaviors with psychopathology, brain maturation and genetic risk for obesity in an adolescent cohort study. Nature. Mental Health, 3(1), 58–70. https://gbr01.safelinks.protection.outlook.com/?url=https%3A%2F%2Ftrack.pstmrk.it%2F3ts%2Fdoi.org%252F10.1038%252Fs44220-024-00354-7%2FNBTI%2FFQe-AQ%2FAQ%2F93c5e214-08b4-4de8-9a78-8fcd4a8e6ab0%2F4%2F8N2yX1KjDj&data=05%7C02%7Cgmsouth.rec%40hra.nhs.uk%7Cbce02d67c68a4de7d1c808ddcf8ded84%7C8e1f0acad87d4f20939e36243d574267%7C0%7C0%7C638894930694112315%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=WuD9OFewRhJ4yp3DSkoc1clT4sq0t3SZFyV28KWu4GE%3D&reserved=0
Yu, X., Robinson, L., Bobou, M., Zhang, Z., et al. (2024). Multimodal Investigations of Structural and Functional Brain Alterations in Anorexia and Bulimia Nervosa and Their Relationships to Psychopathology. Biological Psychiatry, S0006-3223(24)01759-1. https://gbr01.safelinks.protection.outlook.com/?url=https%3A%2F%2Ftrack.pstmrk.it%2F3ts%2Fdoi.org%252F10.1016%252Fj.biopsych.2024.11.008%2FNBTI%2FFQe-AQ%2FAQ%2F93c5e214-08b4-4de8-9a78-8fcd4a8e6ab0%2F5%2FrVeBKe12t-&data=05%7C02%7Cgmsouth.rec%40hra.nhs.uk%7Cbce02d67c68a4de7d1c808ddcf8ded84%7C8e1f0acad87d4f20939e36243d574267%7C0%7C0%7C638894930694125561%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=En8CVYmfuHVeyxD5fqviESr%2Bm8Rna7pPZ1r44GEWWJ0%3D&reserved=0
REC name
North West - Greater Manchester South Research Ethics Committee
REC reference
23/NW/0232
Date of REC Opinion
19 Oct 2023
REC opinion
Further Information Favourable Opinion