Unravelling causal relations in eating disorder symptom networks of people with anorexia nervosa

We aim to investigate the causal structure of eating disorder symptoms in persons with anorexia nervosa using causal discovery analysis. Previous symptom network studies have consistently identified strong associations among core eating disorder symptoms, such as desiring weight loss, restraint, shape and weight preoccupation, shape overevaluation and fearing weight gain. However, these studies are limited to undirected associations and do not allow conclusions about potential causal relationships among symptoms. To address this limitation, the present study applies causal discovery methods to item-level Eating Disorder Examination–Questionnaire (EDE–Q) data from a large sample of inpatients with anorexia nervosa treated at the Schoen Clinic Roseneck (Prien am Chiemsee, Germany). The primary aim is to compare a literature-informed causal discovery approach, which incorporates prior knowledge from previously published cross-sectional Gaussian graphical models of EDE–Q-based symptom network studies, with a purely data-driven causal discovery approach without prior structural constraints. Causal discovery models will be estimated using training data and evaluated using held-out test data to assess generalizability, while model stability will be examined using nonparametric bootstrapping. By comparing literature-informed and data-driven causal models, this study aims to evaluate the extent to which previously reported symptom network structures are consistent with potential underlying data-generating causal models and to assess the robustness and generalizability of different causal discovery strategies. Overall, this project seeks to contribute to a more mechanistic understanding of symptom interactions in anorexia nervosa and to inform future research on symptom-focused and individualized intervention approaches.