Disentangling food addiction in anorexia nervosa: a latent class analysis and its clinical relevance

Background: The construct of Food Addiction (FA) has generated growing interest in the context of Anorexia Nervosa (AN), especially for its potential clinical implications. While AN is traditionally associated with restrictive eating patterns, recent findings suggest that FA symptoms may also be present in this population, complicating the clinical picture. This study aimed to explore FA symptomatology in individuals with AN using a person-centered approach and to identify psychological variables associated with different FA profiles. Methods: A sample of 202 individuals with a diagnosis of AN (including AN-R, AN-BP, and Atypical AN) completed the YFAS 2.0. A Latent Class Analysis (LCA) was performed to identify subgroups based on FA symptoms. Chi-square tests assessed associations between latent classes and AN subtypes. Binary logistic regressions explored psychological predictors of class membership, including eating psychopathology (EDE-Q), depressive symptoms (BDI-II), emotion dysregulation (DERS), and metacognitive abilities (MSAS). Results: LCA identified three latent classes: Class 1 (Addicted, 11.4%), characterized by high endorsement of FA symptoms; Class 2 (High-Risk, 33.7%), displaying moderate FA symptom severity; and Class 3 (Non-Addicted, 54.9%), showing minimal FA symptomatology. No significant association emerged between class membership and AN diagnostic subtype (χ²(4)=3.085, p=.544) and FA was present in 37.6% of the total sample, with comparable prevalence across AN-R (36.3%), AN-BP (44.0%), and atypical AN (38.1%) subgroups (χ²(2)=0.538, p=.764). Logistic regressions revealed that impulsivity significantly predicted membership in the Addicted class (OR=4.119, p<.05), while membership in the Non-Addicted class was associated with lower eating psychopathology, better metacognitive skills, and reduced impulsivity. Belonging to the High-Risk class was predicted by higher eating concern (OR=2.618, p<.05), lower metacognition (OR=.566, p<.05) and greater difficulties with emotional awareness (OR = 1.62; p<.01). Conclusions: These findings highlight the presence of distinct FA profiles within AN, independent of traditional diagnostic categories. The Addicted and High-Risk classes showed greater psychopathology and dysregulation. Notably, the High-Risk group also exhibited impaired metacognition, suggesting potential vulnerability despite moderate symptom expression. In contrast, the Non-Addicted class showed a more adaptive profile. LCA may support the development of personalized treatment strategies by targeting FA-related features, including interventions to improve emotion regulation and metacognitive functioning.