Mindful Eating, Disordered Eating, and Exercise Addiction: A Bayesian Gaussian Network Analysis

Nutrients. 2026 Aug 4;18(15):2524. doi: 10.3390/nu18152524.

ABSTRACT

Objectives: This study examined component-level relationships among mindful eating, disordered eating attitudes, and risk of exercise addiction (REA) in recreational exercisers. Using Bayesian Gaussian network analysis, the study aimed to identify conditional associations and central variables across these constructs. Methods: Data were collected from 303 regular exercisers in Türkiye aged 18-57 years (M = 24.42, SD = 7.10; 57.8% male, 42.2% female). Participants completed the Eating Attitudes Test-26 (EAT-26), the Mindful Eating Questionnaire (MEQ), and the Expanded Exercise Addiction Inventory-3 (EAI-3), with all subscales entered as nodes in a 12-node Bayesian Gaussian network. The estimated network contained 22 nonzero edges among 66 possible connections (sparsity = 0.667), indicating a moderately sparse structure. The robustness of the recovered structure was examined against a regularized frequentist counterpart and against models adjusting for age, sex, and exercise frequency/duration. Results: The strongest conditional associations occurred within rather than between instruments: EAT-26 Restriction (Diet/Calorie Control) and EAT-26 Bulimia and Food Preoccupation, EAI-3 Healthy Relevance and EAI-3 Addiction Tendency, MEQ Eating without Thinking and MEQ Emotional Eating, and EAT-26 Bulimia and Food Preoccupation and EAT-26 Social Pressure. Cross-domain associations were present but consistently weaker, with the mindful-eating and exercise-addiction domains bridged principally by MEQ Eating Discipline-EAI-3 Healthy Relevance and MEQ Conscious Eating-EAI-3 Addiction Tendency. Two negative edges emerged, between EAT-26 Social Pressure and EAI-3 Healthy Relevance and between EAT-26 Bulimia and Food Preoccupation and MEQ External Influences. EAT-26 Bulimia and Food Preoccupation showed the highest standardized strength, whereas MEQ Conscious Eating showed the strongest expected influence, and MEQ External Influences showed the highest betweenness. Conclusions: Exercise-addiction risk was selectively and modestly linked to eating-related components rather than to global disordered eating or mindful eating, with each instrument retaining substantial internal cohesion. The findings highlight the value of component-level Bayesian Gaussian network analysis for identifying candidate components associated with REA, with implications for screening and the design of future prospective and intervention studies. Because the EAI-3 subscales differ in valence, the Healthy Relevance dimension is interpreted as reflecting health-oriented exercise involvement rather than pathology. Finally, since the network is cross-sectional and undirected, the associations are non-causal, and generalizability is limited to recreational exerciser populations in Türkiye.

PMID:42588149 | DOI:10.3390/nu18152524