Personal causal modeling of affect and eating behaviors in bulimia nervosa: implications for personalized treatment.

backgroundAffect regulation models suggest high negative and low positive affect may drive binge eating and purging in bulimia nervosa (BN). While ecological momentary assessment (EMA) studies often support these theories, inconsistent outcomes in affect-targeted interventions suggest causal relations vary across individuals. This study applied causal discovery analysis (CDA) to EMA data to characterize such heterogeneity in person-specific causal models for BN.

methodsEMA data from 118 adult women with BN, collected over 14 days, assessed momentary negative affect, positive affect, binge eating, and self-induced vomiting. Using the Greedy Fast Causal Inference algorithm, we derived individual causal models and estimated effect sizes via structural equation modeling. Heterogeneity was evaluated by the proportion of participants with affect as a causal factor for BN behaviors.

resultsCausal patterns were highly heterogeneous. Elevated negative affect was causal for binge eating in 16.9% of participants, vomiting in 18.6%, and either behavior in 27.1%. Low positive affect was causal for binge eating in 8.5%, vomiting in 11.0%, and either behavior in 15.3%. Behavior-behavior causality was also common: vomiting caused binge eating in 26.3% of participants, and binge eating caused vomiting in 22.0%.

conclusionsCDA revealed marked heterogeneity in causal factors underlying BN behaviors, with some models showing affect-driven behaviors and others indicating behavior-driven patterns. Ultimately, this work indicates that the link between momentary affect and BN behaviors is highly individualized, underscoring the need for precision-targeted interventions rather than one-size-fits-all treatments.