Avoidance-driven Decision Making and Learning in Anorexia Nervosa and Bulimia Nervosa

PROJECT SUMMARY/ABSTRACT Identification of neural and behavioral processes that drive eating disorder (ED) symptomatology is critical for the development of more effective interventions for these serious and sometimes deadly disorders. One candidate transdiagnostic mechanism underlying divergent but often overlapping symptoms in anorexia nervosa

(AN) and bulimia nervosa (BN) is avoidance learning. The ability to learn how to avoid harm is critical for health and survival, yet excessive avoidance learning leads to chronic maladaptive avoidance or compulsive behaviors. Both AN and BN are characterized by elevated trait harm avoidance and maladaptive behaviors to avoid aversive

outcomes such as weight gain, ranging from extreme dietary restriction to repetitive cycles of binge eating and purging (e.g., self-induced vomiting, excessive exercise). This raises the question of whether symptom heterogeneity corresponds to differences in an individual’s strategy to avoid aversive outcomes that fall on a

continuum of anxious-avoidant vs compulsive behavior. Distinguishing between excessive active avoidance learning (learning from successful actions that reduce harm) and passive avoidance learning (learning what to avoid doing to prevent harm), which are thought to differentiate compulsive and anxious-avoidant behavior,

might critically inform etiological models of ED. This proposal tests the novel hypothesis that instrumental avoidance learning is altered in ED, with corresponding differences in corticostriatal and limbic-prefrontal prediction error BOLD response and functional connectivity, and that associations between active vs passive

avoidance learning and ED symptoms might differentiate bulimic-type from restricting behaviors, informing compulsive vs anxious-avoidant mechanisms underlying symptom heterogeneity in ED. Participants (26 AN, 26 BN and 26 healthy controls (HC) ages 18-35) will complete a probabilistic card gambling task during fMRI that

assesses instrumental learning strategies to avoid heat pain. Computational modeling approaches will be used to distinguish active from passive learning bias. Aim 1 will compare avoidance learning bias in AN, BN and HC and will evaluate the association of active and passive avoidance learning with ED symptoms. Aim 2 will examine

whether corticostriatal and limbic-prefrontal function associated with avoidance learning differs in AN, BN and ED, and in particular whether dorsal caudate response associated with active and passive avoidance learning differs in AN and BN and relates to symptoms. Aim 3 will examine group differences in corticostriatal and limbic-

prefrontal PE-related functional connectivity associated with active and passive avoidance learning to better characterize avoidance learning functional neural architecture in ED. Elucidating the relationship between avoidance learning, ED symptoms and brain function to inform mechanistic understanding of the neurobiological

underpinnings of ED is both innovative and significant. Outcomes will be of clinical importance by providing foundational knowledge to identify new behavioral and neural therapeutic targets. Data will provide methodological groundwork for an R01 application examining avoidance learning across a broader ED spectrum.