Developing a Computational Electroencephalogram (EEG) Paradigm to Study Prediction Error in Anorexia Nervosa

Adolescent anorexia nervosa (AN) is an eating disorder associated with intense fear of weight gain, food refusal, and severe weight loss. AN is the third most common chronic illness among adolescent females with a mortality rate 12 times higher than expected for females 15-24 years old. Little is known about biomarkers in adolescent AN.

Neuroimaging studies have repeatedly suggested altered reward processing in AN including in studies using the dopamine associated prediction error (PE) model. The brain PE response is elicited during unexpected receipt or omission of reward stimuli and thought to reflect the functionality of brain dopamine circuits. This is an important research direction as the dopamine system can be manipulated pharmacologically. In ill and recovered adult AN, unexpected or randomly applied sucrose taste stimuli evoked higher insular and striatal responses and unexpected omission or receipt of monetary or taste reward was associated with a similar response pattern in adolescent AN. PE was also inversely related to weight gain in treatment. Thus, PE brain response promises to be an important biological marker for adolescent AN with predictive value for treatment outcome. However, functional brain imaging is costly, prohibitive for instance for individuals with braces or other metal in their body and only available at certain centers. In order to study PE in AN in larger scale studies, a more practical approach and method need to be developed. In this application, we will use the exploratory/developmental R21 mechanism to develop a study protocol using electroencephalography (EEG) to study PE signals in adolescent AN. Recent studies in healthy individuals support that this is a valid approach.

Our primary goal for this study is to test the feasibility of the use of EEG for prediction error and reversal learning studies in AN with the longer term goal of replacing fMRI that is costly and associated with frequent participant rule out.

In Aim 1. we test the feasibility of adapting a computational taste PE reinforcement learning paradigm from fMRI to EEG in adolescents with AN and healthy controls. We expect that we will find internal consistency of taste PE brain response across fMRI and EEG in adolescents with AN as well as age-matched healthy controls, within each group. We further expect that we will find preliminary evidence that the EEG paradigm will be able to discriminate the AN group from the HC adolescents based on feedback related negativity and higher event-related potential amplitudes, which will correlate with fMRI PE brain response.

In Aim 2., we test whether a monetary PE paradigm will show similar EEG brain response as taste PE in Aim 1. to establish the generalizability of EEG taste and non-taste paradigms.

The development of an EEG based reward PE study paradigm will enable us in the future to conduct large-scale studies that will be less costly and independent from brain imaging centers that are only available to a small subset of adolescents with AN.