Neural correlates of taste reward value across eating disorders

Abstract Individuals with eating disorders (ED) make extreme food choices, raising the possibility of altered food-value computation. We utilized an associative taste reward learning paradigm to test whether value signaling differs between participants with EDs vs. healthy controls (HC). We followed up on previous work examining prediction error (PE) signaling, which is a brain response to violation of a learned reward contingency. Expected value (EV) signal is a trial-by-trial assessment of reward significance accounting for error signaling, reward-likelihood, and learning rate. Adult female participants ( N  = 111) performed a temporal difference (TD) fMRI taste task, which is a specific type of associative reward learning paradigm, to determine EV signal: Anorexia Nervosa-ill ( N  = 28), Anorexia Nervosa-recovered ( N  = 20), Bulimia Nervosa (BN) ( N  = 20), and HC ( N = 43). Anatomical region-of-interest (ROI) analyses were performed utilizing EV regressors derived via algorithm, with ROIs based on prior EV analyses: orbitofrontal cortex, anterior cingulate (ACC), amygdala, and striatum. EV signal was elevated in the bilateral ACC in AN-ill vs. HC and BN. Intolerance of uncertainty negatively correlated with EV in AN-ill. BMI and EV were negatively-correlated across groups. Altered ACC EV computation in response to food stimuli could contribute to food restriction in AN-ill.