Computationally-informed insights into anhedonia and treatment by k-opioid receptor antagonism.

Background: Anhedonia, the loss of pleasure, is prevalent and impairing. Parsing its computational basis promises to explain its transdiagnostic character. One manifestation of anhedonia-reward insensitivity-may be linked to limited memory. Further, the need to economize on limited memory engenders a perseverative bias towards frequently chosen actions. Anhedonia may also be linked with deviations from optimal perseveration for a given memory capacity, a pattern that causes inefficiency because it results in less reward for the same memory cost.

Methods: To test these hypotheses, we apply a theory of optimal decision-making under memory constraints that decomposes behavior into a memory component and an efficiency component. We apply this theory to behavior on the Probabilistic Reward Task, a reward learning paradigm validated in anhedonia, and perform secondary analysis of a randomized controlled trial testing κ-opioid receptor (KOR) antagonism for anhedonia (N=24 KOR; N=31 placebo), as well as analyses of three other datasets (N=100, 66, 24 respectively). We fit a resource-bounded reinforcement-learning model to behavior.

Results: Across clinical and nonclinical populations, anhedonia is associated with deficits in efficiency but not memory. The reinforcement learning models demonstrate that deficits in efficiency arise from the inability to perseverate optimally. KOR antagonism, which likely elevates tonic dopamine, increases both memory and efficiency, and the model demonstrates that this arises from increased reward sensitivity and perseveration.

Conclusions: KOR antagonism therefore has distinct cognitive effects, only one related to anhedonia. These findings have potential implications for the applications of KOR antagonists.