Reinforcement Learning Mechanisms of Pharmacological Treatments for Depression (RELMED)

First-line antidepressants target serotonin, dopamine and noradrenaline. A large body of detailed experimental and theoretical work in reinforcement learning (RL) has led to detailed understanding of the roles these neuromodulators play in higher cognitive function, affect and learning, and related them to depression. However, it is unknown if these RL processes are the mechanisms through which antidepressants relieve depression. The translational gap exists because studies in this area have rarely involved patients and have mostly been small cross-sectional rather than substantial longitudinal treatment studies.

We propose to definitely test whether specific RL processes are the mechanisms of action by which different antidepressants work. To achieve this, we propose to run two large trials (n=516x2) in which primary care patients with depression are randomized to escitalopram (serotonergic), bupropion (dopaminergic/noradrenergic) or placebo. RL processes will be assessed using online tasks, and through electroencephalography. The first trial will broadly assess RL domains (spanning instrumental and Pavlovian learning, effort and control). The second trial will build on the results and test specific RL mechanisms. We also aim to use our experience to kick-start a new open science approach to facilitate rapid translation of neuroscience findings into treatment improvements in the longer term.