ABSTRACT
BACKGROUND Cognitive behavioural therapy (CBT) is an effective evidence based treatment for depression. At present there is no reliable predictor of CBT in depression. Although the key to successful CBT in depression lies in altering maladaptive information processing, no previous imaging study has probed predictors of CBT response using pre-treatment neural encoding of information processing. METHODS: Using functional magnetic resonance imaging we scanned 37 unmedicated depressed subjects before and after completing computerised CBT (cCBT). We model the trial-by-trial appraisal of feedback information during a probabilistic learning task by means of a dynamic learning rate. To discriminate response to cCBT we capitalise on the pre-treatment blood oxygen level dependent (BOLD) activity encoding the dynamic learning rate as a function of feedback congruence and valence. Additionally, we probe between-group differences in the learning style encoded in the model’s parameters.
RESULTS We show BOLD activity in the dorsomedial prefrontal cortex (dmPFC) to be encoding the dynamic learning rate. Crucially, responders exhibit greater BOLD activity in the dmPFC during incongruent negative trials but lower BOLD activity during congruent negative trials than non-responders. Additionally, on between-group comparisons of model’s parameter estimates we show responders take relatively greater account of previous feedback history and make comparatively smaller adjustments to the learning rate as a result of outcome surprisingness.
CONCLUSIONS Our findings provide novel and important insights into the cognitive mechanisms underpinning response to cCBT and lend support to the feasibility and validity of neurocomputational approaches to treatment prediction research in psychiatry.