Computational Psychiatry (Oct 2022)

Components of Behavioral Activation Therapy for Depression Engage Specific Reinforcement Learning Mechanisms in a Pilot Study

  • Quentin J. M. Huys,
  • Evan M. Russek,
  • George Abitante,
  • Thorsten Kahnt,
  • Jacqueline K. Gollan

DOI
https://doi.org/10.5334/cpsy.81
Journal volume & issue
Vol. 6, no. 1

Abstract

Read online

Background: Behavioral activation is an evidence-based treatment for depression. Theoretical considerations suggest that treatment response depends on reinforcement learning mechanisms. However, which reinforcement learning mechanisms are engaged by and mediate the therapeutic effect of behavioral activation remains only partially understood, and there are no procedures to measure such mechanisms. Objective: To perform a pilot study to examine whether reinforcement learning processes measured through tasks or self-report are related to treatment response to behavioral activation. Method: The pilot study enrolled 13 outpatients (12 completers) with major depressive disorder, from July of 2018 through February of 2019, into a nine-week trial with BA. Psychiatric evaluations, decision-making tests and self-reported reward experience and anticipations were acquired before, during and after the treatment. Task and self-report data were analysed by using reinforcement-learning models. Inferred parameters were related to measures of depression severity through linear mixed effects models. Results: Treatment effects during different phases of the therapy were captured by specific decision-making processes in the task. During the weeks focusing on the active pursuit of reward, treatment effects were more pronounced amongst those individuals who showed an increase in Pavlovian appetitive influence. During the weeks focusing on the avoidance of punishments, treatment responses were more pronounced in those individuals who showed an increase in Pavlovian avoidance. Self-reported anticipation of reinforcement changed according to formal RL rules. Individual differences in the extent to which learning followed RL rules related to changes in anhedonia. Conclusions: In this pilot study both task- and self-report-derived measures of reinforcement learning captured individual differences in treatment response to behavioral activation. Appetitive and aversive Pavlovian reflexive processes appeared to be modulated by separate psychotherapeutic interventions, and the modulation strength covaried with response to specific interventions. Self-reported changes in reinforcement expectations are also related to treatment response. Trial Registry Name: Set Your Goal: Engaging in GO/No-Go Active Learning, #NCT03538535, http://www.clinicaltrials.gov.

Keywords