Abstract
This study assessed the preliminary effectiveness of a game-based digital therapeutics (DTx) intervention for depression and anxiety using a randomized controlled trial (RCT) design to examine the role of reinforcement learning (RL) personalization. This RCT included 223 individuals with depressive symptoms, aged 18-50, divided into three groups: an RL Algorithm group (personalized treatment), an active control group (fixed treatment), and a no-intervention control group. The intervention combined cognitive bias modification and cognitive behavioral therapy, with outcomes measured by the Patient Health Questionnaire-9 and the Generalized Anxiety Disorder-7. Results showed significantly higher treatment response and recovery rates in the RL Algorithm group compared to the no-intervention group. The game-based DTx intervention, enhanced by RL personalization, effectively reduced depression and anxiety symptoms, supporting its potential for mental health treatment. The study was registered at clinicaltrials.gov (NCT06301555).