Abstract
BACKGROUND: Attention bias modification (ABM) therapy aims to modify threat-related attention patterns via computerized tasks. Despite showing medium clinical effect sizes for anxiety disorders, underlying neural-cognitive mechanisms of change remain unclear. We used visual mismatch negativity (vMMN), an event-related potential sensitive to violations of learned statistical contingencies, to assess therapy-related contingency extraction processes in healthy participants and in patients with social anxiety disorder (SAD). We then assessed whether vMMN amplitude predicts ABM treatment outcome. METHODS: A modified version of the dot-probe task was used to elicit vMMN, in which 80% of trials were standard and 20% were deviant. In study 1, 30 healthy adults were randomly assigned to one of two ABM conditions: one in which threat-congruent targets were deviant trials and threat-incongruent targets were standard trials, and another in which the contingency was reversed. Electroencephalography (EEG) was continuously measured and vMMN analyzed. In study 2, 38 patients with SAD underwent six sessions of ABM therapy. We tested whether rule extraction in the ABM task, indicated by vMMN amplitude, predicts treatment outcome. RESULTS: vMMN clearly emerged over prespecified scalp locations indicating contingency extraction during ABM (study 1). vMMN amplitude predicted clinical improvement after ABM therapy, uniquely accounting for 7% and 14.4% of the variance in clinician-rated and self-reported posttreatment SAD symptoms, respectively. CONCLUSIONS: vMMN emerges as a neural marker for contingency learning in ABM, suggesting a significant role for contingency extraction processes in the clinical efficacy of this therapy.