3 - (OP28) The Effects of Digital CBM-I on Interpretation Bias, Anxiety, and Depression: A Systematic Review and Meta-analysis

Background: Evidence for the efficacy of Cognitive Bias Modification for Interpretation (CBM-I) remains mixed, particularly for symptom outcomes and digital delivery. Given recent large-scale implementation calls, definitive evidence is needed. This review aims to examine the overall effects of digital CBM-I on interpretation bias, anxiety, and depression, and explore potential moderators.

Methods: We systematically searched databases (PsycINFO, MEDLINE, Embase, SCOPUS, Web of Science) and grey literature (PsyArXiv, ProQuest Dissertations) in July 2025 without date or language restrictions. Eligible studies employed digital CBM-I, recruited any samples (excluding neurological diseases, and/or learning/developmental disorders), and reported at least one of three outcomes: interpretation bias, anxiety, or depression. Grey literature was included. Group-level data were extracted by two blinded reviewers. A three-level random-effects meta-analysis (Hedges’ g) estimated overall effects, and multilevel meta-regression with Benjamini–Hochberg correction examined moderators. The study was pre-registered with PROSPERO (CRD42023489316).

Results: The search identified 537 records, leading to 64 unique studies included. In total, 567, 321 and 115 effect sizes were included for bias, anxiety and depression, respectively. Digital CBM-I yielded a medium-to-large effect on post-treatment interpretation bias (g = –0.62, SE = 0.05, 95% CI [–0.72, –0.52]), and small-to-moderate effects on both anxiety (g = –0.31, SE = 0.04, 95% CI [–0.38, –0.23]) and depression (g = –0.27, SE = 0.05, 95% CI [–0.37, –0.17]), all p < .001. All effects remained significant at follow-up, with mean follow-up durations ranging from 11.3 to 12.7 weeks across outcomes (SDs = 7.8–9.7). Heterogeneity was small to moderate across outcomes. Larger effects were observed in clinical than non-clinical samples for anxiety and depression at post-treatment, and for auditory (vs text-based) delivery in depression at both post-treatment and follow-up. Other hypothesised moderators (e.g., dose, imagery, comparator type) were non-significant.
Interpretation: This systematic review and meta-analysis provides the most comprehensive synthesis to date of digital CBM-I effects, demonstrating improvements in interpretation bias and modest, but significant, symptom reductions, particularly in clinical populations, all maintained at follow-up. Future trials should include longer follow-ups, clinically depressed samples, and policy-relevant benchmarks for clinical significance and health economics.