1 - (SYM 26) Testing How Digital Cognitive Behavioral Therapy Works for Anxiety and Depressive Symptoms: Meta-analytic Structural Equation Modeling of 549 Randomized Trials

Cross-cultural demand for scalable mental health treatments has magnified interest in digitally delivered cognitive behavioral therapies (d-CBTs) to improve common mental disorder (CMD) symptoms and functional outcomes. However, the change mechanisms driving d-CBT efficacy remain fragmented and unclear, limiting our ability to optimize these interventions and allocate resources efficiently.

This preregistered meta-analysis synthesized 549 randomized trials (N=134,090) testing d-CBT efficacy on CMD symptoms and functional outcomes in clinical and community samples. Meta-analytic structural equation modeling (MA-SEM) integrated with robust variance estimation evaluated six theory-driven mediator groups as plausible change mechanisms: cognitive processes, behavioral targets, regulation patterns, distress, circadian sleep-wake rhythms, and functional indicators. Comprehensive publication bias analyses, sensitivity analyses, and GRADE criteria examined robustness and certainty of evidence.

Small-to-moderate bias-corrected direct effects emerged uniformly for d-CBT on depressive symptoms (β=-0.22 to -0.10), anxiety symptoms (β=-0.22 to -0.12), and quality of life (β=0.06 to 0.13), but not role impairment. All direct effects remained robust across publication bias corrections and sensitivity analyses. However, all initially significant mediation pathways were consistently nullified after rigorous publication bias adjustments and alternative model specifications. Guidance effects were inconsistent; self-guided d-CBT showed stronger effects for sleep-wake mediators. Gender and age were not robust moderators of mediation pathways.

These patterns indicate that d-CBT yields clinically meaningful improvements in CMD symptoms and quality of life, underscoring its importance as a scalable intervention accessible across diverse populations. However, non-robust bias-corrected indirect effects imply that unique change mechanisms driving d-CBT efficacy remain open to inquiry. This meta-analysis invites theoretical and methodological refinements in process-based research, including preregistered mediation protocols, adequate temporal sequencing (≥3 assessment waves), assessment of unmeasured mediators (engagement, expectancy, alliance), and sufficient power. Future research should focus on experimental manipulation of proposed mediators and expanding transdiagnostic frameworks to illuminate genuine mechanisms of change in d-CBT for CMD symptoms and functional outcomes.