Virtual reality-based therapy for substance use disorder: a systematic review and Bayesian meta-analysis

This Bayesian meta-analysis evaluated virtual reality (VR) therapy for substance use disorder (SUD). A total of 22 studies involving 2648 participants were included. VR-based therapy reduced substance craving compared to controls (standardized mean difference (SMD) = − 0.75, 95% credible intervals (CrI): −1.72 to 0.00, posterior probability (PP) = 83.30%, inclusion Bayes factor (BF) = 5.00, tau(τ) = 1.26). Subgroup analyses showed that VR aversion therapy (VR-AT), the illicit substance use disorder (ISUD) subgroup, and single-session within 30 minutes each demonstrated significant reductions in craving. Craving reductions showed little difference across immersion levels, and meta-regression found no reliable linear/non‑linear relationship between total VR duration and efficacy. Leave-one-out and prior sensitivity analyses confirmed the robustness of the results. VR improved executive function (SMD = −0.53, 95% CrI: −1.89 to 0.00, PP = 79.70%, inclusion BF = 3.94, τ = 0.11) and reduced anxiety symptoms (SMD = −0.33, 95% CrI: −0.72 to 0.00, PP = 85.90%, inclusion BF = 6.11, τ = 0.08). Evidence quality was low for all outcomes. No credible evidence supported its superiority over controls in abstinence rate, self-efficacy, depression, or dropout rate. Substantial heterogeneity across effect estimates highlights the need for additional high-quality RCTs.