Intranasal Oxytocin for Alcohol Use Disorder: A Systematic Review and Multilevel, Bayesian, and Variance Meta-Analyses of Randomized Clinical Trial Data

Background Intranasal oxytocin (OT) has been proposed as a promising adjunctive treatment for Alcohol Use Disorder (AUD), yet randomized controlled trials (RCTs) have yielded mixed and inconclusive findings. To clarify its therapeutic potential, we conducted a comprehensive meta-analysis using frequentist, Bayesian, and variability-based approaches. Methods We performed a multilevel random-effects meta-analysis of six eligible RCTs comparing intranasal OT with placebo for alcohol-related outcomes. Hedges' g values were calculated and winsorised at |g| = 3 to limit leverage from extreme small-sample effects. Moderator analyses assessed outcome domain, OT dose, treatment duration, year of publication, administration frequency, and clinical setting. Publication bias was evaluated using multilevel PET-PEESE with cluster-robust correction, Egger's test, trim-and-fill, and limit meta-analysis. Bayesian multilevel models examined average treatment effects and outcome variability. Results The overall pooled effect was not statistically significant (Hedges' g = 0.34, 95% CI −0.48 to 1.17, p = 0.47), with substantial between-study heterogeneity (Q(48) = 504.40, p <.001). Cook's distance identified Pedersen et al. (2013) as statistically influential; moderator and publication bias analyses were conducted on the remaining five studies. No significant moderation was observed by outcome domain, dose, duration, frequency, or setting. Year of publication showed a nominally significant positive association with effect size in the restricted sample (β = 0.184, p =.042). Multiple publication bias diagnostics converged on the absence of a systematic adjusted effect. Bayesian multilevel analysis confirmed the absence of a credible treatment effect (posterior mean μ = −0.005, 95% CrI −0.53 to 0.52). Robust Bayesian meta-analysis provided moderate support for the null hypothesis (BF₀₁ = 4.74). Variability analyses found no evidence that OT increased outcome dispersion relative to placebo (lnVR = −0.146, p =.153 after robust correction), providing no support for latent responder subgroups. Conclusions Contrary to early expectations, intranasal OT does not confer a consistent therapeutic benefit across alcohol-related outcomes. The lack of evidence for publication bias or increased outcome variability argues against latent differential responsiveness driving effects. Nonetheless, context-sensitive signals - particularly in domains related to social-cognitive processing - underscore the need for mechanistically informed trials. Future research should prioritise precision frameworks to clarify when, for whom, and under what conditions OT may yield meaningful clinical benefit.