353. A MULTIVERSE META-ANALYSIS OF OXYTOCIN ADMINISTRATION STUDIES

Abstract Background Oxytocin is an ancient peptide hormone and neuropeptide that is mainly produced in the hypothalamus and is distributed both centrally and peripherally. Oxytocin is integral to various physiological and cognitive functions. The effects of oxytocin administration on cognition and behavior in humans have been extensively studied, yet some results have been inconsistent. Meta-analyses are commonly used to synthesize research results; however, researchers’ analytical decisions can be a source of bias in meta-analyses. Aims & Objectives We aimed to estimate the influence of analytical decisions on meta-analysis outcomes in oxytocin administration studies. We performed a series of unique meta-analyses using a multiverse meta-analysis approach by varying analytical decision combinations. Method 530 intranasal oxytocin effect sizes were identified from 185 meta-analyses and included in our multiverse meta-analysis. We performed 314 unique meta-analyses by systematically varying three inclusion criteria: study populations (neurotypical population and population with neurodevelopmental/psychiatric conditions), dosing schedules (single and multiple administration), and outcome types (social cognition, eating behaviors, non-social cognition, symptom improvement, and pain perception) and 11 meta-analysis methods: meta-analysis models (fixed- and random-effects models), multiple outcome handling (including estimates with largest and small variance), heterogeneity estimators in random-effects model (Restricted Maximum Likelihood, DerSimonian-Laird, Hedges, Hunter-Schmidt, and Paule-Mandel), and publication bias adjustment methods (PET-PEESE, p-uniform, and limit meta-analysis). Results Effect sizes varied substantially depending on the analytical decision (d = -0.10 to 1.45), corresponding to small-to-large effects in the oxytocin literature. Specifically, larger effects were observed in neurotypical populations and studies using multiple oxytocin administrations. Oxytocin had the largest effects on symptom improvements in neurodevelopmental and psychiatric conditions, while the smallest effects were associated with threat-processing and stress/anxiety levels. While heterogeneity estimators did not impact summary effect sizes, publication bias correction methods had a varied influence on summary effect sizes. Discussion & Conclusions Our findings show that the overall effects of oxytocin administration are generally robust with small to medium effect sizes, while analytical decisions can substantially influence outcomes when synthesizing studies evaluating the effects of oxytocin administration using meta-analysis. This variability emphasizes the benefits of meta-analysis pre-registration, which can increase the robustness of results by pre-specifying analytical decisions.