A multiverse meta-analysis of oxytocin administration studies

Oxytocin administration is a valuable approach for experimentally increasing central oxytocin levels to investigate its impact on brain function and behavior. Despite considerable research interest, results have shown considerable variability across studies. Several meta-analyses have been conducted to help address this inconsistency. However, conducting a meta-analysis requires researchers to make numerous decisions, such as defining inclusion criteria and analytical techniques. In response to this challenge, a 'multiverse' approach can be used to simultaneously explore a range of different potential and realistic scenarios by systematically varying these analytical decisions. In this study, we performed a multiverse meta-analysis to investigate how researcher decisions can influence outcomes in meta-analyses of oxytocin administration studies. We systematically varied inclusion criteria, data synthesis models, and methods for correcting publication bias across 530 effect sizes derived from 185 studies, conducting 256 distinct meta-analyses. Our results demonstrate that summary effect estimates can vary considerably depending on meta-analytic decisions (from d = -0.16 to d = 1.45). Notably, meta-analyses with neurotypical populations tended to yield larger effect sizes than non-neurotypical populations, and multiple administrations studies tended to yield larger effect sizes than single administration studies. Furthermore, summary effect estimates can substantially vary according to the applied publication bias correction method. Our findings also demonstrated that the current evidence mostly supports the presence of the effect of oxytocin across domains, regardless of meta-analytic choices, as over 90% of the observed meta-analyses exceeded the range of bootstrapped meta-analyses assuming a null effect. These results warrant future research on how different contexts may impact the robustness of the effects of oxytocin, while underscoring the value of meta-analysis pre-registration to facilitate the transparent evaluation of meta-analyst decision-making.