Speakers
Description
Currently available methods for mediation analysis in meta-analysis are based on structural equation modeling (MASEM) and require the intercorrelation matrix between causal factor(s), mediator(s), and outcome(s) within the primary studies to be known. This may severely limit the application of this method. Here, we present an alternative method whose perspective lies instead on the between-study level and which thereby allows increasing the number of eligible studies. It is applicable to RCTs, wherein the causal factor is represented by the study arms and both the mediator and the outcome are measured at two time points. Using an illustrative example from intervention research, it is demonstrated how three-level meta-analysis may be used to examine mediating effects of self-reported mindfulness on the treatment efficacy of different types of mindfulness interventions. This extends a previous MASEM study (12 RCTs and non-RCTs; 1,109 participants) to 146 RCTs with 10,979 participants. Relative direct and indirect effects of categorical moderators are discussed and we describe how the indirect effect may be tested for significance.
