Section | Recommendation |
---|---|
Objectives | State whether mediation analysis(es) is/are exploratory or hypothesis-based |
Methods | Specify criteria or statistical tests used to assess mediation, with references Was the goal to categorize mediation as absent, partial or complete, or to estimate exact values for direct and indirect effects? |
 | Detail how exposure, mediator and outcome variables were defined and measured |
 | Detail when exposure, mediator and outcome variables were measured |
 | Describe statistical models used for the mediator(s) and outcome(s), and any assumptions underlying use of such models (e.g. proportionality, rare outcome assumption for Cox Proportional Hazards models) |
 | State whether interaction between exposure and mediator was considered, and how |
 | Reference any software programs used for mediation analysis |
 | If relevant for exposure, mediator, and outcome being considered, state how the following were addressed: - clustering or repeated events - competing risks |
 | Describe assumptions underlying mediation analysis, and methods used to address these (e.g.: randomisation, regression, weighting, stratification, sensitivity analysis) |
Results | Report measures of mediation effect (indirect effect or proportion mediated) accompanied by 95% confidence intervals |
 | Report p-values for mediation hypothesis testing |
Discussion | Discuss limitations of causal inference based on mediation analysis results, including whether underlying assumptions were met Discuss magnitude and direction of any potential bias |