Applied mediation analyses: a review and tutorial

Abstract

In recent years, mediation analysis has emerged as a powerful tool to disentangle causal pathways from an exposure/treatment to clinically relevant outcomes. Mediation analysis has been applied in scientific fields as diverse as labour market relations and randomized clinical trials of heart disease treatments. In parallel to these applications, the underlying mathematical theory and computer tools have been refined. This combined review and tutorial will introduce the reader to modern mediation analysis including: the mathematical framework; required assumptions; and software implementation in the R package medflex. All results are illustrated using a recent study on the causal pathways stemming from the early invasive treatment of acute coronary syndrome, for which the rich Danish population registers allow us to follow patients’ medication use and more after being discharged from hospital.

Use of antimalarial drugs is associated with a lower risk of preeclampsia in lupus pregnancy: A prospective cohort study

“…We also perform a mediation analysis to adjust the use of antimalarial drugs by SLE activity. 30 All analyses used the program package SPSS version 23.0 (IBM).…”

The conduct and reporting of mediation analysis in recently published randomized controlled trials: results from a methodological systematic review

“…For instance, when a logistic outcome model is used for a binary endpoint, then the difference-of-coefficient approach has a systematic tendency to find indirect effects in settings where treatment has no effect on mediator; the product-of-coefficient approach instead delivers direct and indirect effects which may not add up to the total causal effect, thereby failing to provide an effect decomposition [4,5]. To accommodate this, a counterfactual-based framework to MA has recently been introduced, which has the aforementioned approaches as special cases [1,4,5,12]. The literature on this framework is often of a more technical nature, partly because of its focus on nonlinear models and partly because it makes more explicit the unverifiable assumptions on which an MA relies.…”

“…Different, possibly nonlinear, regression models are specified for the mediator and the outcome. The IE of X on Y through M is quantified by using the aforementioned models to simulate (via Monte Carlo simulation) how a change in X would affect M for each individual, how this change would in turn affect Y, and then averaging these results [1,18]. The DE of X on Y is quantified by using the aforementioned models to simulate (via Monte Carlo simulation) how Y would change for each individual if X were changed, but M were fixed at the (simulated) level it would take if X took on some reference level, and then averaging these results [1,18].…”

Performance of methods to conduct mediation analysis with time‐to‐event outcomes

“…However, given the nature of the natural effect models used in the simulation‐based approach, this approach can easily be extended to the AFT model. The simulation‐based approach is based on the following five steps (Lange et al, ). Use the original data set to fit the imputation model, that is, a Cox PH or AFT model in which the outcome is the dependent variable and both the exposure and mediator are independent variables. Create a new data set based on the original data set in which all observations are present twice.…”