Causal mediation analysis in the context of clinical research

Abstract

Clinical researches usually collected numerous intermediate variables besides treatment and outcome. These variables are often incorrectly treated as confounding factors and are thus controlled using a variety of multivariable regression models depending on the types of outcome variable. However, these methods fail to disentangle underlying mediating processes. Causal mediation analysis (CMA) is a method to dissect total effect of a treatment into direct and indirect effect. The indirect effect is transmitted via mediator to the outcome. The mediation package is designed to perform CMA under the assumption of sequential ignorability. It reports average causal mediation effect (ACME), average direct effect (ADE) and total effect. Also, the package provides visualization tool for these estimated effects. Sensitivity analysis is designed to examine whether the results are robust to the violation of the sequential ignorability assumption since the assumption has been criticized to be too strong to be satisfied in research practice.

Keywords: Causal mediation analysis (CMA); clinical research; mediator; sensitivity analysis.

Conflict of interest statement

The authors have no conflicts of interest to declare.

Figures

Figure 1

Causal mediation analysis in its simplest form. X, Y and M are treatment, outcome and mediator variables. c is coefficient linking X and Y (total causal effect), c' is the coefficient for the effect of X on Y adjusting for M (direct effect), b is the effect of M on Y adjusting for explanatory variable, a is the coefficient relating to the effect of X on M. e1, e2 and e3 are residuals

Figure 2

Visualization of results from mediate() function. Estimates for both treated and control group were depicted because they are different in the example. The dashed line represents the control and solid line represents the treated. It appears that most of the total effect is explained by the direct effect

Figure 3

Sensitivity analysis based on residual correlation. The dashed line represents the ACME (0.0735) without correlation (ρ=0), which is computed under the assumption of sequential ignorability. Note that ρ≥0.2 is required for the ACME to become negative.

Figure 4

Sensitivity analysis based on coefficient of determination. Four plots are produced by varying r.type and sign.prod arguments within plot() function. Note that only the first plot that has displayed a series of ACME values. Each of the other plots has displayed the contour of one ACME value, because the native R contour function automatically adjusts the displaying values.

Figure 5

Sensitivity analysis based on coefficient of determination. Series of ACME values are displayed by adjusting levels argument in the plot() function.