Skip to main content

Table 2 Rationale and measures of effect estimated and reported for Causal Mediation Analysis

From: Implementation and reporting of causal mediation analysis in 2015: a systematic review in epidemiological studies

References

Reasona

Measures discussed or reported

Motivation for applicationb

Randomized controlled trials

 D’Amelio et al. [27]

Mediation

Natural direct and natural indirect effects

Emphasized direct effect

Improve understanding to show that above and beyond how the treatment works through the mediator, there is an independent effect

 Freeman et al. [28]

Mediation

Direct and indirect effectsc

Proportion mediated by various factors

Improve understanding of mechanisms

Cohort studies

 Banack et al. [26]

Mediation

Similar to controlled direct effect (with caveat that no manipulation of obesity could actually occur)

Refute/confirm that selection bias drives the obesity paradox in cardiovascular disease

 Jackson et al. [29]

Mediation

Natural direct and indirect effects

Proportion mediated by each medical event

Improve understanding of mechanisms

 Kositsawat et al. [30]

Mediation

Not identified

Not clear

 Louwies et al. [31]

Mediation

Direct and indirect effectc

Improve understanding of mechanisms

 Lu et al. [32]

Mediation

Natural direct and natural indirect effect

Percent excess risk mediated

Natural indirect effect emphasized

Improve understanding of mechanisms

 Mendola et al. [33]

Mediation

Controlled direct effect

Improve understanding

 Messerlian et al. [34]

Mediation

Controlled direct effect

Improve understanding

 Raghavan et al. [35]

Mediation

Direct and indirect effects but only indirect effects reportedc

Proportion of risk mediated through genetic and metabolic factors

Improve understanding of what mediators might be ripe for intervention

Case control studies

 Rao et al. [36]

Mediation

Controlled direct effect

Improve understanding

 Song et al. [37]

Mediation

Effect not mediated

mediated effectc

Proportion mediated through various biomarkers

Improve understanding of mechanisms

 Xie et al. [38]

Mediation

Direct and indirect effectc

Proportion of effect mediated through testosterone

Improve understanding

  1. aReason for applying causal mediation analysis: Mediation, Interaction, or Interference
  2. bMotivation for each application of causal mediation analysis. For mediation (1) improve understanding; (2) confirm/refute theory; (3) intervention refinement. For interaction (1) help allocate resources better; (2) identifying groups in which treatments may be harmful or beneficial (qualitative or cross-over interactions); (3) understand mechanisms; (4) increase statistical power of main effect analysis, and (5) understand which mediator to intervene upon to eliminate most of the effect of primary exposure. For interference (1) quantify spillover effects for cost-effectiveness studies; (2) understand what proportion must be treated to attain population outcomes desired; (3) create knowledge for intervention development and refinement
  3. c“Natural” was not specifically used in the article but appeared to have counterfactual framework and appropriate references