This article provides an overview of recent developments in mediation analysis, that is, analyses used to assess the relative magnitude of different pathways and mechanisms by which an exposure may affect an outcome. Traditional approaches to mediation in the biomedical and social sciences are described. Attention is given to the confounding assumptions required for a causal interpretation of direct and indirect effect estimates. Methods from the causal inference literature to conduct mediation in the presence of exposure-mediator interactions, binary outcomes, binary mediators, and case-control study designs are presented. Sensitivity analysis techniques for unmeasured confounding and measurement error are introduced. Discussion is given to extensions to time-to-event outcomes and multiple mediators. Further flexible modeling strategies arising from the precise counterfactual definitions of direct and indirect effects are also described. The focus throughout is on methodology that is easily implementable in practice across a broad range of potential applications.


Article metrics loading...

Loading full text...

Full text loading...


Literature Cited

  1. Albert JM. 1.  2012. Distribution-free mediation analysis for nonlinear models with confounding. Epidemiology 23:879–88 [Google Scholar]
  2. Albert JM, Nelson S. 2.  2011. Generalized causal mediation analysis. Biometrics 67:1028–38 [Google Scholar]
  3. Alwin DF, Hauser RM. 3.  1975. The decomposition of effects in path analysis. Am. Sociol. Rev. 40:37–47 [Google Scholar]
  4. Avin C, Shpitser I, Pearl J. 4.  2005. Identifiability of path-specific effects. Proc. Int. Jt. Conf. Artif. Intel., Edinburgh, Aug.357–63 [Google Scholar]
  5. Baron RM, Kenny DA. 5.  1986. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J. Pers. Soc. Psychol. 51:1173–82 [Google Scholar]
  6. Daniel RM, De Stavola BL, Cousens SN. 6.  2011. gformula: estimating causal effects in the presence of time-varying confounding or mediation using the g-computation formula. Stata J. 11:479–517 [Google Scholar]
  7. Daniel RM, De Stavola BL, Cousens SN, Vansteelandt S. 7.  2015. Causal mediation analysis with multiple mediators. Biometrics 71:1–14 [Google Scholar]
  8. Ding P, VanderWeele TJ. 8.  2015. Sharp sensitivity bounds for mediation under unmeasured mediator-outcome confounding. Tech. Rep. In press [Google Scholar]
  9. Goetgeluk S, Vansteelandt S, Goetghebeur E. 9.  2008. Estimation of controlled direct effects. J. R. Stat. Soc. Ser. B Stat Methodol. 70:1049–66 [Google Scholar]
  10. Greenland S. 10.  2003. Quantifying biases in causal models: classical confounding versus collider-stratification bias. Epidemiology 14:300–6 [Google Scholar]
  11. Greenland S, Robins JM, Pearl J. 11.  1999. Confounding and collapsibility in causal inference. Stat. Sci. 14:29–46 [Google Scholar]
  12. Hafeman DM. 12.  2011. Confounding of indirect effects: a sensitivity analysis exploring the range of bias due to a cause common to both the mediator and the outcome. Am. J. Epidemiol. 174:710–17 [Google Scholar]
  13. Hong G, Nomi T. 13.  2012. Weighting methods for assessing policy effects mediated by peer change. J. Res. Educ. Eff. (Spec. Issue: Stat. Approaches Stud. Mediat. Eff. Educ. Res. 5:261–89 [Google Scholar]
  14. Hyman HH. 14.  1955. Survey Design and Analysis: Principles, Cases and Procedures Glencoe, IL: Free Press [Google Scholar]
  15. Imai K, Keele L, Tingley D. 15.  2010. A general approach to causal mediation analysis. Psychol. Methods 15:309–34 [Google Scholar]
  16. Imai K, Yamamoto T. 16.  2012. Identification and sensitivity analysis for multiple causal mechanisms: revisiting evidence from framing experiments. Pol. Anal. 21:141–71 [Google Scholar]
  17. Jiang Z, VanderWeele TJ. 17.  2015. Causal mediation analysis in the presence of a misclassified binary exposure. Tech. Rep. [Google Scholar]
  18. Jiang Z, VanderWeele TJ. 18.  2015. Causal mediation analysis in the presence of a mismeasured outcome. Epidemiology 26:e8–9 [Google Scholar]
  19. Jiang Z, VanderWeele TJ. 19.  2015. When is the difference method conservative for mediation?. Am. J. Epidemiol. 1822105–8 [Google Scholar]
  20. Judd CM, Kenny DA. 20.  1981. Process analysis: estimating mediation in treatment evaluations. Eval. Rev. 5:602–19 [Google Scholar]
  21. Lange T, Hansen JV. 21.  2011. Direct and indirect effects in a survival context. Epidemiology 22:575–81 [Google Scholar]
  22. Lange T, Vansteelandt S, Bekaert M. 22.  2012. A simple unified approach for estimating natural direct and indirect effects. Am. J. Epidemiol. 176:190–95 [Google Scholar]
  23. le Cessie S, Debeij J, Rosendaal FR, Cannegieter SC, Vandenbroucke J. 23.  2012. Quantification of bias in direct effects estimates due to different types of measurement error in the mediator. Epidemiology 23:551–60 [Google Scholar]
  24. MacKinnon DP. 24.  2008. Introduction to Statistical Mediation Analysis New York: Erlbaum [Google Scholar]
  25. MacKinnon DP, Warsi G, Dwyer JH. 25.  1995. A simulation study of mediated effect measures. Multivar. Behav. Res. 30:41–62 [Google Scholar]
  26. Martinussen T, Vansteelandt S, Gerster M, von Bornemann Hjelmborg J. 26.  2011. Estimation of direct effects for survival data by using the Aalen additive hazards model. J. R. Stat. Soc. Ser. B Stat. Methodol. 73:773–88 [Google Scholar]
  27. Nandi A, Glymour MM, Kawachi I, VanderWeele TJ. 27.  2012. Using marginal structural models to estimate the direct effect of adverse childhood social conditions on onset of heart disease, diabetes and stroke. Epidemiology 23:223–32 [Google Scholar]
  28. Ogburn EL, VanderWeele TJ. 28.  2012. Analytic results on the bias due to nondifferential misclassification of a binary mediator. Am. J. Epidemiol. 176:555–61 [Google Scholar]
  29. Pearl J. 29.  2001. Direct and indirect effects. Proc. Conf. Uncertain. Artif. Intel., 17th, Seattle411–20 San Francisco: Kaufmann [Google Scholar]
  30. Robins JM, Greenland S. 30.  1992. Identifiability and exchangeability for direct and indirect effects. Epidemiology 3:143–55 [Google Scholar]
  31. Robins JM, Richardson TS. 31.  2010. Alternative graphical causal models and the identification of direct effects. Causality and Psychopathology: Finding the Determinants of Disorders and Their Cures P Shrout 103–58 Oxford, UK: Oxford Univ. Press [Google Scholar]
  32. Robinson L, Jewell NP. 32.  1991. Some surprising results about covariate adjustment in logistic regression models. Int. Stat. Rev. 59:227–40 [Google Scholar]
  33. Sobel ME. 33.  1982. Asymptotic confidence intervals for indirect effects in structural equations models. Sociological Methodology S Leinhart 290–312 San Francisco: Jossey-Bass [Google Scholar]
  34. Strong V, Waters R, Hibberd C, Murray G, Wall L. 34.  et al. 2008. Management of depression for people with cancer (SMaRT oncology 1): a randomised trial. Lancet 372:40–48 [Google Scholar]
  35. Tchetgen Tchetgen EJ, Shpitser I. 35.  2012. Semiparametric theory for causal mediation analysis: efficiency bounds, multiple robustness, and sensitivity analysis. Ann. Stat. 40:1816–45 [Google Scholar]
  36. Tchetgen Tchetgen EJ, VanderWeele TJ. 36.  2014. On identification of natural direct effects when a confounder of the mediator is directly affected by exposure. Epidemiology 25:282–91 [Google Scholar]
  37. Tingley D, Yamamoto T, Hirose K, Keele L, Imai K. 37.  2014. Mediation: R package for causal mediation analysis. J. Stat. Softw. 59:1–38 [Google Scholar]
  38. Valeri L, Lin X, VanderWeele TJ. 38.  2014. Mediation analysis when a continuous mediator is measured with error and the outcome follows a generalized linear model. Stat. Med. 33:4875–90 [Google Scholar]
  39. Valeri L, VanderWeele TJ. 39.  2013. Mediation analysis allowing for exposure-mediator interactions and causal interpretation: theoretical assumptions and implementation with SAS and SPSS macros. Psychol. Methods 18:137–50 [Google Scholar]
  40. Valeri L, VanderWeele TJ. 40.  2014. The estimation of direct and indirect causal effects in the presence of a misclassified binary mediator. Biostatistics 15:498–512 [Google Scholar]
  41. VanderWeele TJ. 41.  2009. Marginal structural models for the estimation of direct and indirect effects. Epidemiology 20:18–26 [Google Scholar]
  42. VanderWeele TJ. 42.  2010. Bias formulas for sensitivity analysis for direct and indirect effects. Epidemiology 21:540–51 [Google Scholar]
  43. VanderWeele TJ. 43.  2011. Causal mediation analysis with survival data. Epidemiology 22:575–81 [Google Scholar]
  44. VanderWeele TJ. 44.  2012. Structural equation modeling in epidemiologic analysis. Am. J. Epidemiol. 176:608–12 [Google Scholar]
  45. VanderWeele TJ. 45.  2013. Unmeasured confounding and hazard scales: sensitivity analysis for total, direct and indirect effects. Eur. J. Epidemiol. 28:113–17 [Google Scholar]
  46. VanderWeele TJ. 46.  2014. A unification of mediation and interaction: a four-way decomposition. Epidemiology 25:749–61 [Google Scholar]
  47. VanderWeele TJ. 47.  2015. Explanation in Causal Inference: Methods for Mediation and Interaction New York: Oxford Univ. Press [Google Scholar]
  48. VanderWeele TJ, Knol MJ. 48.  2014. A tutorial on interaction. Epidemiol. Methods 3:33–72 [Google Scholar]
  49. VanderWeele TJ, Robinson W. 49.  2014. On the causal interpretation of race in regressions adjusting for confounding and mediating variables. Epidemiology 25:473–84 [Google Scholar]
  50. VanderWeele TJ, Tchetgen Tchetgen EJ. 50.  2014. Mediation analysis with time-varying exposures and mediators Work. Pap. 168, Harvard Univ. Biostat. Work. Pap. Ser. http://biostats.bepress.com/cgi/viewcontent.cgi?article=1176&context=harvardbiostat [Google Scholar]
  51. VanderWeele TJ, Valeri L, Ogburn EL. 51.  2012. The role of misclassification and measurement error in mediation analyses. Epidemiology 23:561–64 [Google Scholar]
  52. VanderWeele TJ, Vansteelandt S. 52.  2009. Conceptual issues concerning mediation, interventions and composition. Stat. Interface 2:457–68 [Google Scholar]
  53. VanderWeele TJ, Vansteelandt S. 53.  2010. Odds ratios for mediation analysis for a dichotomous outcome. Am. J. Epidemiol. 172:1339–48 [Google Scholar]
  54. VanderWeele TJ, Vansteelandt S. 54.  2013. Mediation analysis with multiple mediators. Epidemiol. Methods 2:95–115 [Google Scholar]
  55. VanderWeele TJ, Vansteelandt S, Robins JM. 55.  2014. Methods for effect decomposition in the presence of an exposure-induced mediator-outcome confounder. Epidemiology 25:300–6 [Google Scholar]
  56. Vansteelandt S. 56.  2009. Estimating direct effects in cohort and case-control studies. Epidemiology 20:851–60 [Google Scholar]
  57. Vansteelandt S, VanderWeele TJ. 57.  2012. Natural direct and indirect effects on the exposed: effect decomposition under weaker assumptions. Biometrics 68:1019–27 [Google Scholar]
  58. Zheng W, van der Laan MJ. 58.  2012. Targeted maximum likelihood estimation of natural direct effects. Int. J. Biostat. 8:1–40 [Google Scholar]

Data & Media loading...

  • Article Type: Review Article
This is a required field
Please enter a valid email address
Approval was a Success
Invalid data
An Error Occurred
Approval was partially successful, following selected items could not be processed due to error