6.3 Integrating environmental mixtures in mediation analysis

Mediation analysis is a common approach to investigate causal pathways relating the exposure(s) and the outcome of interest. When evaluating environmental mixtures, there are several settings where our mixture of interest is only a component of an even larger picture. For example, we may want to integrate sources of exposures, or evaluate the contribution of environmental chemicals to health disparities. Our mixture, in these cases, is a mediator of a given \(X-Y\) association. In other settings, we might be interested in the mechanisms through which the mixture affects the outcome. The mixture here is the exposure in a mediation model. We can also have several mixtures affecting each other, or potential causal dependencies within the mixture.

In the general framework of exposome analysis, the underlying hypothesis is that a set of multiple exogenous exposures (the external exposome) affects the complex set of biomarkers at the microbiome level (the internal exposome), thus contributing to the development of health effects (Figure 6.1). This structure is explicitly making assumptions in terms of mediation:

The exposome concept as presented in Vrijheid et al. 2014

Figure 6.1: The exposome concept as presented in Vrijheid et al. 2014

The DAG in Figure 6.2 presents a common situation where researchers might be interested in evaluating an integrative framework for environmental exposures (E), lifestyle and behavioral factors (B), and social constructs (X), which may be complex but has the potential to elucidate mechanisms through which diseases are caused. This framework was presented in an introductory publication by Bellavia et al. (2018).

Bellavia et al. Env. Epi. 2018

Figure 6.2: Bellavia et al. Env. Epi. 2018

Integrating methods for environmental exposures into mediation analysis has been the goal of several recent papers, which the reader could refer to for further details (Bellavia, James-Todd, and Williams (2019)), (Blum et al. (2020)), (Devick et al. (2018)). These methods have been largely unexplored in applied studies and may represent a critical tool to further identify the mechanisms through which the exposome affects human health. A recent R function was also developed to integrate BKMR into a mediation analysis context (Wang et al. (2020)).


Bellavia, Andrea, Tamarra James-Todd, and Paige L Williams. 2019. “Approaches for Incorporating Environmental Mixtures as Mediators in Mediation Analysis.” Environment International 123: 368–74.
Bellavia, Andrea, Ami R Zota, Linda Valeri, and Tamarra James-Todd. 2018. “Multiple Mediators Approach to Study Environmental Chemicals as Determinants of Health Disparities.” Environmental Epidemiology (Philadelphia, Pa.) 2 (2).
Blum, Michaël GB, Linda Valeri, Olivier François, Solène Cadiou, Valérie Siroux, Johanna Lepeule, and Rémy Slama. 2020. “Challenges Raised by Mediation Analysis in a High-Dimension Setting.” Environmental Health Perspectives 128 (5): 055001.
Devick, Katrina L, Jennifer F Bobb, Maitreyi Mazumdar, Birgit Claus Henn, David C Bellinger, David C Christiani, Robert O Wright, Paige L Williams, Brent A Coull, and Linda Valeri. 2018. “Bayesian Kernel Machine Regression-Causal Mediation Analysis.” arXiv Preprint arXiv:1811.10453.
Wang, A, KL Devick, JF Bobbs, A Navas-Acien, BA Coull, and L Valeri. 2020. “BKMR-CMA: A Novel r Command for Mediation Analysis in Environmental Mixture Studies.” In ISEE Conference Abstracts. Vol. 2020. 1.