Translational Epidemiologic Approaches to Understanding the Consequences of Early-Life Exposures

Abstract

Prominent developmental theories posit a causal link between early-life exposures and later functioning. Yet, observed associations with early exposures may not reflect causal effects because of genetic and environmental confounding. The current manuscript describes how a systematic series of epidemiologic analyses that combine several genetically-informative designs and statistical approaches can help distinguish between competing theories. In particular, the manuscript details how combining the use of measured covariates with sibling-comparisons, cousin-comparisons, and additional designs can help elucidate the sources of covariation between early-life exposures and later outcomes, including the roles of (a) factors that are not shared in families, including a potential causal effect of the exposure; (b) carryover effects from the exposure of one child to the next; and (c) familial confounding. We also describe key assumptions and how they can be critically evaluated. Furthermore, we outline how subsequent analyses, including effect decomposition with respect to measured, plausible mediators, and quantitative genetic models can help further specify the underlying processes that account for the associations between early-life exposures and offspring outcomes.

Keywords: Causal inference; Cousin comparisons; Developmental origins of health and disease; Fetal growth; Genetically-informed designs; Pregnancy; Sibling comparisons.

Figure 1

Schematic Example of a Causal Diagram for the Structural Relationship between an Early-life Exposure and Later Outcome Note. In the diagram both measured (C₁) and unmeasured (U₁) common causes create back-door paths between the exposure and outcome. A known mediating factor in the causal pathway from exposure to outcome also shares common causes with the exposure (U₂) and the outcome (U₃) respectively. The exposure-mediator association is used to illustrate how the act of conditioning (marked by a box) on a measured covariate (C₂) can block the back door path from unmeasured common causes (U₂), so that C₂ is a confounder of the exposure-mediator association, although the variable itself is not a common cause. Lastly the diagram also serves to illustrate how the presence of common causes of the mediator and outcome (U₃) make the mediator a collider (an effect of both exposure and of U₃). Conditioning on the mediator would, therefore, introduce a spurious association between exposure and outcome through the common causes (U₃).

Figure 2

Framework for Guiding Analyses of Associations with Early Life Exposures Note. Black (thick) boxes represent methods. Blue (thin) boxes represent inferences. Solid black arrows represent results. Shaded gray arrow indicates further steps.

Figure 3

Example of a Causal Diagram for the Relationship between an Early-life Exposure and Later Outcome in a Pair of Siblings Note. For simplicity of illustration, we assume that all measured confounders of exposure (X) and outcome (Y) for sibling pair (i) have been properly controlled for (Aim 1), so that all common causes in this diagram are unmeasured. In addition to sharing some common causes of exposure and outcome (U_i), siblings may also share causes of exposure (U_Xi) and outcome (U_Yi) respectively. All common causes of exposure and outcome that can vary in siblings (non-shared) are captured in U_i1 and U_i2 respectively. The diagram assumes no contagion from the first sibling’s status to the next (i.e. no arrow from X_i1 to X_i2, or from Y_i1 to Y_i2), and no carryover effect from the first siblings’ exposure to subsequent sibling’s outcome (X_i1 → Y_i2). When a comparison is made “within” (or conditional on) siblings, all pathways through shared factors (U_i, U_Xi and U_Yi) will be blocked. However, restriction to differentially exposed siblings (which corresponds to conditioning on one of the siblings’ exposure) will create a spurious association between one sibling’s exposure and the other sibling’s non-shared common causes (U_i1 → X_i1 ← U_Xi → X_i2; where X_i1 is a collider).