Tracing Early-Life Exposures Across the Life Course: Development, Behavior, and Adult Health

Abstract

Understanding how early-life exposures shape health across the life course remains a central challenge in epidemiology. Prenatal and early developmental characteristics influence biological systems in ways that may manifest in childhood behavioral and physical health outcomes, and extend into adult health and well-being, yet the mechanisms linking these stages are not fully understood. Disentangling biological pathways from methodological artifacts, especially in the context of substance exposures, requires innovative analytical frameworks that integrate life-course data, developmental markers, and causal inference methods, alongside novel methodological and conceptual approaches. The unifying arc of this work moves from detecting early biological signals through data-driven approaches, to taking a focused mechanistic view of developmental pathways through causal inference, and ultimately to stepping back and applying a new lens to the evidence base that has shaped the field for decades. Together, these studies not only deepen our understanding of development but also challenge and refine how we interpret evidence on the effects of alcohol and other common early-life exposures. In Chapter One, we assessed whether facial parameters identified in children can help understand the neurodevelopmental impact of prenatal exposures on child behavior. Using data from 9- to 10-year-old children in Generation R Study (N = 2,779), we examined associations between facial traits derived from 3D face photographs using an AI-based 3D autoencoder and attention problems, as well as prenatal exposures. We identified a specific facial trait associated with attention problems and found that prenatal smoking, vitamin D, and folic acid were significantly related to variation in this trait. This trait is characterized by chin retrusion, mild nasal contour variation, nose tip protrusion, and overall facial asymmetry. These findings suggest that facial morphology may serve as a useful, though not yet diagnostic, marker of impaired neuroectodermal development. In Chapter Two, we assessed potential sex-specific pathways linking prenatal alcohol exposure to behavioral symptoms in offspring, with a focus on low birthweight as both a mediator and modifier. Using data from 8- to 10-year-old children in the population-based Adolescent Brain and Cognitive Development Study (N = 7,502), we analyzed associations between prenatal alcohol exposure and externalizing behavioral problems and applied a sex-stratified four-way decomposition framework introducing low birthweight as a mediator and/or modifier. Prenatal alcohol exposure was associated with increased externalizing behavioral problems, with slightly stronger effects observed in females than males. Decomposition analyses indicated evidence of interaction, with substantially higher behavioral problems among males with joint exposure to prenatal alcohol exposure and low birthweight, but not females. We found no evidence of indirect effects through low birthweight. These findings suggest that low birthweight may exacerbate the effects of prenatal alcohol exposure, particularly among males, highlighting the importance of sex-specific vulnerability. In Chapter Three we contributed to the complex debate on whether low-level alcohol consumption confers protective health effects. Commonly reported health benefits of low-level drinking have been attributed to both biological mechanisms and methodological artifacts. Prenatal alcohol exposure, a teratogen with documented mutagenic and neurotoxic effects and no known protective health effects, provides a means to examine whether protective patterns emerge in the absence of known health benefits. Using data from Nurses’ Health Study II (N=32,115), we implemented a negative exposure control design, comparing associations of adult low-level alcohol use and prenatal alcohol exposure, with cancer, cardiometabolic, respiratory, digestive, and other chronic health outcomes over 30 years of follow-up. Negative control analysis exhibited patterns of association closely paralleling those observed for adult alcohol use: modestly elevated odds of select malignancies, while most cardiometabolic, gastrointestinal, endocrine, and other chronic conditions showed inverse associations. E-values indicated that relatively modest unmeasured confounding could explain the observed associations. These findings are consistent with the hypothesis that a portion of the apparent protective associations often attributed to low-level adult alcohol use may reflect persistent bias rather than exclusively causal effects Collectively, these findings illustrate how exposures in the earliest stages of life can have enduring implications for development, behavior and health. Future studies can build on the approaches outlined in the presented work by incorporating more robust multi-modal measures of early-life exposures and novel study designs to further clarify the mechanisms linking early exposures to long-term health. Importantly, this work carries ethical considerations, including risks of stigma, blame, and misinterpretation of findings, underscoring the need for careful, context-sensitive communication that prioritizes supportive and non-punitive public health approaches. We believe this research provides both conceptual and methodological advances for life-course epidemiology and informs strategies to mitigate the impact of prenatal and early-life risk factors.