Looking inside the black box – hybrid epidemiology approaches to identify causal inferences

Abstract Plastic chemicals are numerous and ubiquitous in modern life and pose significant risks to human health. Observational epidemiological studies have been instrumental in identifying consistent and statistically significant associations between exposure to certain chemicals and adverse health outcomes. However, these studies often fail to establish causality due to the complexity of real-world chemical mixtures, confounding factors, reverse causation, and study designs that lack measures reflecting underlying genetic and cellular mechanisms indicating causal pathways to harm. Addressing these limitations requires moving beyond traditional ‘black-box’ epidemiology, which mainly focuses on the strength of associations. We propose adopting hybrid epidemiological methodologies that incorporate genetic susceptibility and molecular mechanisms to uncover biological pathways, combined with machine learning and statistical analysis of chemical mixtures, to strengthen the causal evidence linking exposure to harm. By integrating observational multi-omics data with experimental and mechanistic models, hybrid epidemiology offers a transformative path to improve causal evidence and public health interventions. In addition, machine learning and statistical methods provide a more nuanced understanding of the health effects of exposures to plastic chemical mixtures, facilitating the identification of interactions within chemical mixtures and the influence of biological pathways. This paradigm shift is critical addressing the complex challenges of plastic exposure and protecting human health.