Predictive Modeling of Agrochemical Exposure and Adult Cancer Risk in U.S. Farmers: A Narrative Review
Alex Nnanyelugo Egbuchiem
*
Department of Environmental, Agricultural and Occupational Health, College of Public Health, University of Nebraska Medical Center, USA.
Antwi Edmond Owusu
Department of Agrobiotechnology, Agricultural-Technological Institute, RUDN University, 117198 Moscow, Russia and Department of Agricultural Extension, University of Ghana, Legon, Ghana.
Buabeng Victoria Fosua
Department of Agrobiotechnology, Agricultural-Technological Institute, RUDN University, 117198 Moscow, Russia and Department of Agriculture Engineering, University of Cape Coast, Ghana.
*Author to whom correspondence should be addressed.
Abstract
Agrochemical exposure is a pervasive occupational hazard in U.S. agriculture, where farmers experience repeated and often cumulative contact with pesticides, herbicides, and related chemical agents over extended working lifetimes. Epidemiologic studies have long raised concerns about elevated risks of certain adult cancers in farming populations, yet conventional analytic approaches frequently struggle to capture the complexity of exposure patterns, prolonged latency periods, and interacting occupational and environmental risk factors. In recent years, predictive modeling has emerged as a valuable framework for integrating heterogeneous exposure data and improving cancer risk estimation in agricultural settings. This narrative review synthesizes the literature on predictive modeling approaches used to examine relationships between agrochemical exposure and adult cancer risk among U.S. farmers. Emphasis is placed on exposure assessment strategies, including cumulative and time-varying metrics, the application of traditional statistical and machine learning models, and the cancer outcomes most commonly evaluated. Across studies, models that incorporate multidimensional exposure information consistently demonstrate greater predictive utility than those relying on simplified or binary exposure indicators. Machine learning approaches often achieve improved predictive performance in high-dimensional exposure contexts, although limitations related to interpretability, validation, and generalizability remain. Overall, the evidence suggests that predictive modeling can meaningfully advance understanding of agrochemical-related cancer risk when applied with methodological rigor and transparency. Strengthening exposure data integration, enhancing model explainability, and prioritizing external validation will be essential for translating predictive insights into effective occupational health surveillance, targeted cancer prevention strategies, and evidence-based policy decisions for U.S. farming populations.
Keywords: Agrochemical exposure, predictive modeling, cancer risk, farmers, occupational health, pesticides, agricultural epidemiology