Pesticides, cancer, and oxidative stress: an application of machine learning to NHANES data

Abstract

Abstract Background The large-scale application of pyrethroids and organophosphorus pesticides has great benefits for pest control. However, the increase of cancer incidence rate in recent years has also caused public concern about the health risks of pesticides. Hence, we utilized data from the National Health and Nutrition Examination Survey (NHANES) to assess the association and risk between pesticide exposure and several cancers, along with the comprehensive impact of oxidative stress. In this study, six cancers and six common pesticides were included to analyze their correlation and risk. And the levels of eight oxidative stress marks and two inflammatory markers were used for stratified analysis. Multiple logistic regression analysis was applied to estimate the odds ratio and 95% confidence intervals. Machine learning prediction models were established to evaluate the importance of different exposure factors. Results According to the data analyzed, each pesticide increased the risk of three to four out of six cancers on average. Iron, aspartate aminotransferase (AST), and gamma glutamyl transferase levels positively correlated with cancer risk in most cases of pesticide exposure. Except for demographic factors, factors such as AST, iron, and 3-phenoxybenzoic acid showed high contributions to the random forest model, which was consistent with our expectations. The receiver operating characteristic curve showed that the prediction model had sufficient accuracy (74.2%). Conclusion Our results indicated that specific pesticide exposure increased the risk of cancer, which may be mediated by various oxidative stress mechanisms. Additionally, some biochemical indicators have the potential to be screened for cancer prevention.

Keywords

• Pesticide
• Oxidative stress
• Cancer
• NHANES
• Machine learning
• Predict model