The Relationship Between Metal Exposure and HPV Infection: Evidence from Explainable Machine Learning Methods.

Abstract

HPV is a ubiquitous pathogen implicated in cervical and other cancers. Although vaccines are available, they do not encompass all subtypes. Meanwhile, metal exposure may elevate the risk of HPV infection and amplify its carcinogenic potential, but studies to further elucidate this relationship are insufficient. This study entailed a cross-sectional analysis utilizing data from the National Health and Nutrition Examination Survey (NHANES) 2007-2016. The study sample comprised 2765 women. Multivariate logistic regression was employed to examine the association between single metal exposure and HPV infection, weighted quantile sum (WQS) regression was utilized for assessing the mixed metal exposure effect, and the XGBoost + SHapley Additive exPlanations (SHAP) to evaluate the contribution of metal exposure in HPV infection. Multivariate logistic regression analysis indicated that elevated Co concentration was inversely associated with HPV infection (OR 0.891; 95% CI 0.814-0.975), while elevated Pb concentration correlated with an increased HPV infection (OR 1.176; 95% CI 1.074-1.287). Regression analysis of the WQS for mixed metal exposure suggested that the WQS index was potentially linked to an increased likelihood of HPV infection in the positive direction (OR 1.249; 95% CI 1.052-1.482), with no significant association observed in the negative direction (OR 0.852; 95% CI 0.713-1.017). SHAP analysis prioritized the importance of characteristics: number of sexual partners, marital status, poverty-to-income ratio (PIR), Co, Pb, and alcohol consumption. Exposure to Pb was associated with an increase in the incidence of HPV infection, whereas Co exposure demonstrated an inverse relationship. The composite exposure to multiple metals showed a positive association with the prevalence of HPV infection. These findings indicate that exposure to metals could potentially escalate the prevalence of HPV infection.