Per- and polyfluoroalkyl substances (PFAS) have infiltrated countless everyday products, raising concerns about potential effects on human health, specifically on the cardiovascular system and the development of abdominal aortic calcification (AAC). However, our understanding of this relationship is still limited. This study aims to investigate the effects of PFAS on AAC using machine learning algorithms. Leveraging the power of machine learning technique, extreme gradient boosting (XGBoost), we assessed the relationship between PFAS exposure and AAC risk. We focused on three PFAS compounds, perfluorodecanoic acid (PFDeA), perfluorohexane sulfonic acid (PFHxS), and perfluorononanoic acid (PFNA) through multiple logistic regression, restricted cubic spline (RCS), and quantile g-computation (QGC) models. To get more insight into the underlying mechanisms, mediation analyses are used to investigate the potential mediating role of fatty acids and blood cell fractions in AAC. Our findings indicate that elevated serum levels of PFHxS and PFDeA are associated with the increased risk of AAC. The QGC analyses underscore the overall positive association between the PFAS mixture and AAC risk, with PFHxS carrying the greatest weight, followed by PFDeA. The RCS analyses reveal a dose-dependent increase between serum PFHxS concentration and AAC risk in an inverted V-shape way. Moreover, age and PFHxS exposure are identified as the primary factors contributing to abdominal aortic calcification risk in SHapley Additive exPlanation (SHAP) summary plot combined with XGBoost technique. Although PFAS significantly change the profile of fatty acids, we do not find any mediating roles of them in AAC. Despite strong associations between PFAS exposure and hematological indicators, our analysis does not find evidence that these indicators mediate the development of AAC. In summary, our study highlights the detrimental impact of PFAS on abdominal aortic health and emphasizes the need for further research to understand the underlying mechanisms involved.

中文翻译：

---

全氟烷基和多氟烷基化学物质与中老年人腹主动脉钙化之间的关联

全氟烷基物质和多氟烷基物质 (PFAS) 已渗透到无数的日常产品中，引发了人们对对人类健康的潜在影响的担忧，特别是对心血管系统和腹主动脉钙化 (AAC) 的发展。然而，我们对这种关系的理解仍然有限。本研究旨在使用机器学习算法研究 PFAS 对 AAC 的影响。利用机器学习技术、极限梯度增强 (XGBoost) 的力量，我们评估了 PFAS 暴露与 AAC 风险之间的关系。我们通过多元逻辑回归、限制三次样条 (RCS) 和分位数 g 计算 (QGC) 模型重点研究了三种 PFAS 化合物：全氟癸酸 (PFDeA)、全氟己烷磺酸 (PFHxS) 和全氟壬酸 (PFNA)。为了更深入地了解潜在机制，我们使用中介分析来研究脂肪酸和血细胞组分在 AAC 中的潜在中介作用。我们的研究结果表明，PFHxS 和 PFDeA 血清水平升高与 AAC 风险增加相关。 QGC 分析强调了 PFAS 混合物与 AAC 风险之间的总体正相关关系，其中 PFHxS 权重最大，其次是 PFDeA。 RCS 分析显示，血清 PFHxS 浓度与 AAC 风险之间呈倒 V 形，呈剂量依赖性增加。此外，在 SHapley Additive exPlanation (SHAP) 摘要图结合 XGBoost 技术中，年龄和 PFHxS 暴露被确定为导致腹主动脉钙化风险的主要因素。尽管 PFAS 显着改变了脂肪酸的分布，但我们没有发现它们在 AAC 中具有任何调节作用。尽管 PFAS 暴露与血液学指标之间存在密切关联，但我们的分析并未发现这些指标介导 AAC 发展的证据。总之，我们的研究强调了 PFAS 对腹主动脉健康的有害影响，并强调需要进一步研究以了解所涉及的潜在机制。