HEDGE Index: A Data-Driven Framework Overcoming the Limitations of TEFs for Prioritizing Polycyclic Aromatic Hydrocarbon Risks

Conventional
risk assessment of airborne polycyclic aromatic hydrocarbons
(PAHs), based on carcinogenicity-focused toxic equivalency factors
(TEFs), largely overlooks a broader spectrum of systemic health risks
from inhalation exposure. To address this limitation, we developed
the HEDGE index (Hazard and Exposure Index from Data-Driven Generation),
a data-driven framework integrating real-world exposure with multidimensional
hazard profiles. We classified 16 priority PAHs into three distinct
hazard subgroups (High-, Medium-, Low-Risk) using an ensemble clustering
approach on 16 toxicological and environmental fate descriptors. Subsequently,
an interpretable XGBoost machine learning model was trained to identify
the key structural drivers, such as density, underlying these classifications.
Epidemiological case study in Guangzhou, China, demonstrated that
the HEDGE index exhibited improved model fit for daily nonaccidental
mortality compared to the traditional toxic equivalency quantity (TEQ)
approach (ΔAIC = 6.2). An interquartile range increase in HEDGE-weighted
exposure corresponded to a 14.7% (95% CI: 12.5%, 17.0%) rise in total
mortality, a risk estimate substantially greater than that from TEQ.
The association was robust after adjusting for copollutants. By accurately
capturing a broader spectrum of systemic risks, the HEDGE framework
provides a more scientifically robust and health-protective tool for
regulatory prioritization.

传统上基于致癌性导向毒性当量因子（TEFs）的大气多环芳烃（PAHs）风险评估，很大程度上忽视了吸入暴露所带来的更广泛的全身健康风险。为弥补这一局限，本研究构建了HEDGE指数（数据驱动生成的危害与暴露指数，Hazard and Exposure Index from Data-Driven Generation），这是一种整合真实世界暴露数据与多维度危害特征的数据驱动分析框架。研究基于16种毒理学与环境归趋描述符，采用集成聚类方法，将16种优先管控多环芳烃划分为高、中、低风险三个明确的危害亚组。随后，本研究训练了可解释性极端梯度提升（XGBoost）机器学习模型，以识别支撑该分类结果的关键结构驱动因子（如密度）。针对中国广州开展的流行病学案例研究显示，相较于传统毒性当量（TEQ）方法，HEDGE指数对每日非意外死亡的模型拟合效果更优（ΔAIC=6.2）。HEDGE加权暴露的四分位距升高，对应总死亡率升高14.7%（95%置信区间：12.5%~17.0%），该风险评估结果显著高于毒性当量方法的估算值。在校正共存污染物后，该关联依然稳健。通过精准覆盖更广泛的全身健康风险维度，HEDGE框架为监管优先级制定提供了科学性更强、更利于健康保护的分析工具。