王慧敏, 付萌萌, 吴敏, 刘成娟, 杜娟娟, 聂继盛. 三种统计模型在职业工人多环芳烃暴露与认知水平关联研究中的应用[J]. 环境与职业医学, 2022, 39(5): 478-484. DOI: 10.11836/JEOM21428
引用本文: 王慧敏, 付萌萌, 吴敏, 刘成娟, 杜娟娟, 聂继盛. 三种统计模型在职业工人多环芳烃暴露与认知水平关联研究中的应用[J]. 环境与职业医学, 2022, 39(5): 478-484. DOI: 10.11836/JEOM21428
WANG Huimin, FU Mengmeng, WU Min, LIU Chengjuan, DU Juanjuan, NIE Jisheng. Application of three statistical models in association between polycyclic aromatic hydrocarbons exposure and cognitive level in workers[J]. Journal of Environmental and Occupational Medicine, 2022, 39(5): 478-484. DOI: 10.11836/JEOM21428
Citation: WANG Huimin, FU Mengmeng, WU Min, LIU Chengjuan, DU Juanjuan, NIE Jisheng. Application of three statistical models in association between polycyclic aromatic hydrocarbons exposure and cognitive level in workers[J]. Journal of Environmental and Occupational Medicine, 2022, 39(5): 478-484. DOI: 10.11836/JEOM21428

三种统计模型在职业工人多环芳烃暴露与认知水平关联研究中的应用

Application of three statistical models in association between polycyclic aromatic hydrocarbons exposure and cognitive level in workers

  • 摘要: 背景 多环芳烃(PAHs)作为一类复杂的有机污染物,暴露呈现多种羟基代谢物共同暴露的特征,大部分研究分析了每种PAHs代谢物的独立作用,并将其他代谢物作为协变量调整,未考虑到相互作用,其毒性效应的研究需要合理的统计方法。

    目的 探讨logistic回归、加权分位数之和(WQS)回归和贝叶斯核机器回归(BKMR)在评估外源化学物混合暴露与健康结局相关分析中的适用性,比较三个模型的优点和局限性,提出化学物混合暴露健康效应评价的分析策略以应用于PAHs暴露与认知水平关联的分析中。

    方法 根据相应的纳排标准,收集山西省某焦化厂和水处理厂1 051名参加职工健康体检工人的尿液,使用超高效液相色谱串联质谱法(HPLC-MS/MS)检测11种PAHs羟基代谢物单羟基多环芳烃浓度,使用蒙特利尔认知评估量表评定其认知水平,获取轻度认知功能障碍(MCI)检出率。分别应用logistic回归、WQS回归和BKMR分析二者之间的联系。

    结果 研究对象MCI检出率为21.7%(228/1 051)。2-羟基萘(2-OHNAP)的浓度在11种代谢物中最高,中位浓度为0.30 μg·L−1,其次是9-羟基菲(9-OHPHE)(0.26 μg·L−1)。MCI组与认知正常组2-OHNAP、1-羟基萘(1-OHNAP)、2-羟基芴(2-OHFLU)、9-OHPHE、1-羟基菲(1-OHPHE)和1-羟基芘(1-OHPYR)浓度差异有统计学意义(均P<0.05)。logistic回归显示:2-OHNAP和2-OHPHE与MCI检出有关,2-OHNAP与2-OHPHE浓度每增加10倍,MCI检出的OR(95%CI)分别为1.28(1.01~1.67)和1.27(1.00~1.72)。WQS回归显示:WQS指数与MCI检出率呈正相关(OR=1.37,95%CI:1.10~1.72)。BKMR分析显示:当所有代谢物浓度均处于或高于其第30百分位数时,总体效应具有统计学意义;与均处于其中位浓度相比,当所有PAHs代谢物暴露水平均处于第75百分位数时,MCI检出风险比例上升6%。

    结论 基于三种模型的结果,认为2-OHNAP和2-OHPHE是与认知水平相关的最重要因素。推荐使用传统的logistic和WQS或BKMR任一种相结合的方式进行PAHs与MCI关联研究。

     

    Abstract: Background As a complex organic pollutant, polycyclic aromatic hydrocarbons (PAHs) exposure shares the common exposure characteristics of multiple hydroxyl metabolites. Most studies have analyzed independent effect of each PAHs metabolite and have adjusted for the potential confounding effects induced by other metabolites concomitantly, without considering possible interactions among them. Proper statistical methods are needed to study their toxic effects.

    Objective To explore the applicability of logistic regression, weighted quantile sum (WQS) regression, and Bayesian kernel machine regression (BKMR) in evaluating the correlation between mixed exposures to exogenous chemicals and health outcomes, compare the advantages and limitations of the three models, and propose analytical strategies for evaluating the health effects of mixed chemical exposure for application in the analysis of the association between PAHs exposure and cognition.

    Methods Urine samples were collected of workers from a coke oven plant and a water treatment plant in Shanxi Province, who participated in their routine employee healthexamination. Mono-hydroxylated PAHs were detected by high-performance liquid chromatography with tandem mass spectrometry (HPLC-MS/MS), cognitive function was assessed using the Montreal Cognitive Assessment (MoCA). A cut-off value of MoCA less than 26 was considered mild cognitive impairment (MCI). According to a predetermined inclusion and exclusion criteria, 1 051 cases were included in the final data analysis. Logistic regression, WQS regression, and BKMR were used to analyze the relationship between PAHs metabolites and MCI.

    Results The prevalence rate of reporting MCI among the 1 051 workers was 21.7% (228/1 051). The concentration of 2-hydroxynathalene (2-OHNAP) was the highest among the 11 PAHs metabolites with a median concentration of 0.30 μg·L−1, followed by 9-hydroxyphenanthrene (9-OHPHE) (0.26 μg·L−1). There were significant differences between the two groups in 2-OHNAP, 1-hydroxynaphthalene (1-OHNAP), 2-hydroxyfluorene (2-OHFLU), 9-OHPHE, 1-hydroxyphenanthrene (1-OHPHE), and 1-hydroxypyrene (1-OHPYR) (all Ps<0.05). In the logistic regression, 2-OHNAP and 2-OHPHE were associated with MCI, and theOR (95%CI) for reporting MCI was 1.28 (1.01-1.67) and 1.27 (1.00-1.72) for each 10-fold increase in 2-OHNAP and 2-OHPHE concentrations, respectively. In the WQS regression analysis, the WQS index was positively correlated with the prevalence rate of reporting MCI (OR=1.37, 95%CI: 1.10-1.72). In the BKMR analysis, compared with the median exposure levels of all chemicals, the overall effect was statistically significant when all PAHs metabolites concentrations were at or above their 30th percentile; when all exposures were at the 75th percentile, the risk of reporting MCI increased by 6%.

    Conclusion Based on the results of these three models, 2-OHNAP and 2-OHPHE are the most important factors related to cognitive. It is recommended to use a combination of traditional logistic regression and either WQS or BKMR to study the association between PAHs and MCI.

     

/

返回文章
返回