Background

Environmental exposure to toxic metals is an important risk factor to human health. Traditional methods have examined associations between a health endpoint and exposure to heavy metals by either univariate or multiple regression. In the setting of ubiquitous heterogeneous environmental exposures, statistical methods that incorporate mixed exposures are increasingly relevant and may provide new insight into the association between metal exposure and important cardiovascular, renal and respiratory outcomes.

Objective

The objective of this study was to classify the population of National Health and Nutrition Examination Survey (NHANES) into different exposure subgroups using modern unsupervised clustering methods based on lead, cadmium, mercury, and arsenic measured in urine or whole blood, and to assess the association between the identified exposure groups and twelve important health endpoints.

Methods

We analyzed a sub-cohort of 9662 subjects participating in the 6 cycles (2003–2004 to 2013–2014) of NHANES study. The urine levels of 3 heavy metals (total arsenic, lead, cadmium) and blood levels of 3 heavy metals (lead, cadmium and mercury) were analyzed using a two-step approach. In the first step, we stratified the population into subgroups using unsupervised clustering (k-medoids) based on levels of metals either in urine or in blood. Then, we examine the association between 12 health endpoints and identified exposure subgroups while controlling for age, sex, race/ethnicity, education, smoking status, BMI, and urinary creatinine.

Results

The k-medoids algorithm clustered NHANES population into 2 groups based on either blood or urinary levels of heavy metals. The concentrations of all the three heavy metals were significantly different between the identified groups in blood (p < 2.2e−16) or in urine (p = 0). The group with higher concentrations was defined as the “high-exposure” group, while the group with lower concentrations was defined as “low-exposure” group. Association analysis with health outcomes suggested that the high-exposure group according to either blood or urinary metal levels had significantly higher total mortality (1.63–1.64 times higher, p < 0.0001), mortality caused by malignant neoplasms (2.05–2.62 times higher, p < 0.0002), Gamma-glutamyl transferase (GGT) (1.03–1.05 times higher, p < 0.0001). In addition, the high-exposure group based on blood levels was also significantly associated with SBP, death related to hypertension, heart disease and chronic lower respiratory disease, while the high-exposure group based on urinary concentrations had higher mortality related to nephritis.

Conclusions

We proposed an unsupervised clustering method to stratify the population into high- and low-exposure groups based on the co-exposure of heavy metals. The high-exposure groups, characterized by higher metal concentrations, had significant higher GGT, SBP, DBP, and mortality rates suggesting the detrimental effects of exposure to these heavy metals. The stratification of the NHANES population based on exposure patterns provides an informative method to study the impact of metal exposures on health outcomes.

中文翻译：

---

基于铅，镉，汞和砷的暴露模式及其与心血管，肾脏和呼吸系统结局的关系，NHANES 2009-2014年人口分层

背景

在环境中接触有毒金属是危害人类健康的重要风险因素。传统方法通过单变量或多元回归研究了健康终点与重金属暴露之间的关联。在普遍存在的异质环境暴露中，结合混合暴露的统计方法越来越重要，并且可能为金属暴露与重要的心血管，肾脏和呼吸道结局之间的关联提供新的见解。

目的

这项研究的目的是使用基于尿液或全血中铅，镉，汞和砷的现代无监督聚类方法，将国家健康和营养检查调查（NHANES）的人群分为不同的暴露亚组，并评估确定的暴露人群与十二个重要健康终点之间的关联。

方法

我们分析了参加6个周期（2003–2004年至2013–2014年）的NHANES研究的9662名受试者的亚队列。使用两步方法分析了三种重金属（总砷，铅，镉）的尿液水平和三种重金属（铅，镉和汞）的血液水平。第一步，我们根据尿液或血液中的金属含量，使用无监督聚类（k-medoids）将人群分为亚组。然后，我们在控制年龄，性别，种族/族裔，教育程度，吸烟状况，BMI和尿肌酐的同时，检查了12个健康终点与已确定暴露亚组之间的关联。

结果

k-medoids算法根据血液或尿液中的重金属含量将NHANES人群分为两组。在血液（p <2.2e-16）或尿液（p = 0）中，确定的组之间所有三种重金属的浓度均存在显着差异。高浓度组定义为“高暴露”组，低浓度组定义为“低暴露”组。与健康结局的关联分析表明，根据血液或尿液金属水平的高暴露组，总死亡率显着较高（高1.63–1.64倍，p <0.0001），恶性肿瘤所致死亡率（高2.05–2.62倍，p <0.0002），γ-谷氨酰转移酶（GGT）（高1.03-1.05倍，p <0.0001）。此外，

结论

我们提出了一种无监督聚类方法，根据重金属的共同暴露将种群分为高暴露和低暴露组。以高金属浓度为特征的高暴露人群的GGT，SBP，DBP和死亡率显着较高，表明暴露于这些重金属的有害影响。根据接触方式对NHANES人群进行分层，为研究金属接触对健康结果的影响提供了一种有用的方法。