Endocrine disruptors and insulin resistance: quantifying impacts with a novel exposure burden score

项目总结 超过3400万美国成年人患有糖尿病，这是一种与高发病率和死亡率相关的慢性病。 最近，接触内分泌干扰物(EDCs)，无论是持久性的还是非持久性的，一直是 被认为是胰岛素抵抗和糖尿病风险的潜在贡献者。有证据表明，因为 内分泌细胞影响相似的代谢途径，内分泌细胞的整体影响可能大于个体影响 化学物质对代谢结果的影响。然而，研究人员缺乏一个简单的总结指标来量化暴露 给EDC带来了负担。由于存在大量的EDC，汇总EDC负担指标可能有助于 风险评估、生物监测，并用于糖尿病风险预测模型。在本R03中，我们引入了一个 灵活类别的项目反应理论(IRT)模型，以量化EDC负担分数。我们估计EDC 负担作为一个潜在变量，捕捉了暴露于内分泌干扰物的总体，对两者都进行了测量 和无法测量的化学物质。这一综合指标旨在捕捉内分泌和 其他生理器官系统受到内分泌细胞的干扰或负担。据我们所知，我们是第一个 IRT模型在环境暴露数据中的应用。项目反应理论是一大套很好的 建立了教育测试中常用的潜在变量模型(例如，高考成绩 考试)。这些模型的应用填补了目前混合物研究中缺失的重要空白：1)它们 应对不同化学品组随时间或在 不同的队列，2)它们允许我们将不经常检测到的化学物质包括在负担得分计算中 不需要推卸责任，3)他们没有监督，所以负担分数将是相同的，无论 健康结果，这是生物监测所需的。为了证明这种方法的可行性，我们 将利用国家健康和营养检查调查(NHANES)多年的数据来 获取有关美国成年人内分泌干扰物和胰岛素抵抗的代表性数据。在这些NHANES上 调查年份，测量了不同组的EDC，使用了所有年份的一些常见化学品，这些化学品 必须进行数据协调，以充分利用所有测量的化学数据。在目标1中，我们开发了三个 全氟辛烷磺酸、邻苯二甲酸盐和苯酚/对羟基苯甲酸酯的单独负担分值，以及总体EDC负担 得分。我们将确定不同社会经济群体在幼儿发展负担方面是否存在差异(例如， 性别、种族/族裔、社会经济地位)。在目标2中，我们将调查EDC负担分数是否 与胰岛素抵抗的稳态模型评估所测量的胰岛素抵抗有关。 我们将把我们的发现与其他量化内分泌干扰物混合物的方法进行比较，例如 成分分析和摩尔总和，以及监督混合方法。我们将创建一个R 包、交互式Web应用程序和教程，使环境健康和糖尿病研究人员能够 为他们的研究计算化学暴露负担得分。

PROJECT SUMMARY Over 34 million US adults live with diabetes, a chronic disease associated with high morbidity and mortality. Recently, exposure to endocrine disrupting chemicals (EDCs), both persistent and non-persistent, has been recognized as a potential contributor to insulin resistance and diabetes risk. Evidence suggests that because EDCs affect similar metabolic pathways, the overall effect of EDCs may be greater than effects of individual chemicals on metabolic outcomes. However, researchers lack a simple summary index to quantify exposure burden to EDCs. Because of the large number of EDCs that exist, a summary EDC burden metric could aid in risk assessment, biomonitoring, and be used in diabetes risk prediction models. In this R03, we introduce a flexible class of item response theory (IRT) models to quantify an EDC burden score. We estimate EDC burden as a latent variable that captures the totality of exposures to endocrine disruptors, to both measured and unmeasured chemicals. This summary metric aims to capture the total degree to which the endocrine and other physiological organ systems are perturbed, or burdened, by EDCs. To our knowledge, ours is the first application of IRT models to environmental exposures data. Item response theory is a large set of well- established latent variable models that are commonly used in educational testing (e.g. scoring college entrance exams). Application of these models fill important gaps that are currently missing in mixtures research: 1) They address data harmonization challenges in which different sets of chemicals are measured over time or in different cohorts, 2) They allow us to include infrequently detected chemicals in the burden score calculation without the need for imputation, 3) They are unsupervised so the burden scores will be the same no matter the health outcome, which is needed for biomonitoring purposes. To demonstrate feasibility of this approach, we will leverage multiple years of data from the National Health and Nutrition Examination Survey (NHANES) to gain representative data on endocrine disruptors and insulin resistance for US adults. Over these NHANES survey years, different sets of EDCs were measured, with some common chemicals across all years, which necessitates data harmonization to make full use of all measured chemical data. In Aim 1, we develop three separate burden subscores for PFAS, phthalates, and phenols/parabens, as well as an overall EDC burden score. We will determine if there are disparities in EDC burden for different socio-economic groups (e.g. age, sex, race/ethnicity, socio-economic status). In Aim 2, we will investigate whether EDC burden scores are associated with insulin resistance as measured by the Homeostatic Model Assessment of Insulin Resistance. We will compare our findings with other methods to quantify endocrine disruptor mixtures, such as principal components analysis and molar sum, as well as supervised mixtures approaches. We will create an R package, interactive web application and tutorial to allow environmental health and diabetes researchers to calculate chemical exposure burden scores for their research.