Statistical methods for exhaled breath biomarkers in environmental epidemiology

项目总结/摘要 呼出的一氧化氮（FeNO）是气道炎症的非侵入性生物标志物，其应用于 哮喘的临床评估和环境流行病学。常规FeNO-在流速下评估 50 ml/s（FeNO 50）-是公认的，国际社会的评估和临床指南 解释。一种有前途但尚未完善的评估FeNO的方法是“多流NO分析”， 其使用在多个流速下测量的FeNO来估计“NO参数”，量化气道和 一氧化氮（NO）的肺泡来源，来自下呼吸道的确定性生理模型。 虽然这些生理模型在成年人中相当发达，但用于估计其 参数不是，特别是对儿童而言。大多数研究人员使用基于过度 过于简单的假设（例如，在稳定状态下的固定气道尺寸）和结果的线性化 非线性模型这些方法易于实现，但统计性能较差， 导致儿童气道较小。这些是在这一领域取得进展的主要障碍。 对于这个项目，我们将开发一个层次贝叶斯建模方法，使用马尔可夫实现 链蒙特卡罗方法。贝叶斯方法的一个关键优势是能够将 外部数据进入模型，产生精确的参数估计，并可能完善我们的理解 呼吸道炎症特别是，我们计划纳入两种类型的外部信息： 与气道大小有关，以及关于炎症的潜在决定因素的数据，如环境暴露。 通过将测量的表型数据显式地并入估计过程，得到的参数 估计将更好地调整每个人独特的生理影响。鉴于这些重大变化， 发生在整个青春期，我们的期望是，这种调整将证明特别有用， 将这些模型应用于儿童。该项目有三个具体目标： 目标1.开发方法来估计NO参数在一个修改的确定性2CM，个性化的 每个参与者的气道长度（目标1a）和/或分层内更真实的气道形状（目标1b） NO参数估计的贝叶斯框架。 目标2.开发方法来估计潜在决定因素的关联（例如，环境暴露） 与NO参数从确定性2CM使用分层贝叶斯框架的横截面 (Aim 2a）和纵向多流NO数据（Aim 2b）。 目标3.为了传播R包（Aim 3a）中的最终软件和直接在 浏览器使用新转换的JavaScript数值库（Aim 3b）。 这项工作的结果将是研究气道和肺泡NO在两个细化的统计方法。 环境流行病学和儿童和成人的临床环境。基于Web的开发 软件将增加广泛采用这些方法的可能性。

Project Summary/Abstract Exhaled nitric oxide (FeNO) is a non-invasive biomarker of airway inflammation, with applications in the clinical assessment of asthma and environmental epidemiology. Conventional FeNO—assessed at a flow rate of 50 ml/s (FeNO50)—is well-established, with international society guidelines for assessment and clinical interpretations. A promising but less well-established method of assessing FeNO is “multiple flow NO analysis”, which uses FeNO measured at multiple flow rates to estimate “NO parameters”, quantifying airway and alveolar sources of nitric oxide (NO), from deterministic physiological models of the lower respiratory tract. While these physiological models are quite well-developed in adults, the statistical methods for estimating their parameters are not, especially for children. Most researchers use estimation methods based on overly simplistic assumptions (e.g., a fixed airway size under a steady state) and linearizations of the resultant nonlinear models. These methods are easy to implement, but have poor statistical performance and do not account for the smaller airway size of children. These are major barriers to progress in this field. For this project, we will develop a hierarchal Bayesian modeling approach implemented using Markov chain Monte Carlo based-methods. A key advantage of a Bayesian approach is the ability to incorporate outside data into the model, producing refined parameter estimates and, potentially, refining our understanding of airway inflammation. In particular, we plan to incorporate two types of outside information: measurements related to airway size, and data on potential determinants of inflammation, such as environmental exposures. By incorporating measured phenotype data explicitly into the estimation process, the resulting parameter estimates will better adjust for the impact of each individual’s unique physiology. Given the major changes that occur throughout adolescence, our expectation is that this adjustment will prove particularly useful when applying these models to children. This project has three specific aims: Aim 1. To develop methods to estimate NO parameters in a modified deterministic 2CM that personalizes the airway length for each participant (Aim 1a) and/or more realistic airway shapes (Aim 1b) within a hierarchical Bayesian framework for NO parameter estimation. Aim 2. To develop methods to estimate associations of potential determinants (e.g., environmental exposures) with NO parameters from the deterministic 2CM using a hierarchical Bayesian framework for cross-sectional (Aim 2a) and longitudinal multiple flow NO data (Aim 2b). Aim 3. To disseminate resultant software in an R package (Aim 3a) and a web application, running directly in a browser using a newly converted JavaScript numerical library (Aim 3b). The outcome of this work will be refined statistical methods for studying airway and alveolar NO in both environmental epidemiology and clinical settings for children and adults. The development of web-based software will increase the likelihood of widespread adoption of these methods.