Methods for modeling air pollution effects on exhaled biomarkers

描述（由申请人提供） 环境健康研究的未来将包括更加重视暴露或潜在疾病过程的生物机制和生物标志物。环境健康研究中生物标志物的统计方法可以在这项研究中发挥重要作用，但迄今为止尚未受到足够的重视。此申请适用于 K22 过渡到独立环境健康研究职业发展奖。候选人提出了一个关于统计方法的 K22 项目，用于模拟空气污染对呼出气生物标志物的影响。这将是环境生物统计学独立研究计划的基础研究项目，重点是开发和应用统计方法，以全面了解环境暴露影响的机制。 大多数呼出物生物标志物的高级分析仍然相对较新，但迄今为止研究最好的呼出物生物标志物之一是呼出物一氧化氮 [FeNO]，这是一种指示气道炎症的生物标志物。该项目的重点是在空气污染健康影响研究的背景下对 FeNO 数据进行建模。 然而，对于呼出气体冷凝物和呼出挥发性有机化合物的成分，已经发现了类似的建模问题，并且随着该领域的进步，可能会发现更多问题。流行病学数据表明空气污染暴露与 FeNO 升高有关，导致儿童哮喘发病风险增加。一种以多种流速收集的新方法通过估计下呼吸系统中 NO 交换的两室确定性模型中的参数，可以将 FeNO 分为气道和肺泡来源。 现有的多流分析方法是针对来自少数参与者的良好控制的实验环境中的数据而开发的，但这些方法对大规模流行病学研究的直接适用性值得怀疑。多流 FeNO 数据收集已在多项流行病学研究中收集，其中最大的研究之一是南加州儿童健康研究 (n=1640)。多流 FeNO 流行病学研究中采用的现有统计方法使用数据效率低下，并且忽略了参数估计的不确定性。为了解决这些问题，候选人提出了利用模拟数据和儿童健康研究数据来完成 K22 项目的 3 个目标。目标 1 是开发、评估和应用非线性混合模型，以同时估计两室模型参数，并将这些参数与潜在的决定因素（例如细颗粒物和粗颗粒物空气污染）联系起来，同时适当地 考虑参数估计的不确定性。 目标 2 是开发、评估和应用贝叶斯马尔可夫链蒙特卡罗方法，直接估计简单稳健的双室模型和更复杂的具有轴向扩散的喇叭形模型的参数，然后比较这两个模型与儿童健康研究数据的拟合程度。 目标 3 是确定多个流量测量研究的最佳设计。 候选人是一名博士。她将利用 K22 奖提供的过渡期，从生物医学工程的角度获得生理系统数学建模方面的专业知识，这对于 a) 完成本研究项目和 b) 将拟议的研究项目扩展和推广到她的独立研究计划的下一阶段是必要的。该奖项的主要研究者资助者是 Frank D. Gilliland 博士，他是一位在环境健康研究方面拥有丰富经验的医师科学家，为该奖项组建的多学科顾问/合作者网络包括生物统计学（David Conti 博士）、环境生物统计学（Duncan C. Thomas 博士）、生物医学工程（David D'Argenio 博士）和计算生物学（Paul Marjoram 博士）方面的专家。

DESCRIPTION (provided by applicant) The future of environmental health research will include increased emphasis on biological mechanisms and biomarkers of exposure or latent disease processes. Statistical methods for biomarkers in environmental health research can play an important role in this research, but they have received insufficient attention to date. This application is for a K22 Transition to Independent Environmental Health Research Career Development Award. The candidate proposes a K22 project on statistical methods for modeling the effects of air pollution on exhaled breath biomarkers. This will be the foundational research project in an independent research program in environmental biostatistics, focused on developing and applying statistical methods to gain an integrated understanding of the mechanisms of environmental exposure effects. Advanced analytics of most exhaled biomarkers is still relatively new, but one of the best studied exhaled biomarkers to date is exhaled nitric oxide [FeNO], a biomarker indicative of airway inflammation. This project focuses on modeling FeNO data in the context of air pollution health effects research. However, similar modeling issues have already been identified for components of exhaled breath condensate and exhaled volatile organic compounds, and more will likely be discovered as the field advances. Epidemiologic data have linked air pollution exposure with elevated FeNO, leading to increased risk of incident asthma in children. A newer method of collection at multiple flow rates allows FeNO to be partitioned into airway and alveolar sources by estimating parameters in a two-compartment deterministic model of NO exchange in the lower respiratory system. Existing multiple flow analysis methods were developed for data from well-controlled experimental settings with a small number of participants, but direct applicability of these methods to large- scale epidemiologic studies is questionable. Multiple flow FeNO data collection has been collected in several epidemiologic studies, one of the largest of which is the Southern California Children's Health Study (n=1640). Existing statistical methods employed in epidemiologic studies with multiple flows FeNO use data inefficiently and ignore the uncertainty in the estimation of the parameters. To address these issues, the candidate proposes 3 aims for the K22 project to be completed using simulated data and data from the Children's Health Study. Aim 1 is to develop, evaluate, and apply non-linear mixed models to simultaneously estimate two- compartment model parameters and to relate these parameters to potential determinants (e.g., fine and coarse particulate matter air pollution) while appropriately accounting for the uncertainty in parameter estimation. Aim 2 is to develop, evaluate, and apply Bayesian Markov Chain Monte Carlo methods to directly estimate parameters from the simple and robust two-compartment model and a more complex trumpet-shaped model with axial diffusion, and then to compare the fit of both models to Children's Health Study data. Aim 3 is to determine optimal designs for multiple flow measurement studies. The candidate is a Ph.D. biostatistician, and she will use the bridge period afforded by the K22 award to gain expertise in mathematical modeling of physiologic systems, from a biomedical engineering perspective, which will be necessary: a) to complete this research project and b) to extend and generalize the proposed research project into the next phase of her independent research program. The principal investigator's sponsor for this award is Dr. Frank D. Gilliland, a physician-scientist with a strong track-record in environmental health research, and the multidisciplinary network of advisor/collaborators assembled for this award include experts in biostatistics (Dr. David Conti), environmental biostatistics (Dr. Duncan C. Thomas), biomedical engineering (Dr. David D'Argenio), and computational biology (Dr. Paul Marjoram).