Bayesian Risk Modeling of Racial-spatial Interactions Among Childhood Cancer Hist

项目描述（由申请人提供）：项目概述/摘要。儿童癌症的病因鉴定一直难以捉摸，只有两个例外。地理上的风险，比如靠近污染源，往往会涉及其中，而且在风险方面的种族差异也很大。解释地理暴露和个体特征，如种族，如何在多种癌症类型中起作用和相互作用是目前的僵局。为了克服这一障碍，我们需要一种风险建模方法，该方法既包括地理层面的风险，也包括个人层面的风险，并且可以促进癌症类型的客观汇集或分析。最近的两个模型发展提供了一个很有希望的机会。首先，现在有可能利用分层模型从地理风险因素中分离出个体风险因素。其次，现在单一模型可以包含多种结果的风险，如癌症组织类型。这一应用要求开发一个儿童癌症风险模型，利用这两个最近的进展。本应用程序的目的是建立一种既完全分层又包括多种儿童癌症组织型的建模方法。中心假设是种族，作为个体水平风险因素中最重要的例子，与地理水平的暴露相结合，导致多种儿童癌症组织类型。这一假设将通过两个具体目标来验证：具体目标1是建立儿童癌症风险的小规模空间模式模型。这一具体目标的结果将是能够展示风险模式和风险集群的详细风险表面。这些建模结果将支持针对高风险地点的集中调查，包括集中风险建模。具体目标2是确定儿童癌症的空间特异性种族风险。这一具体目标将利用最近开发的空间变化风险的风险模型。这一特定目标的结果将是多路径分析，并将阐明儿童癌症的因果途径。这个应用程序是创新的，因为它将利用两个新的发展，全条件分层建模和多变量建模。

DESCRIPTION (provided by applicant): Project Summary/Abstract. Identification of the causes of childhood cancer has been elusive with two exceptions. Geographic risks, like proximity to pollution sources, are often implicated and racial differences in risk have been profound. Interpreting how geographic exposures and individual characteristics, like race, act and interact among multiple cancer types is the current impasse. To overcome this obstacle, we need a risk modeling approach that incorporates both geographic-level and individual-level risks and can facilitate the objective pooling or parsing of cancer types. Two recent modeling developments present a promising opportunity. First, it is now possible to disentangle individual risk factors from geographic risk factors utilizing hierarchical modeling. Second, it is now possible for a single model to incorporate risks for multiple outcomes, like cancer histotypes. This application calls for developing a childhood cancer risk model that capitalizes on these two recent advances. The objective of this application is to establish a modeling approach that is both fully hierarchical and inclusive of multiple childhood cancer histotypes. The central hypothesis is that race, as the most important example of individual-level risk factors, and geographic-level exposures combine to cause multiple childhood cancer histotypes. This hypothesis will be tested by two specific aims: Specific Aim 1 is to model the small-scale spatial patterns of childhood cancer risks. The results of this specific aim will be detailed risk surfaces capable of demonstrating risk patterns and risk clusters. These modeling results will enable focused investigation, including focused risk modeling, for high-risk locations. Specific Aim 2 is to identify the spatially specific racial risks for childhood cancer. This specific aim will utilize recently developed risk modeling for spatially varying risks. The results of this specific aim will be multiple path analyses and will clarify causal pathways for childhood cancer. This application is innovative because it will exploit two new developments, fully conditional hierarchical modeling and multivariate modeling. PUBLIC HEALTH RELEVANCE: Project Narrative. The proposed research is significant because what is needed most to advance our understanding of the causes of childhood cancer is a risk model that incorporates both geographic risks and individual risks and applies to the multiple types of childhood cancer. The study will validate a risk modeling approach that will enable the elucidation of the causal mechanisms for childhood cancer.