Collaborative Proposal: Case-Control Studies, New Directions and Applications

病例对照研究可能是统计学家在公共卫生和流行病学领域作出的最重要和最深远的贡献。本研究的主要目的是为病例对照研究在方法学和应用上做出一些新的贡献。特别是，PI提出了一种半参数贝叶斯方法，将纵向数据纳入使用惩罚样条法的病例对照分析中。这里需要新的贝叶斯函数数据分析工具。这些方法将被应用于基于生物标记物的筛查程序，并将提供前列腺癌与过去前列腺特异性抗原测量轨迹之间的相关性的关键评估。这项研究的第二个组成部分是对协变量非单调缺失的两阶段抽样产生的病例对照数据进行分析。这种设计在探索遗传和环境风险因素相互作用的病例对照研究中有许多潜在的应用。利用基因-基因和基因-环境独立性等假设增加了推理的复杂性。这些方法的动机是立即应用于一项基于大量人群的结直肠癌病例对照研究。这项提案的最后一个方面涉及一种统一的方法，该方法基于前瞻性和回溯性模型，对优势比参数提供等效推论。频率法和贝叶斯法都将被考虑。私人投资机构在这一领域有成功合作的记录，并希望在这些新方向上进一步推进/扩展他们的工作。目前有两个科学流派主导着临床医学和公共卫生：侧重于遗传学和新生物标记物发现的分子生物学方法，以及侧重于流行病学的定量方法。这些领域的发展共同为复杂疾病的病因、诊断、预后和治疗的研究做出了基础性贡献。尽管标准的不匹配病例对照研究设计仍然是最流行的流行病学工具之一，但医学科学和基因技术的惊人进步正在引发许多统计学家和流行病学家以前从未遇到过的复杂设计和分析问题。这一建议在于流行病学和统计学的新接口。了解复杂疾病的发病机制，为高危人群设计针对性的干预策略是本世纪科学研究的主要领域之一。目前的提议不仅仅是学术追求，而是为这一科学进程做出贡献的努力。

Case-control studies mark possibly the single most important and far-reaching contribution that statisticians have made in the domain of Public Health and Epidemiology. The prime objective of this research is to make some new contributions to case-control studies, both in methodology and in application. In particular, the PIs propose a semiparametric Bayesian method to incorporate longitudinal data in a case-control analysis using penalized splines. New Bayesian functional data-analytic tools are needed here. The methods will be applied to biomarker based screening procedures, and will provide a critical appraisal of the association between prostate cancer and the past trajectory of prostate-specific antigen measurements. A second component of the proposed research is the analysis of case-control data generated from two-phase sampling with non-monotone missingness in covariates. Such designs have many potential applications in case-control studies that explore interplay of genetic and environmental risk factors. Exploiting assumptions like gene-gene and gene-environment independence adds to the complexity of the inference. The methods are motivated by an immediate application to a large population based case-control study of colorectal cancer. The final aspect of this proposal deals with a united methodology which provides equivalent inference for odds ratio parameters based on prospective and retrospective models. Both frequentist and Bayesian methods will be considered. The PIs have a track record of successful collaboration in this domain, and want to advance/ extend their work further in these new directions. Two scientific streams are currently dominating clinical medicine and public health: the molecular biology approach with an emphasis on genetics and discovery of novel biomarkers, and the quantitative approach with an emphasis on epidemiology. The developments in these areas jointly are making fundamental contributions to the study of etiology, diagnosis, prognosis and treatment of complex diseases. Though the standard unmatched case-control study design still remains one of the most popular epidemiologic tools, phenomenal advancement of medical science and genetic technology is giving rise to many complex design and analysis issues which statisticians and epidemiologists have never confronted before. This proposal lies in that new interface of epidemiology and statistics. To understand the mechanism of complex diseases and to design targeted intervention strategies for high-risk individuals is one of the major areas of scientific research in this century. The current proposal is not a mere academic pursuit but an effort to contribute to this scientific process.