Analysis of Case-Control Follow-Up Studies

描述(申请人提供)：病例对照随访(CCFU)研究是一种混合设计，病例对照和队列或随访研究相结合。在这样的研究中，监测一个队列的早期或前哨事件，这些事件与一些感兴趣的结果密切相关(例如，乳腺癌的诊断)；获得关于所有发生前哨事件的受试者以及其余受试者样本的信息。然后对发生前哨事件的受试者(有时还有其他受试者)进行前瞻性跟踪，以了解感兴趣的事件(例如，死于乳腺癌)。 该设计已被用于癌症筛查和自然流产的有效性研究，但通常被误认为病例对照设计。这个项目的目的是增进对设计的理解，并改进从设计中得出的推论。这将通过1)制定描述这些研究的一般符号和理论；2)应用和开发适当的统计方法来分析这些研究产生的数据；以及3)比较各种分析这些研究的方法。 我们为这些研究确定了两种主要的一般统计方法：一种我们称之为直接方法，另一种我们称之为综合方法。我们建议详细阐述这些方法，然后考虑如何处理每种方法中可能出现的偏差。新的和熟悉的方法将通过模拟，通过应用于几项CCFU研究，以及通过从队列研究中抽样人工生成CCFU研究来评估。

DESCRIPTION (provided by applicant): The case-control follow-up (CCFU) study is a hybrid design, a combination of case control and cohort or follow-up studies. In such studies, a cohort is monitored for early or sentinels events, which are closely associated with some outcome of interest (e.g., diagnosis of breast cancer); information is obtained on all subjects developing the sentinel events, and on a sample of the remaining subjects. Subjects developing the sentinel event (and sometimes other subjects) are then followed prospectively for an event of interest (e.g., death from breast cancer). The design has been employed in studies of the efficacy of cancer screening and of spontaneous abortion but has generally been mistaken for a case-control design. The purpose of this project is to advance understanding of the design and improve inference derived from it. This will be done by 1) developing a general notation and theory for describing these studies; 2) applying and developing appropriate statistical methods for analyzing data arising from these studies; and 3) comparing various approaches for analyzing these studies. We have identified two main general statistical approaches for these studies: one we term the direct approach and the other we call the synthetic approach. We propose to elaborate these approaches, and then consider how to deal with biases that may arise in each of them. New and familiar methods will be assessed through simulation, through application to several CCFU studies, and through artificial generation of CCFU studies by sampling from cohort studies.