Semiparametric mixture methods for the analysis of correlated counting and survival processes with applications to study design

本提案为临床研究的设计开发了有效的方法，用于监测复发事件，例如皮肤或其他癌症复发或癫痫发作。有时不可能或不需要对经历复发事件的个体进行持续监测。在某些情况下，从伦理上讲，根据护理需要，患者不需要接受这种监测，同时，用于持续或过度频繁监测的资金可以得到更好的分配；另一种情况是，事件可能发生得太频繁，以至于无法记录其确切的发生时间。监测复发性事件的研究设计面临的统计挑战在于确定最佳随访时间和所需随访时间，考虑到人类受试者研究特征的众多差异，例如，个体对治疗的反应存在很大的异质性和数据缺失。所提出的研究所采用的方法对产生递归的基本过程提供了最小的假设，使它们具有高度的通用性。

This proposal develops efficient methods for the design of clinical studies that monitor recurrent events, for example, skin or other cancer recurrences or epileptic seizures. Sometimes continuous monitoring of individuals experiencing recurrent events is not possible or imperative. There may be situations where, ethically, patients need not be subjected to such monitoring based on the need for care, and at the same time, funds spent in continuous or overly-frequent monitoring could be better allocated; alternatively, events may occur too frequently for their exact times of occurrence to be recorded. Statistical challenges to the design of studies that monitor recurrent events lie in determining optimal times for followup and the number of followup times required, accounting for the numerous distinctions that characterize studies of human subjects, for example, substantial heterogeneity in responses to treatments amongst individuals and missingness of data. The methods employed by the proposed research afford minimal assumptions on the underlying process that generates recurrences, making them highly generalizable.