GROUP SEQUENTIAL METHODS FOR LONGITUDINAL DATA

点击翻译按钮获取中文摘要

It is now commonplace for interim analyses to be performed in clinical trials. Such monitoring of the clinical trial data is essential to ensure that volunteer patients participating in the trial are treated ethically. Furthermore, the use of group sequential stopping rules can greatly increase the efficiency with which new treatments are adopted. The goal of this research is to broaden the spectrum of studies in which interim analyses can be used, with specific emphasis placed on methods appropriate for clinical trials which involve outcome measures based on measurements made over time. Such longitudinal studies pose unique problems in group sequential monitoring due to the limited timespan covered by the data at the earliest analyses. This variation in length of follow-up over the course of the study can have major impact on the robustness of the analytic methods used for determining treatment efficacy. Specific areas of research to be conducted under this project include: 1) development of testing strategies for survival data that are robust to crossing hazards; 2) development of testing strategies for longitudinal nonsurvival data that are robust to nonlinear treatment effects over time; 3) development of Bayesian methods for clinical trial monitoring that exhibit robustness properties similar to those developed for frequentist analysis of survival and other longitudinal data; 4) generalization of methods developed for multiple endpoints in order to address longitudinal data issues; 5) generalization of flexible monitoring schemes to accommodate robust longitudinal methods; and 6) development of computer software to implement these strategies in a clinical trials module.