The Study Of Interim Analysis And Relative Problems In Clinical Trials

In clinical trials, due to ethical and economic reasons, it is often required to take interim "look" at the accumulating data to monitor the treatment effect. When statistical analyses are performed repeatedly, some adjustment to the boundary of each interim analysis has to be made to maintain the type I error probability at a pre-specified level. Group sequential designs are known to be practical and suitable interim analysis methods. In this thesis, based on some retrospective studies on interim analysis methods, we focus on the group sequential type I error spending function interim analysis method and its uses in clinical trials. The multiple-endpoints and multi-armed interim analysis are explored too. The main problems solved are as follows:1. The definition > history and uses of group sequential procedures are reviewed briefly The rationale of group sequential type I error spending function method is described, with its applicability to a variety of commonly used statistical data, including survival data and longitudinal data is detailed.2. Several commonly used a spending functions are given, several stochastic processes(E process, S process, Z process)are retrospectively described, the role of Brownian motion process is emphasized, the relationof the monitoring of clinical trials to the theory of group sequential analysis is explicated, based on which the boundaries are constructed, through integration method.3. Information and information fraction: The information has two uses in interim analysis, one is for use with the type I error spending function, the other is for determining the correlation of successive test statistics. The information fraction is the role of the a spending function method. The information in a clinical trial depends on the type of data collected, we examine four situations: quantitative data, categorical data, survival data and repeated measures, for the latter two, the estimator of information fraction is often used, at the same time, two information scales are used too.4. Group sequential statistics: The formulations of statistics for different type of clinical data are given. The sequence of the standard test statistics follows the same canonical joint distribution, with the information levels {/,,...,1K }for the parameter #, with this distribution the calculation of the statistics can be simplified.5. The estimation of sample size and power: The estimation of sample size is a critical problem in clinical trials. Group sequential score statistics has the same discrete construction as the Brownian motion process, with the drift parameter? the formulations for calculating sample size in different clinical trial data are given. Meanwhile, we can use the sample size ratio R to estimate the sample size for group sequential design. The diift parameter î–§nd sample size ratio R for different a spending functions are given, which is convenient for using.6. Multiple endpoints and multi-armed interim analysis: In clinicaltrials, we often refer to the other multiple problems: the multiple endpoints and multi-armed analysis. In this paper, we relate above two multiple problems to interim analysis, different methods for multiple endpoints interim and multi-armed interim analyses are discussed, with their boundaries are given.7. All the data collected in this paper were analyzed by SAS software system to computer the statistics, while the sample size, boundaries and nominal significance levels of interim analysis were computed by special software for interim analysis. The group sequential type I error spending function interim analysis has extensive practical uses in clinical trials, the calculation of interim analysis boundaries is simplified by the special software, which supply a practical and useful tool for the application and extension of interim analysis in clinical trials.