Planning and ANalyzing OPTImal Clinical trials with Adaptive Design

点击翻译按钮获取中文摘要

Sample size calculation is an essential biometrical task when planning clinical trials. Adaptive sample size rules provide an attractive alternative to standard rules for fixed sample sizes. Here, the ongoing study is interrupted for one or several interim analysis(es) when the sample size is adaptive based on the data collected so far. While this general idea seems intuitive the question on the “best” choice of the sample size rule is much more complicated. During the ongoing DFG-project ORACLE, sensible and efficient rules for sample size recalculation in adaptive trial designs are developed. Among others, the software package adoptr in the statistical programming language R has been developed. It allows his user to compute the optimal design for an arbitrary scoring criterion. This is highly helpful for clinical and methodological research. On the one hand, adaptive designs provide choosing a sample size rule that cannot be further improved with respect to the selected scoring criterion. Furthermore, the user-specific choice of the objective criterion allows a statistical investigation of the resulting optimal designs.Within the proposed DFG project, it is planned to expand the scope of the research software adoptr and to increase its quality for the usage in clinical research. To this goal, the following working packages are anticipated:1. Further endpoints. Currently, adoptr only provides normally distributed test statistics. This limits a concrete application. For instance, planning clinical trials with binary endpoints (e.g., rate of success) or time-to-event endpoints (e.g., survival time) may be a sensible extension. An implementation of such scenarios in adoptr may increase its applicability. 2. Analysis. Besides planning, the final analysis of a trial is highly relevant. However, applying standard methods to analyze designs with fixed sample sizes is not adequate for adaptive designs. Therefore, it would be highly beneficial to provide the computation of point estimators, confidence intervals, and p-values besides sample size rules. Thereby, adoptr would cover the entire range of biostatistical tasks within clinical trials.3. Quality assurance. Particularly in the regulated environment of clinical trials, the usage of qualified software is of high importance. Since R is an open-source programming language, explicit quality assurance is necessary to allow its application in clinical research. First steps to achieve this goal have already been made. However, those are neither complete, nor do they fulfill regulatory requirements. Therefore, encompassing quality assurance of adoptr is anticipated.