论文部分内容阅读
缺失数据是临床试验中常见但又不可避免的一个问题。缺失数据不仅会降低试验的把握度,还会给试验结果带来偏倚。因此,一方面可以在统计分析中采用合适的缺失数据处理方法,另一方面要特别注意尽可能预防缺失数据的产生。其中,缺失数据的预防应当是第一位的。从数据的角度来讲,首先,应在方案设计、数据采集和数据核查的各个阶段,采取合理措施提高受试者的依从性,减少不必要的数据缺失;其次,对于确认发生的数据缺失,应详细记录缺失数据产生的原因,这对于判定数据缺失的机制和选择合适的缺失数据处理方法 (例如,前一次观察数据向后结转、多重填补和重复测量数据混合效应模型等)具有非常重要的作用。
Missing data is a common but unavoidable problem in clinical trials. Missing data not only reduces the test’s power but also biases the test results. Therefore, on the one hand, appropriate methods of missing data processing can be used in statistical analysis, and on the other hand, special attention should be paid to prevent the generation of missing data as much as possible. Among them, the prevention of missing data should be the first. From the perspective of data, first of all, reasonable measures should be taken to improve the compliance of subjects and reduce unnecessary data loss in various stages of program design, data collection and data verification. Secondly, in order to confirm the absence of data, The reason for the missing data should be carefully documented, which is very important for determining the missing mechanism of the data and for choosing the appropriate missing data processing method (eg, the previous observation data is carried forward, multiple fill and repeated measurement data mixed effects model, etc.) Role.