Discussion: 'Incomplete Data inClinical Studies: Analysis, Sensitivity,and SensitivityAnalysis' by GeertMolenberghs

摘要

The analysis of clinical trial data in the presence of missing endpoint data is a topic that can directly impact the safety and welfare of future patients because of the effect missing data techniques have on the perception of both the risks and benefits of a new product. There are many different perspectives on the selection of appropriate techniques for the analysis of confirmatory (phase 3) clinical trials with missing data. For example, philosophically there is not complete agreement on whether or not data after treatment discontinuation due to treatment toxicity should be considered as missing at all or simply as representing treatment failure. These ideas can be thought of in terms of the differences between the per-protocol (PP) analysis and the intention-to-treat (ITT) analysis. In effect, unlike other areas of application such as surveys, the missing data are often caused by the intervention itself rather than an unknown mechanism that leads to either a respondent declining to participate at all or refusing to answer specific questions. This suggests, as pointed out by Dr. Molenberghs, that great care must be taken in making inferences from clinical trials with missing data.