Approaches to data analyses of clinical trials

There are two types of data analyses of randomized clinical trials (RCTs). The primary analyses are pre-specified in the protocol and the findings form the basis for recommendations and clinical decisions. They typically adhere to the intention-to-treat principle. Secondary analyses are supplemental and of various sorts. Although some may be pre-specified, many are not. We encourage the use of the rich sources of data from large RCTs for these secondary purposes. Depending on the kinds of secondary analyses, whether they are pre-specified, and whether intention-to-treat analysis is used, the results range from being quite conclusive to being hypothesis generating. In this article we answer four questions related to secondary analysis with emphasis on sharing of data primarily from NIH-sponsored trials: Who has access to this information? What questions can be asked? What are the requirements? What are the common challenges?