ABSTRACT
The results of a well-designed and executed clinical trial should be evident from a couple of summary statistics-but stopping at those summaries in a manuscript is rare. It is reasonable to want to see if more can be learned from a long, expensive effort. The temptation to over-interpret secondary analyses can be irresistible. Exploring data and analyzing different endpoints does sometimes lead to new insights; instead, far too often, it leads to implausible hypotheses and faulty conclusions. In this chapter we discuss problems with some common types of analyses which have been used to try to draw treatment conclusions beyond those supported by the study design. In Chapter 10 we discuss methods for exploratory data analysis.
