ABSTRACT
The previous chapters discussed methods in the design and conduct of clinical trials that minimize bias in the comparison of the experimental and control treatments. It is also possible, however, to introduce bias into both hypothesis tests and estimates of treatment effects by flawed analysis of the data. This chapter addresses issues regarding the choice of analysis population, missing data, subgroup and interaction analyses, and approaches to the analysis of multiple outcomes. Methods specific to interim analyses or intended for use after early termination are discussed in Chapter 10 and will not be covered here. Since the underlying goal in most of this material is the minimization of bias in the analysis, we begin with a brief discussion of bias.
