Calculating sample sizes

Having determined the appropriate error rates, baseline results and target difference to detect, you have all the basic information you need to calculate a sample size, or perhaps more usually to consult appropriate tables or software packages (or statisticians) to do the calculation for you. Readers who are happy to omit the technical details can jump to Section 5.7, which gives some guidance on the use of tables and software and some pitfalls to avoid. It is though useful to have an understanding of the underlying principles for the most common types of outcome measure. We therefore review the basic principles for three common types of data: continuous (normally distributed), binary (i.e. simple proportions) and survival data. In fact, the underlying basis of most sample size calculations (including that for some non-randomized phase II trials) is perhaps best illustrated in the case of continuous data. Section 5.4.1 therefore explains the background in some detail, and sample size issues for phase II studies are described in Section 5.4.5.

0 0

Post a comment