ABSTRACT
The importance of establishing the reliability, validity, and responsiveness to change of quality-of-life instruments before they are recommended for use in clinical trial protocols has regularly been emphasized. However, there has been little discussion in the clinical trials literature of the statistical methods to be used in the analysis of trials incorporating quality-of-life end-points. Most commonly, these instruments are used to assess quality of life over time, that is, in a longitudinal manner. In general, there are two types of longitudinal studies. The first outcome of interest is the time until occurrence of an event. The second type of longitudinal study employs the numerical results from quality-of-life questionnaires as outcome measures, which is more commonly used in other disciplines. Univariate repeated measures analysis and the split-plot design are very similar. The two have almost identical analysis of variance models. Both univariate and multivariate approaches rests on the assumption that the population random error components follow a normal distribution.
