On the composite approach to dropouts in clinical trials

A major problem in the analysis of clinical trials is missing data from pat ients who drop out of the study before the predetermined schedule. Shih and Quan (1997) proposed a "composite" (or "pattern but not mix") approach to the problem, which makes simultaneous inference on the conditional mean of the completers and the probability of completion or dropping out of the tri al. Their method involves the combination of test statistics for the condit ional mean and dropout probability. In this paper, we justify the asymptoti c independence of the test statistics used in Shih and Quan.