Analysing change in clinical trials using quasi-likelihoods

In clinical trials, investigations focus upon whether a treatment affects a measured outcome. Data often collected include pre- and post-treatment measurements on each patient and an analysis of the change in the outcome is typically performed to determine treatment efficacy. Absolute change and relative change are frequently selected as the outcome. In selecting from these two measures, the analyst makes implicit assumptions regarding the mean and variance-mean relationship of the data. Some have provided ad hoc guidelines for selecting between the two measures. We present a more rigorous means of investigating change using quasi-likelihoods. We show that both absolute change and relative change are special cases of the specified quasi-likelihood model. A cystic fibrosis example is provided.