ABSTRACT
Analysis of change from baseline is probably the most popular study design in medical and epidemiological research. If the study subjects are randomly allocated to different treatment groups, such as in randomised controlled trials, the comparisons of changes amongst groups are relatively straightforward; a few statistical issues are occasionally discussed, around which methods give rise to unbiased estimates, and some discussions focus on which statistical methods yield the greater statistical power. In contrast, if the subjects come from naturally assigned groups, such as those in observational studies, then making comparisons of changes from the baseline usually requires making some crucial assumptions that are quite often hard to ascertain; this warrants considerably more discussion around the statistical process and the interpretation of results.
