Regression toward the mean in 2 x 2 crossover designs with baseline measurements.

Abstract

Clinical trials often involve the study of patients whose values for some quantitative characteristic are in the extreme of the distribution for the normal population. Subsequent observation of the characteristic of interest will ordinarily tend to regress toward the mean of the distribution even in the absence of a treatment effect. We consider the impact of regression toward the mean in the analysis of data arising in 2 x 2 crossover trials with two baseline measurements. The usual summary statistics can be especially misleading in the presence of regression toward the mean.