Using ANCOVA as a method for adjusting for baseline
Q7.5 I have done a randomised clinical trial with several groups measured before randomisation (baseline) and after randomisation. How do I show a group effect?
A7.5 Provided that the data fulfil the relevant model assumptions (normality, homogeneity of variance) analysis of covariance (ANCOVA) is the best method to use. In this model the post-randomisation variable is used as the dependent variable, the baseline variable as a covariate and the group factor as the explanatory variable whose effect is to be tested. Alternative approaches such as the analysis of change scores (=difference between post and pre-randomisation values) or the analysis of post randomisation scores alone have been proposed but ANCOVA can be shown to be most powerful provided that the correlation between the repeated measures is not small. For more details see
Senn S (1997) Statistical Issues in Drug Development. New York: John Wiley.
Senn SJ (1989) The use of baselines in clinical trials of bronchodilators, Statistics in Medicine8: 1339-1350.
Vickers AJ and Altman DG (2001) Statistics notes: Analysing controlled trials with baseline and follow-up measurements, BMJ 323: 1123-11124.