Uses and abuses of analysis of covariance in clinical trials

Measurement of improvement in clinical trials in chronic diseases commonly compares baseline data to endpoint values by performing t-tests or analysis of variance (ANOVA) on raw gains or percentage changes. This procedure can be misleading and the use of an analysis of covariance (ANCOVA) should be considered. Properly used, ANCOVA increases statistical power in a clinical trial. However, its advantage over t-tests can be nullified by small numbers of patients, violations of assumptions, and incorrect application of the techniques. An evaluation of ANCOVA in chronic disease studies is given, with examples of its strengths and weaknesses as seen in several drug trials in the rheumatic diseases. Recommendations on its use and a decision tree for the nonstatistician are provided.