Introduction

In a crossover clinical trial with two treatment periods each patient starts by receiving one of the treatments in the first period and then crosses over to receive the other treatment during the second period. Referring to the two treatments as A and B, there are thus two groups of patients: those starting with A and crossing over to B (group 1) and those starting with B and crossing over to A (group 2). Equal numbers of patients are allocated at random to the two groups but drop-outs can cause the final numbers to be unequal. Trials like these frequently figure in consultancy work because interpreting the results from them can be difficult, both for the physician and the statistician. On this occasion the client, a pharmaceutical company, asked us to analyse the results from a crossover trial in which the response to treatment took the form of a continuous 24-hour electrocardiogram.

The trial was one of a number, carried out under varying conditions, in which a new drug was compared with a standard treatment. We knew, therefore, that our analysis would not be viewed in isolation but as part of a whole body of evidence, and that it might well form part of a submission to a regulatory authority such as the FDA (Food and Drugs Authority, USA).