Statistical shortcomings in licensing applications.

This paper concerns the statistical work carried out with respect to clinical trials conducted for regulatory purposes. Although the general quality of such work has improved markedly over recent years and is now generally high, a number of shortcomings remain. A few of these arise from failure to follow well established statistical practice. Rather more arise from a poor understanding of areas of known statistical disagreement and from the unsatisfactory use of newer and more advanced techniques. Inadequacies in reporting statistical work are commonplace. Examples of all these shortcomings are provided and emphasis is placed on the value of a statistical contribution to overall summaries such as the clinical expert report.