Statistical principles for clinical trials (ICH E9): an introductory note on an international guideline.

The opportunity to write a truly international guideline on statistics is a rare one indeed. In 1995, the Steering Committee of the International Conference on Harmonisation1 (ICH) took a decision which was of major importance to all medical statisticians involved in the development of new medicinal products. This decision was to add a guideline on statistical methodology (E9) to the collection of ICH guidelines already developed or under development. It was to be based on the corresponding European guideline (previously published in Statistics in Medicine2,3). An Expert Working Group (EWG) including 12 statisticians (see below) was formed to take the task forward. This group consisted of four representatives from each of the three main ICH regions, namely Europe, Japan and the U.S.A. Two from each region represented industry and two represented the regulatory authorities. The U.S.A. led the process by chairing the meetings of the group. It was decided that the guideline should concentrate on principles rather than detailed procedures and that it should attempt to address a broader audience than statisticians alone. Although the European guideline was taken as the starting point, the guideline was also heavily in#uenced by regional guidelines from the U.S.A.4 and Japan.5 The U.S. regional guideline had already had much of its material captured within the ICH E3 guideline6 entitled &Structure and content of clinical study reports'. This ICH E3 guideline is therefore also clearly of special importance for statisticians. The new guideline was completed in 1998 and recommended for adoption in the three regions by the ICH Steering Committee. The adoption process is now complete. The ICH Steering Committee encourages the publication of its guidelines to ensure their widespread availability and the Editors of Statistics in Medicine have kindly agreed to publish E9 in full on the following pages. This guideline (and many others) is also available on a number of websites including: ICH (http://www.ich.org/ich5e.html), EMEA (http://www.eudra.org/humandocs/humans/ICH.htm) and FDA (http://www.fda.gov/cder/guidance/index.htm). Although this document was drafted by the EWG listed below, a very large number of statisticians and other scienti"c experts from all three regions had a considerable in#uence on its contents through an extensive consultative process. Some attended EWG meetings, some attended special professional meetings to debate the content of early drafts and some provided valuable written comments on those early drafts. The resulting document does not pretend to represent unanimous agreement but it certainly does represent a very wide consensus of views, perhaps one of the widest achieved on matter of statistical methodology. The guideline has