Statistical Issues in Drug Development

Rating:
Summary: heavy on issues, provocative and with minimal mathematics
Review: Senn is a great writer. He has written an excellent text on cross-over trials that raises many issues about when such design can be used and what their limitations are. This book covers the gamut of issues in drug development concentrating on important and sometimes subtle issues in clinical trials including design and analysis, intention to treat principle, multiple testing, Bayesian and frequentist approaches and interpretations, meta analysis, regulatory issues and ethics. It also covers cross-over designs, pharmacokinetics, pharmacodynamics and pharmacoeconomics.

The introduction gives you a feeling for the approach in the book and how it splits into two parts. Part I, consisting of chapters 2-5, provides some history of the development of statistical methods and some introductory topics that are fundamental to the discussion in Part II.

Part II is the heart of the book where the practical statistical issues in clinical trials are raised. The text is intended for non-statisticians who work in the pharmaceutical industry but to quote part of Senn's preface he states "Although addressed to the life-scientist it is my hope that many statisticians, in particular those studying medical statistics or embarking on a career in drug development, will also find it useful. Above all I hope that it will help communication between the disciplines: a process by which the statistician stands to benefit as much as any other professional in drug development."

I can really appreciate what Senn has done. He explains the issues of intention-to-treat, washout, multiplicity and other problems that I have had to wrestle with and try to explain to MDs and clinical managers. But even more importantly to me than helping me communicate the many issues that I was aware of, he also raises many subtle issues that I was not aware of. This includes questions of bioequivalence, the use of baseline information and particularly percentage change from baseline versus covariate adjustment, problems of inference regarding measurements taken after titration and issuesw in N of 1 trials. I even learned a few new techniques (e.g. Taves minimiization and Atkinson's generalization of it for allocating patients to treatment groups).

The only complaint I can see is that there is not enough detail. However, remember the text was not designed for statisticians and so much of the mathematics and technicalities are deliberately left out.

But Senn does provides a detailed list of relevant references at the end of each chapter that allows the reader to find in texts and journal articles all the detail one might need. Also to aid with communication there is a large glossary of terms at teh back of the book.

This is a great reference for scientists and statisticians as well!