Applied Statistics in the Pharmaceutical Industry with Case Studies Using S-PLUS. (Book Reviews)
Eastwood, Brian J., Journal of the American Statistical Association
Steven P. MILLARD and Andreas KRAUSE (Eds.). New York: Springer-Verlag, 2001. ISBN 0-387-98814-9. vi+513 pp. $79.95 (H).
This book is a collection of 18 chapters by more than 30 statisticians covering virtually all aspects of the pharmaceutical research and manufacturing aspects of the business. These chapters are preceded by an overview chapter that puts them in context. The book also has a companion website, where all the S-PLUS code and most of the datasets may be downloaded.
The book is reasonably divided across the phases of drug discovery, with a slight overemphasis on Phase I material: one chapter on preclinical discovery, two chapters on preclinical toxicology, six chapters on Phase I studies, six chapters on Phase II/III studies, one chapter on Phase IV (postmarketing) studies, and two chapters on manufacturing and production. The six chapters on Phase I seem excessive; there is much repetition between the chapters, especially because one of the preclinical toxicology chapters also covers the analysis of pharmacokinetic data (in animals). This material is also the most dated material, because the impact of imaging technology will substantially change the way Phase I studies are conducted in the future.
The coverage of topics within each phase is quite variable. The Phase II/III material seems reasonably complete. However, the chapter on drug discovery presents only one very traditional study. As the introductory chapter indicates, genomics, molecular biology, high-throughput screening, and molecular modeling are revolutionizing discovery research. Unfortunately, instead of these topics, the chapter is devoted to a one-factor parallel-groups animal efficacy study. As the author indicates, statisticians do not typically analyze these studies. In large pharmaceutical companies, probably 20-50 of these types of studies are analyzed daily throughout all of preclinical discovery research; so by necessity, the scientists producing the data also analyze them. S-PLUS is an unlikely choice for analysis by a nonstatistician. The study also looks rather overanalyzed (to emphasize S-PLUS features) rather than presenting the analysis that is typically needed for that type of study.
Most of the chapters are written from a procedural perspective, describing the task to be done, showing the S-PLUS code required to conduct the task, and then displaying the S-PLUS results. From a statistical content perspective, most of the material is not advanced, and should be understandable to anyone with a graduate degree in statistics. Some chapters that deal with survival analysis and/or time series analysis do assume that the reader is familiar with those subjects. The analysis usually emphasizes S-PLUS's strengths- customized graphics with the ability to create built-in functions to create the necessary summary or inferential analysis. The sample S-PLUS code is reproduced at the end of each chapter, and this material can be downloaded from the companion internet site.
Some chapters do not follow the format described above; these usually present more technical material. …