 |
 |
 |
 |
 |
 |
 |
|
|
|
|||||||||
|
|||||||||||
Professor Faculty of Pharmaceutical Sciences The University of British Columbia Associate Member Department of Pharmacy Children's and Women's Health Centre of British Columbia Vancouver, British Columbia, Canada
| By Shein-Chung Chow and Jen-Pei Liu. Published by John Wiley
& Sons, Hoboken, NJ, 2004. ISBN 0-471-24985-8. Clothbound, xiii + 729 pp.
(26 x 18.5 cm), $130.
www.wiley.com
|
Published Online, August 3, 2004. www.theannals.com, DOI 10.1345/aph.1E017
The book is divided into 15 chapters, with the first 7 initially introducing the subject and statistical concepts and then dealing with considerations and methods for the design of clinical trials. Chapters 8–12 focus on methods for statistical analysis of clinical trial data, sample size determination, and specific issues in efficacy evaluation. Chapter 13 focuses on assessing the safety of drug products, and the final 2 chapters deal with protocol development and management of study data. While one may question the order of presentation of topics, the fact that this book is most likely to be used as a resource makes the order less important than the ease of access to information on particular questions. In this respect, the chapter topics and index serve the purpose well. It is somewhat surprising, however, to find that there are frequent grammatical errors, which can occasionally be irritating.
A strength of this book for clinical investigators is the detailed introduction to good clinical practice (GCP) in clinical research and the movement toward international standards through the work of the tripartite (Europe, Japan, US) International Conference on Harmonization of Technical Requirements for the Registration of Pharmaceuticals for Human Use (ICH). The ICH guidelines serve as the benchmark for GCP in clinical trials. While the authors note that only the participating regions are affected, in fact other jurisdictions are adopting the ICH guidelines as their standard. For example, when clinical trials are audited by Health Canada, the ICH "Good Clinical Practice: Consolidated Guideline" (www.ncehr-cnerh.org/english/gcp/) is now used as the standard, and Canadian investigators are thus becoming aware that their clinical trial practices must conform to the guideline.
In addition, clinical investigators involved in planning clinical trials will find much useful information related to specific issues in trial design in this book. For example, Chapter 4 provides a useful and very detailed discussion of the principles and methods for randomization and blinding. As is done throughout the book, the authors provide useful examples from the literature to illustrate their points and include critical commentary when comparing different methods. On occasion, statements are made that may mislead clinical researchers who have little statistical background. For example, the authors state that randomization ensures that the intervention and control groups are similar at baseline (p. 165). When properly done, randomization ensures that selection bias will not occur; however, only when the sample size is sufficiently large will randomization ensure that the groups are well balanced at baseline. For trials with relatively small sample sizes, important imbalances between groups will often occur by chance.
There are also other areas in which non-statisticians may misunderstand important basic concepts. In the section on hypothesis testing, the authors indicate that, when the null hypothesis (H0) is rejected, then the alternative hypothesis (HA) is accepted and "we conclude that there is a significant difference between the drug product and placebo" (p. 73). This is technically incorrect and may serve to perpetuate the misunderstanding among clinical investigators of the interpretation of statistical tests and the relation of hypothesis testing to scientific inference. When an investigator rejects the null hypothesis, the result is considered not to have been due to chance. The logical alternative to the null is that the observed difference between groups is thought to have occurred due to some non-chance difference between groups. This may have been due to imbalance at baseline, loss of blinding, selective loss of patients prior to completion of the trial, or numerous other confounding factors. If these can be satisfactorily ruled out as explanations for the non-chance difference between groups, then the scientific inference is that the drug product appears to be effective relative to placebo. This book discusses the possible confounding factors in the analysis of trial data in considerable and excellent detail; however, clinicians reading the basic statistics sections may not appreciate this important principle. It is essential for good, critical evaluation of research, as well as evidence-based clinical practice, to explain to readers that statistically significant differences do not lead directly to the conclusion that the drug works.
It is stated in the preface to the first edition that this book is written for readers with minimal mathematical or statistical backgrounds. This is actually not the case. Some of the sections on trial design (eg, cluster randomized designs, p. 175; placebo-challenging design, pp. 202–8) and those involving data analysis (Chapters 8–13) require statistical understanding and familiarity with notation that are considerably beyond the backgrounds of many clinical investigators. However, these sections (including new ones on cluster randomization, group sequential designs, equivalence/noninferiority trials and others) contain information and approaches that will be very useful to statisticians and students involved in clinical trial design and analysis. Readers of the first edition will likely appreciate the addition of chapters and sections on cancer trial design, protocol development and implementation, and data management. Prospective new readers should be aware, however, that this is not an introductory or "how to" book for beginning clinical investigators, such as Designing Clinical Research by Hulley et al. (2nd edition, Lippincott, Williams and Wilkins, 2001).
Finally, one surprising omission is the lack of any discussion of Bayesian methods in trial design or analysis. In fact, the term "Bayesian" does not appear in the index. In light of the growing discussion of Bayesian methods for clinical research, the authors may wish to consider including Bayesian applications in future editions.
While this book may not achieve the authors' ambitious aim of filling "the gap between clinical and statistical disciplines," there is a wealth of useful information in it that will be of value to select members of either group. The authors do, however, admirably achieve their objective of providing a comprehensive reference book for scientists, statisticians, and others involved in clinical research. This book will also serve as a useful resource (rather than a textbook) for academics teaching graduate courses involving principles, methods, or data analysis for clinical trials.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
 |
 |
 |
 |
 |
 |
