PREFACE TO THE FIRST EDITION

This book was written for those involved in clinical research and who may, from time to time, need a guide to help demystify some of the most commonly used statistical methods encountered in our profession.

All too often, I have heard medical directors of clinical research departments express frustration at seemingly cryptic statistical methods sections of protocols which they are responsible for approving. Other nonstatisticians, including medical monitors, investigators, clinical project managers, medical writers and regulatory personnel, often voice similar sentiment when it comes to statistics, despite the profound reliance upon statistical methods in the success of the clinical program. For these people, I offer this book (sans technical details) as a reference guide to better understand statistical methods as applied to clinical investigation and the conditions and assumptions under which they are applied.

For the clinical data analyst and statistician new to clinical applications, the examples from a clinical trials setting may help in making the transition from other statistical fields to that of clinical trials. The discussions of 'Least-Squares' means, distinguishing features of the various SAS® types of sums-of-squares, and relationships among various tests (such as the Chi-Square Test, the Cochran-Mantel-Haenszel Test and the Log-Rank Test) may help crystalize the analyst's understanding of these methods. Analysts with no prior SAS experience should benefit by the simplifed SAS programming statements provided with each example as an introduction to SAS analyses.

This book may also aid the SAS programmer with limited statistical knowledge in better grasping an overall picture of the clinical trials process. Many times knowledge of the hypotheses being tested and appropriate interpretation of the SAS output relative to those hypotheses will help the programmer become more efficient in responding to the requests of other clinical project team members.

Finally, the medical student will find the focused presentation on the specific methods presented to be of value while proceeding through a first course in biostatistics.

For all readers, my goal was to provide a unique approach to the description of commonly used statistical methods by integrating both manual and computerized solutions to a wide variety of examples taken from clinical research. Those who learn best by example should find this approach rewarding. I have found no other book which demonstrates that the SAS output actually does have the same results as the manual solution of a problem using the calculating formulas. So ever reassuring this is for the student of clinical data analysis!

Each statistical test is presented in a separate chapter, and includes a brief, non-technical introduction, a synopsis of the test, one or two examples worked manually followed by an appropriate solution using the SAS statistical package, and finally, a discussion with details and relevant notes.

Chapters 1 and 2 are introductory in nature, and should be carefully read by all with no prior formal exposure to statistics. Chapter 1 provides an introduction to statistics and some of the basic concepts involved in inference-making. Chapter 2 goes into more detail with regard to the main aspects of hypothesis testing, including significance levels, power and sample size determination. For those who use analysis-of-variance, Appendix C provides a non-technical introduction to ANOVA methods. The remainder of the book may be used as a text or reference. As a reference, the reader should keep in mind that many of the tests discussed in later chapters rely on concepts presented earlier in the book, strongly suggesting prerequisite review.

This book focuses on statistical hypothesis testing as opposed to other inferential techniques. For each statistical method, the test summary is clearly provided, including the null hypothesis tested, the test statistic and the decision rule. Each statistical test is presented in one of its most elementary forms to provide the reader with a basic framework. Many of the tests discussed have extensions or variations which can be used with more complex data sets. The 18 statistical methods presented here (Chapters 3-20) represent a composite of those which, in my experience, are most commonly used in the analysis of clinical research data. I can't think of a single study I've analyzed in nearly 20 years which did not use at least one of these tests. Furthermore, many of the studies I've encountered have used exclusively the methods presented here, or variations or extensions thereof. Thus, the word 'common' in the title.

Understanding of many parts of this book requires some degree of statistical knowledge. The clinician without such a background may skip over many of the technical details and still come away with an overview of the test's applications, assumptions and limitations. Basic algebra is the only prerequisite, as derivations of test procedures are omitted, and matrix algebra is mentioned only in an appendix. My hope is that the statistical and SAS analysis aspects of the examples would provide a springboard for the motivated reader, both to go back to more elementary texts for additional background and to go forward to more advanced texts for further reading.

Many of the examples are based on actual clinical trials which I have analyzed. In all cases, the data are contrived, and in many cases fictitious names are used for different treatments or research facilities. Any resemblence of the data or the tests' results to actual cases is purely coincidental.

Glenn A. Walker May 1996