Statistical evidence in medical trials: what do the data really tell us? Simon SD (2006) ISBN: 0198567618; 216 pages; £25.00, $44.50 Oxford University Press; http://www.oup.com/

Authors


This book grew out of the author's teaching at the Children's Mercy Hospital in Kansas City, where a common response to the question why they were taking the class was that they wanted to understand the statistics used in medical journal articles. Accordingly, this book is written for healthcare professionals who are reading and evaluating medical publications. In addition, non-medical professionals who sometimes have to read medical journals, like journalists and lawyers, are part of the target audience. As the author expresses it, the book is written not for producers but for consumers of research. One could also say that it is not written for producers but for consumers of statistics. As such, the book avoids formulas and technical language.

A brief overview of the book's content is as follows. Chapter 1, ‘Apples or Oranges?’, discusses the internal validity of a study by considering selection and quality of the control group. The next, ‘Who Was Left Out?’, is devoted to the external validity of a study by examining exclusions, refusals, and dropouts, while ‘Mountain or Molehill?’ treats the clinical importance of the results from a study. ‘What Do the Other Witnesses Say?’ discusses how to find additional corroborating evidence outside the journal article one is reading, and ‘Do the Pieces Fit Together?’ examines the use of systematic overviews and meta-analyses. ‘What Do All these Numbers Mean?’ explains in non-technical language some statistical measures and terms, such as p-value, confidence interval, odds ratio, relative risk, and number-needed-to-treat. Finally, ‘Where is the Evidence’ gives an overview of how to search for information in databases, literature, and the Internet.

A characteristic feature of this book is the extensive use of real-world medical examples to illustrate the topics discussed, which I found very useful. Another useful feature is that the author, when possible, uses examples from and refers to open-access journals. This makes it easy for the reader to get immediate access to the full text. Every chapter ends with an ‘On Your Own’ section, usually containing abstracts or excerpts from a few published articles. These are also taken from open-access articles, as are all graphs and figures in the book. This is something that I think should be encouraged. Another interesting feature of this book is that most chapters have a section called ‘Counterpoint’, where the author tries to balance the arguments given in the chapter by presenting counter-arguments pointing at the weaknesses of the methods, with the conclusion that the given methods can sometimes be overrated.

The text is generally well written and quite humorous at sometimes without losing focus of the topic and the respect of the reader. However, it contains quite a lot of misspellings and grammatical shortcomings, and does not handle references well: quite often one finds a reference in the text, but searches in vain to find it in the bibliography. This is very frustrating when one reads an interesting example that one wants to look up.

To conclude, although it has some shortcomings, this is a very interesting and useful book, especially for consumers of research with a limited knowledge of statistics, but even for producers of research. The main message is the importance of aspects of statistics other than formulas or calculations, such as selection of the right control group and avoidance of bias, which make the formulas more or less useless according to how they are implemented. Most practical statisticians in the pharmaceutical industry will find this book very useful.

Ancillary