In Statistical Evidence in Medical Trials: What Do the Data Tell Us? author Stephen Simon attempts to introduce readers with limited statistical backgrounds to some of the major methodological issues of interest in the medical literature. The book is divided into seven major sections, which cover selection of the control group, trial exclusion criteria and patient refusal/dropout, the practical clinical importance of research findings, comparison of study results with other available literature, systematic reviews and meta-analyses, a brief overview of commonly used statistical methods, and places to look for information regarding medical topics (e.g., PubMed). Simon's text does a wonderful job of presenting and explaining the relevance of statistical issues in a manner understandable even to those with no statistical training whatsoever. For those seeking a painless overview of some large-picture statistical topics of interest and for those seeking a better understanding of statistical issues that impact published medical findings, this book is a good find. It is likely to be of little use, however, to readers seeking an introductory “how to” on statistical methods.
Simon well motivates and explains such important items as the role of randomization, clinical vs statistical significance, selection bias, subgroup comparisons, and type I and II errors. Each topic is supported by actual examples gleaned from published medical research. Simon avoids burdening the reader with the mechanics involved in statistical computations and tries to get readers to grasp the big-picture topics that can impact a study's validity. For example, rather than running through the details of computing a Kaplan-Meier curve, the author discusses the impact that informative censoring can have on survival studies. In addition, readers are introduced to systematic reviews and meta-analyses, topics that are often omitted from an introductory book. When a systematic review is not available, Simon presents practical approaches for comparing a given research finding to the available literature. In these instances, the author recommends that the reader ask a series of questions, which include, “Is there a strong association? Is the association consistent? Is the association biologically plausible?”
The qualities of this book notwithstanding, further details regarding statistical pitfalls that are often seen in the literature would have been beneficial. This is particularly the case with chapter 6, which provides a brief (and equation-free!) introduction to some commonly used statistics. For example, after introducing the correlation coefficient, the author neglects to point out the pointless and too-often-used test statistics constructed around this measure 1. One also wishes that the author would have provided a basic introduction to multivariable modeling as a way to build predictive tools or to adjust treatment/exposure effects for confounders. Finally, while noting that many of the statistical details provided in a paper are only there for those wishing to reproduce the results, Simon does not point out how many times such details are regrettably omitted. This is particularly worrisome as such omissions should always be a warning sign for readers.
Statistical Evidence in Medical Trials does accomplish its main goal—to provide a brief introduction to statistical issues in medical research that is readable for nonstatisticians. Readers should be aware that this book does not explain the details of any particular statistical method, nor does it discuss ongoing statistical controversies (readers will learn nothing of the Frequentist/Bayesian debate, for example). However, if you are not planning to become an amateur statistician but do wish to become a better-informed reader, this book is for you. As Simon states in his introduction, the book is written “for consumers of research, not producers of research.”