International Journal of Epidemiology 2002;31:697
© International Epidemiological Association 2002
Book Review |
Systematic Reviews in Health Care. Meta-analysis in Context.
M Egger, G Davey Smith, Doug Altman (eds). London: BMJ Books, 2001, pp. 487, £50.00. ISBN: 072791488X.
This book has been launched as the second edition of the book Systematic Reviews by Iain Chalmers and Doug Altman published by the BMJ Publishing Group in 1995. The first edition counts 8 chapters and 119 pages, this second edition 26 and 487. It really is a new book, and rightfully so, because so much has happened since 1995 in this field of interest.
Both its size and the detailed level of the contents mean that it is not bedtime reading material to be finished in one go. Rather it is a work that anyone who is involved in preparing systematic reviews should have on their shelf for reference to specific elements of interest. In many but not all chapters, the level of detail goes far beyond what a user of systematic reviews, or a beginner about to start their first review is looking for; it is the methodologists who will read the book from cover to cover. It excels in the quality and extensiveness of referencing. Most chapters are at the brink of the methodological developments and some even take a risk and go beyond that. However, the users of systematic reviews are not forgotten and six chapters in part V of the book deal with issues of interest to various users. The focus, as demonstrated by the examples used, is mainly on systematic reviews in a medical context; those involved in systematic reviews in other fields such as psychology or social science will have to make a mental translation step to examples in their field of interest. However, its methodological principles and the emphasis on prevention of bias in all aspects make the book applicable to any area in which systematic reviews are being used.
The authors have pursued the laudable aim of clarifying what systematic reviews are and how they relate to meta-analysis (statistical pooling). Unfortunately, this is not realized completely, probably caused by the fact that there are so many authors of different chapters. One chapter mentions meta-analysis throughout and actually succeeds in using the words ‘systematic review’ only once and that is when referring to another chapter!
The fact that the book consists of a collection of separate chapters also shows itself in other ways. For example, two excellent chapters on evaluation of prognostic variables and on evaluation of diagnostic and screening tests are organized according to the process of preparing a systematic review; whereas the preceding chapter on observational studies focuses on the problems of such studies with various biases and meta-analysis. In general, there is a lot of emphasis on the importance of exploring heterogeneity and bias and on helping the reader understand the limitations of meta-analysis. This is perhaps one of the most important messages this work has: too often reviewers go on to do meta-analysis of data that are unsuitable and where it clearly does not make any sense.
Still a lot of space is used to discuss the statistical details of meta-analysis, including references to software, examples and data sets (see www.systematicreviews.com). Certainly many readers will appreciate that information. It would, however, have made the book more complete if it had also included a chapter about how to summarize data in situations where meta-analysis is not sensible and/or feasible. Granted, this is probably an even bigger challenge than clearly explaining the statistical details of pooling, in which the authors have succeeded very well.
One chapter discusses the assessment of quality (internal validity) of randomized trials that are included in the systematic reviews and the chapters on prognosis and diagnosis also touch upon quality assessment for those particular studies. The approach in these chapters is very academic and methodological (and valuable as such) as opposed to practical; possibly a main reason why the debate about how to assess quality has gone on for over a decade now. A more practical approach might start by considering the various reasons why quality is being assessed in a systematic review, and then follow on with the options of how to do it for different particular purposes.
One other theme that is consistently stressed throughout the book is the importance of meta-analysis based on individual patient data from randomized trials. The number of occasions where this currently is feasible might just be a few per cent or even less of all systematic reviews. We could not agree more about stressing the potential importance of such analyses and the importance of trialists making such data available for analysis in systematic reviews.
So what do we make of this book? This is probably the first and only ‘textbook’ on systematic reviews in the sense of its volume and the level of detail of most of the contents. It has been written by leading practitioners of the science of reviewing biomedical research and will be a valuable resource to those preparing, appraising and using systematic reviews. Highly recommended.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||