Читать книгу Understanding Clinical Papers - David Bowers - Страница 10

Preface to the Second Edition

We received a great many favourable comments from those who used the first edition of this book – for which many thanks. Why then a second edition? The reason is the usual one in these circumstances – we think we can make the book even better. When we set out to write the first edition we had a good idea of what we wanted to include, but inevitably there was some jostling for the available space. In the end some things that we might have included had to be left on the cutting room floor. With this second edition we have now been able to include most of that excluded material. We have also taken the opportunity to respond to some helpful suggestions from readers. In addition to these changes we have now added a considerable amount of completely new material.

Thus this second edition includes a new chapter on measurement scales, and new or significantly expanded sections on the following: ethical considerations; abstracts; consent to randomization into trials; pragmatic and explanatory trials; intention‐to‐treat analysis; elements of probability; data transformation; non‐parametric tests; and systematic review, among others.

Moreover, there is a lot of new material in the chapters on regression – including more on variable selection and model building, and on Cox regression. A good deal of the material in the middle chapters of the book has been re‐arranged and improved to make for a better and more lucid flow (the treatment of dummy variables has been brought forward a chapter, for example).

We have all taken the opportunity to update many of the extracts from clinical papers which we use to illustrate the various ideas and procedures we describe, and also to revise much of the text in the book to improve clarity and understanding. We remain more than willing to receive any constructive comments and suggestions from readers. Otherwise we are confident that this is now an even better book than the original.

SOME NOTES ON STATISTICAL SOFTWARE

There are several statistical packages, of varying levels of sophistication and complexity, which can be used to analyse clinical data. Among the most widely used are the following:

 CIA (Confidence Interval Analysis)

 EPI‐Info

 Minitab

 SPSS (the Statistics Package for the Social Sciences)

 STATA

 S‐PLUS

In our opinion Minitab is the simplest and friendliest statistics package for the general user. SPSS is not quite as easy to use but handles cross‐tabulation of two variables rather better and has a wider range of the more sophisticated types of analyses. The choice of types of analysis and their outputs are perhaps easier to understand in Minitab than in SPSS. Each application has, of course, its limitations. To the best of our knowledge, Minitab does not do the McNemar test, nor does it have a clinically‐based survival analysis programme, nor allow for a direct calculation of Spearman's correlation coefficient (the data needs first to be ranked). On the other hand, SPSS does not allow a chi‐squared test to be done directly on a contingency table in the columns of the data sheet, nor does it provide confidence intervals for the difference between two proportions, or with the Mann–Whitney or Wilcoxon tests, all of which Minitab does. But as we have said, SPSS has a wider range of applications.

CIA, as its name implies, only does confidence interval calculations (but in this respect is very useful). EPI‐Info is a combination database and epidemiological tool, which originates from the Centre for Disease Control (CDC) in the USA. It has the advantage of being free (it can be downloaded from the internet along with a user's manual).

Most professional clinical statisticians will probably use either STATA or S‐PLUS; both more powerful and versatile than either Minitab or SPSS (but rather less easy to use).

We would not recommend Excel as a statistics programme since it is fundamentally a spreadsheet tool and thus has an extremely limited range of statistical functions – and in any case these are not set out in a way that is well‐suited to clinical research.

WRITING PAPERS FOR CLINICAL JOURNALS

Those of you who envisage writing up your research and submitting a paper to a clinical journal may find the following web site addresses (URLs) useful. They contain detailed advice and instructions to authors on what is required prior to submission: for example, how to contact the journal, what should be in the paper (order and content of sections), information on the required style, editorial policies and guidelines, and so on.

The first URL directs you towards a set of instructions to authors for each of over 3500 health and life‐sciences journals, worldwide. The second and third URLs relate specifically to the British Medical Journal, but contain a huge amount of detailed and splendidly informative material related to the preparation and submission of clinical papers, and collectively provide a set of desirable standards which anyone who is contemplating the submission of such a paper to the BMJ or any other journal should aspire to.

http://mulford.mco.edu/instr

http://bmj.bmjournals.com/advice

http://bmj.bmjjournals.com/advice/article_submissiom.shtml

David Bowers, Allan House, and David Owens School of Medicine, University of Leeds,

Autumn, 2005