Article Text

Download PDFPDF
Chapter 5: The tools of evidence-based medicine
  1. Phil Wiffen1,
  2. Tommy Eriksson2,
  3. Hao Lu3
  1. 1Pain Research Unit, Churchill Hospital, Oxford, UK
  2. 2Department of Clinical Pharmacology, Laboratory Medicine, Lund University, Lund, Sweden
  3. 3Department of Pharmacy, Beijing United Family Hospital, Beijing, China
  1. Correspondence to Professor Phil Wiffen, Pain Research Unit, Churchill Hospital, Old Road, Oxford OX3 7LE, UK; phil.wiffen{at}

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

One of the exciting developments in evidence-based medicine has been the emergence of concepts that help to quantify and graphically illustrate how well a medicine works. This has brought understanding to many who failed to grasp statistics during their training and therefore have avoided any statistical analysis ever since. This paper examines some of these new concepts. We need to start with confidence intervals (CIs) as these put results into context.

Confidence intervals

Most pharmacists will be aware of p values in terms of whether or not an answer is significant (in a statistical sense). However, the use of p values is now largely being replaced by the concept of CIs.

The CI, usually set at 95% limits (although others may be used), is a way of describing the confidence that can be placed on the ‘result’. By describing the CI the range of the result can be clearly seen, in this case with a confidence of 95% that the ‘result’ is correct.

CIs figure largely in many systematic reviews, meta-analyses and for describing numbers-needed-to-treats (NNTs). For example, the NNT for a particular drug in moderate or severe postoperative pain may be 3.6 (95% CI 3.0 to 4.4). While it is tempting to concentrate on the single figure of 3.6, in reality the answer lies somewhere between 3.0 and 4.4.

Systematic review or meta-analysis?

Generally, the term ‘systematic review’ refers to the publication of a review carried out systematically in a way that minimises bias. Within the review there may be an analysis known as a meta-analysis, presenting data graphically often in the form of a Forest plot (see below). In North America the term ‘meta-analysis’ is sometimes used to describe the review.

Forest plots

Some studies touch or even cross over the vertical line of no difference. These studies are not statistically significant in terms of a …

View Full Text


  • Competing interests None.

  • Provenance and peer review Commissioned; internally peer reviewed.