Skip navigation


We are often asked how items are chosen for Bandolier . The answer is pretty straightforward. We have all come to recognise that systematic reviews of randomised trials present the highest quality of evidence for treatments. So each month we search MEDLINE and (increasingly) PubMed using the terms systematic review and meta-analysis. We hope that the new on-line version of the Cochrane Library will make it easier to access information from the 400 or so reviews now on the database.

This process gives rise to about 50 to 100 abstracts in each category for the previous 30 days or so, and with PubMed especially the information is bang up-to-date, noting papers published in the last month. Then it's a question of sorting the wheat from the chaff by looking at abstracts, and going on to read those papers which seem to be the most pertinent, or interesting, or which shed light on methods.


This month Bandolier concentrates on NNTs. Reviews have been chosen which allow the calculation of NNTs from information given in the papers. We have chosen nicotine replacement , from a terrific Cochrane review, antibiotic prophylaxis for colorectal surgery from an HTA review, NNTs for new antiepileptic drugs , some calculated NNTs from recent meta-analyses on tamoxifen , and NNHs from the Cochrane review of albumin in critically ill patients.

Bandolier had to get the information from the papers, and calculate the NNTs itself, as well as making some L'Abbé plots. Doing this certainly helped us understand what these papers were trying to say, and get a sensible handle on the information. Odds ratios just don't cut it, let alone some of the more obscure statistical outputs. And while we're on the subject, some of our statistical methodologists might like to scan some of the reviews being published to see how many of them use pukka statistical methods. We have a sneaky suspicion that some papers (none in this issue) are being published which just get it wrong!

next story in this issue