Identifying reports of controlled trials in the BMJ and the LancetBMJ 1996; 313 doi: https://doi.org/10.1136/bmj.313.7065.1116 (Published 02 November 1996) Cite this as: BMJ 1996;313:1116
- Steven J McDonald, research assistanta,
- Carol Lefebvre, information specialista,
- Michael J Clarke, overviews coordinatorb
- a UK Cochrane Centre, Oxford, OX2 7LG
- b Clinical Trial Service Unit and ICRF Cancer Studies Unit, Radcliffe Infirmary, Oxford, OX2 6HE
- Correspondence to: Steven McDonald.
- Accepted 29 July 1996
In 1994 the European Union BIOMED programme awarded £330 000 to a project, coordinated by the UK Cochrane Centre, to identify reports of randomised controlled trials in general health care journals in Europe. The rationale is that reports of randomised controlled trials are difficult to identify through sources such as Medline.1 For example, a study of the first six months of Medline for 1993 identified over 400 reports of randomised controlled trials which were not coded as such, despite having the word random or randomised in the title or abstract.2 A systematic review in 1993 indicated that, on average, searches of Medline identify only 50% of trials.3 We present here the results of handsearching the BMJ and the Lancet from 1948 onwards to identify reports of trials.
Methods and results
Twenty handsearchers were trained to identify reports of clinical trials where random allocation, or some quasirandom method of allocation, such as alternation or date of birth, was definitely or possibly used to assign individuals (or other units) prospectively to one of two (or more) alternative forms of health care. Handsearchers coded these reports as randomized-controlled-trial (RCT) or controlled-clinical-trial (CCT), the Medline publication type terms introduced in 1991 and 1995 respectively. They also noted any reports about which they were uncertain. All articles, editorials, letters, and news items in the BMJ and the Lancet were examined from 1948 to 1994. The most experienced handsearchers were assigned to the period 1948–65, when the quality of methodological reporting made identifying trials more difficult.
All identified reports were verified and recoded where necessary by an experienced clinical trialist (MC). The relevant electronic records for 1966–94 were taken from Medline and submitted to the US National Library of Medicine through the Baltimore Cochrane Center, which coordinates this activity on behalf of the Cochrane Collaboration. They were retagged with the correct publication type terms and have been available in Medline since January 1996 (fig 1). For those reports not on Medline (missed issues, pre-1966 reports, etc) electronic records have been generated.
For the period 1948–65, 1916 reports were identified, 956 (255 RCTs; 701 CCTs) in the BMJ and 960 (214 RCTs; 746 CCTs) in the Lancet. These 1916 reports, for which no Medline records exist, are now available in The Cochrane Library4 and will also be included in an ancillary database of the National Library of Medicine, currently under development. For 1966–94, 5347 reports of trials were identified, 2153 (1451 RCTs; 702 CCTs) in the BMJ and 3194 (2074 RCTs; 1120 CCTs) in the Lancet. Of these, 4093 reports are now identifiable in Medline using the publication type terms randomized-controlled-trial or controlled-clinical-trial. These are in addition to the 1050 which were previously identifiable. A further 204 reports with no Medline records have been added to The Cochrane Library.
This project builds on work undertaken at the UK Cochrane Centre in 1994, when about 100 000 Medline abstracts from 1985 to 1993 were read to identify reports of trials: as a result roughly 19 000 additional reports of randomised controlled trials were retagged in Medline for 1995. Of these, 185 were in the BMJ and 282 in the Lancet.
Having identified reports of trials in the two richest UK sources, other UK general health care journals are now being handsearched. This activity is part of an international exercise to identify all reports of trials in health care journals and databases; over 600 journals are being handsearched within the Cochrane Collaboration. This exercise will help identify randomised controlled trials for health care decision making, and in particular will help the Cochrane Collaboration to prepare, maintain, and disseminate systematic reviews.
We thank the staff of the Baltimore Cochrane Center, libraries in the Oxford region, Iain Chalmers for his comments on the manuscript, and all the handsearchers.
Funding The BMJ, the Lancet, the NHS research and development programme, and the European Union BIOMED Programme.
Conflict of interest None.