This paper attempts to explain why systematic reviews of randomized controlled trials, based on as high a proportion as possible of the relevant studies, are so important in generating reliable information for evidence-based decision making within health care. The preparation, maintenance and dissemination of such reviews is the challenge which has been taken up by the Cochrane Collaboration. The first phase of data collection is the identification of relevant studies. Currently, bibliographic databases are inadequate for this task. MEDLINE searches identify on average only about half of the relevant studies, and until 1994 there were no suitable indexing terms in EMBASE to identify randomized controlled trials. Co-operation between the Cochrane Collaboration and both the National Library of Medicine and Elsevier, however, is already transforming this situation. From January 1994 a new indexing term has been added to EMBASE to help identify randomized controlled trials. From January 1995 a new indexing term will be added to MEDLINE to help identify controlled trials where the method of allocation to treatment or control cannot be described with certainty as being randomized. Also from January 1995 an additional 20,000 reports will be identifiable as randomized controlled trials in MEDLINE. Progress during the first 2 years of the Cochrane Collaboration has been encouraging, but much remains to be done if users of health services are to benefit from the evidence-based health care which they deserve.