The case for registering all clinical trials—first advanced a decade ago1—is now unanswerable. The public has the right to know what research is being funded. Researchers and research funders don’t want to waste resources repeating trials already under way. And those conducting systematic reviews need to be able to identify all trials begun on a subject to avoid the problem of publication bias. Otherwise, clinicians may be deceived on what the evidence shows. Next week the Lancet, the Association of the British Pharmaceutical Industry, and the BMJ Publishing Group will hold a joint conference to promote the registering of trials.
Each year a vast financial investment is made by national funding agencies, medical research charities, and drug and device manufacturers in randomised controlled trials. Unfortunately the process is chaotic and takes little account of concurrent research. Several case studies have shown how the manipulation of trial data can provide a seriously misleading picture of an intervention’s effectiveness. In a systematic review of trials using ondansetron to treat postoperative nausea and vomiting Tramer et al2 found that “a false impression of ondansetron’s efficacy may arise because a quarter of all relevant published reports are duplicates.” Huston and Moher found it almost impossible to complete a systematic review of risperidone’s efficacy in schizophrenia for the same reason.3 These studies show that we have to find better ways of identifying and tracking clinical trials.
The history of this effort shows much good intention but only limited progress. One attempt to link research to practice in the setting of an entire health service began in the United Kingdom in 1991 with the launch of the NHS research and development initiative.4 That programme placed the systematic collection of data from randomised trials at its intellectual centre. The Cochrane Collaboration has been its most important and successful partner and has focused its work on published clinical trials. But this leaves untackled the large amount of unpublished trials.5 Chalmers famously described this underreporting of research as scientific misconduct,6 and publication bias remains a pervasive problem. The medical editors’ trials amnesty tried to flush out that evidence, with only partial success.7
Rather than treat the problem of hidden research retrospectively, a more sensible approach might be to prevent it.1 Based on their original investigations of publication bias, Dickersin and Min have argued that one “possibility is to require registration of all clinical trials prior to initiation. While this is widely agreed to be a good approach, widespread registration has not yet been effected….Who will take the lead?”8
Apart from the NHS national research register and the Cochrane controlled trials register, the most significant recent lead has been taken by the pharmaceutical industry. For example, Schering Health Care and GlaxoWellcome have committed themselves to registering information about their own trials. Richard Sykes (chairman of GlaxoWellcome) argued that he and his colleagues understood “the value of information, and we want to create a climate of openness where the evidence for prescribing our products is clear.”9 Not all in the pharmaceutical sector agree, and Sykes has been ridiculed by some who see his step as opening up a window of vulnerability in GlaxoWellcome’s commercial armour. But how can this be so when all that GlaxoWellcome is doing is releasing administrative information about continuing work (objective of the trial, end points, numbers, groups, and expected data of closure), not the actual data?
Editors also have a part to play. During peer review, editors increasingly find themselves requesting copies of the original trial protocol to check against the final submitted report. That “protocol culture” has led one of us to begin (and the other to plan) a protocol registration scheme.10 Editors are unwilling to fill their journals with promises of what might be, but they can publish these protocols on their web sites, perhaps linking them to a central registry.
Publishers could also help this process by collaborating with one another to construct such a free online database. The lead here has been taken by Current Science, which launched a metaregister of randomised controlled trials in October 1998. Trials depend on patient participation and are often funded with public money. Publishers make money from reprints of clinical trials, so it is reasonable to expect them to contribute to an initiative from which they ultimately benefit. A valuable partner might be PubMed Central, a project launched by director of the National Institutes of Health to create a free electronic archive of biomedical research.11
The pressure to register trials will rise when research ethics committees, medical research charities, and drug and device manufacturers start to encourage trialists to register, especially since the responsibility for not publishing trial results seems to rest more with investigators than editors.8 A further challenge is to devise an internationally agreed method for assigning each trial a unique identifier. One such scheme is being piloted in cancer, with the help of the Cochrane cancer network.
Taken together, these efforts might bring shape to a presently formless clinical research enterprise. Such a structure should help to deliver high quality evidence to the clinical setting.