In 1999, Iain Chalmers, director of the UK Cochrane Centre, and Doug Altman, of the ICRF Statistics Group in Oxford, called for journals to play a more radical role in the prevention of poor medical research [1]. Most useful, they suggested, might be the publication of protocols for proposed or ongoing research, and in particular, protocols of randomised controlled trials. Taking up this challenge, BioMed Central now invites trialists and other researchers, including systematic reviewers, to publish their full protocols on line (see table). Two have already been published: a case control study of autism and MMR vaccination (http://www.biomedcentral.com/1471-2458/1/2) [2], and the CRASH trial, a multicentre randomised controlled trial of steroids in head injury (http://www.biomedcentral.com/1471-227X/1/1) [3].

A protocol is a crucial part of any study. It is a statement of intent. It provides a detailed account of the hypothesis, rationale, and methodology, and is a plan for all the investigators to follow. But protocols are rarely available to people outside the study. Once a study is completed, the protocol may be filed and forgotten, if not lost. And because of journals' severe space constraints, most reports of studies contain only a shortened summary of the methods.

Chalmers and Altman list several key reasons why publication of protocols will help to tackle what Altman has called the scandal of poor medical research.[4] First, publication while the protocol is still in draft form allows reviewers and readers to suggest improvements to be made to the study before it begins. This idea of early remedial intervention was one motive behind the Lancet's protocol review scheme[5] [6]. So far all protocols reviewed by the Lancet [7], and both protocols published by BioMed Central, have been finalised before submission, leaving no scope for improvements to the study design. But journal review of finalised protocols can still improve the presentation, and in the case of a randomised controlled trial, ensure that it conforms to the CONSORT guidelines (http://www.biomedcentral.com/1471-2288/1/2) [8].

Second, protocol publication allows easier comparison between what was originally intended and what was actually done. It reduces the potential for "data dredging" - where associations are sought or stumbled upon during data analysis rather than hypothesised a priori. It also reduces the potential for unacknowledged or post-hoc revision of the study aims, design, or planned analyses. Such practices are not only detrimental to the advancement of medical research, they are ethically unsound since they may result in patients receiving inappropriate care. A randomised controlled trial recently published in the BMJ was, by coincidence, reviewed for the journal by the same person as had reviewed the protocol for the funding body. The reviewer noticed discrepancies in the power calculation that turned an inconclusive result into a strongly negative one [9]. As emphasised by commentaries published with the trial, and by the CONSORT guidelines, departures from the protocol should be clearly documented. Rather than relying on accidental discovery, such discrepancies could be routinely checked for if protocols were routinely published.

Protocol publication is perhaps even more important for observational studies. They have even greater scope for data dredging and for post-hoc revisions. Recent correspondence in the Lancet criticised a report on the Barker hypothesis on foetal origins of diseases for, among other things, failing to specify the original study hypothesis [10]. An accompanying editorial regreted that the report did not make explicit the "sharp shift in thinking" that had occurred since the study protocol was finalised [11]. When data from the MMR/autism case control study are reported, the prior publication of the protocol will leave no such room for doubt [2].

Thirdly, publication of protocols lets people know what studies are underway. Those likely to be interested in ongoing studies include funders of research, other researchers, and patients who may wish to take part in trials. Protocol publication can reduce wasteful duplication of research effort. It can make it easier for systematic reviewers to find trials, which may in turn reduce distortion of the evidence from publication bias. And it can boost recruitment into trials, speeding up evaluation of interventions and increasing the chances of a trial's success. The CRASH trial protocol links through to information about how to join the trial (http://www.biomedcentral.com/1471-227X/1/1) [3].

Finally, publication of protocols helps to promote prospective registration of trials. The arguments in favour of prospective registration mirror those for protocol publication: to improve opportunities for collaboration, to reduce duplication of research effort, to boost recruitment into trials, and to reduce bias in the reporting of evidence from trials. These arguments have been elaborated in recent articles calling for a centralised database of ongoing trials [12, 13, 14, 15]. Such a database now exists in the metaRegister of Controlled Trials (http://www.controlled-trials.com), which contains summaries of over 6000 randomised controlled trials from 20 trial registers worldwide.

An important activity linked to the metaRegister is the unique numbering scheme for randomised controlled trials. The scheme will mean that each trial can be identified unambiguously, allowing it to be tracked throughout its life cycle. The aim is to reduce confusion caused by over-reporting of trials (a Cochrane systematic review of olanzapine found 158 reports of a single trial [16]) and to reduce bias caused by under-reporting, where data known to have been collected fail to reach the public domain. The numbering scheme is now being piloted with help from the UK Medical Research Council (MRC) and other groups. As part of the pilot, the CRASH trial, which is funded by the MRC, has received its International Standard Randomised Controlled Trial Number (ISRCTN), which is displayed in the title of the trial protocol. (http://www.biomedcentral.com/1471-227X/1/1). [3] Once the scheme is fully operational, ethics committees, funders, and journals can help to encourage uptake by asking trialists to obtain an ISRCTN before a trial can be approved, funded, or published. The BMJ and the Lancet have already made initial commitments to do this [7, 15].

With all its potential benefits, widespread protocol publication has become possible only because of the internet. The on line journals within BioMed Central provide open peer review, rapid publication, unlimited space, worldwide, barrier free dissemination, and the ability to link protocols to recruitment initiatives, trial reports, full data sets, and systematic reviews (see table). They also provide, through PubMed and PubMed Central, a searchable and permanent record of the protocol for citation. We hope that researchers will take advantage of these new opportunities, and in so doing, make their own crucial contribution to the prevention of poor medical research.

To submit your protocol to BioMed Central go to http://www.biomedcentral.com/manuscript