Since July 2005, the International Committee of Medical Journal Editors (ICMJE) has recommended that all journals subscribing to their guidelines should require clinical trials to be registered with a public registry (e.g. http://clinicaltrials.gov) prior to commencing the study (http://www.icmje.org/publishing_10register.html).1 Moreover, BJOG also supports the WHO recommendation that from July 2008, any research study that prospectively assigns human participants or groups of humans to one or more health-related interventions, to evaluate the effects on health outcomes, must have been registered prospectively (this includes phase-I and phase-II trials and non-randomised studies of interventions) (http://www.bjog.org/view/0/authorInformation.html).
The arguments for this editorial policy are compelling. It results in greater transparency in the execution and reporting of such studies, because it allows first the editorial staff, and then the journal readership, to check whether there has been any deviation between the study protocol reported in a submitted paper and the study design when the trial began. It also safeguards against selective reporting of outcomes that just happen to be statistically significant. Irrespective of the results of the study and its publication fate, a copy of the details of the study design is kept as a public record, allowing clinicians, patients and healthcare policy makers open access. Finally, it also allows grant funding bodies and researchers to identify whether a similar study on the same topic has already been conducted or is in progress, thereby avoiding any unnecessary duplication of research work.
Following the implementation of this policy for prospective studies, a similar initiative has been launched for the conduct and reporting of systematic reviews. The International Prospective Register of Systematic Reviews (PROSPERO) is an open access database of systematic reviews administered at the Centre of Reviews and Dissemination, University of York, UK (http://www.crd.york.ac.uk/prospero). Registration is conducted online, is free, and is anticipated to take less than an hour to complete. Within the first few months of opening, there have been over 260 systematic reviews registered.
Registration involves the submission of the following information about the review, derived from a Delphic approach (iterative consultation) between international experts: the review question(s); the search strategy; the type of studies included; the study inclusion criteria for the review; a brief description of study participants, interventions, comparator(s) and control; the primary and secondary clinical outcomes; the method of data extraction from the primary studies; the assessment of study quality; the strategy for data synthesis; the dissemination plans; the anticipated commencement and completion dates of the review; the source of funding; and the contact details of the primary reviewer. There are a total of 22 mandatory fields and 18 optional fields to be completed before a review can be accepted for publication in the registry. The review will then be allocated a unique registration number to simplify the retrieval of the record as well as to link it with any subsequent report(s). The database is intended to include any systematic reviews of healthcare interventions and preventive or diagnostic strategies with a clinical outcome, but not scoping reviews or reviews of reviews. Reviews on methodological issues and reviews on animal studies are also currently not eligible. However, the long-term aim is for the registry to include details of all continuing systematic reviews with a health-related outcome in the broadest sense.2
The reasons why BJOG should support a public register on systematic reviews are similar to those outlined above for randomised controlled trials. In addition, registration with PROSPERO will encourage researchers to comply with the PRISMA statement (http://www.prisma-statement.org) when reporting their findings in a peer review publication. Similar to the registration of randomised controlled trials, the registration of systematic reviews with PROSPERO is intended to be undertaken when the review protocol has been completed, but before the commencement of screening of primary studies for inclusion into the review. Unlike the situation with trials, however, it is not currently suggested that all systematic reviews must be prospectively registered before they can be considered for publication in a peer review journal. Nonetheless, a mandatory policy remains a future possibility.
The BJOG editorial policy is to encourage registration. We would like systematic reviews to be registered as soon as the review protocol has been completed and before the literature search begins. However, reviews submitted to BJOG that have not been prospectively registered will still be considered for publication. The rationale for this decision is as follows. Firstly, there is no universal agreement amongst biomedical journals that mandatory prospective registration for systematic reviews should be implemented at this stage. Secondly, the issue of reporting biases with deviation from the original protocol might appear to be less relevant with systematic reviews as compared with randomised controlled trials. However, large studies of systematic reviews have indicated that there is still a real risk of selective reporting bias. In one study, over a fifth of protocol–review pairings were found to contain a discrepancy in at least one outcome measure, mainly the primary outcome measure. There was potential bias in a third of these reviews, with changes being made after knowledge of the results. Outcomes that were promoted in the review were more likely to be significant than if there was no discrepancy (relative risk 1.66; 95% CI 1.10–2.49; P = 0.02).3 We acknowledge that it might often be necessary and desirable to modify the original review protocol,4 including the questions a review addresses, in light of the evidence accumulated in the process of conducting the review.5 For example, new relevant clinical outcomes might need to be included in the list of studied outcomes, and this may only come to light when the full transcripts of the included primary studies are examined in closer detail. Indeed, in one analysis the greatest variation between protocols and reviews was in the methods section, in which two-thirds of reviews had undergone a major change.6 In such instances, PROSPERO allows such changes as part of the registration record, and makes them transparent to reduce the risk of data-driven questions generating false conclusions based on spurious results.5
Thirdly, the issue of avoiding duplication of research is valid for both systematic reviews and randomised trials. This is best achieved by prospective registration. Even between submission for publication and the appearance of the paper in a journal, there is the possibility of someone inadvertently duplicating the study. However, please note that at present retrospective registration with PROSPERO is not possible. There may be occasions when another systematic review on the same topic is justified (perhaps with different methodology, for example), but this should at least be carried out in the knowledge that a prior study is underway or has been completed.
It should be noted that the ICMJE is now encouraging researchers (although this is not currently a mandatory requirement) to deposit a structured abstract of the study after its completion if the results do not appear in a peer-reviewed publication after 24 months.1 The publication of such a brief structured abstract does not constitute duplicate publication should the full manuscript be subsequently published elsewhere, and may reduce publication bias (as many studies with negative results currently end up not being published). However, contrary to primary studies, a lack of resources or organisational support, but not the statistical significance of the results, has been shown to be a consistent reason for not publishing systematic reviews.7,8 Thus, the true value of publication of such abstracts for systematic reviews remains (as yet) unknown.
In conclusion, BJOG have decided to recommend registration with PROSPERO for all systematic reviews. This will improve the transparency and rigour of secondary research. We will not make it a requirement until the wider scientific community is educated on the advantages of registration for systematic reviews, and further evidence or consensus becomes available. We invite our readers to comment on this addition to our editorial policy, the benefits of registration with PROSPERO, and the perceived barriers and facilitators to registration of systematic reviews.