We would like other editors’ opinions as to whether adhering to the journal’s policy on trial registration may contribute towards the non-publication of trial results (and thus bias in the literature).
All of our journals have the same policy on trial registration—for studies started before July 2005, we permit retrospective registration (providing it was done before submission) but for trials started after July 2005, we require registration to have been done before enrolment of the first participant.
In recent years we have tried to enforce this policy strictly and have rejected many papers reporting trial results where the trial was not registered in line with our policy. A very quick audit reveals about 20 or so studies which have been rejected in the past year or so due to non-compliance. For these studies, we have informed the authors of our policy and advised them they can retrospectively register, for free, at clinicaltrials.gov, where they are also able to deposit results (should they be unable to secure journal publication).
We are now also considering removing the ‘grandfather’ clause for old trials (pre-2005) (ie, that we would be unable to consider these for publication unless they had also been prospectively registered). One specific case we handled recently relates to a large cancer screening trial, which we felt was likely to be methodologically sound, and addressed a question where there was little data in the peer reviewed literature. The results would undoubtedly have informed future clinical practice and may have had the potential to save lives—the trial results suggested benefit for a screening approach which is not currently routine practice. The trial had been done after July 2005 but was only registered some months after enrolment started. We rejected purely on the basis that the study had not been properly registered in line with our (and other journals’) policies. We did not think that selective reporting was an issue with this particular study.
However, would other editors have a concern about strict application of the registration rule (ie, that editors have a duty not just to apply policy regardless but to consider their responsibilities to the integrity of the evidence base in a more flexible way, which in some cases may be achieved by overriding their own policies)? However, we are concerned that having some laxity in our policy (eg, with the grandfather clause for older trials) may encourage triallists to think that prospective registration is not mandatory (and thus avoid registration in the future).
The advice from the Forum was that, in general, it is probably best to judge each paper on a case by case basis. Some argued for leniency in the case of trials published in and around 2005, as many authors may not have been aware of the new requirements for registration. However, whether or not a trial is registered has little bearing on the quality or ethics of the study, and so it is up to the editor to decide whether or not a study should be published. All authors should now be aware of the regulations so there is no excuse for not registering a trial. However, it was pointed out that many journals do not require trial registration and so the authors can always have their work published elsewhere. Editors need to balance the benefits of having a strict policy requiring prospective registration (which should encourage trial registration) against the harms of non-publication of individual studies that have not been properly registered but may contain valuable information.
The advice from COPE was very helpful, and we are still handling new cases which relate to this problem. Our journals have not changed their policies. We still have a strict policy requiring prospective registration for trials done since mid-2005, although we are more lenient still and permit retrospective registration for trials done before mid-2005. Very sadly, the journals continue to reject some numbers of trials which were not registered in line with our policies. We continue to refer authors to clinicaltrials.gov to deposit their results along with their registry details.