Elizabeth Loder: The persistent problem of unregistered clinical trials

How can we get to zero unregistered clinical trials? Elizabeth Loder discusses.

elizabeth_loderIn 2004 the International Committee of Medical Journal Editors (ICMJE) announced that clinical trials commencing after 1 July 2005 would be subject to a new trial registration policy. In order to be published in member journals, trials would have to be registered in an approved trial registry before enrollment of the first participant. The registry entries, in publicly accessible databases, would provide a reliable record of a trial’s existence and planned outcomes. The requirement for prospective trial registration was intended as a safeguard against well-documented and pervasive problems of selective reporting and non-publication of clinical trial results. No one should doubt that “biased under-reporting of clinical trials kills people and wastes money.”

Things have improved dramatically since then, but we are nevertheless a long way from realizing the full potential of trial registration. Timely registration is a requirement that some researchers and journals still do not take seriously. Unregistered and retrospectively registered trials continue to be published, especially by small journals whose editors do not uniformly embrace or understand the need for trial registration. Even the Committee on Publication Ethics takes a soft line on late trial registration, suggesting that “it is probably best to judge each paper on a case by case basis.” Unfortunately, this sends the message that although prospective trial registration is required, no one really means it. Perhaps such permissive attitudes explain why a recent analysis of data from the largest trial registry, Clinicaltrials.gov, shows that about a third of trials from 2012 through 2014 were registered late, defined as more than three months after their start.

This is far worse than the situation at The BMJ, but these findings and our experience both demonstrate a persistent problem. Since the spring of 2013 I’ve kept an informal list of clinical trials submitted to The BMJ that were rejected because they were unregistered or were registered late (retrospectively).  As of this writing my list contains 104 entries. That’s a small number in relation to the thousands of research papers we’ve received over four years, but it’s big in relation to zero, which should be the goal. And of course the situation at The BMJ and other major journals is likely to be a best-case scenario.

What would it take to make unregistered or retrospectively registered clinical trials a thing of the past? One solution might be to treat them as the research equivalent of medical “Never Events.” These are “particularly shocking medical errors (such as wrong-site surgery) that should never occur.”

Like medical Never Events, unregistered trials are “of concern to patients, policy makers, and healthcare professionals and providers.” They thwart a system designed to ensure the honest reporting of medical research. Also like medical Never Events, late or unregistered clinical trials are “clearly identifiable and measurable (and thus feasible to include in a reporting system).” Audits of various kinds can be undertaken to assess compliance with registration and outcome reporting. Such audits are relatively easy to perform—I’ve done one with some of my clinical colleagues—and arguably already serve as a kind of an informal reporting system.

Unregistered clinical trials are likewise “of a nature such that the risk of occurrence can be reduced by establishment of protocols, policies, and procedures.” Indeed, almost nothing lends itself so well to protocols and procedures as the simple act of matching a registration number to an online record in a registry or including trial registration on a checklist of items to be completed before funding is dispensed or a trial is allowed to proceed.

Finally, as with medical Never Events, late or nonexistent trial registration tells us “something fundamental about the quality, care, and safety processes in an organisation.” Attention to detail is the hallmark of good research. Small errors are more common in retracted papers compared to others, suggesting that such mistakes might be “an early and accessible signal of unreliability in clinical trial reports.” It is reasonable to wonder whether a research team unable to comply with registration requirements might have overlooked other vital details.

The two things—failure to register trials and medical Never Events—thus have much in common, suggesting that their solutions might too. In the hospital world, medical Never Events trigger a harsh response. US government health programs and many private insurers will not pay for additional costs associated with Never Events. Many authorities require public reporting of such events or levy fines when they occur. Such consequences ensure that avoidance of Never Events is uppermost in the minds of those in the best position to prevent them.

We could do something similar with late or unregistered clinical trials. Funders and ethical approval bodies might make full payment or final approval of trials contingent on prospective registration. Journal editors should continue to refuse to publish such studies and notify institutional or company authorities of the problem when possible. Trial results can be publicly posted, for example on Clinicaltrials.gov or sponsor websites, and thus need not be lost to posterity even if no journal will publish them. Institutional discipline could take many forms, including additional oversight of subsequent research projects or requiring that such omissions be considered in decisions about academic promotion.

Getting to zero is only possible if all parts of the research community work together and make prospective registration of clinical trials a top priority.  A main goal of the requirement for trial registration was to restore public trust in medical research. Trustworthy clinical trials = prospectively registered trials that faithfully report their prespecified outcomes. Unregistered clinical trials are serious, preventable events that should never happen—and must come first on any list of Research Never Events.

Elizabeth Loder is the head of research, The BMJ.

  • What about “conditional ethics approval” for trials? So rather than grant blanket ethics approval up front (maybe asking PIs nicely to registered their trials at the time), issue a conditional ethics approval that only becomes valid from the moment the trial is registered.

    So any PI starting a trial without pre-registering it first would be running a trial that lacks ethics approval and is thus illegal. This could be a powerful “nudge” – or not?

    Would be interested in your thoughts on this. Thank you!
    Till Bruckner, http://www.TranspariMED.org