When it comes to clinical trials, publication bias is a well-acknowledged problem. Statistically speaking, trial sponsors are more likely to publish studies that show positive findings, while burying trials that return negative or inconclusive results. This can add a misleading skew to the literature and impede clinicians’ understanding of a drug.
The detriments are huge, not least when it comes to the likelihood of conducting an accurate meta-analysis. In recent years, the pharmaceutical community has faced calls to address the situation and report all trial results irrespective of outcome.
For the last decade, certain prominent medical journals have asked trials to be registered in a public database as a precondition for publication. This became a legal mandate in 2007, when the US Food and Drug Administration Amendments Act came into force. Since then, US sponsors have been required to register their trial at ClinicalTrials.gov, posting summarised results within a year of study completion.
Seven years later, however, compliance remains low, and proposed legislation in Europe has yet to come to fruition. As pressure continues to mount, a new study has been published in the August edition of the journal PAIN, which sheds light on the scale of the problem.
The ICTRP in action
In this study, entitled ‘RReACT goes global: Perils and pitfalls of constructing a global open-access database of registered analgesic clinical trials and trial results’, investigators detail their experience using the World Health Organization’s International Clinical Trials Registry Platform (ICTRP). In theory, this platform should provide good access to data: it opens a portal to 15 different clinical trials registries including ClinicalTrials.gov and the EU Clinical Trials Register. Each has free public access and government support, and gives salient information about study design.
As clinical trials continue to grow in size and complexity, more and more companies are adopting software-as-a-service applications.
Unfortunately, the investigators faced substantial challenges along the way. Led by Dr Michael Rowbotham, scientific director of the California Pacific Medical Center Research Institute in San Francisco, they found that access to unbiased trial results was inadequate and in need of a shake-up.
As Rowbotham told the American Association for the Advancement of Science: “A distressingly large number of trials have no published results at all or are mentioned only in sponsor press releases. Recent analyses have found that only 25-35% of clinical trials required to post studies on ClinicalTrials.gov actually do so.”
The team had been using the platform to build the Repository of Registered Analgesic Clinical Trials (RReACT) database. Focusing on three chronic pain disorders – fibromyalgia, post-herpetic neuralgia and painful diabetic peripheral neuropathy – their database initially drew upon randomised trials registered on ClinicalTrials.gov, but was later expanded to include all the primary clinical trial registries in the ICTRP. In total, they identified 447 relevant trials, 86 of them listed on multiple registries.
Their aim – to create a global database of registered trials and trial results – was simple enough in theory but proved surprisingly difficult in practice.
Perils and pitfalls
As Dr Rowbotham explained: “We identified several perils and pitfalls of using the ICTRP. Manual searches are necessary, as ICTRP does not reliably identify trials listed on multiple registries. Searching ICTRP as a whole yields different results from searching registries individually. Outcome measure descriptions for multiply-registered trials vary between registries. Registry-publication pairings are often inaccurate or incomplete.”
He added that, ideally speaking, searching on PubMed for the trial registration number would reveal all articles pertaining to that study. Unfortunately, a recent analysis has shown that about 40% of journal publications miss out the registration numbers.
Finally, any so-called ‘grey’ literature results (informally published written material such as reports) are impermanent. This means that trial-specific press releases, company statements or information found on pharmaceutical company websites may be difficult to trace.
The key challenge here lies in synthesising information across the ICTRP’s numerous registries. It’s a job that requires a great deal of care and attention: far from being a simple matter of inputting a search term, investigators need to pay regard to a wide range of formatting differences. Because of inconsistencies between registries, along with incomplete data fields and poorly updated trial records, it can be hard to tell whether you are viewing two separate trials or a single trial listed twice.
The investigators commented that, while creating a single global registry would undoubtedly solve this problem, it would be a challenging and perhaps untenable goal. Because the ICTRP offers a search portal into 15 different regional or national registries, any
This being the case, the investigators’ suggestions for improvements were non-prescriptive and open-ended.
Doctors and researchers are seeking ways to treat severely obese individuals without the drastic and expensive option of gastric bypass surgery.
“Compliance might improve, especially for difficult-to-publish ‘negative’ studies, if posting results on trial registries could be made simpler and uniform,” said Rowbotham.
“Alternative solutions to the problems of publication bias and selective reporting should also be explored. These might involve including journals specialising in publishing ‘negative’ results, creating user-friendly and publicly available databases to publish results, and raising the awareness of authors, reviewers, and editors about these issues.”
The difficulties associated with publication bias are unlikely to vanish overnight. A dispiritingly low proportion of sponsor companies comply with, or perhaps are even aware of, the legal requirements surrounding trial registration and reporting. And of those that do report, not all include the trial registration number. In consequence, it can be extremely difficult to ensure that search results can be accurately paired with a trial.
On the other hand, for all its flaws, the overall aims of the ICTRP are laudable. As the study in PAIN makes clear, a few concrete changes to the portal and the way it is used could make a world of difference.