What is new? In registry records of published noninferiority trials, the noninferiority design was not mentioned in 60% of studies. The noninferiority margin was not reported in most registry records of noninferiority trials. This is in line with the finding that records from publicly accessible registries are often incomplete. Trial registries should search for ways to improve the reporting of noninferiority trials.
The International Committee of Medical Journal Editors (ICMJE) and others have been promoting the registration of clinical trials to prevent selective publication of clinical trials depending on their results, that is, publication bias and increase the transparency of performing and reporting clinical trials [1]. Since 2005, the journals that are members of the ICMJE require registration in a public trials registry as a condition of consideration for publication [2], which was followed by a substantial increase in the number of clinical trial registrations [3]. At present, authors are required to complete a web-based form, which must cover 20 key items, including a description of the study design [1].
Noninferiority trials are increasingly used in the licensing of new drugs and published in the medical literature [4]. The choice of a noninferiority margin, which determines the boundary for deciding whether the result of a trial demonstrates noninferiority of an intervention, is crucial for noninferiority trials [5], [6], [7]. There are no fixed rules that determine the choice of these margins [8], [9]. However, it is essential that it is known which margins were used a priori in the planning phase of a trial and that they are not chosen or modified post hoc at analysis stage [10]. We examined the reporting of noninferiority studies in publicly accessible trial registries.
We used a dual and complementary search strategy and identified registry records by starting both from publications and registry:
We examined the registry records of published noninferiority trials.
We also directly identified registry records of noninferiority trials in two large publicly accessible databases.
The search strategy identified 133 noninferiority trials published from 2005 onward (Fig. 1). For 88 published noninferiority trials, an accompanying protocol in a trial registry was found. One study was registered in EudraCT, the clinical trial database of the European Medicines Agency, which is not publicly accessible. We identified 153 records from searching the two trial registries, of which one record did not relate to a noninferiority trial. Three records were excluded because they were
Trial registration serves the goal of creating a record of all clinical trials, independently of publication, and thus to prevent publication bias. Moreover, the recording of essential methodological details early on allows the scientific community to detect and judge later changes in key aspects of the study design [12], [13]. The study design is one of 20 items covered by the minimal registration data set [1]; however, we found that the noninferiority design and margin were often not included
A survey of 21 clinical trial registries found that only 11 provided guidelines for registration, and the compliance of trial registries with the 20-item minimum data set suggested by the World Health Organization (WHO) was a median of 14 [9]. Other studies found that noninferiority design [5] and subgroup analyses [10] are frequently not reported in registry records. Our study adds that trial status information is frequently misleading in case of discontinued trials.
Fig. 2 documents the disposition of the review process. We located 38 articles that analyzed the effectiveness of interventions to prevent or reduce publication bias [20–57]. The included research provides evidence regarding: prospective trial registration; the peer review process; disclosure of conflicts of interest (CoIs); and open-access publishing.
The Consolidated Standards of Reporting Trials (CONSORT) statement has been extended to improve the reporting of such trials [7]. The integrity of the NIT cannot be affirmed if authors do not accurately report the prespecified noninferiority margins and the relevant confidence intervals [11]. Authors must document the margins selected during the planning phase, and ensure that these margins are not chosen or modified post hoc, during analysis [4].
The complete and detailed registration of primary and secondary outcomes in trial registries would greatly support the appropriate interpretation of results reported in journal articles and help prevent outcome-reporting bias. Records from publicly accessible registries are, however, often incomplete [10,22]. A study of randomized clinical trials from cardiology, rheumatology, and gastroenterology indexed in MEDLINE in 2008 found that less than half of trials (147 of 323, 46%) were registered before the end of the trial, with the primary outcome clearly specified [10].
Conflict of interest statement: The authors declare that they do not have any conflict of interest.
Funding: This study was funded by intramural funds of Leiden University Medical Center and University of Bern.
