Date Published: March 7, 2018
Publisher: Public Library of Science
Author(s): Jessica Fleminger, Ben Goldacre, Omid Beiki.
Trial registries are a key source of information for clinicians and researchers. While building OpenTrials, an open database of public trial information, we identified errors and omissions in registries, including discrepancies between descriptions of the same trial in different registries. We set out to ascertain the prevalence of discrepancies in trial completion status using a cohort of trials registered on both the European Union Clinical Trials Register (EUCTR) and ClinicalTrials.gov.
We used matching titles and registry IDs provided by both registries to build a cohort of dual-registered trials. Completion statuses were compared; we calculated descriptive statistics on the prevalence of discrepancies.
11,988 dual-registered trials were identified. 1,496 did not provide a comparable completion status, leaving 10,492 trials. 16.2% were discrepant on completion status. The majority of discrepancies (90.5%) were a ‘completed’ trial on ClinicalTrials.gov inaccurately marked as ‘ongoing’ on EUCTR. Overall, 33.9% of dual-registered trials described as ‘ongoing’ on EUCTR were listed as ‘completed’ on ClinicalTrials.gov.
Completion status on registries is commonly inaccurate. Previous work on publication bias may underestimate non-reporting. We describe simple steps registry owners and trialists could take to improve accuracy.
Trial registries have been developed over the past three decades to address ongoing structural problems in evidence-based medicine including non-publication of trial results [1–3], selective reporting of results in trial publications , and duplication of research . Trial registration is required by law for some trials in various territories including the United States (US) (Food and Drug Administration Modernization Act (FDAMA), Food and Drug Administration Amendments Act (FDAAA)) and the European Union and European Economic Area (European Medicines Agency (EMA)). Registration is also the subject of guidance from organisations including the International Committee of Medical Journal Editors (ICMJE) and the World Health Organisation (WHO), which provide international standards on the types of information to be included in a clinical trial registries.
Using the OpenTrials database we created a dataset of all trials registered on both ClinicalTrials.gov (as of 02/06/2017) and EUCTR (as of 15/06/17). The ClinicalTrials.gov and EUCTR data available in the OpenTrials database were generated from a simple download of both registries and therefore contained all trials registered on these registries, as of the date that they were accessed. We required any one of three conditions to be met in order to classify a trial as registered on both registries: 1) the ClinicalTrials.gov ID was listed in the EUCTR entry; 2) the EUCTR ID was listed in the ClinicalTrials.gov entry; or 3) the trial title was an exact match on both registries. Trials were matched from EUCTR to ClinicalTrials.gov using the ‘US NCT (ClinicalTrials.gov registry) number’ field in EUCTR records. Trials were matched from ClinicalTrials.gov to EUCTR using the ‘Other Study ID Numbers’ field in ClinicalTrials.gov. In addition to matching using secondary IDs, we also matched trials using the text of the title (‘Official Title’ in ClinicalTrials.gov and ‘Full title of the trial’ in EUCTR). In order to limit the number of incorrect matches, we only matched trials if the full text of the title was exactly the same on the two registries. We excluded any one-to-many or many-to-many matches. As all trials registered on EUCTR are interventional clinical trials, the final dataset of matched trials only included interventional clinical trials.
As at 15/06/2017 EUCTR contained 30,492 trials, 3,141 (10.3%) of which contained a ‘US NCT (ClinicalTrials.gov registry) number’. As at 02/06/2017 ClinicalTrials.gov contained 237,230 studies, 63,166 (26.6%) of which contained at least one secondary ID in ‘Other Study ID Numbers’. Matching from ClinicalTrials.gov to EUCTR using the ClinicalTrials.gov ‘Other Study ID Numbers’ data field yielded 8,003 matches. Matching from EUCTR onto ClinicalTrials.gov using the EUCTR ‘US NCT (ClinicalTrials.gov registry) number’ data field yielded 2,913 matches, of which 1,250 (43.0%) were already matched by the preceding method. Matching on title yielded 5,318 matches, of which 2,996 (56.3%) were already matched by the preceding two methods. In total this yielded a dataset of 11,988 dual-registered trials (see Fig 1). Of the 11,988 dual-registered trials, EUCTR provided a matching ClinicalTrials.gov ID for 24.3%, ClinicalTrials.gov provided a matching EUCTR ID for 66.8% and 44.4% had identical titles for matching.
In a dataset of 10,492 dual registered trials, we found that 16.2% were discrepant on completion status. The majority of these discrepancies (90.5%) were due to a trial being described as ‘completed’ on ClinicalTrials.gov, but still marked as ‘ongoing’ on EUCTR; since a registry entry is manually changed to ‘completed’, this suggests that the data on EUCTR are more commonly incorrect. Of the dual-registered trials that were ‘ongoing’ on EUCTR, 33.9% were ‘complete’ on ClinicalTrials.gov; therefore in a third of cases the status appears to be incorrect, and the positive predictive value (the likelihood that a trial marked as ‘ongoing’ is actually ‘ongoing’) is only 66.1% for EUCTR. By comparison, only 5.1% of the dual-registered trials marked as ‘ongoing’ on ClinicalTrials.gov were ‘complete’ on EUCTR, making the positive predictive value 94.9% for ClinicalTrials.gov.
Trial completion status on registries is commonly inaccurate. 33.9% of dual-registered trials listed as ‘ongoing’ on EUCTR are listed as ‘completed’ on ClinicalTrials.gov. Previous work on publication bias may have underestimated the rate of non-reporting. Registry owners should undertake simple cross-checks of data to ensure completion status is accurate.