Background: As of September of 2005, all prospective trials need to be registered in a public trial registry at the beginning or before the onset of participant enrollment, as defined by the International Committee of Medical Journal Editors (ICMJE). Trial registries therefore serve as an excellent resource for identifying ongoing trials with evidence of completing yet unpublished trials suggesting potential publication bias. Thus trial registries, as recommended by the PRISMA checklist, should be searched for the detection of missing data to ensure a complete evidence body.
Methods: Based on a written a priori protocol we performed a comprehensive search in MEDLINE for systematic reviews (SRs) published in 10 high-impact, general medical journals (e.g. Journal of the National Cancer Institute, Journal of Clinical Oncology, NEJM, Lancet, BMJ, etc.) over a 5-year period (2012/01- 2016/12). In addition, we identified all cancer-related Cochrane reviews from the same period using the Cancer filter in the Cochrane Database of Systematic Reviews (CDSR). Two review authors extracted relevant data in duplicate and independently.
Results: We identified 178 high-impact SRs and 356 Cochrane reviews, which met our inclusion criteria. The majority of these addressed topics related to cancer in general (25.8%). Overall, the percentage of systematic reviews using trial registries to identify unpublished, ongoing or completed trials was 52.6% (281 of 534 reviews). Out of 356 Cochrane reviews, 72.9% (266 of 356 reviews) used trial registries, while the percentage of high-impact SRs using trial registries was 8.4% (15 of 178 reviews). 50% (268 of 534) of SRs using non-IPD data used trial registries, while 10.2% (6 of 59) IPD-data SRs conducted trial registry searches.
Conclusions: This study suggests that the majority of SRs published in high-impact journals do not include trial registries in their search strategies; this is in contrast to reviews published in the Cochrane Library where this is common practice. In light of the tremendous value of trial registries to reduce the risk of publication bias, it appears critically important to raise awareness of this issue.