Document Type

Poster

Publication Date

9-26-2016

Comments

Poster presented at L3 Conference in Philadelphia PA.

Abstract

BACKGROUND:

ClinicalTrials.gov (CT) is an increasingly important resource for systematic reviewers attempting to identify published and unpublished clinical studies. In addition to clinical studies, however, some searches of the CT database also return systematic reviews (SRs). When I inquired about the SRs appearing in the results, the NLM Help Desk responded that “We do not recommend that systematic reviews be entered in ClinicalTrials.gov, since we only want the results of a clinical trial entered once. However, we will not refuse them if they are entered.” I wanted to find out how many SRs are included, describe their characteristics, and suggest search strategies for those wishing to exclude them.

METHODS:

Conduct a CT search for “systematic review” without limiting by field in case an SR was not explicitly titled as such. Screen the results for those records representing SRs as opposed to, e.g., mentioning one in the background to a clinical trial. Identify the total number of SRs. Test strategies for their ability to exclude them and calculate sensitivity, precision and specificity.

RESULTS:

I ran a search for “systematic review” (in quotes) in the advanced search > Search Terms (field) on July 14, 2016, and applying no other limits, downloaded 181 results for analysis from among the 220,113 total number of records in the CT database. Of the 181 records, 47 (26%) were systematic reviews. All 47 were listed as Study Type: Observational. The remaining 134 records that were not SRs included a mix of Observational (21, 15.7%) and Interventional (113, 84.3%) study types. Title searching offers an effective way to avoid SRs: all but two true SRs had “systematic review” or “meta-analysis” in the Brief or Official Title. So in the expert search you could add the filter: NOT ( "systematic review" [TITLES] OR "metaanalysis" [TITLES] ). This filter has a sensitivity of 94.8%, precision of 96.9%, and specificity of 91.5%.

CONCLUSION:

The number of systematic reviews registered in CT is small at this time. They can be accurately avoided if you are looking for interventional studies by using the Study Type field, but not if you are looking for observational studies. Using the proposed title searching filter offers an effective way to avoid them. Librarians should advise their teams to register systematic reviews in appropriate sources such as PROSPERO (http://www.crd.york.ac.uk/PROSPERO/), but not ClinicalTrials.gov.

L3SupplementalData.csv (146 kB)
CSV file with supplemental data

Share

COinS