Identifying RCTs for Systematic Reviews: An Evaluation of the Accuracy of Clinical Trial Register and Electronic Bibliographic Database Searches

Article type
Authors
Kaul T1, Colombijn JM2, Vernooij RW2, Spijker R3, Idema DL3, in ‘t Veld LF3, Damen JA3, Hooft L3
1Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht University, Utrecht
2Department of Nephrology and Hypertension, University Medical Centre Utrecht, Utrecht University, Utrecht
3Cochrane Netherlands, Julius Centre for Health Sciences and Primary Care, University Medical Centre Utrecht, Utrecht University, Utrecht
Abstract
Background:
Cochrane guidelines recommend broad electronic bibliographic database (EBD) searches and additional clinical trial register (CTR) searches to inform systematic reviews (SRs) of randomized clinical trials (RCTs) involving interventions. This two-step approach limits the number of unidentified RCTs to prevent the synthesis of biased results. Clinical trial registration is performed prospectively for RCTs; thus, the publication status does not interfere with the availability of the CT registration within the register. Performing broad CTR searches by themselves may be a more efficient approach to identify RCTs for SRs of RCTs involving interventions.

Objectives:
To evaluate how broad CTR searches compare to broad EBD searches in identifying RCTs for SRs of RCTs involving interventions.

Methods: We performed identical searches in the CTRs of ClinicalTrials.gov and the International Clinical Trials Registry Platform, as well as the EBDs CENTRAL, EMBASE, and MEDLINE, and compared whether we could identify the same RCTs. Our search was based on a subset of RCTs from an SR that studied the underrepresentation of people with chronic kidney disease from cardiovascular trials.

Results:
We included 92 completed RCTs. Of these, 80 had results available. Both sources identified 64 completed RCTs with available results (80% agreement [95% CI 80 – 87]). We identified eight completed RCTs with results exclusively by CTR search (10% [95% CI 5 – 19]) and eight exclusively by EBD search (10% [95% CI 5 – 19]). 12 RCTs were completed but lacked results. Also, we identified 42 eligible ongoing RCTs: 16 by both sources and 26 exclusively by CTR search. Lastly, we identified four RCTs of unknown status.


Conclusions:
CTR searches identify a greater number of completed RCTs than EBD searches. Both searches missed some completed RCTs. To ensure comprehensive identification of all RCTs, reviewers should search both CTRs and EBDs.

Patient, public, and/or healthcare consumer involvement: No involvement. Our study has implications for the methodology of SRs of RCTs involving interventions.