Descriptive analysis of non-randomized studies included in Cochrane Reviews regarding their availability in PubMed

Article type
Authors
Metzendorf M1, Richter B1, Bandeira-Echtler E1, Hausner E2, Waffenschmidt S2
1Cochrane Metabolic and Endocrine Disorders Group
2IQWiG
Abstract
Background:
Randomized controlled trials (RCT) are relatively easy to identify due to the availability of well-performing, validated search filters. This does not hold true for non-randomized study (NRS) designs. There are many uncertainties regarding the identification of NRS in bibliographic databases within the context of systematic review elaboration. One uncertainty relates to the bibliographic databases in which NRS can be predominantly identified.

Objectives:
To describe the percentage of NRS indexed in PubMed and analyze the results according to topic and type of intervention by investigating a comprehensive set of Cochrane Reviews (CR) including NRS.

Methods:
We identified CR evaluating NRS via the Cochrane Database of Systematic Reviews up to 20 October 2016. For the generation of the reference set, we screened CR evaluating NRS according to predefined inclusion criteria. For example, the CR had to evaluate an intervention on a health-related question and had to include study types beyond RCTs or controlled clinical trials (CCT). In addition, we considered only CR that included fewer than 65 studies in order to avoid bias related to a few CR containing many studies. We identified the citations of the studies included in the CR via the reviews’ bibliographies and the corresponding PubMed identification numbers (PMID) extracted from PubMed. The studies were classified following the study design classification scheme by Hartling 2011.

Results and conclusions:
A total of 271 CR from 41 different Cochrane groups were eligible for data extraction. The reference set included 2873 studies classified as NRS. We are currently calculating the percentages of studies indexed in PubMed, and will present results according to topic and type of intervention during the Cochrane Colloquium.

Patient or healthcare consumer involvement:
No patients or consumers have been involved, as this is a project of methodological nature relating to information retrieval.

Reference: Hartling L, Bond K, Santaguida PL, Viswanathan M, Dryden DM: Testing a tool for the classification of study designs in systematic reviews of interventions and exposures showed moderate reliability and low accuracy. J Clin Epidemiol 2011, 64(8):861-71.