Identifying and assessing study filters in searches for non-randomised intervention studies




Poster session 2 Thursday: Evidence synthesis - methods / improving conduct and reporting


Thursday 14 September 2017 - 12:30 to 14:00


All authors in correct order:

Hausner E1, Metzendorf M2, Richter B2, Bandeira-Echtler E2, Lotz F1, Rodenhäuser L1, Waffenschmidt S1
1 IQWiG, Germany
2 Cochrane Metabolic and Endocrine Disorders (CMED) group, Germany
Presenting author and contact person

Presenting author:

Elke Hausner

Contact person:

Abstract text
Background: Randomised-controlled trials show the least uncertainty in results as long as their methods are adequate and they are conducted in a manner suited to answer the study objective. However, in some cases, for example, where dramatic effects or very rare diseases are to be investigated, evidence from non-randomised intervention studies (NRS) might need to be included. In order to identify NRS in bibliographic databases it should thus be clarified whether it is possible by means of search filters to restrict the search to certain study types. NRS include a wide range of study types (e.g. controlled clinical trial, before-after study, cohort study). These studies may be difficult to identify in the literature, as “study design labels are not used consistently by authors and are not indexed reliably by bibliographic databases” (1).

Objectives: The aim of the present analysis was the validation of existing search filters to identify NRS in MEDLINE using the relative recall approach on the basis of Cochrane reviews.

Methods: In an initial step we determined NRS study types (2) and existing study filters for NRS; for this purpose, we screened among others the website of the InterTASC Information Specialists' Sub-Group (3). Cochrane reviews evaluating NRS were identified via the Cochrane Database of Systematic Reviews. To create a reference set, the information on these NRS was extracted, together with the information on the study type, and the corresponding citations were identified in Medline. The search filters were examined with regard to the accuracy measures 'sensitivity' and 'specificity'.

Results and Conclusions: A total of 391 eligible Cochrane Reviews were identified. Data extraction and the testing of the search filters are currently being performed. The results and conclusions of our analysis will be presented at the Summit.

1. Reeves BC. Chapter 13: Including non-randomized studies. URL:
2. Hartling L. J Clin Epidemiol 2011; 64(8): 861-871.
3. InterTASC Information Specialists' Sub-Group. URL: