6.2.1 What and where to search
The MECIR guidance on the conduct of new Cochrane intervention reviews states in the section Searching for studies (C24-38) that it is mandatory to search the Cochrane Review Group’s specialised register. CENTRAL and MEDLINE should be searched either specifically for the review or through the Group’s specialised register. Embase should also be searched if it is available to the Group.
It is also mandatory to search trials registries for Cochrane intervention reviews. These should include ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform.
It is highly desirable to search appropriate national, regional and subject specific bibliographic databases. Examples of these include Literatura Latino Americana em Ciências da Saúde (LILACs, a database covering the Latin American region), and PsycINFO (a subject specific database covering psychology literature)
The Centralised Search Service (part of the Evidence Pipeline project) aims to search several databases for references to RCTs and quasi RCTs, and publish them in CENTRAL. The databases searched include Embase. You may therefore want to consider whether a separate search of Embase is required.
6.2.2 Searching regulatory databases
An editorial on the Cochrane Library by Jeppe Schroll and Lisa Bero concluded that "searching regulatory data from the EMA and the FDA should be part of any Cochrane Review of drug interventions". This recommendation was based on research showing "that including unpublished data from regulatory agencies changed the results of the original meta-analysis".
The European Medicines Agency (EMA) and USA Food and Drug Administration (FDA) databases should be searched if all three of the following circumstances apply to a research question
- If the drug/device is newly approved by EMA or FDA (last 5 years) or is utilised for a new indication (last 5 years)
- If the drug/device has been approved for use for the condition in question
- When the drug/device is compared to placebo for the comparisons in the review; although searches for reviews of other drug comparisons might also yield results
Searching the databases under the above conditions could provide useful study level data on drug efficacy. These sources are difficult to search and the search interfaces change regularly. The data can be difficult to analyse, therefore CISs should discuss with their author teams whether the above circumstances apply to the review question, and whether to search these sources.
For more information, see the paper on "Guidance on searching regulatory databases" on the Information Specialists Portal.
6.2.3 Study design search filters
Filters are pre-written search strategies designed to retrieve a particular type of record. The Cochrane Highly Sensitive Search Strategy (CHSSS) for identifying randomised controlled trials in MEDLINE has been developed and tested and can be added to your search strategy to retrieve RCTs and quasi-RCTs.
- For full details of search filters, see the Cochrane Handbook 4.4.7 Search filters
- A filter has been developed for Embase by the team involved in the Centralised Search Service. For more information, see the Cochrane Library help pages (scroll down to Set 2: Records retrieved by the new search).
- See an example of a MEDLINE search strategy for a Cochrane review (Appendix 6).
- See an example of a CENTRAL search strategy for a Cochrane review (Appendix 7).
6.2.4 Other methods of finding studies
Other methods of identifying relevant studies include:
- checking the references of relevant studies (from the Cochrane review and other related reviews), together with citation searches
- the use of the ‘related articles’ feature in electronic databases
- ‘citing alerts’ in electronic journals
6.2.5 Design and structure of search strategies
For information about designing search strategies, see the Cochrane Handbook 4.4.2 Structure of a search strategy
Designing a Cochrane search strategy usually involves providing sets of terms to search for the:
- health condition of interest (i.e. the population) e.g. pregnant women or people with melanoma
- the intervention(s) e.g. drug or behaviour
- the types of study design to be included (typically a filter for randomized controlled trials, as described above) i.e. RCT / quasi-RCT
These terms are usually combinations of database thesaurus terms (i.e. MeSH terms), and text words (e.g. terms found in the title or abstract of a record) and a ‘filter’.
6.2.6 Peer review of search strategies
See Press Forum for information on how to get your search strategy peer reviewed by other information specialists.