Methods used in reviews

Search strategies

Access to specialised register by authors

The specialised register is available for all authors to consult through the Cochrane Library. However, to avoid duplication of searching activities and to provide authors with a comprehensive search of the database, the Trials Search Co-ordinator will liaise with authors to construct and develop search strategies for each review, to be carried out at the editorial base. Searches for updating reviews will be carried out at the editorial base by the Trials Search Co-ordinator on an annual basis.

Additional search strategies

The Trials Search Co-ordinator will work with the author to create a specific search strategy for each author, which will then be run in EMBASE and MEDLINE in addition to the Cochrane Library. In order to identify unpublished trials, experts in the field and pharmaceutical companies will be contacted for information, and, where applicable, the Internet will also be searched. Reports of trials found by these methods will be added to the UGPD register of trials. Authors should search citations in each trial report for additional trials.

Study selection

The UGPD Group recommends that the trials included are randomised, pseudorandomised or controlled clinical trials which compare the test intervention with placebo or standard treatment. Other types of trials can be used where necessary.

Selection of studies for inclusion in a review should be performed independently by more than one author. The editor assigned to the review will work to resolve differences in study selection between authors.

Assessment of methodological quality

Procedures for the assessment of methodological quality are under development.

Advice on standard criteria for assessing quality will be given by the editorial team. Examples of quality assessment checklists are available. Methods are described in the Cochrane Authors Handbook which is available from the editorial team, on the Cochrane library and via the Cochrane Web sites.

In general:

  • An accepted method of quality assessment should be used.
  • Quality should be independently assessed by more than one author and the level of agreement should be reported in the review.
  • The editor responsible for the development of the review will resolve difference in quality assessment between authors.
  • Quality assessment will be reported in the methods and results sections of the review.

Data collection

The UGPD group recommends that the extraction of data is done independently by more than one author. The editor responsible for the development of the review will resolve difference in data interpretation between authors. Data verification with the person responsible for the study will not normally be required other than where the data is unpublished or confirmation of results are required.

Data from RCTs that have not been published will be eligible for use in systematic reviews prepared by the UGPD group, subject to verification of data by the primary investigator. The UGPD Group will not routinely collect and analyse data on rare adverse events collected from non-RCTS.

The UGPD group will request copies of data extraction forms (for included studies) to be submitted to the group when a draft review is submitted for peer referee.   Data extraction forms will be retained by the group, for reference or to aid a future review updater, should an initial review author not be able to update a review.  Data extraction forms will not be published nor used in any other way by the group or its members.


Statistical guidance is available from the editorial base (Statistical Editor:  Noori Akhtar-Danesh).
Data entry to RevMan should be done using the double data entry facility which allows more than one author to independently enter dataPolicies on statistical methods are under development. 

These will incorporate guidance derived from Section 9 of the Authors Handbook on analysing data and undertaking meta-analyses.

Heterogeneity of trials and issues such as crossover trials will be addressed.

Reporting of reviews

Discussion and conclusions section

The strength of the evidence should be categorised using the hierarchy of evidence scale detailed in CRD report 4, available from the editorial base.

The applicability of the results should be commented on taking into account the applicability of the trials to use of the intervention in standard practice for treatment of the disorder. Cost benefit analysis will not be routinely performed.

The use of non-RCT derived data when discussing results and drawing conclusions should be commented on in this section. Where applicable, other reviews will be cross-referenced in this section.

Tables and figures

Information in the excluded and excluded trials tables should be brief and structured to include the Study Identifier, Methods of the Trial, Participants, Interventions, Outcomes, and further Notes.

Each included study should ideally include a "Risk of bias" table including information on each study's and/or individual outcomes for bias from: sequence generation; allocation concealment; blinding; incomplete data, selective reporting or other means.

Studies in the excluded trials table should consist of those trials which were initially selected for assessment, but which later proved to be non-RCTs or ineligible for other reasons. Advice is available from the editorial team on the validity of trials for inclusion into the review.

Table of comparisons

Policies for the structure or order of outcomes are under development and depend to some extent on the outcomes we select as 'default' for this group.

The order of trials in the tables will be alphabetical, then by date. Trials will be named preferably by author surname (e.g. Smith 1998) or, where this is not possible, by trial group identifiers (e.g. Oesophageal Cancer Trials Collaborators Group OCTCG 1997).

There may be multiple publications from one trial. Such reports should be cross referenced to the original study, for example, a publication by Jones et al reporting data from the Oesophageal Trials Collaborators Group study of 1997, should be reported as Jones 1999 (OCTCG 1997).

Any factors which could be perceived as conflict of interest should be stated.