11.8  Writing an abstract

All full reviews must include an abstract of not more than 400 words. The abstract should be kept as brief as possible without sacrificing important content. Abstracts to Cochrane reviews are published in MEDLINE and the Science Citation Index, and are made freely available on the internet. It is therefore important that they can be read as stand-alone documents.

 

The abstract should summarize the key methods, results and conclusions of the review and should not contain any information that is not in the review. Links to other parts of the review (such as references, studies, tables and figures) may not be included in the abstract. A hypothetical example of an abstract is included in Box 11.8.a.

 

Abstracts should be targeted primarily at healthcare decision makers (clinicians, informed consumers and policy makers) rather than just to researchers. Terminology should be reasonably comprehensible to a general rather than a specialist healthcare audience. Abbreviations should be avoided, except where they are widely understood (for example, HIV). Where essential, other abbreviations should be spelt out (with the abbreviations in brackets) on first use. Names of drugs and interventions that can be understood internationally should be used wherever possible. Trade names should not be used.

 

The content under each heading in the abstract should be as follows:

Background: This should be one or two sentences to explain the context or elaborate on the purpose and rationale of the review. If this version of the review is an update of an earlier one, it is helpful to include a sentence such as “This is an update of a Cochrane review first published in YEAR, and previously updated in YEAR”.

Objectives:  This should be a precise statement of the primary objective of the review, ideally in a single sentence, matching the Objectives in the main text of the review. Where possible the style should be of the form “To assess the effects of  [intervention or comparison] for [health problem] for/in [types of people, disease or problem and setting if specified]”.

Search methods: This should list the sources and the dates of the last search, for each source, using the active form ‘We searched….’ or, if there is only one author, the passive form can be used, for example, ‘Database X, Y, Z were searched’. Search terms should not be listed here. If the CRG’s Specialized Register was used, this should be listed first in the form ‘Cochrane X Group Specialized Register’. The order for listing other databases should be the Cochrane Central Register of Controlled Trials, MEDLINE, EMBASE, other databases. The date range of the search for each database should be given. For the Cochrane Central Register of Controlled Trials this should be in the form ‘Cochrane Central Register of Controlled Trials (The Cochrane Library 2007, Issue 1)’. For most other databases, such as MEDLINE, it should be in the form ‘MEDLINE (January 1966 to December 2006)’. Searching of bibliographies for relevant citations can be covered in a generic phrase ‘reference lists of articles’. If there were any constraints based on language or publication status, these should be listed. If individuals or organizations were contacted to locate studies this should be noted and it is preferable to use ‘We contacted pharmaceutical companies’ rather than a listing of all the pharmaceutical companies contacted. If journals were specifically handsearched for the review, this should be noted but handsearching to help build the Specialized Register of the CRG should not be listed.

Selection criteria: These should be given as ‘[type of study] of [type of intervention or comparison] in [disease, problem or type of people]‘. Outcomes should only be listed here if the review was restricted to specific outcomes.

Data collection and analysis: This should be restricted to how data were extracted and assessed, and not include details of what data were extracted. This section should cover whether data extraction and assessments of risk of bias were done by more than one person. If the authors contacted investigators to obtain missing information, this should be noted here. What steps, if any, were taken to identify adverse effects should be noted.

Main results: This section should begin with the total number of studies and participants included in the review, and brief details pertinent to the interpretation of the results (for example, the risk of bias in the studies overall or a comment on the comparability of the studies, if appropriate). It should address the primary objective and be restricted to the main qualitative and quantitative results (generally including not more than six key results). The outcomes included should be selected on the basis of which are most likely to help someone making a decision about whether or not to use a particular intervention. Adverse effects should be included if these are covered in the review. If necessary, the number of studies and participants contributing to the separate outcomes should be noted, along with concerns over quality of evidence specific to these outcomes. The results should be expressed narratively as well as quantitatively if the numerical results are not clear or intuitive (such as those from a standardized mean differences analysis). The summary statistics in the abstract should be the same as those selected as the defaults for the review, and should be presented in a standard way, such as ‘odds ratio 2.31 (95% confidence interval 1.13 to 3.45)’. Ideally, risks of events (percentage) or averages (for continuous data) should be reported for both comparison groups. If overall results are not calculated in the review, a qualitative assessment or a description of the range and pattern of the results can be given. However, ‘vote counts’ in which the numbers of ‘positive’ and ‘negative’ studies are reported should be avoided.

Authors’ conclusions: The primary purpose of the review should be to present information, rather than to offer advice or recommendations. The Authors’ conclusions should be succinct and drawn directly from the findings of the review so that they directly and obviously reflect the main results. Assumptions should generally not be made about practice circumstances, values, preferences, tradeoffs; and the giving of advice or recommendations should generally be avoided. Any important limitations of data and analyses should be noted. Important conclusions about the implications for research should be included if these are not obvious.