4.5    Main text

The text of the review should be as succinct and readable as possible.  Although there is no formal word limit on Cochrane reviews, review authors should consider 10,000 words an absolute maximum unless there is a special reason to write a longer review (for example, where a large number of index tests are evaluated). The majority of reviews should be substantially shorter than this.  A review should be written so that someone who is not an expert in the area can understand it, in light of the following policy statement, reported in Cochrane News 1999; 15: 14:

 

“The target audience for Cochrane reviews is people making decisions about health-care. This includes health-care professionals, consumers and policy makers with a basic understanding of the underlying disease or problem.

 

It is a part of the mission and a basic principle of The Cochrane Collaboration to promote the accessibility of systematic reviews of the effects of healthcare interventions to anyone wanting to make a decision about healthcare. However, this does not mean that Cochrane reviews must be understandable to anyone, regardless of their background. This is not possible, any more than it would be possible for Cochrane reviews to be written in a single language that is understandable to everyone in the world. It is important to translate the content, or elements of the content, of reviews into different languages and formats targeted at different audiences including healthcare professionals, consumers and policy makers in a variety of circumstances.

 

Cochrane reviews should be written so that they are easy to read and understand by someone with a basic sense of the topic who may not necessarily be an expert in the area. Some explanation of terms and concepts is likely to be helpful, and perhaps even essential. However, too much explanation can detract from the readability of a review. Simplicity and clarity are also vital to readability.  The readability of Cochrane reviews should be comparable to that of a well-written article in a general medical journal.”

 

 

The text of a Cochrane review contains a number of fixed headings and subheadings that are available in the RevMan5 document structure.  Additional optional subheadings are also included in the structure, but authors should not be limited by these and should add their own subheadings where appropriate.  Authors are encouraged to use the optional subheadings included in the structure where possible, but not if they make individual sections needlessly short.

Text cannot be placed immediately under the following Level 1 headings: ‘Abstract’, ‘Methods’, ‘Criteria for considering studies for this review’, ‘Results’, or ‘Authors conclusions’.  Text for these sections starts below the first subsequent subheading.

 

Background                                                                             [fixed, level 1 heading]

Well-formulated review questions do not appear out of thin air. They occur in the context of an existing body of knowledge. This context should be explained in the background section of the review. The background helps set the rationale for the review, and should explain why the questions being asked are important and why closely related questions are not being covered. The background section of the review should inform the readers of the review about why the review is being done, particularly where there are existing systematic reviews published outside The Cochrane Collaboration. This can generally be done within 1 to 1.5 printed pages.

 

The background section contains 4 optional subheadings which can be used to help structure the text:

 

Target condition being diagnosed                                             [optional, level 2 heading]

A description of the target condition of interest (frequency, severity, prognosis and possible treatments). If there are Cochrane reviews of interventions for the target condition they should be cross-referenced here.

 

Index test(s)                                                                               [optional, level 2 heading]

A description of the index tests that are being evaluated in this review, whether they are currently used in clinical practice, and the roles being considered for the tests (for example, as replacement for a comparator test because of lower cost and invasiveness).

 

Alternative test(s)                                                                      [optional, level 2 heading]

A description of the possible diagnostic tests and strategies that could (and are) used in clinical practice, irrespective of whether they are evaluated in this review. This will help place the review in context of all available diagnostic options.

 

Rationale                                                                                   [optional, level 2 heading]

The background helps set the rationale for the review, and should explain why the questions being asked are important. It might also mention why this review was undertaken and how it might relate to a wider review of a general problem.

 

Objectives                                                                                [fixed, level 1 heading]

The review question should be clearly stated. The primary objective should be described and include the index test(s), the target condition verified by the reference standard (‘gold standard’) and the study population. Where possible the style should be ‘To determine the diagnostic accuracy of [index test] for detecting [target condition] in [patient description]’. It should also be clear from the objective whether the authors want to investigate the accuracy of an index test as triage for the current normal test regime, as a replacement for it, or in combination with it (see Chapter 5). The term ‘comparator’ is specifically used to refer to what is thought to be the most commonly used current normal test regime. Where the aim of the review is to make comparisons between tests these should be clearly stated.

 

Secondary objectives                                                                [optional, level 2 heading]

Complex reviews that investigate multiple questions may categorise their objectives as ‘Primary Objectives’ and ‘Secondary Objectives’. For example, the primary objectives may be to compare test accuracy between two tests; the secondary objectives may estimate test accuracy for each test at pre-specified thresholds.  However, secondary objectives related to investigating heterogeneity between study results should not be listed under this subheading but under the next subheading.

 

Investigation of sources of heterogeneity                                  [optional, level 2 heading]

Heterogeneity investigations explore factors which may affect diagnostic accuracy, and are essential because they may provide a framework by which the expected heterogeneity may be explained a priori and to provide a more clinically useful review. For example, the test accuracy of troponin may vary depending upon when the test is done after the start of symptoms. Providing a primary objective only, for example a comparison of troponin against Creatine Kinase-MB for patients with chest pain to detect myocardial infarction, is not sufficient for clinical decision making when patients present at different time points after the onset of chest pain, and the test performance of troponin may vary accordingly.  Alternative subgroups can be aspects of study design and execution, age groups, setting, differences in the presence or degree of clinical characteristics or differences in the operation of either the index test or the reference test.

 

Methods                                                                                   [fixed, level 1 heading]

The Methods section in a protocol should be written in the future tense. Cochrane reviews are updated as new evidence accumulates, therefore methods outlined in the protocol should generally anticipate a sufficiently large number of studies to address the review’s objectives (even if it is known this is not the case).

 

The Methods section in a full review should be written in the past tense, and should describe what was done to obtain the results and conclusions of the current version of the review. Often a review is unable to implement all of the methods outlined in the protocol, usually because there is insufficient evidence. In such circumstances, it is recommended that the methods that were not implemented still be outlined in the review, so that it serves as a protocol for future updates of the review. Some Cochrane review groups (CRGs) have policies on this issue, and these should be available from the Review Group Co-ordinator (RGC). Examples include adding an additional subsection at the end of ‘Methods of the review’, or including the methods for future updates in an additional table.

 

In the methods section, the authors should clearly describe the selection criteria for considering studies for the review (Chapter 5 and Chapter 6), the methods used to identify relevant studies (Chapter 7), the process used for selection of studies and collecting data (Chapter 8) and how the methodological quality of the included studies is assessed (Chapter 9). A statistician may best write the section for describing the statistical analysis and data synthesis (Chapter 10).  In addition, information about how to investigate sources of heterogeneity and any pre-planned sensitivity analyses should be described clearly here (Chapter 10).

 

Criteria for considering studies for this review                               [fixed, level 2 heading]

The eligibility criteria required for studies to be included in the review must be clearly stated. 

 

Types of studies                                                                                                        [fixed, level 3 heading]

Eligible study designs should be stated here, along with any thresholds for inclusion based on the conduct or quality of the studies.  For example, ‘All consecutive series of patients and case-control studies’ or ‘All study designs”.  Exclusion of particular types of studies (for example, case-control studies, retrospective studies) should briefly be justified. Restrictions based on the use of particular reference standards should not be listed here.

 

Participants                                                                                                                 [fixed, level 3 heading]

Specify the participants for whom the test would be applicable, including any restrictions on diagnoses, age groups and settings.  Planned subgroup analyses related to participant characteristics should not be listed here.

 

Index tests                                                                                                                   [fixed, level 3 heading]

Specify the test or tests under evaluation. 

 

Comparator tests                                                                                                [optional, level 3 heading]

This subheading is optional.  The comparator test is the testing regime used in practice, which the index test may be seeking to replace.  Not all reviews will have a comparator test, hence it is an optional subheading.  If a comparator is stated in the review question it should be stated here.  There is potential to confuse the comparator with the reference standard.  However the reference standard used in the context of research to establish test accuracy will not usually be the same as the test or test combination used in day-to-day practice. 

 

Target conditions                                                                                                      [fixed, level 3 heading]

The target condition is a particular disease or disease stage that the index test is intended to identify.  Tests may occasionally be used to differentiate between several target conditions – if this is the case they should all be listed here.

 

Reference standards                                                                                                [fixed, level 3 heading]

Describe the clinical reference standards (‘gold standard’) that are considered appropriate to establish the presence or absence of the target condition in the tested population.  If particular reference standards are commonly used but considered inadequate they should be stated here as exclusion criteria.

 

Search methods for identification of studies                                   [fixed, level 2 heading]

 

Electronic searches                                                                                                  [fixed, level 3 heading]

The methods used to identify studies should be summarized.  Further details of the content of these sections are discussed in Chapter 7.  The bibliographic databases searched, the dates and periods searched and any constraints, such as language, should be stated.  The full search strategies for each database should be listed in an appendix to the review.

 

Searching other resources                                                                             [optional, level 3 heading]

List grey literature sources, such as reports and conference proceedings. If journals are specifically handsearched for the review, this should also be noted. List people (for example, researchers, experts) and/or organisations who were contacted. List any other sources, which may include, for example, reference lists, the World Wide Web or personal collections of articles. 

 

This text may be organised under the following four subheadings:

·         Grey literature,

·         Handsearching,

·         References lists and

·         Correspondence. 

 

These subheadings are not included in the RevMan5 structure, so if required an author will need to create them.  They can be used either in place of ‘Searching other resources’ or as subheadings to it.

 

Data collection and analysis                                                           [fixed, level 2 heading]

Selection of studies                                                                                                  [fixed, level 3 heading]

The method used to apply the selection criteria should be described stating whether the criteria were applied independently by more than one author, and how any disagreements were resolved (see Chapter 8).

 

Data extraction and management                                                                        [fixed, level 3 heading]

State the method used to extract or obtain data from published reports or from primary authors of the included studies (for example, using a data extraction/data collection form).  Any reanalysis of individual patient data should be described.  Whether data are extracted independently by more than one author should be stated, along with how any disagreements are resolved.  If relevant, methods for processing data in preparation for analysis should be described (see Chapter 8).

 

Assessment of methodological quality                                                            [fixed, level 3 heading]

Assessment of methodological quality involves describing both the tool, and the method by which it was applied.  The tool(s) used (i.e. QUADAS) should be described or referenced, with an indication of how quality assessments were incorporated into the interpretation of the results (see Chapter 9).  Operational definitions of items within the quality assessment tool should be stated (possibly using an additional table).  For example, the QUADAS scale requires authors to explicitly state what they consider to be ‘appropriate spectrum’ and an ‘appropriate reference standard’ before applying the tool.  The method used to assess methodological quality should be described, stating whether the tool was applied independently by more than one author and how any disagreements were resolved.

 

Statistical analysis and data synthesis                                                            [fixed, level 3 heading]

In this section, the descriptive and inferential statistical methods should be described.  Descriptive methods consist of tabulation, graphical displays of estimates of diagnostic accuracy (for example sensitivities and specificities), and plotting the study results in ROC space (see Chapter 10).  Inferential statistical methods are used for the estimation of summary ROC curves and average operating points, testing of differences between tests, and investigations of heterogeneity.  Details should be given of the statistical method and model used, what parameters were estimated, whether random effects were estimated, and the software used.

 

Investigations of heterogeneity                                                                    [optional, level 3 heading]

Indicate how the sources of heterogeneity listed in the objectives were investigated (see Chapter 10). 

 

Sensitivity analyses                                                                                          [optional, level 3 heading]

Pre-planned sensitivity analyses should be stated here.  These could include restricting analyses to a particular subgroup of patients, or to studies without a particular methodological shortcoming for example verification bias or review bias (see Chapter 10). 

 

Assessment of reporting bias                                                                       [optional, level 3 heading]

If any tests or investigations were undertaken to detect reporting biases the methods used should be explained here (see Chapter 10).

 

Results                                                                                     [fixed, level 1 heading]

Results of the search                                                                      [fixed, level 2 heading]

The results section of the full review should start with a summary of the results of the search, reporting the number of citations identified by the electronic searches, the number for which full reports were retrieved, the number of citations that were finally included in the review, and the number of unique studies that they report.  Brief details should be given of the occurrence of duplicate reports citations of the same studies.  Similar details should be given for the results of searching other sources where possible (for example handsearching, correspondence and reference lists). 

 

An overview of the selection process should be given.  This can be summarized using a flow diagram, and it is recommended that authors consider including one as an additional Figure (see Chapter 7).

                                                                                                                                                                                      

The number of included studies should be clearly stated, together with the numbers of participants and the number who have the target condition.  If a review evaluates more than one test these numbers should be given per test.  Where pairwise comparisons are based on direct (within study) comparisons, the number of studies and patients available for each comparison should be stated.

 

Individual details of the studies are tabulated in the ‘Characteristics of included studies’ table.  A succinct summary of the key characteristics of the design, participants, index tests and other methodological issues presented in this table should be given.  In some circumstances an additional table may be useful to give a tabular summary of the different index tests that were encountered, or other aspects of the study.  If such a table is produced a link to it must be created at an appropriate place in this section.

 

The number of excluded studies should be mentioned here and refer to the information contained in the ‘Characteristics of Excluded Studies’ table, providing a succinct summary of why studies were excluded from the review.

 

Methodological quality of included studies                                    [fixed, level 2 heading]

A figure summarising overall methodological quality, and a table of the methodological quality ratings for individual studies can be produced automatically in RevMan5 (see Figures below), and should be linked to the text in this section.  The text should summarize the general quality of the included studies, its variability across studies and any important flaws in individual studies which will threaten the validity of the results.  The criteria that were used to assess the methodological quality bias should be described in the ‘Methods’ and not here.  For reviews that evaluate several tests in separate studies, reporting of methodological quality by test should be considered if there are important differences.

 

Findings                                                                                         [fixed, level 2 heading]

The main findings on the diagnostic accuracy of the index tests studied in the review should be presented, together with the results of the planned comparisons between the index tests, or between the index tests and the comparator tests.  The section should directly address the objectives of the review rather than list the findings of the included studies in turn.  The focus should be on reporting the pattern of results across all the included studies.  Numerical summaries and results of statistical analyses may best be summarized in additional tables and a link to the table placed in the text where they are discussed.  Links to relevant forest plots and summary ROC plots may also be placed in the text. 

 

Subheadings are encouraged if they make reading easier (for example, for each index test if a review addresses more than one).  Any exploration of heterogeneity, sensitivity analyses and investigations of possible biases that were undertaken should be reported.

 

Inferences should be avoided in the results section.

 

Discussion                                                                               [fixed, level 1 heading]

A structured discussion can aid the systematic consideration of the implications of the review (Docherty 1999).

 

Summary of main results                                                               [fixed, level 2 heading]

The authors should summarize the main findings of the diagnostic accuracy of the index test(s).  This description should mirror the summary of results table (it may be easiest to create this table before writing this section).  The authors should summarize the relevance of the findings of investigations of heterogeneity.  There should also be comment on the homogeneity and methodological quality of the evidence, and the completeness of the evidence in terms of whether it has been possible to address all the objectives of the review and the degree to which uncertainties remain.  Consideration should be given to the amount of data available to address the primary hypotheses.

 

Strengths and weaknesses of the review                                         [fixed, level 2 heading]

The author should state the strengths and limitations of the review with regard to preventing bias.  These may be factors within, or outside, the control of the review authors.  The discussion might include whether all relevant studies were identified, whether all relevant data could be obtained, or whether the review methods used (for example, searching, study selection, data extraction, and analysis) could have introduced bias.  The degree to which robust conclusions can be drawn if the study results are notably heterogeneous should be discussed. 

 

Comments on how the included studies fit into the context of other published reviews might be included, stating clearly whether the other evidence was systematically reviewed. 

 

Applicability of findings to clinical practice and policy                   [fixed, level 2 heading]

To assess the applicability of their findings authors consider whether the studies that were identified were sufficient to address all of the objectives of the review.  They should consider the relevance of the patients, tests and settings that were included in the review to the review objectives, and identify the limits of the situations to which the evidence from the included studies definitely applies.  For example, studies from secondary care may have limited relevance to primary care, or the evaluations located may have only evaluated an outmoded version of the test technology.  Comments on how the results of the review fit into the context of current practice should also be included here, although authors should bear in mind that current practice might vary internationally.

 

Authors’ conclusions                                                             [fixed, level 1 heading]

The primary purpose of the review should be to present information, rather than to offer advice.  Conclusions of the authors are divided into two sections:

 

Implications for practice                                                                 [fixed, level 2 heading]

The implications for practice should be as practical and unambiguous as possible.  They should not go beyond the evidence that was reviewed and be justifiable by the data presented in the review. 

 

Implications for research                                                                [fixed, level 2 heading]

This section of Cochrane reviews is used increasingly by people making decisions about future research, and authors should try to write something that will be useful for this purpose.  As with the ‘Implications for Practice’, the content should be based on the available evidence and should avoid the use of information that was not included or discussed within the review.

 

In preparing this section, authors should consider the different aspects of research, perhaps using types of studies, index tests, study population and target condition verified by the reference standard as a framework.  Implications for how research might be done and reported should be distinguished from what future research should be done.  It is important that this section is as clear and explicit as possible.  General statements that contain little or no specific information, such as “Future research should be better conducted” or “More research is needed” are of little use to people making decisions, and should be avoided.

 

Acknowledgements                                                                 [fixed, level 1 heading]

This section should be used to acknowledge any individuals or organisations who have contributed but who are not listed among the authors.  This would include any previous authors of the Cochrane review and might include the contributions of the editorial team of the CRG.  Permission should be obtained from persons acknowledged.

 

Contribution of authors                                                         [fixed, level 1 heading]

The names and contribution of the present co-authors should be described in this section.  One author, usually the contact author, should be identified as the guarantor of the review.  All authors should discuss and agree on their respective descriptions of contribution before the review is submitted for publication in the Cochrane Database of Systematic Reviews.  When the review is updated, this section should be checked and revised as necessary to ensure that it is accurate and up to date.

 

The following potential contributions have been adapted from Yank et al (Yank 1999).  This is a suggested scheme and the section should describe what people did, rather than attempt to identify within which of these categories someone’s contribution falls.  Ideally, the contributors should describe their contribution in their own words:

 

 

Declarations of interest                                                          [fixed, level 1 heading]

Authors should report any conflict of interest that might be perceived by others as being capable of influencing their judgements, including personal, political, academic and other possible conflicts, as well as financial conflicts.  Authors must state if they have been involved in a study included in the review.  Details of The Cochrane Collaboration’s policy on conflicts of interest appear in Chapter 2 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2006a).

 

Financial conflicts of interest cause the most concern, and should be avoided, but must be reported if there are any.  Any secondary interest (such as personal conflicts) that might unduly influence judgements made in a review (concerning, for example, the inclusion or exclusion of studies, assessments of the validity of included studies or the interpretation of results) should be reported.

 

If there are no conflicts of interest, this should be stated explicitly, for example, by writing ‘None known’.

 

Differences between protocol and review                            [fixed, level 1 heading]

Authors should report and clearly explain why the differences between the methods reported in the protocol and the full review exist. 

 

Published notes                                                                     [fixed, level 1 heading]

These will be published in the Cochrane Database of Systematic Reviews.  They may include:

 

The published notes must be completed for all withdrawn publications to give the reason for withdrawal.  Only the cover sheet and published notes are published for withdrawn protocols and reviews.