Each Overview should contain one or more tables using the format shown in Figure 22.3.b to summarize its results. This format has been designed to reflect (as much as possible) the format of ‘Summary of findings’ tables: see Chapter 11 (Section 11.5) for additional guidance. If the Overview addresses more than one clinical population (e.g. groups that differ by stage or severity of disease, co-morbidities, or other factors likely to affect the outcomes under study) then separate tables should be used for the different clinical populations. Clearly the exact form may vary with review topics but each table should include both beneficial and harmful outcomes, the frequency or severity of these outcomes in the control groups, estimates of the relative and absolute effects of the interventions, indications of the risk of bias (which may vary by outcome and comparison), and any comments.
Figure 22.3.b provides a template for an ‘Overview of reviews’ table. The intention is to make the format for this table as similar as possible to that used for ‘Summary of findings’ tables. If the recommended format for ‘Summary of findings’ tables changes, the recommended format for this table will change as well.
The rows should be organized by outcome, beginning with the primary outcome of interest. Within each outcome a series of rows should provide the results from the various intervention or comparison pairs for which data are available. Generally, one or more rows for adverse outcomes should be included, even if the included reviews did not report results for these.
The main beneficial and harmful outcomes should be listed (those most relevant to participants, preferably determined prior to completing the results of the Overview to avoid the potential of selection of reported outcomes based on significance and not clinical importance). The number of outcomes should not exceed seven. Important outcomes for which no data are available may be listed in the table as well.
If there are multiple interventions being compared, the table should be primarily organized by outcome, with rows included in each outcome subsection that present data comparing the results of two interventions regarding that outcome.
Representative comparator group risks should be provided for each row. These might be obtained from control group risks as reported in the included Cochrane reviews. If there is important variation in control group risks, two or three representative rates should be included for each row of the table – representing a low risk, moderate risk and high risk population. Whenever possible, indicate the types of participants to which a given control group risk may apply in this column, in the comments column or in a footnote.
This column is intended to show the expected absolute risk upon intervention at the one, two or three assumed comparator risks cited in the previous column. The numbers can be calculated by applying the relative effect to each assumed risk for the same row (see Chapter 11, Section 11.5.4).
For dichotomous outcomes, the risk ratio or odds ratio should generally be used. So far as possible the summary statistic that is used should be standardized across included reviews even if different reviews used different summary statistics in their analyses. The 95% confidence interval should be included to provide a measure of uncertainty. This may be calculated using either a fixed or random-effects model; however, the same model should be used for all results relative to a given outcome.
In many cases, the number of studies and participants for whom data are available for a specific outcome and treatment comparison will be less than the total number of studies and participants reported in the Cochrane review from which the data are extracted (because the Cochrane review may include studies that did not report on a specific outcome or a specific comparison). If so, the number of studies and participants reported in this column should reflect only the subset providing data for the comparison and outcome of interest.
Comment on the quality of the evidence for each row of the table (note that, because different rows may contain data extracted from different Cochrane reviews or from different studies within an individual Cochrane review, the quality of evidence may vary from row to row). Use of the specific evidence grading system developed by the GRADE group (GRADE Working Group 2004) is recommended and is incorporated in the software available to authors of Cochrane reviews for preparing ‘Summary of findings’ tables. The system and methods employed to grade quality of evidence should be described in the Methods section of the Overview.
The aim of this field is to provide additional comments to help interpret the information or data identified in the row. For example this may be on the validity of the outcome measure or effect modification. Important caveats about the results should be flagged here. Not all rows will need comments, so it is best to leave a blank if there is nothing of importance to comment on.
Continuous outcome measures can be shown in the Overview table, but should be clinically meaningful. This requires that the units are clear and that these units are readily interpretable, for example days of pain or frequency of headache are readily interpretable. However, many scales are not readily interpretable by non-specialist clinicians or patients, for example points on a Beck Depression Inventory or quality-of-life score. For these, a more meaningful presentation might be to express results in terms of risks (e.g. of a 50% improvement) where possible, as discussed in Chapter 12 (Section 12.6).
The labelling of the outcomes should also be kept simple. For example, ‘ability to perform everyday functions’ would be preferred to ‘functional status’. If specific details of outcome definitions are required, these might be added as footnotes.
A detailed discussion of heterogeneity generally should not be part of the summary table. However, if either (i) heterogeneity made important changes to the clinical or statistical significance; or (ii) there were important effect modifiers, then these should be reported in the Comments column. Occasionally an important effect modification may require a separate row or separate table to describe, for instance, difference in effect of endarterectomy for different grades of stenosis.