Ask the question: Can the results apply in my situation?

People in many different situations will use your review. It is important that you write the review in a way which will allow individuals to decide whether your review applies to them.

First, a user has to decide whether the review provides valid information about the potential benefits and harms that may be important to them. Then they need to decide whether the participants and settings in the included studies are reasonably similar to their own situation. It is often helpful for users to consider asking themselves whether there are any good reasons why the evidence should not be applied in their situation. Some of these reasons might be related to:

• Biological and cultural variation
• Variation in compliance
• Variation in the baseline risk

Now try to think of some examples where these factors might influence the applicability of the findings from your own review and write these down. An example would be that if the trials in your review are done in a developed country with good access to diagnostic equipment, and the reader of your review is a doctor trying to decide to whether to implement the treatment in a resource poor country where the equipment available to determine diagnosis is not available, and there are no trials in your review performed in participants similar to the people she treats, the results of your review may not help her make this decision.

If you have come up with any factors that might limit the applicability of your own review these might be worth highlighting in the Discussion.

How consistent are the results?

There are many reasons why the results of studies included within a review may vary (even if there is no heterogeneity detected with statistical testing). For example, in a review of giving smoking cessation advice, we found a considerable variation in the results depending on the intensity of the advice and follow-up provided. Other reasons why results may vary include differences in the participants (for example, their age, gender, or presence of some biochemical marker) or differences in their underlying disease status. You should have an idea of the possible reasons why the results of the trials in your review might vary from thinking about it in the module on heterogeneity. If these differences might be clinically relevant they would be worth highlighting when discussing the evidence in your review.

A word of caution

Some chance variation between different subgroups is inevitable. It is often a trap to try to explore variation in results by undertaking separate analyses of different sub-groups. You should avoid doing this unless there was a good prior reason to believe that a particular sub-group might respond differently to the intervention.

Other useful information

Often there may be some other information that might be useful to consider when discussing the results of a review. For example, it might help to include some information about the size or frequency of a particular health care problem that the review is addressing. However, you need to be quite careful because this information may be very context specific and might be better addressed at a more local level. Much as it would be good if we could provide users of a review with all the information they require to make a decision, this is rarely (if ever) possible.

There are more formal ways of using the evidence and considering benefits, harms, patient preferences and other factors specific to your particular setting. These tools, often termed decision analyses or decision trees, are outside the scope of this material, but often draw on evidence provided by Cochrane Reviews.

A final check ....

At the end of the discussion section it is worth re-reading and asking yourself:

• Have all the main outcomes been considered?
• Have data been presented about the absolute change as a result of the intervention for all possible outcomes?
• Have I considered any factors that might limit the application of these results in different situations?
• Are these results consistent across the included studies or do they vary for some reason?
• Have I avoided making value judgments about how to interpret my findings?