Starting at the end ....

We know that many people who read a review begin at the end, by looking at the conclusions and, often, the brief conclusions in the abstract. It is probably human nature as much as anything else to want a 'bottom line' from a research study or a review. Unfortunately, it is often very difficult or indeed impossible to provide this bottom line.

Whose conclusions?

As we have already seen in previous modules, users in different settings will often interpret the same evidence from the same review quite differently.

Within a review, any conclusions drawn are only those of the reviewers. For this reason Cochrane reviews specifically refer to Reviewers' Conclusions in order to highlight where these conclusions come from.

Usually there are three aspects to drawing conclusions:

• does the intervention work at all? (as covered in module 16)
• what have we learned from this review that can be applied to clinical practice?
• what have we learned from this review about the need for further evaluation and research?

What are the implications for clinical practice

The first component of a conclusion is a bit like putting all the evidence you have obtained from the review on a balance scale and trying to see which way it tips and how confident you are about this interpretation. In essence, you need to decide, after taking all the possible outcomes into account, which of the alternatives (on balance) is likely to do more good than harm (and should be applied in practice) or more harm than good (in which case it should be abandoned from practice). However, more often than not your conclusion will probably be somewhere in between these two scenarios. It may be that there are important trade-offs between different outcomes (such as in the example of treating sore throats with antibiotics that we used in Module 17) and your conclusion needs to reflect this.

Uncertainty reigns

Alternatively, you may find that there is insufficient evidence to arrive at a firm conclusion of any kind. In situations where there is still uncertainty (for example not enough or no trials, trials of poor quality or contradictory evidence from a group of trials), you may be wise to conclude that there is a need for further research in order to gain sufficient evidence to assess the effects of an intervention.

What are the implications for further research

While there may sometimes be a need for more research in order to gather sufficient evidence to address the original objectives of the review, it is often the case that the further research needs to be targeted to specific issues that have arisen out of the review. For example, in a review of nicotine replacement that has been maintained regularly since 1994, it was established that different nicotine replacement delivery systems (such as nicotine gum, patch, and inhaler) were more effective than placebo. However, further primary research was required to establish the effectiveness of a combination of different delivery formats. This was subsequently done and has now been included in updates of the review.

Each time a review is updated some questions that previously required further research might have been answered, and new questions may arise. This is why updating reviews regularly is so important.

Categorising interventions according to evidence

There have been several attempts to try to categorise interventions according to whether there is sufficient evidence to provide clear guidelines for clinical practice and, if not, whether further evaluation is required. One of these was developed several years ago as part of a successful pilot project in Pregnancy and Childbirth. It has since been further developed by the Cochrane Pregnancy and Childbirth Group. Stop and read it now in Section 9.6 of the Reviewers' Handbook.

Then try to classify the interventions in your review according to these categories.