Reviews which limit the literature search to published studies risk misrepresenting the evidence, given that studies which report significant results are published faster and cited more often. Reviews which do not practice duplicate data extraction - where two people do this independently of one another - run the risk of distorting the evidence, because single data extraction leads to more inaccuracies and omissions. Choosing narrative synthesis over meta-analysis is also found to give less accurate results.

In each case, there is a tendency to present conclusions that are more positive than the evidence warrants. Traditionally, it has been hard to explain to funders that the complexity of the review process is necessary, and that a high-quality review involves both time and resources. This made it all the more heartening to hear in Dhaka that policymakers are increasingly aware of this problem. Representatives of the UK's development agency stated that DFID had decided not to compromise on quality. The challenge now is to ensure that more genuinely systematic reviews are conducted and completed. How about that for a New Year's Resolution?

Eamonn Noonan, CEO