Include trials once only

Individual trials are often reported several times, and it is important to ensure that each eligible trial is included only once in the systematic review and meta-analysis [34]. In practice, it can be quite difficult to determine when multiple publications refer to the same trial because authors may change, different numbers or subgroups of patients may be analysed and reported and the trial may be described differently. As it is likely to be the results of trials favouring new treatments that are reported and presented most often, this could exaggerate the influence of publication bias. It is therefore important to scrutinize publications extremely carefully to spot papers that are updates or duplicate reports of trials already represented in the systematic review.

In full papers, clues may be obtained by matching the trial accrual dates, lists of participating centres and details of treatment scheduling as well as the authors, patient numbers and general descriptions. However, it is almost impossible to do this with abstracts. Since trials are very likely to be presented at several meetings this can make things difficult and it is advisable to seek clarification from authors if trials appear similar. In an IPD meta-analysis, direct contact and collaboration with trialists should identify duplicate publications especially as they will ultimately need to supply the underlying dataset.

In future, if the unique identifying schemes currently in development (Box 11.1), become adopted widely then the problem of duplicate publication bias should be solved. As all publications should include the unique trial identifying number it will make it easy to spot at a glance where reports refer to the same trial.

0 0

Post a comment