Spotting duplicate reporting.(In the News)

In tallying up the evidence for or against a particular intervention, meta-analyses and systematic reviews of clinical trials rely on a crucial assumption--that each set of data was reported only once. For example, suppose there are three publications showing a positive result for treatment X, but one showing no effect; thus, the evidence seems weighted toward concluding that the treatment is effective. But if the three positive reports turn out to be based on the same dataset, then suddenly they are no longer three independent studies, but one single study. In that case, the evidence seems equally balanced on either side, one positive vs. one negative study. Thus, when results from the same trial are reported multiple times in the literature, this can create a "publication bias" that may in turn distort clinical recommendations. Ultimately, such biases can have serious consequences for patients by providing a misleading picture of the evidence.

Detecting duplicate publications poses an enormous challenge, however, given the size and rapid growth of the medical literature. But a software program developed by researchers at the University of Texas Southwestern Medical Center in Dallas may provide a useful tool for uncovering such duplication. The team, led by computational biologist Harold Garner, recently tested its program on a sample of citations from the Medline database, the National Library of Medicine's (NLM) biomedical literature database that is accessed through PubMed. Based on the …