Identifying duplicate publications: primum non nocere.

Errami and Garner published in the January 24, 2008 issue of Nature an article entitled “A Tale of Two Citations” that deals with the timely subject of publication of duplicate papers (1 ). With the increase in the number of scientific journals and the ongoing pressure to publish in them, inappropriate and unethical practices such as plagiarism, unauthorized “cosubmission” of a paper to two or more journals simultaneously, and duplication of a previous report may also be on the rise. The authors of the Nature Commentary focused on two important reports to document this problem (2, 3 ). The first was from a large study that used textmatching software to mine 280 000 entries in arXiv, an open-access database for publications in mathematics, physics, biology, statistics, and computer science (2 ). Of the examined articles, 0.2% of them were suspected of plagiarism and 10.5% were suspected of being duplicate publication by the same authors. The second study was from an anonymous survey of 3247 American biomedical researchers (3 ). Here, 1.4% admitted to the act of plagiarism and 4.7% to repeated publication of the same results. With more that 17 million citations in the Medline database, the above-mentioned Figs. imply that the number of articles suspected of plagiarism ranges from 34 000 –238 000 and of duplicate publication from approximately 800 000 –1.8 million. Armed with this information, Erami and Garner embarked on the challenging task of documenting the trend in suspected duplicates in the biomedical literature over the past 30 years (1975–2005) and sought to determine whether the publication of duplicates is a country-dependent phenomenon. The authors searched online databases such as Medline using textsimilarity software and the eTBLAST search engine, which is freely available online. The potential duplicates were deposited in Déjà vu, a publicly accessible database (http://spore.swmed.edu/dejavu). Over a 30-year period, the authors found that the annual number of biomedical publications increased from approximately 250 000 – 600 000 while the percent of suspected duplicates had risen disproportionately from 0.018% to 0.9%. Furthermore, the duplication rates for countries were proportional to the numbers of contributions from the countries, with the exception of China and Japan, which had twice the expected rates of duplicates for the numbers of their contributed publications. Although these findings are interesting, some aspects of the conclusions are concerning. The authors acknowledged that manual verification of articles listed in Déjà vu as duplictes is currently ongoing and results should be interpreted with caution. However, the rate of suspected duplicates appears to be high and the issue of false-positive identification is not well addressed in the overall message of the report. To examine the false-positive rate in the Déjà vu database, we checked the suspected duplicates in 3 journals, the New England Journal of Medicine (NEJM), Clinical Chemistry and The Lancet, since 1975. During this period and according to ISI (http://portal. isiknowledge.com), NEJM published 11 779 original and review articles. According to Déjà vu 14 were suspected duplicates. After careful examination, we found none of those articles in NEJM to be a duplicate publication. In Clinical Chemistry, 8867 original and review articles were published with 27 identified as suspected duplicates. Of those, one article could be viewed as a duplicate and may require a corrective action, a second could not be verified as we have not been able to access the other journal/newsletter, and a third was republished in a language other than English. This latter point requires examination. Perhaps the republication of a review article or a case study in a language other