Correcting the Literature—Retraction and Republication

In this issue of JAMA, Ascherio and colleagues¹ request retraction of their article titled “Multiple Sclerosis and Epstein-Barr Virus,”² which was published in the March 26, 2003, issue. As the authors explain in their letter to the editor,¹ an inadvertent error occurred in sorting of one of the data files used in their nested case-control study. This error led to incorrect matching of the serologic findings from the laboratory analysis and the clinical data, including incorrect assignment of several dates of blood collection. Once the authors discovered this error and reanalyzed the data with the correctly matched samples, it was apparent that many of the data points reported in the original article were incorrect. Even though the main finding of the study was essentially unchanged, correcting the data errors completely would have required publication of an extensive correction, including republication of correct data points in the abstract and throughout the text, along with republication of corrected tables and figures.

In our view, publishing such an extensive correction would have been more confusing than clarifying. Therefore, to correct the literature and to present the findings as clearly as possible, the authors submitted a manuscript with the properly matched data, along with an updated analysis based on the corrected blood collection dates. This newly submitted manuscript has been peer reviewed and revised satisfactorily and is published in this issue of JAMA.³ This article presents the findings of the corrected data and analyses clearly and demonstrates the significant association between anti–Epstein-Barr virus nuclear antigen titers and risk of multiple sclerosis as originally reported, along with evidence that this association is modified by age.

Retractions of articles from major medical journals are unusual events but have occurred based on serious grounds, such as falsification and fabrication of data⁴ or other major ethical breaches, such as misrepresentation.⁵ Publication of incorrect data, as occurred with the article being retracted from JAMA, is also a serious issue. However, in this case, the errors that led to the retraction were unintentional, and the authors have been forthcoming in disclosing this error and diligent in ensuring that the most accurate information from this study is communicated clearly to clinicians, physician-scientists, and other researchers. This is a good example of an optimal outcome when inadvertent errors occur in biomedical research; after all, to err is human, to correct, almost divine.