Cancer researcher Robert Mandic of University Hospital in Marburg ENT at the time must probably have been pretty upset. He had just completed an article on the subject of rare head and neck tumours, as he learned that he had been working all the time with a cervical carcinoma cell line. He informed the journal, which then withdrew the article. According to estimates, about one in four tumour cell lines in the large cell banks in the U.S., Germany and England is not what it is assumed to be. And the worst part is that hardly anyone finds out about how much research has already failed because of it, or how many results in the literature are on account of this highly unreliable.
Withdrawn contributions: a steep upward trend
Each week in the database Web of Science about 27,000 new journal articles appear. 200 of these “change” over the course of years because of results or authorships being corrected. A handful of them are declared by the editors of the magazine to be invalid. Nevertheless, even years later countless new technical papers turn up which cite these dubious publications. In other words, the results are based on information which the authors or reviewers have declared null and void.
The mass of reports on studies or experiments no longer being valid is increasing: while in the nineties it was still around 30, the number has now risen to around 400 per year. Media consultant Grant Steen of North Carolina has encountered 180 retracted pieces of work in human research for the period 2000-2010. In the field of Medicine, American Barbara Redman counted some 330 publications in the period 1995 to 2004. In this area as well, the number is constantly on the rise. Relative to the approximate 5 million articles over the investigation period, this is not much – but the consequences are often huge.
Lancet: Late deletion with consequences
Journals with a high ‘impact factor” which carry a great reputation are affected more often than little, less significant media. A period of about two years runs on average from publication until a “retracted” notice is inserted. Englishman Andrew Wakefield in 1998 published in the “Lancet” a link between MMR vaccine and the incidence of autism. Although the results were refuted a long time ago, the magazine first withdrew the article in 2010. Meanwhile, the vaccination rates in Europe fell, several measles epidemics occurred and the described connection still serves as the basis for the argument of many vaccine skeptics. According to a study by Jeffrey Furman from Boston, subsequently withdrawn publications are even more often quoted in the first year than publications of comparable content. Steen has counted the citations following withdrawal: works published later mention the results around eleven times, without making reference to the lack of reliability.
Is there more cheating, deception, or simply sloppily production these days? “I do not think that there’s a sudden boom in fraudulent or defective work”, says John Ioannidis, researcher at Stanford University on health policy. With the Internet reaching a much larger audience, pdf files used can also easily be checked with the appropriate software for plagiarism or tampering. Soon, experts hope, these instruments in the hands of editors should however prevent that such publications are even met by the eye of the public.
Detective work on dubious publications: Retraction Watch
Meanwhile the reader, if he or she is lucky, learns via the respective watermark on the pdf file linked to the website of the journal, or via a note with the article, the history of the publication. Or – with various magazines – it may not happen at all. The situation seems similar also with databases for literature search. Sampling has verified that Medline has performed a very thorough job, while EMBASE identified during the same examination only one retraction in 17.
Anyone interested in backgrounds to and reasons for which an author, an institution or a journal retracts a paper can find lots of material in “Retraction Watch”. Ivan Oransky of Reuters Health and Adam Marcus of the journal Aneasthesiology News founded the blog in August 2010. About 150,000 page views per month and hundreds of references to retracted publications in the past year and a half show the importance of this monitoring tool. Excerpts in German language are also found in Laborjournal.
“That has nothing to do with you!”
But those who approach publishing editors with questions on the withdrawn publications often receive quite different answers. In response to requests from Retraction Watch made to the editor of Annals of Thoracic Surgery Henry Edmunds: “That’s none of your damn business”. Even Nature was refused an answer by Edmunds. More often an enquiry receives very vague answers, such as “At the request of the authors the article was withdrawn.” Such a branded contribution is liable to see the loss of reputation, of reliability and trust.
Malicious forgeries, fabricated data or bending the results to someone’s liking (such as those of the sponsor) have also increased proportionately and are now ahead in their frequency of experimental and calculation result errors. Daniele Fanelli from the University of Edinburgh put a lot of studies on the topic of “accidental or malicious data tampering” under the microscope and found in his meta-study that about two percent of all scientists have sometime admittedly fabricated or falsified data – with a high number of unreported cases. The American Medical supervisor FDA also arrives at a similar incidence.
Crossmark: Validity check by clicking
But how can one now prevent that pieces of data continue to live even after their “funeral”? Crossmark, developed by the same network of editors who came up with the DOI-labeling literature, offers a newly developed ability to make individual checks. The bookmark on the pdf file of the publication indicates via click whether the respective work was later changed and has experienced corrections. Meanwhile, publishers have had this opportunity to test some 20,000 documents. For several weeks Crossmark has now been open to all journals.
More important, however, are probably still talks between institutions and publishers and dialogue with the research community. Only when researchers and specialist journals recognise that errors are a part of producing research results – as are corrections and, at worst, withdrawn papers – then this information should also not survive in small journals for years. Even more annoying than working with false information is finding out that others have already long known about it.