Delaying the Inevitable?
What’s behind paper retractions? (12)
by Adam Marcus and Ivan Oransky, Labtimes 06/2012
Photo: zettberlin / Source: PHOTOCASE
Journals and authors should stop taking so long to retract papers once it’s clear data is dodgy.
In October 1995, following an investigation by the U.S. Office of Research Integrity (ORI), Weishu Weiser, a former researcher at Harvard, “agreed to submit a letter” to the Journal of Immunology to retract two papers she published in 1991 and 1993.
Weiser – who recently retired after five years as managing director of the German biopharmaceutical company Scil Technology – never wrote to the Journal of Immunology to retract the papers articles, or if she did, the journal has no record of it, according to the retraction notices in its September 15, 2012 issue, published 17 years after the ORI agreement.
We couldn’t get a straight answer out of the authors who did sign the notice, nor from the journal’s editor, about why the retraction took so long. One author said he had no idea why the immunology journal was acting now. The editor merely referred us to the notice, which said nothing about the timing. The most we learned, from the other author who signed, was a cryptic “This is due to the value of organisations like yours.”
We’ll choose to look at that the charitable way, but in either case, here’s why such delays matter: Each of the articles was cited about 40 times, even after the ORI report came out.
Weiser and the Journal of Immunology are not alone. The Journal of Dermatology recently retracted an 18-year-old paper for duplication – a.k.a. self-plagiarism. And in June, the FASEB Journal retracted a 15-year-old paper for duplication after the authors’ institutions declined to investigate the matter – a move by the journal that should be applauded, despite the delay.
These impressive time frames, however, are not the record. As best we can tell, the longest time between publication and retraction is 25 years, a distinction held – perhaps ironically, in the first case – by I.E. Swift and B.V. Milborrow. In 2005, the pair retracted four papers about plant compounds after the work of other groups suggested that the “original claims are not supported”. One of the papers, in the Biochemical Journal, was originally published in 1980.
Twenty five years is clearly an outlier and 17 years may be closer to two standard deviations from the mean than the median. One study found that for retractions published between 1995 and 2004, the average time to retraction was 20.75 months. For those involving the ORI, it was about three years, a lag explained in part by the fact that the agency is at the mercy of university and institutional investigations, which operate on what often seems like deep geologic time. A recent study found that more than half of those studies retracted in that same time period – with an updated dataset – were retracted within two years of publication.
Of course, that means half weren’t retracted within that time frame. And yet another study found evidence that the lag is growing. Some of that, as Grant Steen, the author of that study notes, may be because editors are reaching further back in time. That’s a good thing, if it means the literature will be a better reflection of scientific understanding. And another caveat to all these findings is that we don’t usually know the lag between “questions raised” and retraction.
No matter what the reasons, as the Journal of Immunology cases show, these delays have a real impact on the literature. The argument that “everyone knows which papers are dodgy” just doesn’t hold up.
Unfortunately, retraction delays are in keeping with other stubborn behaviour on the part of journals, authors and institutions. They distort the scientific record. Perhaps we need a public registry – with authors and papers anonymous, if necessary – showing how long various journals take to investigate papers once they’ve learned of potential problems.
We’re pretty sure 17 years is too long.
(The authors run the blog Retraction Watch: http://retractionwatch.com)
Last Changed: 10.11.2012