Journal Retractions Rising:
Vaccine-Autism Paper Highlights Growing Trend
Article Outline
Arthur Kellermann, MD, MPH, received his baptism in the swirling, vitriolic controversy between physicians and the vocal community that believes vaccines cause autism last fall, when he made a fairly innocuous comment to National Public Radio (NPR) for a story on public concerns about the H1N1 vaccine. “It is shocking to me when I hear a news story about nurses or doctors not getting vaccinated,” Dr. Kellermann, then an Emory University professor of emergency medicine, told NPR. His remarks elicited something of a response, such as an e-mail, with the subject line “Nazi Scumbag” claiming vaccines are a “bio-engineered scam for money” and “have killed billions and are being used to destroy the human race …”
The civility of the missive went downhill from there, but this is not a story about epithets on the Internet, nor even so much about the vaccine controversy. This bitter battle over vaccine safety is a direct byproduct of a 1998 article published by The Lancet concerning autism and vaccines, which turbocharged the anti-vaccine movement in the United Kingdom, the United States and elsewhere. The article's recent retraction comes as the number of retracted medical journal articles is rising, and editors are grappling with how to ensure their retractions actually weed the relevant data out of the medical literature. But as this paper and others illustrate, retraction does not erase a paper's impact either on the public or even the scientific community.
On February 2 The Lancet “fully” retracted a 1998 article by Andrew Wakefield and others that found the mumps, measles and rubella vaccine may be unsafe. The retraction came after the British General Medical Council's multi-year assessment of Dr. Wakefield's scientific methods and financial conflicts. A hero to some parents of autistic children who blame their offsprings' illness on vaccines, the journal's retraction could paradoxically raise Dr. Wakefield's standing among this passionate group of parents, many of whom view the journal's actions as further evidence of the medical community's will to suppress evidence that vaccines cause harm.
Two weeks after the retraction, for example, prominent Canadian actor Jim Carrey, a celebrity spokesman for the non-profit Generation Rescue organization that believes the link between vaccines and autism, published the following message on the social networking site Twitter: “Dr. Andrew Wakefield's studies r being unfairly supressed. His newest vacs vs unvacs study MUST BE PUBLISHED (sic).” Nonetheless Dr. Wakefield has stepped down as executive director of the Thoughtful House, an Austin, TX-based autism center, following the retraction.
Since the publication of The Lancet article in 1998, the rate of measles, mumps and rubella vaccinations has fallen from more than 90% to as low as 61% in some parts of England,1 and measles correspondingly went from being a nearly eradicated disease, with just 56 cases in 1998, to 1,370 cases in 2008. In the meantime, scientists have tested and re-tested the conclusions drawn by the Wakefield paper, finding them wanting, and 10 of the paper's 13 authors have disavowed its findings. Nevertheless the paper's influence remains broad, especially as a cultural touchstone.
The Lancet paper has had less impact on vaccines in the United States, but it's still an issue that comes up in emergency departments.
“From time to time we do get parents who say they haven't had their children vaccinated, or don't want to for safety reasons,” said Lou Hampers, MD, MBA, Section Chief of Pediatric Emergency Medicine at The Children's Hospital in Denver and an Associate Professor of Pediatrics at the University of Colorado School of Medicine.
Fabricated Data
Dr. Wakefield's paper joins some infamous ranks. One of the more prominent recent retractions in the medical literature concerned the Massachusetts anesthesiologist Scott Reuben, MD, who fabricated data for 21 medical studies that claimed to show benefits from painkillers such as Vioxx, Celebrex and Lyrica. Reuben even went so far as to urge the US Food and Drug Administration, based upon his own safety and efficacy data, not to restrict usage of some of the painkillers he studied.
The Springfield, MA-based Baystate Medical Center asked several journals to retract the articles in 2009 after a routine audit of Reuben's work showed discrepancies.
Three of the articles were published in the journal Anesthesiology, which promptly retracted them. In an editorial to his journal's readers, Anesthesiology editor in chief James Eisenach noted, “PubMed retraction alone is unlikely to affect continued citation of these fabricated data.”2 Therefore, he wrote, the editors of his journal would search the citations of all submitted manuscripts as part of the peer-review system for both Dr. Reuben's work, as well as work that cited Dr. Reuben's papers, to ensure the articles were no longer cited in Anesthesiology.
“This is obviously a time-consuming task, but we felt it was necessary to remove his work from the subsequent literature,” Dr. Eisenach said in an interview. “If not, retracted articles can affect the scientific and lay community for a long time.”
Following the Reuben retractions Dr. Eisenach said his journal is also considering requiring all authors of an article to attest the veracity of the data in the publication, although it's not clear that would have prevented the falsifications in the painkiller papers.
Viral Vestiges
Dr. Eisenach had good reason to worry about trying to remove the vestiges of Dr. Reuben's work from published studies. Nigel Drury, BM(Hons), MRCS, a surgeon in the Department of Cardiothoracic Surgery at University Hospital Birmingham in Great Britain, almost cited a retracted article in 2008. This prompted his curiosity about the prevalence of retracted articles. He searched MEDLINE for retracted publications in Cardiothoracic Surgery Network journals, with which he is most familiar. Of the 36,000 articles published between 1990 and 2008 he found just 10 articles.
“This is, hopefully, a credit to the integrity of the scientific community,” Dr. Drury said.
Of the 10 articles identified by Dr. Drury, 5 were retracted by the editor, 4 by the authors and one by the publisher (due to accidentally being published twice in consecutive issues of the same journal). Each of the articles had a published retraction in the journal within a median of 8 months. An analysis of the 9 articles retracted by editors and authors showed they had been cited 40 times, and these articles, in turn, had been cited 263 times. Of the 40 primary citations, 31 of these citations came more than one year after the date of the retraction.
Ironically, he said, the other publication withdrawn by the journal through no fault of the authors has been cited 15 times, compared with just once for the earlier version of record. Clearly, Dr. Drury said, the current measures to retract articles are inadequate.
“The impact of ongoing citation of retracted articles is difficult to judge,” he said.
“At best, it is poor research practice that propagates anything from an honest mistake to an attempt to intellectually defraud the scientific community. At worst, it could lead to the use of an inappropriate or unproven treatment to the detriment of patient care. Fortunately, the most significant and therefore high profile cases are likely to avoid ongoing citation in proportion to their potential impact. However, the content of all retracted work should be effectively erased from the scholarly record and the growing use of digital resources should make this easier to achieve in practice.”
Once an article is withdrawn, Dr. Drury noted that all articles which have previously cited it can be identified through the Thomson Reuters Web of Knowledge database. These articles, he said, should then have the reference marked as withdrawn in both the .html and .pdf versions to avoid further propagation as many authors take references from secondary sources without necessarily going to the primary source article and would otherwise be unaware.
“This may be an important mechanism for preventing the citation of retracted articles but is not currently in practice to my knowledge,” Dr. Drury said.
The ineffectual nature of retractions has not escaped the editors of medical journals. Although the overall number of retractions in medical journals remains small–Dr. Eisenach said he has only had to deal with the Reuben papers and one other duplication that required a retraction by another journal as editor in chief – the percentage is growing.
Last year Thomson Reuters analyzed peer-reviewed science journal articles from 1990 to 2008 extracted from the Thomson Reuters Web of Science citation database. The company, not surprisingly, found that the number of journal articles has doubled, from 690,000 3 decades ago to 1.4 million in 2008. But the number of retractions has increased even more, by a factor of 10 when taking the increase in publications into account. The company found just 5 retractions in 1990, and 95 in 2008.
Similar analyses of MEDLINE have come to like conclusions, said Elizabeth Wager, a freelance medical writer and editor. In 2000 there were just 23 retractions out of 526,490 articles in Medline, or 0.004 percent. In 2009 there were 279 retractions out of 835,759 articles, or 0.03 percent.
“It's impossible to tell whether this increase reflects an increase in misconduct, or honest error, or whether editors, or even authors, are more aware of their responsibility to retract flawed work,” Wager said.
How to COPE
This striking rise prompted the Committee on Publication Ethics (COPE), a group that studies the integrity of peer-reviewed publications and whose members are mostly editors in chief of scientific journals (Annals is a member), to publish retraction guidelines in November 2009.3 As chair of COPE, Wager was the first author of the new guidelines.
“There was no particular event that spurred the guidelines, but COPE had grown increasingly aware, through cases brought by its members, that journal editors weren't sure what to do and therefore weren't acting consistently,” Wager said. “Some editors seemed reluctant to retract articles or were under the misapprehension that the only people who can retract an article are the original authors. Others seemed too 'gung ho' and suggested retraction when it wasn't appropriate–for example when there was an authorship dispute but no question that the findings reported in the article might not be reliable. We knew, also, that some publishers were worried about authors suing them if their article was retracted. We therefore felt that guidelines would be helpful.”
Wager said the response to the new guidelines, which differentiate retractions from corrections by stating they should be reserved for publications that are so seriously flawed that their findings or conclusions should not be relied upon, have been well received by member editors and journals.
Reasons for fraud vary. Wager said she and Peter Williams of University College London analyzed more than 300 retractions from PubMed between 1988 and 2008 to determine the cause of retractions. Honest research errors (28%) led the list, followed by redundant publication (17%), plagiarism (16%), non-replicable findings (11%), data fabrication (5%), disputed authorship (5%), data falsification (4%) and inaccurate or misleading reporting (4%).4
As part of their research Wager and Williams conducted interviews with medial journals, who indicated they would like more guidance for retractions so there is more consistency in the field. With the new guidelines coming out in advance of The Lancet's retraction of the Wakefield paper it seems unlikely COPE will revisit the issue. And the International Committee of Medical Journal Editors (ICMJE) will maintain its existing Uniform Requirements for Manuscripts, said Christine Laine, MD, editor of Annals of Internal Medicine, and a leader of ICMJE.
Too Much Information
That doesn't mean, however, that the issue of journal retractions won't be thrust into the public spotlight. And it's already caused The Lancet to reconsider its approach to scientifically interesting but speculative research. The threat of a high-profile retraction may make the journal more conservative.
“We used to think that we could publish speculative research which advanced interesting new ideas which may be wrong, but which were important to provoke debate and discussion. We don't think that now,” said the The Lancet editor Richard Horton, MD, in an interview with NPR's Bob Garfield after the retraction.
“The problem is that that's just impossible today. The 19th century days where you could sit in the salon at the Royal Society and have a private conversation amongst your fellows, it just doesn't exist anymore. So I think, yeah, too much information in this particular case is a bad thing, which seems to go against every kind of democratic principle that we believe in. But in the case of science, it seems to be true.”
References
- Measles-associated encephalitis in children with renal transplants: a predictable effect of waning herd immunity?. The Lancet. 2003;362:832
- . Data fabrication and article retraction: how not to get lost in the woods. Anesthesiology. 2009;110:955–956
- Retractions: guidance from the Committe on Publication Ethics (COPE). http://publicationethics.org/files/u661/Retractions_COPE_gline_final_3_Sept_09_2_.pdfAccessed April 23, 2010
- . Why and how do journals retract articles?. http://www.ama-assn.org/public/peer/abstracts_2009.html#112Accessed April 28, 2010
Section editor: Truman J. Milling, Jr, MD
Funding and support: By Annals policy, all authors are required to disclose any and all commercial, financial, and other relationships in any way related to the subject of this article that might create any potential conflict of interest. The author has stated that no such relationships exist. See the Manucript Submission Agreement in this issue for examples of specific conflicts covered by this statement.
PII: S0196-0644(10)00378-1
doi:10.1016/j.annemergmed.2010.04.012
