Sunday, October 31, 2010

Biomedical Text Similarity

Science Daily alerted me to this publication on PLoS
Systematic Characterizations of Text Similarity in Full Text Biomedical Publications
Sun Z,
Errami M, Long T, Renard C, Choradia N, et al. 2010 Systematic Characterizations of Text Similarity in Full Text Biomedical Publications. PLoS ONE 5(9): e12704.doi:10.1371/journal.pone.0012704
The authors of eTBLAST, a text-similarity search engine, have expanded their work that at first investigated text- and author-similarity on PubMed abstracts (CSP article from 2008). They have now accessed full-text articles to dig deeper into text similarities.

They investigated over 70.000 full papers, and determined that abstract similarity is a good predictor of full text similarity. They caution, however, that the automatic identification of possible cases of plagiarism must be checked by hand to determine if indeed plagiarism is present. They only uncovered 34 highly similar papers, and all were updates or multi-part articles that did indeed share larger sections of text.

However, they note that many of the currently uncovered plagiarized publications, for example in Chile and Peru [1], were translations and these are not included in the PubMed database.

[1] Sources given in the article about the Chilean and Peruvian cases:
  1. Arriola-Quiroz I, Curioso WH, Cruz-Encarnacion M, Gayoso O (2010) Characteristics and publication patterns of theses from a Peruvian medical school. Health Info Libr J 27(2): 148–154. 
  2. Salinas JL, Mayta-Tristan P (2008) [Duplicate publication: a Peruvian case]. Revista de Gastroenterologia del Peru 28: 390–391. 
  3. Rojas-Revoredo V, Huamani C, Mayta-Tristan P (2007) [Plagiarism in undergraduate publications: experiences and recommendations]. Revista Medica de Chile 135: 1087–1088. 
  4. Reyes H, Palma J, Andresen M (2007) [Ethics in articles published in medical journals]. Revista Medica de Chile 135: 529–533.

Advertising and Peer-Review in Medical Journals

The German online news site Telepolis reports in October 2010 about the pharmaceutical company Wyeth, which belongs to the Pfizer concern, hiring a public relations company (DesignWrite) to inject ghostwritten advertising into peer-reviewed articles that appear in closed-access medical journals.

The PR company carefully wrote articles that in diction and appearance seemed to be scientific articles that surveyed the literature on the topic of hormone replacement studies for menopausal women. The surveys downplayed the negative side effects, which include a higher risk of breast cancer, and praised the positive side effects, such as lowering the probability of dementia, of the substances. Later studies have shown that these hormone replacement medicines actually increase the probability of dementia. These surveys were given to researchers, who "edited" the articles and submitted them for publication to peer-reviewed journals.
This blog noted the article published in 2009 in the New York Times. Adriane J. Fugh-Berman, an associate professor for Complementary and Alternative Medicine at Georgetown University, has now published her study of the incident in the open-access journal PLoS: The Haunting of Medical Journals: How Ghostwriting Sold “HRT”.
Fugh-Berman has published a meticulously linked and documented article that demonstrates the depths of this relationship between the ghostwriters. One sample:
One co-author seemed puzzled by the concept that she was to author, but not write, an article [34]: “From what you have written, I would be more of an ‘editor’ rather than the major writer—that is, you guys would be writing the versions—with me ‘altering, editing, etc.? Is that correct?’” 1This query was in response to an e-mail from Karen Mittleman (a DesignWrite employee who supervised medical writers) that stated: “The beauty of this process is that we become your postdocs! … We provide you with an outline that you review and suggest changes to. We then develop a draft from the final outline. You have complete editorial control of the paper, but we provide you with the materials to review/critique” [34].
This would also suggest that the "normal" way of writing is to have the postdocs do the work and the PI publish the paper. And even when a peer review tried to question a paper, documents show that DesignWrite responded to the reviews (and not the supposed authors), at times scolding the reviewers for misusing the peer-review process! 

Fugh-Bermann summarizes:
Acceptance of ghostwriting, euphemistically termed “editorial assistance,” may be so widespread that it is considered normal. This could explain why several authors of ghostwritten articles have defended their involvement.
As a researcher for complementary and alternative medicine, it is of course to her advantage to demonstrate that what we have considered to be "hard science" in the area of medicine up until now has degenerated into an advertising circus. But her results are not based on just a single case - the list goes on and one. Fugh-Bermann: "Medicine, as a profession, must take responsibility for this situation. Naïveté is no longer an excuse."

Sunday, October 17, 2010

Lies, Damned Lies, and Medical Science

The Atlantic has an interesting article by David H. Freedman in its November 2010 issue, "Lies, Damned Lies, and Medical Science", about John Ioannidis, one of the world’s foremost experts on the credibility of medical research.

Much of what medical researchers conclude in their studies is misleading, exaggerated, or flat-out wrong. So why are doctors—to a striking extent—still drawing upon misinformation in their everyday practice? Dr. John Ioannidis has spent his career challenging his peers by exposing their bad science.

Saturday, October 9, 2010

How to prevent scientific misconduct

The Scientist has an opinion piece by Suresh Radhakrishnan on preventing fraud in research. The author was fired as a senior research associate from the Mayo Clinic in Rochester, Minn. for scientific misconduct in May 2010. I find it very good that the author reflects on what caused him to falsify data and to propose solutions for preventing this in the future.