September 22, 2009

In anticipation of sweet sales, American shopkeepers are busy decorating for Halloween. Duff Wilson and Natasha Singer are supplying the ghosts (*). "According to a study [...] by editors of The Journal of the American Association", they write, "responding authors reported a 10.9 percent rate of ghostwriting in The New England Journal of Medicine", a figure which would make Auguste Maquet proud. Today who knows the latter ought to share the fame of The Three Musketeers with Alexandre Dumas (1)?

Ghosts are supposed to be frightening. As Natasha Singer and Duff Wilson explain in more details (**), "one of the underlying assumptions of the health care overhaul effort is that money can be saved and medical care improved by relying more heavily on research showing which drugs and procedures are the most effective." Unfortunately, in exchange for money and anonymity, ghostwriters may be too eager to support the conclusion which favors their patrons. In medical jargon, their presence is a risk factor for bias.

Perhaps fears of bias are overblown? Isn't modern medicine a science and scientific truth a matter of verifiable facts, as opposed to say religious convictions? The problem with experimental sciences is that so many results rest on raw data, an input highly susceptible to manipulations. Had he lived in the Middle Ages, Dr Scott Reuben would have manufactured relics. In the XXIth century, he simply made up medical research data (2).

Less extreme but easier to hide is selecting data to fit the theory, a point raised in a dispute relative to Vioxx, a drug no longer on the market (3). My truncated quote for instance would support the view JAMA was out to skewer its competition. Actually JAMA also accused itself of "a ghostwriting rate of 7.9 percent", similar to the rate of NEJM given the protocol followed. Truth to tell, most authors of medical papers carefully acknowledge how their access or lack of access to raw data created an artificial selectivity which may have biased their findings despite their best efforts(4).

A further difficulty comes from the reliance of much medical research on pattern recognition, a discipline which accepts errors as inevitable. Positive correlations may have predictive value, they do not necessarily uncover a causal link. Take ghostwriting itself. It may be a risk factor for bias but some ghostwriters are as committed to professional ethics as named authors. Bias may also occur without ghostwriters. In this respect Dr Scott Reuben did make one repeatable experiment which shows unwitting co-authors help to get published (5). Shall we call such dupes ghostsigners?

These difficulties arising from data and method change the nature of medical literature. Rather than being a pure channel of scientific truth, it often conveys expert opinions whose truth is based on their authors' personal power of conviction. This does not alter the importance of the channel. Authority is one of the three sources of truth. But authorities must expect contradiction and will benefit from recommendations. The role of medical journals is to provide such a recommendation for their authors.

Authoritative medical opinions themselves are often cast as recommendations rather than proofs, for instance when presenting trade-offs. Drug X is found to be more potent than drug Y but also the source of some harmful side effects (6). This leads some journals to weigh in the implied discussion with an editorial as an independent opinion (7). For those of their readers in practice, medicine remains an art as well as a science.

Looking at medical literature as a source of expert opinions coupled with a recommendation mechanism rather than a fount of science is not without consequences. But before proceeding, I must disclose a personal interest as ePrio, the company I have founded, may benefit from this perspective. After all its technology supports the provision of communication channels over the Internet with these same characteristics.

One focus of these fillips has been to study recommendation mechanisms (8) and they may contribute useful observations to today's topic. For starter they acknowledge how vital and yet how difficult it is to build such a mechanism and how hard it is for recommenders to earn an honest living. For each specific case, nothing short of a systematic review will do. Concerning medical journals, the danger would be to use the current ghostwriter controversy as a red herring to avoid the pain which would attend such a review. Doesn't William L. Lanier, Editor-in-Chief of Mayo Clinic Proceedings, concur in his thought-provoking article about "conflicts of interest involving industry and medical journals" (***)?

Thus introduced by William L. Lanier, Laurence J. Hirsch makes two very valid points to even out the playing field for the industry (****). Medical journal editors should treat every contributor equally. Indeed there exist "serious conflicts of interest not related to the pharmaceutical industry", arising for instance when a paper bears on a subject related to its author's role as an expert witness for the plaintiff in some legal case.

This fillip does not intend to substitute for the systematic review called for. Still the ghostwriter controversy does raise a new twist on the problem of recommender's independence from the object of the recommendation. Even credit rating agencies may no longer enjoy unchecked their lucrative dependency, as Joanna Chung reports on "proposals [by the SEC] that would tighten supervision" (*****). But dependency needs not be financial.

I have proposed the job recommendations teachers routinely make on behalf on their best students as the ideal business model, because their main academic duties free these recommenders from looking for a quid pro quo. It would be natural to seek a similar ideal for the medical literature. Wouldn't it be the case if its authors were released from all financial entanglements? Would the stifling side effects on creativity feared by William L. Lanier and Laurence J. Hirsch be an acceptable trade-off?

The issue is that, contrary to a job recommendation, a paper is a vehicle for both giving a recommendation, e.g. in favor of a prescription drug, and for receiving a recommendation, i.e. one more publication in a peer-reviewed journal. To the author who competes for advancement and prestige, and who doesn't, each recommendation is already paid in kind even in the absence of any other consideration.

"Investigators and authors who wish to bypass any obstructing and behaviorally inappropriate editors and journals can use Internet resources." William L. Lanier should explore the full potential of this solution. Should anonymous reviewers publicly accept the responsibility for recommending specific papers? Surely, in the age of social networking, the abetting of inevitable back-scratching cliques inside academia could be contained by its very publicity and by opening recommendations to all qualified experts. Should then journals stop recommending papers and instead concentrate their resources on recommending authors, vetting their personal authority and integrity beyond the nature of their compensation?

Had Internet been invented two centuries ago, it is hard to believe The Three Musketeers would have been denied both popular and critical acclaim. But as the authors are concerned, maybe Alexandre Dumas would have been left with more money and Auguste Maquet with more fame.

Philippe Coueignoux

• (*) ......... Ghostwriting Is Called Rife in Medical Journals, by Duff Wilson and Natasha Singer (New York Times) - September 11, 2009
• (**) ....... Unmasking the Ghosts, by Natasha Singer and Duff Wilson (New York Times) - September 18, 2009
• (***) ..... Bidirectional Conflicts of Interest Involving Industry and Medical Journals: Who Will Champion Integrity?, by William L. Lanier (Mayo Clinic Proceedings) - September, 2009
• (****) ... Conflicts of Interest, Authorship, and Disclosures in Industry-Related Scientific Publication: The Tort Bar and Editorial Oversight of Medical Journals, by Laurence J. Hirsch (Mayo Clinic Proceedings) - September, 2009
• (*****) . Regulator seeks to tighten its grip on credit ratings agencies, by Joanna Chung (Financial Times) - September 18, 2009
• (1) see Auguste Maquet in the Wikepedia
• (2) The Scientific American has called him "a medical Madoff", see Scott Reuben in the Wikepedia
• (3) see Vioxx in the Wikepedia
• (4) e.g. Do NSAIDs Affect Changes in Knee Cartilage?, by Changhai Ding et alii (The American Journal of Medicine) - September, 2009
• (5) see Former Army Doctor Accused of Research Fraud Takes Leave from University, by Duff Wilson and Barry Meier (New York Times) - May 23, 2009
• (6) e.g. Tigracelor versus Clopidogrel in Patients with Acute Coronary Syndromes, by Lars Wallentin et alii (The New England Journal of Medicine) - September 10, 2009
• (7) e.g. Tigracelor - Is There Need for a New Player in the Antiplatelet-Therapy Fiels?, by Albert Schömig (The New England Journal of Medicine) - September 10, 2009
• (8) see "recommendation mechanisms" in the list of Major Themes of these fillips.