by David Healy
In 1999, having agreed to testify in a case involving homicide and suicide on Prozac, I learned of company documents showing that Lilly had been aware for years that Prozac might make some people suicidal and indicating the lengths to which the company went to avoid warning of the risk. Although the documents were in the public domain, few people were aware of them. My immediate thought was to write an article outlining the material for the British Medical Journal.
After the question of Prozac-induced suicide was first raised in 1990 and the first legal actions filed against Lilly, in 199i the British Medical Journal carried an article by Charles Beasley and a series of Lilly-only authors. Despite the fact that the data in the article show a 1.9-fold increased risk of a suicidal act on Prozac compared to placebo, this article was widely spun as evidence that there was no risk from the treatment.36 It drew a response from one academic who had a history of difficulties with the pharmaceutical industry: “The BMJ is a journal of distinction and, dare I say it, perhaps also of some innocence. At a time when in the United States the manufacturer of fluoxetine [Prozac] is facing litigation, the corporate defense attorneys will be pleased by the journal having published a piece authored wholly by the manufacturer's employees.”37 But even he seemed unaware the article had appeared in the very week Lilly was facing hearings at the FDA about the risk of suicide on Prozac.
The initial response from the then-editor of the British Medical Journal, Richard Smith, to my draft article appeared to recognize that the journal had had a role in the “mistake” in 1991. He seemed to welcome my submission and suggested reframing it as a discursive debate article rather than a data-driven research article. The reframed article was sent to a reviewer, who was apparently not told that it was an article about company behavior and not an evidence-based assessment of the case for Prozac-induced suicidality. The reviewer suggested the article had not established the case for treatment-induced problems— which it had never attempted to do. The editor rejected the piece on this basis. Mystified at the mismatching messages, I appealed, but in vain: in a phone call Smith stated that no matter what revisions I made nothing would be published.38
This article was published unaltered in the International Journal of Risk and Safety in Medicine, whose editor, Graham Dukes, also editor of the most authoritative book on drug-induced side effects in the field,39 commented: “It seems to me your approach is original and fair…. I also agree with your remarks about the failure of the present overall research approach to elicit a reliable picture of adverse effects and the sometimes unrealistic defenses put up by industry when their products are the subject of injury litigation.”40
A year later, in 2000, having conducted a blind and randomized trial in healthy volunteers, in which two volunteers had become suicidal on Zoloft, I again contacted the British Medical Journal about a submission. Nobody had ever accused Richard Smith of being in the pocket of the pharmaceutical industry; there are in contrast many journals whose editors have previously been employees of the pharmaceutical industry. For those who think conflict of interest is at the heart of all therapy-related problems, the prediction is clear—the British Medical Journal should be much more likely to take this article than a journal whose editors have been so closely linked to industry. In fact, even without review, the British Medical Journal declined to take the article. So I then sought publication instead in a journal whose editors had a pharmaceutical company background, which seemed to make them less rather than more nervous about offending industry. The paper was peer reviewed and rapidly published.41
What is going on? The following examples help clarify the possibilities. In Spring 2000, the Hastings Center Report, a leading bioethical journal, published a special issue, “Prozac, Alienation and the Self,” featuring five articles on the drug.42 Two of these articles suggested that Prozac was so wonderfully effective that it would be a mistake to restrict its use simply to people who are depressed; it should be available to anyone who responded to it. Two further articles argued that even though Prozac was very effective, its use should be restricted to people who are depressed. The fifth article, by me, argued that impressions of Prozac's efficacy stemmed in great part from the fact that negative trials were left unreported, that data on the hazards of the drug were concealed, and that this state of affairs was linked in part to the fact that articles on Prozac and on other psychotropic drugs were increasingly ghostwritten.
Following the publication of this issue of the journal, Lilly, at the time the biggest single outside funder of the Hastings Center, withdrew support, charging that the Center had “published articles that Lilly felt contained information that was biased and scientifically unfounded and that may have led to significant misinformation to readers, patients and the community.”43
Subsequently, in 2003, using the Current Medical Directions portfolio of Zoloft articles noted earlier, I submitted an article to the British Journal of Psychiatry on ghostwriting, whose key finding was that a majority of articles that deal with pharmaceutical products in leading journals are likely to be ghostwritten and that these articles do not faithfully report the data on hazards. A clearly nervous journal staff used at least five reviewers (instead of the usual two) and had the revised article re-reviewed—making up to ten reviews. The article was subsequently referred to the journal's legal department and after that the journal's copyeditors pored over the final version for longer than any other article I've ever written. It was finally published after a change in editors.44
A bit earlier, in 2001, Contemporary Psychology requested a review of Joseph Glenmullen's Prozac Backlash, a book that outlined some of the hazards of treatment with SSRIs. My review outlined the book's key points, without endorsing or countering the position of the author. I also noted that I was in possession of five highly critical reviews of the book written by distinguished American psychiatrists, disseminated by public relations agencies working for Lilly who urged media outlets not to feature the book. I sent the review and the accompanying documents to the editors. Having initially been accepted, the review failed to appear. On enquiring I was told that the journal could not find a balancing reviewer and that they thus could not carry my review. The response made little sense.
In 2004 two journals, Open Minds and Young Mind, requested articles on the subject of antidepressant-induced suicidality in pediatric populations. Both journals then declined to publish what I submitted on the basis, I was told, of legal advice. They said they did not have the resources to handle any difficulties that might arise with pharmaceutical companies as a result of the articles, and this could put them out of business.
The issues here do not just affect academic journals. In 2005 the Times Higher Education Supplement (THES) featured a series of articles on the case of Aubrey Blumsohn, the man described above who had “blown the whistle” on Proctor and Gamble over the ghostwriting of articles and concealment of data on patient responses to Actonel, their blockbuster treatment for osteoporosis.45 I wrote to THES to make it clear that Blumsohn's case was not unique. In the new world of medicine, data were routinely withheld and companies threatened to sue periodicals that reported on treatment risks. THES offered to publish an amended version of the letter, but the changes they wanted stripped the letter of its meaning. When I suggested their revisions made the letter pointless, THES responded, “We have also had to run these letters past our lawyers as this is, as you are aware, a very sensitive issue, and there are certain legal amendments we had to make.” The key amendment seemed to be that the letter could not state that anything the journal published on this issue was constrained by threats of legal action. They did not publish my letter.46
A more striking instance took place in 2007, when I was approached by Index on Censorship for a piece outlining evidence “that pharmaceutical companies are not transparent and that medical journals allow this to happen…. I think to an outsider who has certain expectations of science (that data is widely available and that access to data is fundamental in
terms of any credibility) it's a baffling and shocking state of affairs.”
The resulting article covered the evolution of ghostwriting and the lack of access to clinical trial data, focusing on Study 329, the biggest study undertaken of the use of Paxil in children who were supposedly depressed. After much back and forth, the editor commented, “Our lawyer's just taken a look at your piece—and I do need to ask you for more chapter and verse on some points.” I complied, and eventually heard this: “I realize this is taking up more of your time than you bargained for and do apologize—lawyers must make you weary by now—but am sure you'll understand that it's necessary.” The process ended with this assertion: “The documents made interesting reading— and certainly answered the concerns—along with the cuts. But I've still got worries about running the piece…. I regret how things have turned out very much. I've appreciated all your help in finding documents and in cooperating with all my requests. As I've said before—it's a hugely important subject and we should be covering it.” Index on Censorship self-censored.
Leemon McHenry from Los Angeles and Jon Jureidini from Adelaide in Australia also tried to publish a paper on Study 329, “Clinical Trials and Drug Promotion: Selective Reporting in Study 329,” initially in the Lancet and later in the British Medical Journal. The Lancet sent the paper to GlaxoSmithKline, who objected, and the journal backed down. The British Medical Journal then contacted them saying they had heard of their paper and wanted to fast-track its publication. Six months later, the British Medical Journal indicated their lawyers still had concerns and they would not publish. They finally published a heavily modified version several years later.47
In 2005, the British Medical Journal had a new editor and I submitted an article on how the data on suicide and antidepressants had been manipulated. The peer reviews I received back were longer than the original paper. After answering all queries from two rounds of peer review, the paper was accepted, and proofs of the paper due to appear in the journal two weeks later were sent to me. In the middle of correcting the proofs, I received an email from the editor: “Thank you very much for all your hard work on this article. I'm afraid we've run into a legal wall with our libel lawyer reluctant for us to publish your piece…. I remain supportive of publication but obviously can't do this against legal advice.”
I spent over a year answering legal queries and providing supporting documentation; finally, a year and a half later, the article was published.48 Where the original had implicated both companies and regulators in concealing the problem, the wording had been altered so that the new version emphasized the failings of the FDA for the corrupted data in the public domain and deemphasized any company failings.
Where SmithKline & French once had to threaten to sue, journals now self-censor. Views not favorable to company interests seldom if ever appear in medical journals or social science journals or in any journals or other outlets. Across the board the hazards of blockbusters are notably underreported and in particular, reports of these hazards do not appear in the most influential journals. Even reviews of company documents conceding the hazards of drugs that have become available in the public domain seldom appear.
Self-censorship and ghostwriting go curiously well together—ghosts traditionally can be detected by their lack of a mirror image. Where there should be articles on the benefits of treatment mirrored by articles on the hazards, there are a rapidly diminishing number of articles on either the benefits or hazards actually authored by medical academics. When medical writers contact the offices of our most prestigious journals on behalf of corporate-sponsored projects, neither the writers nor the journals are inhibited by any considerations about incurring a legal action.
Quite the contrary. In 2006 the editor of JAMA, Catherine DeAngelis, tackled the issue of why leading journals could not simply ban further articles from academics who had been publicly linked to ghostwritten or other tainted articles:
Leveling sanctions against an author who fails to disclose financial interests by banning publication of his or her articles for some time period would only encourage that author to send his or her articles to another journal; it cleans our house by messing others. So what about all editors, or at least a group…agreeing to share the information and jointly to ban the offending authors? Those who suggest this approach have not considered the risk of an antitrust suit.49
This statement all but concedes that our leading medical journals will, in effect, ignore transgressions of their own rules when it comes to publishing drug company-related articles. Even though it is an essential legal and ethical duty for physicians to report on treatment hazards, journals will rarely take such reports if the material is inimical to the interests of a pharmaceutical company. In contrast, the material that increasingly fills our journals does not conform to the basic norms of science, namely to make available the full complement of data on which claims are based.
These failures to publish are not isolated anomalies. The articles cited above were all articles central to regulatory hearings or legal actions against drug companies. When Study 329 came to light, New York State took a fraud action against GlaxoSmithKline on the basis that the apparent science they had promoted led doctors to prescribe this drug and children to receive it, and the State to pay for these prescriptions, although it was likely to be ineffective. It was this money they were now claiming back. GlaxoSmithKline settled, further illustrating the failure of nerve of medical editors.
SCIENCE EX MACHINA
In striking contrast to these publication difficulties, when marketing sertraline (Zoloft), Pfizer's efforts were geared to producing an average of two to three articles per month in significant journals,50 many of which appear to have been ghostwritten.51 In the case of the three leading SSRIs combined, this would mean six to nine articles per month—two per week. In the case of Lilly's Zyprexa, the four clinical trials that brought this drug on the market gave rise to 234 publications, all advocating the efficacy of the compound with none containing data on the increases in glucose or cholesterol levels or rates of suicide found in these trials that have since become the subject of legal actions.52
As a simple matter of probabilities, therefore, when it comes to any public-domain difficulties companies may have to manage, whether arising from legal cases of treatment-induced injury or from regulatory hearings, they can almost always parade apparently new studies that portray their product in a good light. In addition to simple probabilities, companies are much more aware of the dates of trials and regulatory hearings concerning products than academics, more aware of the publication timelines of journals than academics, and unlikely to be held up by a journal feeling they need to undertake a legal review.53 When it comes to real-time debate in the public domain, therefore, it is almost inconceivable that patients or academics will have the data or resources necessary to engage with a company.
Statements from professional bodies can be targeted with even greater precision to coincide with regulatory hearings or legal trials than standard journal articles. This is because such statements rarely go through editorial or peer review processes. In the week before the first regulatory hearing on suicidal behavior among children taking antidepressants in 2004, the American College of Neuropsychopharmacology (ACNP) sent out a press release of a then as-yet-unpublished position statement apparently from a distinguished set of academics that concluded that antidepressants did not cause suicidality in children.54 This position statement was widely covered in the media at the time. The paper was written by a public relations agency based in Washington, DC.
Having personally written a position paper for another professional association and attended many consensus statement meetings, as outlined in the next chapter, I can vouch for the ease with which the timing of the appearance of statements from bodies like these can be controlled. Typically such statements arise from small working groups within an organization. They may be proposed by an individual, who then selects the working group. Once a position statement
is formulated it is easy for a small group of individuals to target publication to an opportune date. Alternatively if a final statement is agreed upon but publication delayed, a draft can be issued directly to the media, as with the ACNP position paper.
From the time of their launch in 1987 through the end of 2009 there were four legal cases involving SSRIs, three of them centered on suicide and one, the Kilker case, on birth defects. In the weeks prior to each of the suicide cases at least one article had appeared outlining the benefits of drug treatment and claiming that antidepressants reduced rather than increased the risk of suicide55 or proclaimed the benefits of antidepressants in pregnancy.56
Since 1991, there have been four regulatory hearings on SSRIs and suicidality. Two in 2004 involved children, while one in 1991 and one in 2006 involved adults. In addition to the ACNP statement that came out prior to the February 2004 FDA hearings, a paper appeared in July, prior to the September 2004 hearings, suggesting that there was no risk linked to SSRIs.57 The Beasley paper appeared in the week of FDA hearings on Prozac and suicide in 1991,58 while prior to the 2006 FDA hearings regarding adult suicidality on antidepressants, a series of articles was published in the American Journal of Psychiatry suggesting that warnings added to SSRI labels had themselves led to an increase in the rate of suicide, inferring that it would be a mistake to further extend the warning from children to adults.59
In close proximity to every single SSRI jury trial or regulatory hearing, therefore, one or more articles favorable to the drug company view has been published in a major journal. Pharmaceutical companies, it would seem, have refined the process of managing “science” to the point of being able to turn up papers on demand in an attempt to save the day if their product is in difficulty or under scrutiny.
DISAPPEARING SCIENCE
Revelations about ghostwriting, conflicts of interest, and hidden studies have prompted calls for reform. In the United States, Republican senator Charles Grassley has taken a lead in calling universities and professional associations to account for the undeclared conflicts of interest some of their academics have with the pharmaceutical industry60 and for allowing members of their staff to claim authorship for articles that have in fact been ghostwritten.61 He has sponsored a Sunshine Act in the Senate to shed light on these practices62—but there is no mention of shedding light on the clinical trial data without which we can never know if payment has corrupted an academic or not. The Sunshine Act focuses on rotten apples in the barrel rather than on the barrel and risks therefore suiting pharmaceutical companies rather than troubling them.