Originally published in March 2010
By Marilyn T. Zivian, Ph.D. (Psychology)
Senior Scholar, Associate Professor Emeritus
York University, Toronto, Canada
Board Member, Breast Cancer Action
San Francisco, CA
It has long been known that most clinical trials of prescription drugs are supported by grants from pharmaceutical companies. A recent New York Times article (see Singer, N., New York Times, August 5, 2009; August 18, 2009) reported that pharmaceutical companies were hiring ghostwriters to write favorable scientific review articles about their drugs. But what effect do these practices have on the actual research performed? A close look at a set of studies of aromatase inhibitors, which came to my attention entirely fortuitously because I am a breast cancer patient, is a case study of the insidious role of conflict of interest in medical research.
Over the past few years, a search of the research literature on aromatase inhibitors (three drugs used to treat breast cancer) has led me to believe that for at least 20 years journals have been publishing papers which masquerade as objective scientific studies, but are, in fact, little better than advertisements for pharmaceutical companies that make the drugs. Conflicts of interest, spinning, and the inability or unwillingness of publishers and editors of scientific journals to ensure proper peer review of submitted manuscripts have converged to allow pharmaceutical companies to commandeer purportedly objective sources of information. Until rules of publication and peer review are rigorously applied, sources of conflicts of interest are eliminated, and ghost-writing sponsored by pharmaceutical companies is prohibited, neither physicians nor patients will be able to make informed decisions about health care. Nor will the quality of health care be as good as it should be.
Twenty-two years ago I was diagnosed with breast cancer and treated with surgery, radiation, and chemotherapy. Twelve years later I was diagnosed with breast cancer bone metastases. To date, the metastases have been successfully controlled by tamoxifen, a drug which until recently had been the drug of choice for many women with breast cancer. One of the major fears about prolonged use of tamoxifen is that at some point it may cease to be effective. Should that occur, I would probably need to switch to one of the three aromatase inhibitors (anastrozole, exemestane, or letrozole), drugs that are quickly becoming first- and second-line treatments of choice (instead of tamoxifen) for post-menopausal women with hormone-receptive breast cancer.
I am a psychologist with extensive training in experimental methods and statistical analyses. Therefore, a few years after the second diagnosis, I began to search through the research literature on aromatase inhibitors. I was looking for information that would help me choose between them. If I had to take an aromatase inhibitor, I wanted it to be the one with the highest rate of overall survival and the fewest debilitating side effects.
Because the design, follow-up period, and research questions of the large aromatase inhibitor clinical trials were different in each study, it seemed impossible to compare the drugs to one another directly. So, when I came across a paper entitled, “Are all aromatase inhibitors the same? A review of controlled clinical trials in breast cancer,” (Berry, 2005), I thought I might finally find the evidence I was seeking.
The Berry (2005) Review Paper: Conclusions and Conflict of Interest
Two Conclusions Favoring Letrozole
From my first, cursory, read-through of the Berry (2005) article, it seemed that one of the aromatase inhibitors, letrozole, might be better than other two. I was most impressed by the following two conclusions: (1) that letrozole was more effective than anastrozole as a neoadjuvant treatment, and (2) that patients taking letrozole reported fewer side effects and chose letrozole as the aromatase inhibitor they would like to continue taking.
Conflict of Interest
These conclusions were called into question, however, when I discovered that Novartis Pharmaceuticals, the manufacturer of letrozole, had supported the publication of the article and that Phase Five Communications, Inc. had provided editorial assistance. At that time, I was unaware of the ghost writing practices reported in the recent New York Times articles. My immediate concern was that conflict of interest might have skewed the conclusions in the Berry (2005) review.
According to the International Committee of Medical Journal Editors (ICMJ, 2008), a “conflict of interest exists when an author (or the author’s institution), reviewer, or editor has financial or personal relationships that inappropriately influence (bias) his or her actions” (ICMJ, 2008). The ICMJ definition is similar to that used in the psychological literature: “Conflicts of interest occur whenever there is substantial risk that one’s professional judgments or actions regarding a primary interest will be unduly influenced by a secondary interest” (Dana, 2009, p. D-1).
Perhaps the most worrisome aspect of conflict of interest reported in the psychological literature is that “when researchers stand to gain by reaching a particular conclusion, they tend to unconsciously and unintentionally (emphasis added) weigh evidence in a biased fashion that favors that conclusion” (Dana, 2009, p. D-1). The ICMJ (2008), too, warns that “the potential for conflict of interest can exist regardless of whether an individual believes that the relationship affects his or her scientific judgment (again, emphasis added).”
The effects of conflict of interest are insidious. Even if researchers do not want to be affected by– and wholeheartedly believe that they are not being affected by– the vested interests of those who support their research, they are. It is precisely because the effects of conflict of interest can be unconscious and unintentional that it is so difficult, if not impossible, to control them.
Editors of some medical journals have attempted to deal with the effects of conflict of interest by asking authors to disclose all possible conflicts of interests (e.g., a pharmaceutical company’s financial support for a trial of the company’s drug or ownership by any of the authors of the pharmaceutical company’s stock).
Unfortunately, disclosure is often ineffective and can sometimes even make things worse for the reader. As it did in my case, disclosure increases readers’ uncertainty about the value of an article’s conclusions. Should the article’s findings and conclusions be completely discounted? Even though the authors have private incentives, their conclusions may still be legitimate (Dana, 2009, pp. D-9 & D-10). One does not know.
Conclusion 1: Neoadjuvant Therapy
I reread the Berry (2005) article, looking for possible instances of conflict of interest. I found that Dr. Berry’s conclusion that letrozole was more effective than anastrozole as a neoadjuvant treatment was critically undermined by evidence presented in the text of the article. Neoadjuvant therapy consists of giving women diagnosed with breast cancer either tamoxifen or one of the three aromatase inhibitors before surgery. The goal is to reduce the size of tumors in hopes of reducing the extensiveness of the surgery needed to remove them.
In the article’s text, Dr. Berry reports that research has shown that the longer the duration of the neoadjuvant therapy, the better the outcome. During months 0 to 3 the median reduction in tumor volume is 52%; during months 3 to 6, there is an additional 57% reduction; and during months 6 to 12, a further 66% reduction (Berry, 2005, p. 1675). The data purportedly showing the superiority of letrozole is presented in Table III (Berry, 2005, p. 1674). The two neoadjuvant trials for anastrozole each ran for a total of 12 weeks, whereas the trial for letrozole lasted four months—approximately a 45% increase in length of treatment for those who took letrozole (122 vs. 84 days).
Thus, we do not know if the better results were due to letrozole’s superiority or to the fact that letrozole was taken for a considerably longer period than anastrozole. Length of therapy is confounded with choice of drug, invalidating Dr. Berry’s conclusion that letrozole is more effective than anastrozole in the neoadjuvant setting. If he wrote the article himself (see below), Dr. Berry must have been fully aware of the positive correlation between length of treatment and reduction in tumor size, yet he ignored the difference in length of therapy between the two drugs, interpreting the data to support the superiority of letrozole. Such an oversight, whether deliberate or accidental, is highly likely to be due to conflict of interest.
Conclusion 2: Quality of Life
In order to check the validity of the second conclusion, that patients taking letrozole reported fewer side effects, I read the two papers [Thomas (2003) and Thomas et al. (2004)] cited by Dr. Berry in support of this conclusion. I emailed Dr. Thomas requesting copies of the two articles. A few weeks later I received two papers. But the Thomas (2003) paper was not one of them. The papers I did receive included a draft of the Thomas et al. (2004) paper and a pdf file of Thomas et al. (2008), a paper I had not requested. I did eventually obtain pdf files of the Thomas (2003) review and Thomas et al. (2004) research papers.
The Thomas (2003) review paper and the Thomas et al. (2004) research paper: duplicate publications?
The bulk of the Thomas (2003) paper (12 of 17 paragraphs and all 6 tables) describes the same research as the Thomas et al. (2004) paper. Although the style of writing and introductions to the two papers differ, the data, tables, and conclusions are, with minor variations, the same. I was astonished. To me, the two papers appeared to be duplicate publications.
The major problem with duplicate publication is over-counting—the same data used more than once to support a particular conclusion, making that conclusion look stronger than it is. Duplicate publication is expressly forbidden in the publication guidelines of Clinical Oncology. It is possible that Dr. Thomas informed the editor of Clinical Oncology about the similarities between the Thomas (2003) paper and the manuscript being submitted for publication, and that the editor decided that there was not enough overlap for them to be duplicate papers. However, such a conclusion would have been unwise, because both papers contain the same data. And, in fact, the publication of the same data first in the American Journal of Clinical Oncology (Thomas, 2003) and then in Clinical Oncology (Thomas et al., 2004) resulted in both papers being cited in the Berry (2005) paper, a hidden over-counting which made letrozole look better than it is—an over-counting that may be due to conflict of interest.
The Thomas (2003) review paper: no peer-review and a difficult-to-find acknowledgement of Novartis support.
I emailed Dr. Thomas asking him about the duplicate publication and why, given the similarity between the two papers, the educational grant from Novartis Pharmaceuticals that he acknowledged for his 2004 study was not acknowledged at the end of his 2003 paper. Dr Thomas did not answer these questions. He did reveal though that the experiment reported in both papers was part of a “recently completed trial by Novartis” and that his 2003 article had not been peer-reviewed—something I never expected. Peer review, at its best, is an unbiased and independent critical assessment of scholarly work, including scientific research (ICMJE, 2008). Scientists and journal editors count on peer reviewers to point out inconsistencies, confounds, illogical conclusions, and overgeneralizations. It is one of the ways that scientists and journal editors have of ensuring that only high-quality and trustworthy research is published.
I emailed Dr. Thomas once more. I again asked about the possible duplicate publication and, having learned that his 2003 paper had not been peer-reviewed, I asked why. Since Dr. Thomas still did not answer my questions, I decided to look for answers elsewhere.
Since in its Online Submission and Review System instructions to authors, the American Journal of Clinical Oncology describes itself as “a peer-reviewed, multidisciplinary journal…,” I emailed Dr. David Wazer, the current editor-in chief, to see if I could find out how a paper that had not been peer-reviewed was published in his journal. He responded that he could not help since he had not been the editor of the journal in 2003 and advised me to contact the publisher.
So I emailed David Myers, the only person listed under “Publication” on the American Journal of Clinical Oncology website. He responded that the supplement in which the Thomas (2003) article appeared was “probably a Novartis supplement,” but that he had “no way to check,” since the publisher’s “warehouse does not retain issues that old,” and that “Pharma (sic) sponsored supplements are generally not peer reviewed.”
In order to verify the Novartis sponsorship, I found the American Journal of Clinical Oncology supplement in which the Thomas (2003) paper was published at the University of California (San Francisco) Medical Library. The publication of the entire supplement was supported by an “unrestricted educational grant from Novartis Oncology.” So not only was the Thomas (2003) paper not peer-reviewed, but the other five papers published in the supplement to the American Journal of Clinical Oncology had probably not been peer-reviewed either.
It appears that since Novartis support was acknowledged for the entire supplement, it did not have to be specifically acknowledged in any of the individual articles. Journal editors hope to alert readers to possible conflicts of interest by asking authors to reveal all sources of financial support. However, when a potential conflict of interest is not acknowledged in each individual article, but only for an entire publication or supplement, unsuspecting readers, who receive reprints of individual articles, are not forewarned.
Finally, I emailed Dr. Hoskin, the current editor of Clinical Oncology. I forwarded copies of the Thomas (2003) and Thomas et al. (2004) papers, told him why I was concerned that they might be duplicate publications and asked if the person who had been the editor of Clinical Oncology at the time the Thomas et al. (2004) was submitted for publication had been aware of the overlap between the content of the two papers. Dr. Hoskin wrote back, “I can assure you that the editor was not aware of this potential conflict” and that he would “look further into it.” As of yet, I have not heard further from Dr. Hoskin.
The Thomas et al. (2004) results not replicated.
In contrast to the Thomas et al. (2004) results, other studies [Zivian & Salgado (December, 2006; June, 2008), and Thomas et al. (2008)] which directly compared letrozole to other aromatase inhibitors did not find significant differences in the number and severity of side effects reported by patients.
Zivian and Salgado (2006; 2008), in their two reports of a large online survey of women taking aromatase inhibitors, found no significant and reliable differences in the number and severity of side effects between the three aromatase inhibitors. Thomas et al. (2008), a study similar in design to the Thomas et al. (2004) experiment, also found no significant differences between letrozole and exemestane in quality of life, side effects, and patient preferences. (No financial support from Novartis was acknowledged for the Thomas et al. (2008) study, and the online survey was not supported by a pharmaceutical company.)
Although the Zivian and Salgado (2006; 2008) and Thomas et al. (2008) studies are not true replications of the Thomas et al (2004) experiment, their results cast doubt upon the reliability of the Thomas et al (2004) finding. Rather than sponsor a replication themselves or wait for future replications by others, Novartis Pharmaceuticals, in a flurry of publications and presentations (see below), promoted the finding that letrozole provides a better quality of life than one of the other aromatase inhibitors—a finding that, in fact, has not been replicated in later studies.
In my search on the web for information linking Dr. Thomas, his research, and Novartis Pharmaceuticals, I found a website containing “confidential” information “for internal use only.” The document appears to have been addressed to a group of Novartis drug representatives. It extols research presented by Dr. Thomas at the third European Breast Cancer Conference in Barcelona, Spain in March, 2002 (Thomas, Makris, & Bloomfield, 2002) as “the first and only (emphasis in the original) head-to-head comparison evaluating tolerability, quality of life and patient choice between Femara (letrozole) and Arimidex (anastrozole),” and announces that “Dr. Thomas will also be presenting these results at our investigator meeting at ASCO (American Society of Clinical Oncologists) meeting in May.”
Thus, the potential for developing conflict of interest occurred repeatedly. To begin with, the original data were obtained as part of a trial supported by Novartis. Then, the results were presented in March 2002 in Barcelona, in May 2002 at an investigator meeting (ASCO) sponsored by Novartis in Orlando, Florida, and in June 2002 at a Novartis symposium in San Paulo http://www.cancernet.co.uk/robertthomas.htm). These presentations were followed in 2003 by a non-peer-reviewed publication as part of a supplement to the American Journal of Clinical Oncology sponsored by Novartis, in 2004 by a peer-reviewed publication in Clinical Oncology in which the support by Novartis was acknowledged, and finally, in 2005, the Thomas (2003) and Thomas et al. (2004) papers were cited and referenced in a review article by Dr. Berry in Clinical Therapeutics—yet another paper whose publication was supported by Novartis.
In addition to the support provided by Novartis, Dr. Berry also acknowledged editorial assistance provided by Phase Five Communications, Inc—editorial assistance that appears to have been paid for by Novartis. The advertising slogan for Phase Five Communications Inc. is: “Spin your science into gold” (http://www.phase-five.com/). It is “a full-service medical education company partnering with pharma (sic) for close to 20 years.” One of the company’s goals is to “increase the value of your publication plan” by developing a “strategically driven publishing timeline (which) ensures that data release is linked to marketing objectives,” i.e. getting the “right data to the right journal at the right time.”
Phase Five Publications, “a specialized division” of Phase Five Communications, Inc., is “dedicated to the development of…publications in the support of pharmaceutical products and issues (emphasis added).” “On-staff scientific writers…work directly with authors to ensure: scientifically rigorous submissions, bias-free, conflict-free conclusions, and compliance with International Committee of Medical Journal Editors’ guidelines.”
Whatever Phase Five Publications contributed, it is unlikely to have ensured that the Berry (2005) paper would reach bias-free, conflict-free conclusions. The business of Phase Five Communications Inc. is to “spin your science into gold.” Spinning is anathema to science, being a form of propaganda that, in this case, attempted to paint the most favorable picture of a pharmaceutical company’s product. Objectivity is not Phase Five Communications’ mandate. Although, the Berry (2005) publication does not contain a note acknowledging that Phase Five Publications’ editorial assistance was supported by Novartis, it seems clear from the information on Phase Five Publications’ web site that its contracts for editorial assistance are made with pharmaceutical companies, not individual authors.
The flurry of conference presentations and journal publications sponsored by Novartis appears to serve the same purpose as pharmaceutical advertisements on television and in the newspapers. They are attempts to influence practicing physicians, who alone can write prescriptions. Perhaps the most important sentence in television and newspaper advertisements begins with some variant of, “Tell/ask/discuss with your doctor…” When patients present their physicians with information from a pharmaceutical company’s advertisement, they are, in essence, serving as unpaid drug representatives. When medical researchers present the same results in multiple conference presentations around the world and publish them repeatedly in different journals, they too are acting, whether or not they are aware of it, as drug representatives, helping the pharmaceutical companies reach as many practicing physicians as possible.
The disguised advertising in journals and conference presentations, the duplicate publications, the multiple presentations, the use of editorial assistants to spin results to make them look better than they are, and the publication of non-peer-reviewed articles in journals: all of this is the result of conflict of interest. It would be one thing if the research could be trusted, but it cannot. The blatant, but apparently unseen, confound between two sets of data (one showing that length of treatment leads to increasingly better outcomes and the other that letrozole had been given to patients for a significantly longer interval than anastrozole) —a confound that invalidates Dr. Berry’s conclusion that letrozole is more effective than anastrozole in the neoadjuvant setting—is an excellent example of how research might be compromised by conflict of interest. The ICMJE (2008) warns that: “Financial relationships…are the most easily identifiable conflicts of interest and the most likely to undermine the credibility of the journal, the authors, and of science itself.”
Ultimate Effects of Conflict of Interest
The implications of widespread conflict of interest like that of the Berry (2005) paper are staggering. Beginning researchers, before they could even start to think about embarking upon new research, would have to either replicate relevant past research, or obtain the authors’ data in order to check the accuracy of data analyses and verify conclusions. The results of articles that may be biased by conflict of interest or spinning, especially if they have not been properly peer-reviewed, cannot be trusted to form a firm foundation for new research.
Practicing physicians, who, unlike beginning researchers, have neither the means nor the time to replicate experiments and reanalyze data, must rely on the findings contained in articles published in medical research journals or papers presented at conferences, findings that may be compromised. As a result, through no fault of their own, they will be prevented from providing their patients with the best care possible.
Ultimately it is patients, not pharmaceutical companies, who pay for the advertisements masquerading as scientific publications, the duplicate publications, the over-counting, and the spinning. They pay with their money, their health, and their lives. Conflicts of interest may undermine the credibility of journals, authors, and science itself, but it is patients who, in the end, suffer the consequences.
American Journal of Clinical Oncology http://journals.lww.com/amjclinicaloncology/_layouts/1033/oaks.journals/informationforauthors.aspx
Berry, J. (2005). Are all aromatase inhibitors the same? A review of controlled clinical trials in breast cancer. Clinical Therapeutics, 27(11), 1671-1684.
Clinical Oncology http://www.elsevier.com/wps/find/journaldescription.cws_home/623018/authorinstructions
Dana, J. (2009). How psychological research can inform policies for dealing with conflicts of interest in medicine. In Institute of Medicine (ed.), Conflict of Interest in Medical Research, Education, and Practice, The National Academies Press, Washington, DC. Note: This is a prepublication copy (uncorrected proofs).
International Committee of Medical Journal Editors http://www.icmje.org/publishing_4overlap.html
Phase Five Communications http://www.phase-five.com/
Singer, N. (August 5, 2009) Medical papers by ghostwriters pushed therapy. The New York Times.
Singer, N. (August 18, 2009) Ghosts in the journals: a call to put a halt to medical articles drafted by drug companies. The New York Times.
Thomas R. (May, 2002). Patient choice and AIs. Paper presented as part of a Novartis symposium at the meeting of the American Society of Clinical Oncology, Orlando, FL, USA. (It is unclear whether there were other authors; none were named in Dr. Thomas’s CV) http://www.cancernet.co.uk/robertthomas.htm
Thomas R. (2003). Examining Quality of Life Issues in Relation to Endocrine Therapy for Breast Cancer. American Journal of Clinical Oncology (CCT) 26(4 Suppl 1): S40-S44.
Thomas, R., Godward, S., Makris, A., Bloomfield, D., Moody, A.M., Williams, M. (2004). Giving patients a choice improves quality of life: a multi-centre, investigator- blind, randomized, crossover study comparing letrozole with anastrozole. Clinical Oncology. 16: 485-491.
Thomas, R., Makris, A., Bloomfield, D. (March, 2002) Empowering patients to make informed treatment decisions based on tolerability, quality of live and patient preference. A comparison of letrozole and anastrozole in a multicenter, randomized, single blind crossover study (Paper presented at the Third European Breast Cancer Conference, Barcelona) [Abstract 171]. European Journal of Cancer, 38 (Supplement 3):S83.
Thomas, R., Williams, M., Marshall, C., & Walker, L. (2008). Switching to letrozole or exemestane improves hot flushes, mood, and quality of life in tamoxifen intolerant women. British Journal of Cancer, 98: 1494-1499.
Zivian, M.T. & Salgado, B. (December, 2006). Side effects revealed: women’s experiences with aromatase inhibitors (corrected January, 2007). Breast Cancer Action, San Francisco, CA.
Zivian, M.T. & Salgado, B. (June, 2008). Side effects revisited: women’s experiences with aromatase inhibitors. Breast Cancer Action, San Francisco, CA.