The Role of Conflict of Interest in Reporting of Scientific Information FREE TO VIEW

When I began my tenure as the Editor in Chief of CHEST on July 1, 2005, I thought that I was prepared to carry out the duties of the position¹ and that nothing would surprise me. While I was aware that scientific misconduct occurred in the reporting of scientific information,² I have been surprised at how much we have uncovered at CHEST and how unethical behavior is often not intuitively obvious to those who perpetrate it. These realizations have stimulated me to write this commentary that will review what we know about the role of conflict of interest (COI) in the reporting of scientific information and challenge those of us in educator roles to do a better job in mentoring our trainees, junior faculty, and associates on what is right and wrong. It is my hope that this article and its contents can serve as a stimulus for the development and incorporation of an educational series in all training programs on what is ethical and unethical in the conducting and reporting of scientific studies.

The concepts of COI and trust are inextricably linked. According to Wiktionary, a wiki-based, open content dictionary,³ “A conflict of interest is a situation in which someone in a position of trust … has competing professional or personal interests.” “A conflict of interest exists even if no unethical or improper act results from it.” “A conflict of interest can create an appearance of impropriety that can undermine confidence ….” While the research of Hall et al⁴ on medical trust defines the interactions of physicians and their patients, we contend that the same principles describe the interactions of authors/investigators and their readers. According to Hall and Weinfurt,⁵ “Medical trust is the willingness to be vulnerable, with the expectation that the care-taker [authors/investigators] intends to and will protect the patient's [and readers] interest.” Patients and readers have every reason to expect that physicians and authors/investigators possess the same principles of medical trust that include fidelity (ie, implying strict and continuing faithfulness to an obligation, trust, or duty), competence, and honesty.⁶

Based on a plethora of reports in the news media as well as scientific publications,^6–11 it has become increasingly clear that real or perceived COI situations pervade our society, medicine, and the reporting of scientific information. Moreover, the public has also become aware of studies that have revealed that physicians and researchers often do not know when they have a conflict and that many physicians believe that they are more immune to industry influence than other physicians, their colleagues.¹² Other reports^13–15 suggest and, in some cases, show the public that physicians can be bought (Fig 1).

Because of these revelations, the public trust in research has been eroded and there is a perception that professional fidelity and honesty on the part of investigators and clinicians has deteriorated.

The public perceives that scientific misconduct is and has been pervasive because it is not a new phenomenon, not limited to the pharmaceutical industry, and not limited to the clinical scientific industry. Two major scandals regarding suppression by the pharmaceutical industry of adverse results that came to light in 2004 received an enormous amount of media attention.^7,16 From this, the public came to understand that there had been suppression of an increased suicide risk in children taking some selective serotonin reuptake inhibitor antidepressants¹⁶ and the delay of up to 4 years in revealing an increased cardiovascular risk in adults taking the cyclooxygenase-2 inhibitor, rofecoxib.¹⁷ The public also came to realize that these events were merely a tip of an “iceberg” of scientific misconduct that included the disclosure of faked skin transplants in white mice by investigators at the Sloan-Kettering Institute in New York in 1974¹⁸ and the revelation of fake cloning studies in Korea.¹⁹ It has also become common knowledge that industry-funded trials, whether they involve trials of medicines or surgical interventions, are more likely to be associated with statistically significant proindustry findings²⁰; this has suggested self-serving manipulation of the literature on the part of industry.

Moreover, journals and their editors, and regulatory agencies cannot be consistently counted on to protect patients. This creates a fundamental threat to the credibility of journals and to science. Some editors and journals have failed to disentangle themselves from commercial biases due to financial pressure (eg, advertising) or editors' COI. For example, the Editor in Chief of Neuropsychopharmacology published a favorable article²¹ on a controversial new treatment, vagus nerve stimulation, for depression resistant to conventional therapy without disclosing that early drafts of the article were prepared by a professional writer hired by the company and that all nine authors (which included the editor in chief who was the lead author) had financial relationships with the company sponsoring the trial.²² Many in the United States, including the Institute of Medicine, are advocating for regulatory reform of the Food and Drug Administration (FDA) because the agency has failed to adequately protect patients^23,24 (Fig 2).

Design bias and publication bias constitute violations of ethical principles. Unfortunately, much data exist that suggest that our literature has been contaminated by studies that are seemingly designed to be biased.

When randomized, controlled clinical trials are designed with a high likelihood of being positive in favor of the sponsor's drug, they are flawed by design bias. This type of bias is not consistent with the concept of equipoise (ie, the uncertainty principle) that is a central ethical principle of clinical research. The uncertainty principle is supposed to ensure that a subject may be enrolled in a phase 3 randomized, controlled clinical trial only if there is a true uncertainty about which trial arm is most likely to benefit the patient. Focused, scholarly surveys^20,25–33 of approximately 1,000 randomized, controlled clinical trials published from 1990 to 2005 showed that systematic bias has favored products made by the company funding the research. Studies³² funded by drug companies were more likely to have outcomes favoring the sponsor than were other studies (odds ratio, 4.05; 95% confidence interval, 2.98 to 5.51). An analysis^32,34 of many of these studies suggested that there likely was an intent to “stack the deck” in favor of the sponsor's product because there was an inappropriate use of a comparator drug (eg, used in a dose that was unlikely to be effective). Similar surveys²⁰ of studies published from 1999 to 2001 revealed similar proindustry trends and included surgical as well as medical trials. In a study³⁵ that sought to determine the frequency with which equipoise conditions were met in industry-sponsored randomized, controlled trials in rheumatology, the authors studied abstracts accepted for the 2001 American College of Rheumatology meeting; in every trial (45 of 45 trials), results were favorable to the sponsor, indicating that the results of the trials could have been predicted in advance solely by knowledge of sponsorship (p < 0.0001).

When sponsors prevent studies unfavorable to their products from being published or selectively or redundantly report favorable studies, this behavior is referred to as publication bias.³⁴ Because subjects enroll in studies to benefit medical science and society, and negative studies further medical knowledge, it is unethical to fail to submit negative studies for publication. It is also unethical to participate in other forms of publication bias such as publishing funded research in supplements as proceedings of symposia when peer review is less intense or does not occur at all; promoting interpretive spin of data that exaggerates the benefits or minimizes the risks of positive results; and duplicating the publication of positive results. In the same scholarly surveys^20,25–33 of approximately 1,000 randomized, controlled clinical trials mentioned above that favored products made by the company funding the research, potential publication bias as well as design bias were also suggested. For example, among cardiovascular trials funded solely by for-profit organizations published in the Journal of the American Medical Association, the New England Journal of Medicine, and the Lancet between 2000 and 2005, 67.2% favored the newer treatments.³³ During the same time period, among cardiovascular trials funded solely by not-for-profit organizations published in the same three high-impact journals, 51% did not favor the newer treatments.³³ These latter data do not support the perception that journals contribute to publication bias because they are reluctant to publish negative trials.

A particularly egregious example of publication bias that involved the promotion of gabapentin by Parke-Davis is important to revisit because it reveals the complicity of investigators with industry.⁸ It is by no means the only example.¹⁵ An analysis of internal documents of the gabapentin story showed that publications in the medical literature were used to advance marketing goals for the drug. Parke-Davis employed a publication strategy to use research not to gain FDA approval for new indications but to “disseminate the information as widely as possible through the world's medical literature.” Documents expressed the concern that negative results could harm promotional efforts, and indicated the intention to publish and publicize results only if they were favorable. The company contracted with medical education companies to develop review papers, original articles, and letters to the editor about gabapentin for $13,750 to $18,000 per article including honoraria for the physician or pharmacist authors; some publications were favorable, some were neutral, none were negative. Most of these articles qualified as duplicate publications.

A duplicate publication is defined as follows on page 148 in the AMA Manual of Style²: (1) the simultaneous or subsequent publication of the major components (or more) of an article two or more times in one or more forms (ie, print or electronic); (2) is present when there is substantial overlap in one or more elements of an article (eg, design, material and methods, tables, graphics, discussion, or letters to the editor). While likely underestimated, duplicate publication has been uncovered 1.4 to 28% of the time; there has been no reference to the original report 5 to 32% of the time.² Most duplicate articles appear within 1 year of the original publication.²

These examples of violations of ethical principles qualify as COIs in the reporting of scientific information because readers of the scientific literature have every reason to trust that those who contribute to the literature (eg, industry, authors, investigators, and editors) are acting faithfully, competently, and honestly in our obligation to publish the truth.

While there have always been multiple theoretical ways to mitigate COIs in the reporting of scientific information (Table 1), there is now a concerted effort on the part of many stakeholders from multiple disciplines to limit/eliminate scientific misconduct that has seemingly been spiraling out of control by some accounts.¹⁵ For example, medical societies such as the American College of Chest Physicians (ACCP) that develop clinical practice guidelines have instituted policies and firewalls to remove as much as possible COIs related to funding of their guidelines.^36,37 The ACCP has also developed policies to avoid allowing potentially conflicted authors to chair guideline committees or sections of guidelines. The Health and Science Policy Committee that oversees the clinical practice guideline activities of the ACCP not only does the third-party vetting evaluations of the chairs and writing committees before the work of the committees begins,³⁸ but also, while the work is ongoing. By these mechanisms, medical societies and journals have the ability to remove potentially conflicted committee members and authors from writing guidelines and review articles or editorials from their list of potential invitees.

An increasing number of journals, including CHEST, are requiring more robust disclosures by having authors not only state and list their potential personal and financial COIs but also acknowledge the specific roles of each of the authors in the performance of the study and writing of the paper to justify their authorship; authors are also being asked to state if they had access to all data generated during the performance of their study.³⁹ Because industry-performed statistical analyses may be subject to bias, these same journals are also requiring that the companies that have sponsored randomized, controlled clinical trials have independent statistical reviews performed of their data to ensure an unbiased analysis. Reviewers are being asked to consider recusing themselves if they believe they have a potential COI with the funders of a trial or the authors of the study.

There are signs everywhere of an increasing awareness of the scientific misconduct problem and the need for more intense adherence to codes of ethics in performing and reporting research. For example, behavior that is considered unethical has not only been spelled out in the Declaration of Helsinki, the best known policy of The World Medical Association that was first developed in 1964 and amended most recently for the sixth time,⁴⁰ but also in the AMA Manual of Style,² the uniform requirements for manuscripts submitted to biomedical journals,⁴¹ and individual journal instructions to authors. Moreover, these issues are covered in specific terms on the Web sites of the International Committee of Medical Journal Editors (ICMJE),⁴² the Council of Science Editors,⁴³ and the World Association of Medical Editors.⁴⁴ The Committee on Publication Ethics (or COPE), established in 1997 and composed mostly of Editors in Chief of scientific journals,⁴⁵ “provides a forum for publishers and editors to discuss issues relating to the integrity of the work submitted to or published in their journals.” It provides 17 practical flowcharts in algorithmic form for editors to follow when publication misconduct is suspected.⁴⁶ Watchdog groups such as the Integrity in Science Project sponsored by the Center for Science in the Public Interest⁴⁷ have also sprung up to, among other activities, raise awareness of problems related to corporate funding, investigate and publicize COIs and acts of potentially harmful influences of industry-sponsored science, and encourage journalists to ask authors and scientists the tough questions regarding COIs and report what they uncover.

We, as editors, are able to uncover plagiarism and image falsification by utilizing software to uncover these examples of scientific misconduct. In addition, we have been impressed that coauthors, study coordinators, and collaborators have felt the need to be whistleblowers,⁴⁸ and reviewers and readers have increasingly been the groups that have first identified unethical issues. With the help of these groups, we have uncovered the following types of scientific misconduct at CHEST in 2008: duplicate publication including self-plagiarism; undisclosed pharmaceutical company funding; falsifying research data before and after publication; using another's ideas without permission or credit; inadequate record keeping that might permit misconduct to occur or make it difficult to investigate; ghost and guest authorship and ghost writing; simultaneous submission to two or more journals; authors of an abstract indicted for fraud and murder⁴⁹; and uncovering that some other journals are still requiring cross-referencing of their own journal articles to pad their impact factor.

Multiple regulatory reform initiatives have been launched. In response to real and perceived failures of the integrity of the US drug safety system, the Institute of Medicine comprehensively reviewed and issued recommendations for reform of the FDA.²⁴ While negotiations concerning these recommendations and other options for reform are ongoing, there has already been progress. While the 1997 FDA Modernization Act required registration by law of efficacy trials testing drugs for life-threatening diseases/conditions and led to the ClinicalTrials.gov registry administered by the National Library of Medicine, the 2007 FDA Amendment Act expanded the types of trials to all except phase 1 clinical trials and small device feasibility studies.⁵⁰ The 2007 Act also required a timeline of registration requirements over the subsequent 3 years, as follows: posting of basic study results not later than 1 year after enactment of the law (ie, September 2008); posting of serious adverse events or most frequent adverse events by March 2009; and posting of expanded study results by March 2010, including a summary in lay terms of the results, the protocol, and a description of quality-related issues used through the study. The 2009 and 2010 processes have not been finalized. The FDA requires that registration information be updated at least every 12 months and within 30 days of any changes in recruitment status or completion of the study.

While FDA registration is required by law, clinical trial registration as initially advocated by the World Health Organization (WHO) and members of the ICMJE is also required for journal publication. The potential benefits of this initiative^6,51 were as follows: (1) eliminate violation of the ethical concept of equipoise by helping editors and others to understand the context of study results; (2) eliminate the underdeclaring of negative studies and/or selective or redundant reporting of favorable studies; (3) ensure the integrity of evidence-based medicine; (4) improve the efficiency of clinical research by reducing unnecessary duplication of research and informing the planning of new trials by knowing what others are already doing; (5) ensure that results are not lost to the body of general medical knowledge; and (6) promote innovation. While it is too soon to know whether these benefits have been realized, we are optimistic.

September 2004 is a watershed date because that is when members of the ICMJE published a policy statement⁵² that they will consider a phase 3 trial for publication only if it has been registered before enrollment of the first patient. It applied to trials that started recruiting on or after July 1, 2005; for trials that were ongoing at the inception of the policy, registration had to be completed by September 12, 2005. The ICMJE embraced the WHO minimal registration data set of 20 items and emphasized that all 20 fields must be filled out⁵² for the registration to be acceptable. An additional very important requirement was that the data must be electronically searchable and accessible free of charge to the public. In an attempt to make the act of clinical trial registration more easily understandable, Zarin and Keselman⁵³ wrote a “Medical Writing Tip” on the topic in 2007 in CHEST. An update on trial registration by Tse et al appears in this issue.⁵⁴ Tse et al⁵⁵ have also written an article in this issue of CHEST that summarizes the current requirements for reporting results and describes how to submit data elements of the results to the ClinicalTrials.gov results database that will satisfy US law.

The requirements of trial registration for publication have continued to evolve. Effective July 1, 2008, the ICMJE required the registration of all clinical trials, including phase 1 and pharmacokinetic trials.⁵⁶ Moreover, once a trial is registered, the ICMJE, just like the FDA (see above), requires that registration be updated⁵³ every 6 months during the recruitment phase; moreover, records should be amended as soon as recruitment status or protocols are changed, or when other information becomes available (eg, links to publications). While ClinicalTrials.gov continues to be the most comprehensive registry, there are an ever-increasing number of clinical trial registries that participate in the WHO International Clinical Trial Registry Platform⁵⁷; the ICMJE accepts registration in those that require the minimal registration data set of 20 items.⁵⁶ In support of trial registration, CHEST has replicated the language of the WHO and the ICMJE in our “Instructions to Authors.”⁵⁸

In an attempt to start to manage the high number of physician-industry relationships,⁵⁹ state legislation, as of 2007, had been enacted in five states (Minnesota, California, Maine, Vermont, and West Virginia) and the District of Columbia that mandated disclosure payments made to physicians by pharmaceutical companies⁶⁰; the disclosures in Vermont and Minnesota were intended to be made public. While an analysis of the public records in Vermont and Minnesota has revealed considerable payments of $100 or more to physicians, the laws in these states have failed to provide the full disclosures to the public.⁶⁰ While these so-called sunshine laws have allowed some light to shine on the problem, they need to be refined to allow more transparency.^61,62 Perhaps, the Physician Payments Sunshine Act⁶² sponsored by Senators Grassley and Kohl that is being considered by Congress will provide the needed transparency.

In order to mitigate COIs in the reporting of scientific information, we believe that a multipronged combination approach (Table 1) is needed. We have also come to appreciate that disclosure is not enough. Studies⁶³ suggest that some authors are relieved of guilt in perpetrating unethical acts because they have disclosed a potential COI, implying that the journal, reviewers, and readers should have known that the data were biased.

Marcia Angell recently stated⁶⁴ the following: “It would be naïve to conclude that bias is only a matter of a few isolated instances. It permeates the entire system.” “Physicians can no longer rely on the medical literature for valid and reliable information.” “Clinicians just do not know anymore how safe and effective prescription drugs really are, but the products are probably nowhere near as good as the published literature indicates.” “It is more than a matter of perception or appearances; it is a matter of public health.” While Dr. Angell has summarized in stark terms where we have been, we believe that corrective strategies are in place and that things will improve. However, we should not delude ourselves into thinking that transparent disclosures of COIs is enough or that unethical behavior is a priori intuitively obvious to those who perpetrate it.