Until recently, EPIDEMIOLOGY maintained a straightforward policy on conflicts of interest, asking authors to report their place of employment and the study’s sources of funding, as well as any other information that could raise questions about a paper’s credibility if disclosed later. “We will be minimalists,” explained an editorial in 2006.1 Authors were required to identify those conflicts that might be “relevant to their work and interpretation of their findings,” but the journal did not define “relevance.” Instead, the editor noted, “We put that responsibility on our authors, where we think it belongs.”
As of January 2012, a new policy was imposed on the journal by the publisher, Wolters Kluwer. Authors must now complete and submit an extensive questionnaire based on the International Committee of Medical Journal Editors’ (ICMJE) “Uniform Requirements for Manuscripts Submitted to Biomedical Journals.” Authors must check through several pages of questions about potential conflicts of interest (COI), including financial, consultant, institutional, and other relationships.
The impact of COI on research is extensive and well documented: industry-sponsored research produces overwhelmingly pro-industry conclusions,2 and there is substantial evidence of selective reporting, suppression of negative findings, and other forms of publication bias.3 Moreover, these problems may be amplified through systematic reviews or meta-analyses, most of which fail to report or take into account COI disclosures in their component studies.4
It is therefore easy to understand the superficial appeal of more elaborate COI disclosure rituals. Disclosure purports to provide relevant information to health scientists, patients, and journal editors, to help them judge the veracity and reliability of research findings. This contention led to a major revision of the ICMJE COI guidelines a decade ago5 and also underlies the “sunshine” provisions of the Patient Protection and Affordable Care Act, mandating the publication of “transparency reports” that publicly disclose industry payments to physicians and teaching hospitals.6
Although the ideals of transparency and empowerment of health information consumers are laudable, some of the appeal of disclosure can also be viewed as self-serving. For many stakeholders, disclosure serves as an inoculation against more systemic change—an endurably burdensome alternative to real critique and reform, such as intolerance of the conflicts in the first place.7 As James Surowiecki noted in reference to the financial industry, “Transparency is well and good, but accuracy and objectivity are even better. Wall Street doesn’t have to keep confessing its sins. It just has to stop committing them.”8
In the case of biomedical research, disclosure allows journals to continue to publish sponsored research, thus avoiding taking any kind of stand against the more fundamental dilemmas of commercial science, a system that buoys journal impact factors and revenue.9 Disclosure also places the burden of responsibility solely on the consumers of information: if they choose to believe or act on research with a stated COI, they (and not the researchers or the publishers) are solely to blame for any disappointments.7
As members of an evidence-based profession, epidemiologists might legitimately ask: What is the evidentiary basis for this new policy? Does disclosure actually deliver on its promises? And if so, are higher doses of this medicine necessarily better, and free from negative side effects? The answer to the latter two questions is a qualified “No.”
There has been remarkably little research on the impact of written COI disclosures on consumers of medical research. In one randomized trial, a sample of BMJ readers rated articles with declared competing interests as being less credible than the same articles with no such declaration.10 A follow-up trial found that readers rated articles with a “financial” conflict (eg, employee or stockholder who may benefit financially) significantly lower than articles with no stated conflict or with a “grant” conflict (eg, funding from a company).11 In contrast, a more recent study of 253 physicians found that, although they generally believed they should discount for a stated COI, providing this information on a single study had no impact on their likelihood of prescribing a new drug.3
More troubling is the evidence that disclosure may have perverse consequences. Cain and colleagues7 conducted an experiment in which participants were assigned one of two roles.6 “Estimators” were asked to determine the total value of a jar of coins. “Advisors” sat close to the jar and provided suggestions to the estimators, who sat farther away. One subset of advisers was incentivized to ensure that estimators guessed as accurately as possible. Two other subsets of advisers had a COI: they were incentivized to make the estimators’ guesses as high as possible. One of these groups of advisers disclosed this fact, while the other did not. When advisors had no COI, their suggestions were close to their own personal estimates. Advisors’ suggestions were higher than their personal estimates in the undisclosed COI scenario, however—and higher still when their COI was disclosed. The same pattern held for the estimates made by the estimators, which were least accurate in the disclosed COI scenario.
Although this face-to-face interaction might not be directly analogous to the consumption of scientific research, these results are nevertheless worrisome evidence regarding expanded disclosure policies.7,12,13 Accommodating the disclosed COI seemed to be extremely difficult for the estimators—how much should they adjust the advisers’ information, or should they take it into account at all? In the experiment, disclosure did not lead estimators to entirely disregard advisers’ biased advice; they adjusted their guesses, but not by enough to offset the advisers’ bias. Disclosure may have functioned paradoxically to disarm a reasonable skepticism, allowing advisors to cloak biased recommendations in the veneer of honesty.
Disclosure may also have encouraged the advisers to engage in two additional behaviors that amplified rather than mitigated their bias: “strategic exaggeration” (providing more biased advice to offset the estimators’ corrections) and “moral licensing” (justifying bias because the advisees had been warned). COI statements in practice may thus have a perverse effect, not of reminding authors of their duty to present evidence as objectively as possible, but rather of giving permission to exaggerate their claims under the assumption that the readers have been warned. And readers may not sufficiently adjust their interpretations to offset this extra bias.
The idea of “moral license” is applicable in a broader sense. Disclosure purports to offer a solution to the problem of conflict of interest, and more extensive disclosure is naturally seen as the path to even greater purity of knowledge. In fact, the most important function of disclosure may be to absolve the research enterprise as a whole, licensing continued biases in publication of research while forestalling serious, systemic change. To paraphrase Surowiecki, researchers, editors, and publishers would not have to stop committing sins if they just keep confessing them. Ultimately, the empirical research, what there is of it, suggests that the confessional model of disclosure may have more negative than positive consequences and that increasing the dosage may well exacerbate these effects.
The change in policy at EPIDEMIOLOGY is one that authors and editors will have to learn to endure, but it bears noting that this was instated by corporate policy without consideration for its potential negative impact. The previous system may have been just as effective at eliciting important COI disclosures, while at the same time imposing substantially less burden on authors than the new and more intrusive confessional checklist. If we are to take the problem of COI seriously, we should be looking for solutions, not absolution.
1. Wilcox AJ. On conflicts of interest. Epidemiology. 2006;23:241
2. Bekelman JE, Li Y, Gross CP. Scope and impact of financial conflicts of interest in biomedical research: a systematic review. JAMA. 2003;289:454–465
3. Silverman GK, Loewenstein GF, Anderson BL, Ubel PA, Zinberg S, Schulkin J. Failure to discount for conflict of interest when evaluating medical literature: a randomised trial of physicians. J Med Ethics. 2010;36:265–270
4. Roseman M, Milette K, Bero LA, et al. Reporting of conflicts of interest in meta-analyses of trials of pharmacological treatments. JAMA. 2011;305:1008–1017
5. Davidoff F, DeAngelis CD, Drazen JM, et al. Sponsorship, authorship, and accountability. Ann Intern Med. 2001;135:463–466
6. Steinbrook R, Ross JS. “Transparency reports” on industry payments to physicians and teaching hospitals. JAMA. 2012;307:1029–1030
7. Cain DM, Loewenstein G, Moore DA. The dirt on coming clean: Perverse effects of disclosing conflicts of interest. J Legal Stud. 2005;34:1–25
8. Surowiecki J. The talking cure. The New Yorker. December 9, 2002 New York
9. Lundh A, Barbateskovic M, Hróbjartsson A, Gøtzsche PC. Conflicts of interest at medical journals: the influence of industry-supported randomised trials on journal impact factors and revenue—cohort study. PLoS Med. 2010;7:e1000354
10. Chaudhry S, Schroter S, Smith R, Morris J. Does declaration of competing interests affect readers’ perceptions? A randomised trial. BMJ. 2002;325:1391–1392
11. Schroter S, Morris J, Chaudhry S, Smith R, Barratt H. Does the type of competing interest statement affect readers’ perceptions of the credibility of research? Randomised trial. BMJ. 2004;328:742–743
12. Cain DM, Loewenstein G, Moore DA. When Sunlight Fails to Disinfect: Understanding the Perverse Effects of Disclosing Conflicts of Interest. J Consum Res. 2011;37:836–857
13. Loewenstein G, Sah S, Cain DM. The unintended consequences of conflict of interest disclosure. JAMA. 2012;307:669–670