A conflict of interest arises by having two conflicting goals in one's research. The primary goal of research relevant to public health is to produce impartial evidence on health hazards for humans. Several entities – including industry - may have public health as a goal among others, but this is not their primary goal. Primary goals are in those cases profit or career, that conflict with the goal of health. It is a role of the State to foster research whose primary goal is impartial evidence on factors affecting population health. Disclosure of conflicts of interest is not enough: the view that disclosure solves all problems amounts to say that a declaration of having produced unbiased evidence is a self-fulfilling guarantee that the evidence will not be affected by conflicts of interest. This concept is seriously misleading. A conflict of interest arises from the circumstances in which research occurs and does not exist only in the opinion of some people or groups (or the authors of a paper).
- Outcome Research Evaluation
Statistics from Altmetric.com
As highlighted by the debate section in the August 2009 issue of this journal,1 recognition has been growing, especially in the last decade, that conflicts of interest can have a material impact on the results of epidemiological research, be it in the form of original works or of meta-analyses and literature reviews. However, this awareness may still not be matched by the basic recognition of what substantiates an actual case of conflict of interest. A few recent examples of papers can illustrate the point (in spite of occasional mention of clinical medicine, in this paper we exclusively concentrate on public health).
The first is an article by Chang et al2 on the relation between exposure to Agent Orange/2,3,7,8-tetrachloro-dibenzo-dioxin (TCDD) and prostate cancer. The article has been sponsored by two of the largest chemical corporations in the world, Dow Chemicals and Monsanto. Their financial support is clearly acknowledged at the end of the paper, and the authors declare in detail that Dow Chemicals and Monsanto did not influence the methods or materials used by the authors in conducting their independent review and synthesis of the epidemiological literature or the interpretation of the results, nor were they involved in the preparation or review of the manuscript prior to submission.
The second example is a meta-analysis and review paper by Boyle et al3 on sweetened carbonated beverage consumption and cancer risk. The study was funded by an unrestricted research grant from the Coca-Cola company and the authors declare that there are no conflicts of interest as none of the authors have had any financial benefit, or benefit in kind, for working on the project (and none have any stock ownership of relevance to any of the products discussed in this study).
Other examples concern exposure to the pesticide glyphosate and to a herbicide-tolerant genetically modified canola.4 ,5 The first paper is funded by Monsanto and is partly based on internal Monsanto reports but does not acknowledge any conflict of interest; while the second reports funding by Du Pont but explicitly states that there were no conflicts of interests. One can find many similar examples.
Scientific journals have variable policies concerning review papers sponsored by industries producing the agent(s) of which the health effects are evaluated, including the option of not publishing them.1 Whatever the policy, the underlying premise consists of the following question: does such sponsorship substantiate in itself a conflict of interest, even if the conclusions of the review are not demonstrably biased? We imply that scientific rigour is a pre-requisite that we do not discuss in the present paper. All our argumentation is based on a ‘coeteris paribus’ condition, that is, on a comparison of studies with and without a conflict of interest, the rest (including scientific rigour) being equal. We do not enter into a discussion of the quality of the studies mentioned above as all of them overcame a threshold making them acceptable for publication.
Identifying obvious bias is often difficult at the level of an individual paper, but it becomes possible when considering the aggregated totality of evidence. Very well-known, but not isolated, is the case of the distribution of the conclusions of the studies on Environmental Tobacco Smoke by source of funding: studies funded by industry almost always concluded for lack of association and studies funded publicly almost always concluded for the presence of an association with lung cancer.6
Two common problems
We are often confronted with two problems in epidemiological research, the increasing difficulties in avoiding conflicts of interest and the mixing of science and advocacy. Both problems can compromise, but in very different ways, the credibility and integrity of epidemiological research.
At the risk of sounding pedantic, we recall some basic principles of integrity in research. Integrity should be interpreted literally: research needs to remain ‘integer’, namely ‘one’ in its primary goal, not distorted by secondary goals. Public and private institutions fund research of which the primary goal is to answer scientific questions that are relevant to the health of citizens. Thus the primary goal of publicly or privately funded research is to establish ‘reasonable truth’ around a scientific issue, such as the carcinogenicity of a chemical. A conflict arises when the pursuit of truth (“Is chemical x carcinogenic?”) is potentially hampered or corrupted by secondary goals that can in principle distort or anticipate an answer to the question (by anticipation we mean the implicit tendency to consider one result more likely than another). There are several pathways to distorting or anticipating results: motivations are usually the pursuit of benefits, either career advancement or profits, or the support (hidden or overt) of a specific scientific hypothesis, theory or political agenda. To identify a conflict of interest it is not necessary to detect evidence of distortion, because the conflict arises from a clash between the two goals just mentioned.
Career concerns and support of specific scientific hypotheses or theories are inherent normal motives in the job of any researcher, whatever his/her position in public or private organisations, but they may become distorting when uplifted to primary goals, as revealed by the extreme cases of scientific frauds. Profit and financial incentives are, however, extraneous goals that frequently lead to distorting truth in a number of ways, as typified by the long record of misconduct of the tobacco industry.7
Also, ‘advocacy’ can lead to biasing scientific research. Although it is pertinent that the scientist has an opinion on the societal, public health implications of his/her research, the tax-payers (who fund most research) and the society at large fund him/her to ‘pursue a reasonable truth’ (we explain later why we use this expression) without anticipating or inflating the results.
Defining conflict of interest
Let us look at a few commonly accepted definitions of conflict of interest. The brand new Dictionary of Epidemiology8 gives the following definition: “Conflict of Interest. Compromise of a person's objectivity when that person has a vested interest (e.g., in peer review , in the outcome of a study). It occurs when the person could benefit financially or in other ways (e.g., promotion, prestige) from some aspect of a study, report, or other professional activity. Conflicts of interest are not limited to research settings, nor to the health sciences.” Among journals, The BMJ has a similar and rather detailed statement of the conflict of interest.9 According to the policy of the US National Academy of Sciences, the term ‘conflict of interest’ means “any financial or other interest which conflicts with the service of the individual because it (1) could significantly impair the individual’s objectivity or (2) could create an unfair competitive advantage for any person or organization” (however, we would prefer to see the word ‘impartiality’ rather than ‘objectivity’). This definition is used in particular for the selection of eligible members of scientific committees.
According to this and similar definitions, a conflict of interest involves not simply having, for example, two employers at a time, one public and one private, but having two conflicting goals in one’s research. The primary goal of research relevant to public health is to produce impartial evidence on health hazards for humans. Industry may or may not have this as a goal among others, but it is not by definition its primary goal. Industry works towards the public interest by providing scientific evidence and producing goods such as drugs, but its primary legitimate interest is profit. It is instead the role of the State to foster research of which the primary goal is impartial evidence on factors affecting population health.
By State we mean the totality of institutions that have as a goal the regulation of society at large, including protecting citizens from harm and threats to their health. Such goals are acknowledged even by the strictest libertarians. The State also includes publicly funded research institutions. Because of a State's multiple functions conflict of interest might arise when, for instance, a scientific investigation on health is commissioned by the Department of Defence, whose primary goal is by definition not centred on health.
All cited definitions use the conditional ‘could’ (eg, could significantly impair) to distinguish between the circumstances or, in the definition of Stark,10 the objective antecedents that constitute the conflict and the consequences from the outcome behaviour that ‘could’ derive. The antecedents, the relationship with industry, are constitutive elements of conflict of interest even in the absence of evidence of the ‘outcome behaviour’. It is crucially important to keep clear the distinction between conflict of interest and scientific misconduct: for a conflict of interest to exist it is not necessary for the results to be demonstrably distorted in a context of sponsorship from industry—that would constitute scientific misconduct. It is the relationship with industry itself (or with the Department of Defence if this is the case) that creates a conflict because of the objective clash between a primary and a secondary goal, with the uncomfortable consequence that the conflict of interest affects the confidence with which the research results can be judged to be ‘reasonably true’.
Conflicts of interest are not an opinion
There is often confusion about what is the meaning of expressions such as conflict of interest or scientific integrity. For example, a recent article in Science11 describes alleged cases of conflict of interest, including the recent controversy surrounding the case of a senior epidemiologist appointment. The article is unclear in that it suggests that the conflict of interest is a matter of opinion. In relation to the specific case, it argues that working as an epidemiologist for a public institution was judged incompatible with working for the industry by some commentators interviewed by the journalist, while for others this did not raise any moral dilemma. The reader is left with the impression that it is all a matter of purely subjective opinion, or, worse, that there is on one side a ‘politicized’ (advocate) standing that stigmatises the conflict of interest, while on the other side there are independent scientists who are—to use an overworked language—so neutral and objective as to be able to put themselves at disposal of any customer.
Similar to the theory that conflicts of interest are a matter of opinion, the idea that disclosure exhausts the problem is also too simplistic. In principle, a review sponsored by the food industry on the health effects of a food item can be looked at with greater attention (compared with publicly funded research) if the reader is made aware of the conflict through disclosure by the authors. But disclosure itself is not enough: the view that disclosure solves all problems amounts to saying that when a researcher is receiving funds from industry a declaration of having produced unbiased evidence is a self-fulfilling guarantee that the evidence cannot be affected by conflicts of interest, which therefore are regarded as irrelevant to the point of non-existence. This concept is erroneous and seriously misleading. A conflict of interest is an objective datum, it derives from the circumstances and does not exist only in the opinion of some people or groups.
Related to this is the consideration that objectivity is not a realistic goal of research, though sometimes those who disclose their conflicts of interests also pretend they are objective and thus untouched by the consequences of the conflict. We have purposely used the expression ‘reasonable truth’ because ‘objectivity’ is an unattainable goal for an individual scientist (science can attain objectivity only as a collective enterprise). What an epidemiologist can aim at is a comprehensive and impartial description and analysis of observations accompanied by a statement of his/her interpretation and of all methodological limitations. This is ‘reasonable truth’. Believing that we attain objective truth is presumptuous and leads to the wrong view of a scientist who is apparently ‘objective’ but in fact is simply unaware of his/her own prejudice and bias.
Can we avoid secondary goals completely?
A realistic aim is not avoiding secondary goals completely, which is sometimes impossible. There are current rules proposed for dealing with sponsorship bias that are widely agreed on and implemented, such as financial disclosure, study reporting standards or study registries, though they apply more in clinical research than in observational epidemiology. However, these rules often inadequately address the mechanisms by which conflict of interest operates and impacts on research and threatens the confidence in research results. We do not imply that researchers should never accept funding from such sources as industry or Departments of Defence. In fact, the environmental crisis requires a great deal of investments into creative and innovative industries: we refer, for example, to new technologies for the production of clean energy, or chemicals that do not pose carcinogenic hazards. The economic crisis and restriction of government funding may even reinforce the position of such research investments. In addition, scientists in public institutions, notably some universities, are strongly encouraged to establish partnerships and seek funding from industry sources. Also, some kind of studies, typically in occupational epidemiology, can only be conducted with a degree of collaboration from industry.
But the relative roles of researchers (individuals and institutions) and of industry must be very explicitly defined and documented, and the primary goal clearly stated, to impartially answer scientific questions relevant to people's health. An example is the creation of industry-sponsored consortia that fund research projects, selected by independent panels of scientists on the basis of competitive grant calls, targeted to important public health issues (an example is research on electromagnetic fields in France, for which grant attribution is managed by a national agency, ANSES).
We close our considerations with two messages, one negative and one positive.
On the negative side, we warn against often heard self-assured and pretentiously reassuring statements of ‘independence’ unaccompanied by other information: if secondary goals are not rendered explicit (they are explicit in the papers we referred to, thanks to the full disclosure of interests), there is a definite risk that profit-driven or politically driven advocacy agendas become treated in a cavalier way as a mere matter of private opinion rather than public problems objectively interfering with the path towards ‘reasonable truth’.
On the positive side, we recognise the complexity of the issues we have been touching on and we solicit further discussion of them within the epidemiological community carried not as coffee-time entertainment (however pleasant that may be) but at the same level of documentation, depth and rigour as the best of our epidemiological studies.
Conflicts of interest are important and may distort the epidemiological literature.
They originate from an objective clash between a primary goal (protecting public health by providing epidemiological evidence in an impartial way) and secondary goals such as profit or career.
Disclosure of conflicts of interest is not sufficient, but multiple actions are needed to prevent them.
The authors are grateful to Elisabete Weiderpass, Miquel Porta, Béatrice Fervers and Eduardo Seleiro for thoughtful comments on previous versions of this paper.
Competing interests This paper was not supported by grants. The present authors have collaborated in the past with some the authors of two of the quoted reviews (refs. 2 and 3), with whom they have co-authored several peer-reviewed papers. The authors of the present paper do not identify any potential conflict of interest.
Provenance and peer review Commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.