Financial conflicts of interest and Peer Community In
PCI forbids financial conflicts of interest (fCOIs). The PCI code of conduct (https://peercommunityin.org/ + CODE OF CONDUCT) states that:
“Authors should have no financial conflict of interest (see the definition below) relating to the articles they submit to PCI.”
It also provides a definition of fCOIs:
“A financial conflict of interest occurs when the authors, recommenders or reviewers:
– Receive (or have received, during the last four years) salaries, reimbursements, fees, fellowships, grants, or funding from an entity with financial interests potentially affected by the results of the research presented in the article,
– Have shares or stocks in an entity with economic interests potentially affected by the results of the research presented in the article or
– Hold patents linked to the research presented in the article”
In certain scientific communities, many researchers have large numbers of fCOIs. This is a common situation, and it may also be expected by their employer. These researchers may see an outright ban on fCOIs as extreme, insulting and unfair. They may also see it as a means of pointing the finger at them, accusing them, and calling their ethics into question. A number of researchers have, therefore, complained that it is impossible for them to submit preprints without fCOIs. We therefore provide, below, an explanation of the reasons for the ban on fCOIs at PCI (and, thus, at Peer Community Journal).
How are financial conflicts of interest handled at PCI?
Each journal has the right to impose its own rules regarding COIs, regardless of the dispositions of national law on this subject. Thus, some journals choose not to consider the issue of COIs, others require authors to declare any COI, and others prohibit all COI. When we founded PCI in late 2016, we decided, as its founders and the drafters of its statutes and code of conduct, to ban fCOIs.
Hence, the PCI Code of Conduct requires authors to declare any non-financial COI and prohibits fCOIs of authors. This is made clear on a specific page of all PCI websites (e.g. https://zool.peercommunityin.org/about/ethics) and on the homepage of the general PCI website (https://peercommunityin.org/ + CODE OF CONDUCT).
The founders of each new PCI have agreed to this rule, which is common to all thematic PCIs. When a user registers on the website, a reviewer reviews a paper, a recommender takes on the evaluation of a preprint, or an author submits a preprint, they must accept this rule from the outset.
This rule has been in place since the foundation of PCI. It was introduced: 1) to ensure a high level of scientific integrity and reproducibility, 2) to prevent accusations of poor scientific practice against PCI, 3) to avoid complicated situations arising during the reviewing of papers.
But why ban fCOIs, rather than simply insisting that authors declare them?
What are the effects of financial conflicts of interest?
A COI exists when a researcher’s primary interest (establishing scientific knowledge) conflicts with a secondary interest (e.g. receiving funding from a company with an interest in the outcome of the research undertaken). A COI exists, by definition, if there are conflicting interests. It is not necessary to demonstrate the effect of the conflict or to demonstrate manipulation or malpractice on the part of the researchers or the company for a COI to exist.
Many studies have investigated the fCOIs of researchers and their consequences. See, for example, the numerous articles by Sheldon Krimsky, a specialist on the subject.
Most conflicts of interest arise in the medical field, in which pharmaceutical companies fund research projects for drug development, and finance the testing of products and clinical trials. These companies also have other financial links with researchers, whom they employ as advisors or experts or support through funding for conferences, or with gifts, for example.
One frequent major consequence of fCOIs is funding bias: statistically, results with a positive impact on the interests of the company are more frequently obtained in studies funded by the company concerned. A similar situation is observed when researchers have financial ties to companies. There is a strong significant statistical relationship between the presence of a fCOI and the type of results presented in articles.
Many studies on this bias have been published: such associations between fCOI and study outcome have been reported in food science [Food Policy. 2011. 36:97-203. – https://doi.org/10.1016/j.foodpol.2010.11.016], tobacco [JAMA. 1995;274(3):248-253. doi: https://doi.org/10.1001/jama.1995.03530030068036], biomedical [JAMA. 2003;289(4):454-465. doi: https://doi.org/10.1001/jama.289.4.454], nuclear [Sci Eng Ethics. 2011; 17: 75±107. doi: https://doi.org/10.1007/s11948-009-9181-y], sugar [PLoS Med. 2013; 10: e1001578. doi: https://doi.org/10.1371/journal.pmed.1001578] and pharmaceutical [BMJ 2003; 326 doi: https://doi.org/10.1136/bmj.326.7400.1167] studies.
Statistical biases due to fCOIs almost never involve a conscious manipulation of experiments, data, analyses, or interpretations. In most cases, there is, therefore, no “cheating” involved. The bias may be due to error, unconscious manipulation, interpretations extending slightly beyond the reasonable, unconscious self-censorship, and so on. There is generally no reason to accuse the authors of bad practice. A reason for a funding bias may also be that companies choose to fund studies that are likely to generate positive results for them, rather than ones that are likely to generate a negative result (because that is in their interests).
These effects can be detected statistically. In general, it is impossible to demonstrate the causality of the effect of a fCOI on a particular outcome. In most cases, reviewers cannot detect these effects because they work on the basis of the elements provided by the authors. For a reviewer, it is virtually impossible to know whether some of the data were not included in the results, that many statistical analyses were performed before one that “goes the right way” was obtained, or that the assumptions underlying the paper were formulated after the results were obtained (HARKING), for example. The review process cannot reveal the effect of a fCOI on a particular outcome.
Such effects can be excluded only in certain articles with a very high level of evidence: articles based on pre-registrations (e.g. registered reports), articles that can be formally reproduced with self-executing simulation or analysis programs (executable papers), etc. Unfortunately, papers with such very high levels of evidence remain rare.
Until papers with such high levels of evidence become more common, we believe that the traditional peer-review process will remain unable to ensure that a fCOI has no effect on the outcome of the paper.
We also know that fCOIs are not the only types of COI to generate statistical bias and that other, non-financial COIs (political, symbolic, religious, career, etc.), can be problematic. However, fCOIs are special in that they are highly visible, are easy to identify, and published studies indicate that they may have a very strong effect.
What are the solutions?
Scientific publishers have dealt with this issue in different ways.
Most life science and medical journals require authors to declare their interests publicly. Reviewers and editors are often required to declare their interest privately, but not publicly. Some fields of science are “behind the curve”.
COI reporting is often obligatory, but this does not abolish the statistical effects of COIs on results. COI reporting simply makes the COI public, providing readers with a paradoxical piece of information: “the article you are reading has the particular property of being written by scientists who have an interest (usually unconscious) that may result in their content being not completely accurate.” The effect of such public declarations of COI is unpredictable: some readers will reject the conclusions of the article, whereas others will give the authors a total blank check, precisely because the COI has been declared, i.e. 2 totally opposite effects [The Unintended Consequences of Conflict of Interest Disclosure JAMA. 2012;307(7):669-670. doi: https://doi.org/10.1001/jama.2012.154].
Some journals go further. For example, The New England Journal of Medicine, The Lancet, Annals of Internal Medicine, and JAMA announced in 2004 that they would only publish research on drugs funded by pharmaceutical companies if the study designs had been pre-registered.
Some go further still. For example, the Cochrane Database of Systematic Reviews (CDSR), a reference in high-level ethical science that aims to organize medical information systematically, has not published any articles from authors with fCOIs since 2014 (https://training.cochrane.org/online-learning/editorial-policies/coi-policy/coi-policy-cochrane-library#funding_and_interests).
PCI simply applies a rule similar to that of the Cochrane Organization.
Q: If fCOIs prevent us from publishing a large part of our research, are we not censuring a large body of knowledge of the scientific value that should be considered in the evaluation of science?
A: If PCI continues to prohibit fCOIs, it is not to prevent some researchers from using PCI, but rather to ensure that PCI can evaluate articles properly and ensure that it emits good scientific recommendations. The problem of researchers being able or unable to submit is different from the problem of ensuring high-quality science. Note also that we do not censure science as many journals do evaluate and publish work with fCOIs. We would prefer our colleagues with fCOIs to submit their work elsewhere. PCI does not seek hegemony. If researchers cannot submit their work to PCI, they can still submit their work elsewhere.
Q: Does focusing exclusively on biases introduced by fCOIs make us forget that there are many other sources of bias relating to social issues, political positioning, gender issues and so on, all of which can have severe consequences too?
A: Indeed, there are plenty of other sources of bias and COIs, but many are difficult to avoid or detect. By contrast, fCOIs are easy to identify and we prefer to ban them from PCI because we cannot be sure that they do not introduce a bias.
Q: Isn’t it good for science that private funding exists?
A: PCI has no problem with evaluating/recommending preprints of studies funded by private companies, provided that the funding is not associated with fCOIs.