Improving researchers’ conflict of interest declarations

Poor accessibility

Most articles contain conflict of interest statements in the online and PDF version. In others, however, the statements are only available online, as supplementary files, or on separate web pages, or are unavailable because links to disclosure statements are broken or missing. Some journals provide links to a PDF of the authors’ uploaded ICMJE forms instead of providing complete summaries and some include only the links with no disclosure in the article at all. The longest supplementary PDF we found totalled 94 pages (for 31 authors).7 Occasionally, articles have more than one conflict of interest disclosure statement published in different locations, and these statements may be inconsistent.89

Inconsistency

Journals use diverse headings to identify conflict of interest statements (box 1), suggesting that they have not implemented a clear and consistent definition. This makes it challenging to automate extraction of disclosures for analysis, even from well structured web pages. Since March 2017, some publishers have included conflicts of interest in the metadata used by PubMed.10 This could be a way to improve accessibility by centralising where disclosures are recorded.

Conflict of interest statements are often conflated with statements about the funding of the research presented, making it difficult to assess individual authors’ conflicts of interest and study sponsorship separately. We argue that these disclosures should be reported separately because some evidence suggests that funding source and author conflict of interest are independently associated with risk of bias,1112 though the likelihood of confounding is high.13

When conflict of interest disclosures are identified, substantial variability in the language used to make disclosures makes it difficult to interpret the meaning. We identified 130 different ways of stating “no conflicts of interest” across the 637 articles whose authors declared no conflict of interest,6 ranging from one word (“nil” or “none”) to a 62 word statement:

“The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties. No writing assistance was utilized in the production of this manuscript.”

Some of the variants change the meaning of the statement, suggesting that conflicts of interest might exist but are not disclosed (table 1).

Missing information

Compounding the problems of accessibility and semantics is the problem of relevance and who should judge this. Current reporting practices rely on authors to judge the relevance of any interests to the work under consideration, though journal policies and procedures and editorial discretion may also influence what gets reported. For example, the ICMJE form asks authors to “disclose interactions with any entity that could be considered broadly relevant to the work.”

Journal editors, peer reviewers, and readers should be able to judge the relevance of disclosed interests to the work to evaluate whether a conflict of interest exists. However, the necessary contextual information is typically lacking in statements and forms. The result is that although conflict of interest disclosure is widespread, disclosures are not meaningful.

Obfuscation

Authors sometimes seek to define relevance in disclosure statements using adjectives such as “potential,” “relevant,” and “financial” to modify “conflict of interest” or contextualising statements by stating that disclosures pertained to “this work” or “this manuscript.” In some cases, authors disclose seemingly relevant financial ties but make statements that obfuscate whether the disclosed relationship constitutes a conflict of interest.9 For example, in an angiography clinical trial, the authors reported receiving speakers’ fees and unrestricted grants from multiple medical device manufacturers with related products under the heading “Sources of funding” but included nothing under “Disclosures.”14

In other cases, authors provide disclosures that include personal and professional biographical information, funding information that should be in another section of the article, or ambiguous descriptors that may obscure key signals of risk of bias such as financial relationships with industry. For example, one article with 17 authors had a 706 word disclosure statement that was located only in the 63 pages of supplementary ICMJE forms; it included numerous conflicts of interest, as defined by the ICMJE, among disclosures of receipt of public funding and honorariums from not-for-profit organisations (eg, public universities, hospitals); one of the authors disclosed receipt of personal fees from 42 bodies outside of the submitted work, including 23 drug and medical device companies.15 Designations such as “unpaid consultancy” may mask the nature of the relationship with industry and fail to make transparent other transfers of value such as paid travel.16

Enforcement

The variability in how conflicts of interest are reported is probably related to journals having differing requirements for statements rather than a consequence of author intentions. Currently, the onus is on authors to disclose, which makes a breach of this honour system a serious violation of trust. Journal editors often do not have the information or resources needed to identify or verify authors’ conflicts of interest.23 The ICMJE recently revised its policy to include purposeful non-disclosure of conflicts of interest as a form of scientific misconduct.4

However, 14% of articles published in journals claiming to adhere to the ICMJE recommendations still do not routinely include conflict of interest statements.6 ICMJE has limited ability to enforce its guidance but, at minimum, organisations such as COPE should not permit membership to journals that do not conform to these standards.

Assessment of relevance

Though the ICMJE recommends that authors report details on the funding source and the specific role of the sponsor in the design, conduct, and publication of the research, there is no equivalent requirement for authors to provide such structured detail in their declarations of conflicts of interest.4 For example, when an author discloses personal fees from multiple entities, there is rarely information about the scope or extent of the relationships, how the scientific work relates to a company’s products, or the specific reason for the receipt of payment.

We propose that authors make use of public databases, where these exist, to report and maintain a structured and comprehensive list of their disclosures, and that these complete disclosures are made publicly available through a link in the article. Separately, we suggest that authors state which interests are relevant to the manuscript and why, that this statement is peer reviewed and adjudicated by editors and is visibly and succinctly included in all forms of the manuscript under the standard heading “conflicts of interest.” Box 2 provides a sample statement.

Enforced and structured reporting of conflicts of interest would enable large scale retrospective studies of the association between conflicts of interest and research methods, results, and conclusions. Future work should also consider randomised trials of structured reporting systems to assess author, editor, peer reviewer, and consumer usability and acceptability, and the effect on assessment of risk of bias in published work.

Conflict of interest disclosures should help us to evaluate risk of bias in biomedical research, but currently they only muddy the waters. Instead of providing an obvious and clear signal to readers of biomedical research, current practices obfuscate the underlying relationships or flood the signal with noise. To improve consistency, and as a necessary step towards achieving greater independence for healthcare research, we need consensus around the definition of conflict of interest and harmonisation of practices across journals and publishers.