A literature review of non-financial conflicts of interest in healthcare research and publication | BMC Medical Ethics

1. 1.

   Do competing non-financial interests constitute COIs like FCOIs?

The vast majority of authors in our review consider competing non-financial interests to be COIs. Many articles named different forms of competing non-financial interests and assumed they could bias an individual involved in the research or publication process. However, several explanations were given for the categorization of NFCOIs as COIs, including: NFCOIs might introduce bias which skews interpretation of the data to be published [91]; patient and public perception of COIs can be as important as the actual presence of COI [48, 60, 89, 92,93,94,95,96]; NFCOIs might encourage an individual to perform research in a hasty, faulty, or dishonest manner to reach publication as quickly as possible [68]; NFCOIs might result in “spin,” influencing a researcher to use the analyses most likely to produce desired results [68, 97], include only those results aligning with their view, and/or downplay the limitations of their work [97]; author affiliation influences what is emphasized when data is reported [98]; and empirical data shows that disparate forms of NFCOIs can lead to bias for authors [3, 35, 98,99,100,101], reviewers [3, 84, 101], and others involved in the publication process [102]. Despite these specific concerns, an overarching theme across many authors is that NFCOIs are more difficult to define and track than FCOIs. This may partially explain their relative neglect in medical ethical discourse.

Two common themes arose in the articles arguing against the conception of competing non-financial interests as COIs. The first is that such interests are intrinsic to research and impossible to control or eliminate [103, 104]. The second is that the definition of NFCOI is so broad and nonspecific that it loses any intelligible meaning [105, 106] or just becomes another phrase for bias [104]. Additionally, individual articles made arguments that the direction of bias produced by supposed NFCOIs is inconsistent [103], they only affect discrete situations [103], and there is a lack of empirical studies confirming their impact on research [105].

A point raised by articles both for and against considering competing non-financial interests as COIs is the lack of practical, objective criteria by which one would determine when a competing interest rises to the level of a COI. Most articles that discussed a “cutoff” for NFCOIs argued for some form of a “reasonable person” standard, whereby competing non-financial interests deserve attention if they might make a reasonable person doubt the integrity of an academic work, or if they might embarrass a person if they were revealed after publication rather than disclosed beforehand [58, 74, 83, 94].

One article offered an approach to combat the potential broadness and subjectivity of NFCOI declarations, arguing the need to disclose should be based on the relevance of competing interests as determined by the criteria of pertinence, substantiality, and immutability [44]. For example, although religion can fall into the category of NFCOI, it need not be declared unless it is pertinent, substantial, and immutable. In the case of medical assistance in dying, religious views may be pertinent, may result in substantially different outcomes, and may be immutable to change, and thus should be declared. Others offered a spectrum-model of NFCOI. Within this framework, the goal is to have the fewest NFCOIs in a given project, accepting that zero NFCOIs is impossible [107].

1. 2.

   Do NFCOIs in research require management?

A large majority of papers argued that NFCOIs are of equal moral relevance as FCOIs, and must be managed accordingly. Some argued that NFCOIs are more influential than FCOIs [80]. Justifications for this stance included that: NFCOIs reduce the trustworthiness of a given study [83]; the bias that NFCOIs introduce can impact “grant awards and renewals, appointment to positions, promotion, and tenure;” [58] society’s trust in the scientific enterprise to some degree hinges upon the management of such ethical issues [48]; NFCOIs lead to unethical behaviors including selective publication (aka “salami slicing”), duplicate publication, plagiarism, and digital image enhancement [108]; and the bias NFCOIs may cause in primary studies [109] and systematic reviews [109, 110] undermines the work of policymakers [22, 110], advocacy groups, patients [110], physicians [97, 110, 111], and the legal system [97] to the degree that they rely on this evidence.

Those pushing back against the idea of addressing NFCOIs in the same manner as FCOIs offered practical and logical reasons. As detailed in the results, many noted that NFCOIs might divert attention away from the more significant concern of FCOI [78, 103, 104, 112], thereby reducing our ability to address the tangible and consequential threats posed by FCOI. One argument hinged on the meaningfully different impacts of NFCOIs and FCOIs, as NFCOIs “may provide reasons to suspect cognitive bias but they do not typically involve a loss of trust in a social role. The same cannot be said for [FCOIs.]” [112] Another line of argumentation implied that attempts to manage NFCOIs might lead to exclusion of individuals with varied perspectives [10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106, 113], thus hurting the quality of science. Finally, several authors agreed that NFCOIs are properly termed COIs but warrant no management, explicitly stating that FCOIs are of higher ethical priority [89, 112].

1. 3.

   Should NFCOIs be managed with disclosure or with other strategies?

This was the most heterogeneous result with a narrow majority endorsing management strategies in addition to disclosure. Debate about the role of disclosure ran the full logical spectrum. Some authors treated disclosure as sufficient, arguing that it would provide readers the information necessary to assess the validity of the claims within an academic work [58, 93]. Others argued that, although lackluster, disclosure is the best (or only) realistic option [114]. Most argued that disclosure is necessary but insufficient to address NFCOIs [115]. A small minority argued that disclosure may be inappropriate, and has significant drawbacks that should make us hesitant to rely on it as a strategy for NFCOI management [112, 116,117,118,119,120,121].

The most common argument in favor of disclosure as a management strategy for NFCOIs was that it gives readers the information necessary to evaluate the merits of the research for themselves [58, 91, 93, 122]. In the “necessary but not sufficient” camp, some argued that disclosure allows editors and reviewers to determine whether an author is too conflicted and must be recused or even rejected [42, 59]. Beyond these examples, explanations were rarely given by authors endorsing disclosure, treating it as the de facto management strategy.

Those who did not endorse disclosure as sufficient offered several distinct arguments. First, as a direct counter to disclosure giving readers the information they need to evaluate the research, many articles argued that readers do not necessarily use disclosures to weigh the biases of authors as effectively as we might believe [112, 123]. Additionally, disclosure of certain types of NFCOI might result in invasion of privacy [24, 105], for example with regards to conflicts involving religious affiliations, sexual orientation, gender identity, disability status, or illnesses suffered by members of one’s family. Articles posited that disclosure can be an empty proclamation, doing nothing to address the underlying bias that concerns us [97, 115]. Worse, disclosure may have a moral licensing effect [116,117,118,119], whereby individuals feel less inclined to check their own biases once they have made a declaration. Additionally, the “noise” generated by including disclosure of NFCOIs might drown out FCOI disclosures [115].

Many authors discussed the challenges of enforcing disclosure policies. Policies that do exist are often poorly defined [124]. There is little consensus on an objective benchmark for enacting these policies [123]. Moreover, disclosure policies rely on the honesty and integrity of the authors themselves [107, 125, 126], as it is not feasible for editorial staff to investigate and police all submissions. Author self-disclosure is further complicated by requiring self-awareness. COIs can be present without conscious awareness [102]. Additionally, journals often lack clear procedures to verify disclosure [12, 13] and to respond if incomplete disclosure is discovered [12, 124]. Procedures described for management of undisclosed NFCOIs discovered after publication were inconsistent: published corrections/letters [48, 127], “warnings, retractions, statement of lost confidence, notification of the author’s primary institution, and exclusion from publication in the journal for a specified time frame” [128]. One prominent article discussed “registries of interests,” but conceded that these are not feasible to create and maintain [129].

Comparing and contrasting with Wiersma et al

Our methods were similar to Wiersma et al. [9] in that we used similar keywords and Medical Subject Headings (MeSH). Our methods differed in that we focused our search exclusively on healthcare through the PubMed database while their search strategy included articles from Embase, SCOPUS, and Web of Science in addition to PubMed. Our final dataset included 206 articles, whereas Wiersma et al. had 190, likely because we included search terms specific to the types of competing non-financial interests, such as “intellectual,” “political,” and “religious.” A final difference in search strings is that Wiersma et al.’s inclusion criteria allowed for articles that discussed NFCOI outside of research and publication, such as in medical education and practice. In sum, our search represents a deeper dive into NFCOI’s impacts on healthcare research and publication, whereas Wiersma et al.’s speaks to NFCOIs across broader contexts.

Comparing our results with Wiersma et al., 72 articles were shared between the datasets, 134 were unique to ours, and 118 were unique to theirs. Despite mostly different datasets, we had similar distributions of journal types and articles’ fields of focus (biomedical, healthcare, science, and non-science). NFCOI in research and publication constituted the most common healthcare context of both datasets. Our dataset’s second and third most common healthcare contexts were policy statements and guideline development. In contrast, Wiersma et al.’s second and third most common healthcare contexts were practice and education (Fig. 3).

Distribution of articles and journals. The number of articles in “Field of Focus” and “Healthcare Context” exceed the total number of articles because some articles fulfilled multiple categories

Remarkably, Wiersma et al.’s results and discussion were organized around a similar three core questions as our review: definition and legitimacy of NFCOIs as COIs, whether NFCOIs require management, and what that management should entail. Despite unique search strategies yielding different datasets, both our research groups came to similar conclusions for each of the three core questions.

In line with our results, Wiersma et al. found the preponderance of articles supported the conception of NFCOIs as COIs, with a small number of prominent voices arguing the contrary. They identified many of the same NFCOI types, including career-related, interpersonal, status related, and belief or viewpoint based.

Regarding management of NFCOIs, 73% of our total articles^{Footnote 1} versus 55% of Wiersma et al.’s endorsed the need for some form of intervention. Only 4% of our total articles rejected the need for management versus 11% of Wiersma et al.’s. Arguments for and against the need for management were similar across datasets. Arguments in favor included that NFCOIs are of equal or greater importance to FCOIs, and that they disrupt research integrity. Arguments against were led by concerns that management of NFCOIs would distract from FCOIs, and that NFCOI is defined too broadly to be managed.

Although Wiersma et al. did not summarize recommended NFCOI strategies quantitatively, they acknowledged disclosure as a common strategy. Like us, they found frequent discussion that disclosure is necessary but not sufficient, and can have significant drawbacks. Alternative strategies identified by both our studies include open discussion, reflexivity, management according to severity, use of scientific methods, balancing competing interests, registries, and policies.

Strengths and limitations

This review has three unique strengths. First, it independently corroborates the findings of Wiersma [9] et al., having independently conducted the literature search, data analysis, and conclusions prior to the publication of their article. Moreover, two reviewers independently read and coded each article at the 621 and 206 stages, with a third, senior author adjudicating disagreements. Taken together, Wiersma et al. and our study produced similar results from complementary datasets. Second, this review quantifies the number of articles including arguments for and against our three major questions. Third, Wiersma et al. cite their own NFCOI as a limitation given their history of argumentation towards their review’s conclusions. Without such NFCOIs of our own on this topic, our study even more strongly supports Wiersma et al.’s conclusions that NFCOIs are a serious problem warranting management.

We identify several limitations. First, we limited our search to the PubMed database focusing on research and publication. This resulted in a deep dive into NFCOI in healthcare research and publication. Although this was an intentional choice, additional databases may have revealed additional articles and entire fields with differing perspectives. Acknowledging this, Wiersma et al.’s similar findings with the use of multiple databases makes this unlikely.

Second, our study is limited by a search strategy that could be biased for articles favoring the view that competing non-financial interests are COIs requiring management. In searching the literature for COIs that are non-financial, articles that do not even include the term COI would have been missed. This could be one explanation for our results heavily favoring treating NFCOIs as COIs requiring management. Furthermore, our search string was designed to retrieve only articles indexed with the ‘Conflict of Interest’ MeSH term, with no ‘Conflict of Interest’ keyword included in our search string. This strategy prevented creation of an initial dataset with an overwhelming number of irrelevant articles containing the phrase ‘Conflict of Interest’ in their disclosures. However, this led to omission of relevant articles (e.g. Resnik DB) [130] that discuss NFCOI but had not been assigned the MeSH term at the time of our search.

Third, we found that many different terms refer to NFCOIs. Subtypes, such as “Intellectual COI,” “Personal COI,” and “Academic COI” are sometimes used interchangeably with the term “non-financial COI.” Although we included eleven subtypes of NFCOI, our search may have missed articles containing other variations. Nevertheless, our inclusion of subtypes may help explain our larger final dataset as compared to Wiersma et al.