Because of highly publicized cases of financial conflicts of interest leading to a lack of scientific integrity in published articles 
, journals have been urged to adopt comprehensive COI disclosure requirements for authors 
. For disclosure to be effective, reviewers must be able to critically assess to what extent a conflict of interest exists, to judge the impact that COI may have on the validity of a submitted manuscript, and to incorporate those judgments into a recommendation regarding publication. In one of the only empirical studies of the influence of COI disclosures on readers' perceptions of an article, Schroter et al 
completed a randomized control trial in which 900 British Medical Journal (BMJ) readers were provided an article with one of three possible COI disclosures. The “financial statement” declared the authors to be employees or owners of stock,” the “grants statement,” declared the authors to have received research funding and the “none statement” declared no COI. The importance, relevance, validity, and believability ratings of the articles were significantly lower for the article in which “authors are employees and potentially own stock” than in the “none declared” group. Building on these findings but focusing our population of interest on peer reviewers, who are directly involved with decisions regarding the publication of research in contrast to the general readership, we carried out an empirical study with experienced reviewers of medical research from a journal that has an explicit COI policy for authors, reviewers, and editors. We had 4 major findings.
First, the vast majority of respondents perceive of roles on speakers bureaus or as consultants to be aligned with company marketing goals. Even activities condemned by the International Committee of Medical Journal Editors (ICMJE), World Association of Medical Editors (WAME), and Committee on Publication Ethics (COPE), such as ghost writing or authoring articles with only limited access to data, are believed to be likely occurrences 
Second, the majority of our respondents did not believe that the bias attributed to these roles has a minimum monetary threshold. Over one-half of respondents believe that bias most likely exists with any honorarium, regardless of monetary amount; while over 90% of respondents believe that bias most likely exists even if authors received up to only 17% of income from a pharmaceutical company.
Third, in contrast to respondents without financial ties, those who disclosed having any personal financial ties to industry were less likely to attribute bias to speakers bureau and consultant roles, and less likely to believe that honoraria of any monetary value, no matter how small, introduces bias (). There are several possible explanations for this discrepancy between respondents with and without financial ties to industry.
Comparison analysis of reviewers with or without personal financial ties to industry.
Respondents who serve on speakers bureaus or act as consultants may have found, in their direct experience, that companies did not expect their lectures to be aligned with marketing goals and that access to data was unlimited. In addition, these reviewers may have more concrete, inside knowledge about the actual activities and benefits of working as a consultant or on a speakers bureau. This interpretation is supported by our finding that reviewers with personal financial ties to industry were more likely to accurately state that companies must ensure speakers bureau presentations mention only FDA approved uses. However, our study can neither confirm nor refute this explanation.
Basic psychological research suggests “that when individuals stand to gain by reaching a particular conclusion, they tend to unconsciously and unintentionally weigh evidence in a biased fashion that favors that conclusion.” 
Based on this psychological vulnerability to unintentional bias, the discrepancy we found between those with and without personal financial ties to industry may be present because respondents with financial relationships to pharmaceutical companies project their belief that they are immune to undue influence or discount the potential influence in order to prevent cognitive dissonance. This interpretation raises the concern that reviewers with personal ties to industry, regardless of the relevance to a particular article under review, may themselves be vulnerable to minimizing the possible effects of financial relationships with industry to prevent cognitive dissonance when assessing an author's potential of bias.
Finally, most respondents reported that they read articles more carefully and consider the credibility diminished if the author discloses a financial relationship with the manufacturer of the study drug. About two-thirds of respondents would not change their recommendation regarding publication if they found no design or statistical concerns regardless of the author's financial disclosures. Thus reviewers express skepticism of industry relationships but report no overt discrimination against industry-funded manuscripts; rather, their answers suggest that they base their decision for publication on the perceived scientific merit of the manuscript.
Although this is reassuring, it should be interpreted in light of possible social desirability bias as the results rely on self-reported attitudes and behaviors. Additionally, it should be interpreted with an understanding of reviewers' known limitations, including reviewers' variable capacity to find statistical and design flaws within manuscripts under review. In a study using a fictitious manuscript with purposeful errors, 68% of reviewers did not realize that the conclusions of the work were not supported by the results 
. Similarly, a subsequent investigation reported that reviewers found only an average of three out of nine, major, deliberately induced errors in papers manipulated for the study 
. Although our reviewers reported reading papers more carefully and critically, it remains unknown whether after reading author disclosures of financial relationships to the study sponsor or drug manufacturer, they are in fact better able to detect flaws in a manuscript and to recommend revisions to mitigate those flaws.
Our response rate was 54% despite multiple attempts to contact potential participants via email. Nevertheless, no significant differences were found between responders and non-responders or between responders and the larger reviewer pool in reviewer experience, reviewer volume, review quality rating, age, or gender. Even in the unlikely situation that all the non-respondents had answered differently than responders, there would still have been a large proportion of reviewers sharing the views reported here.
As with any survey relying on self-reported attitudes and behaviors, stated preferences are highly subject to social desirability bias. Because of the time, effort, and logistics required, recruiting and randomizing reviewers to review fictitious manuscripts that reveal or omit conflict disclosures was not feasible, particularly since no antecedent intermediate studies have been published to date. Such a study of fictitious manuscripts should be considered the next step in research of this topic.
Our survey also relies on questions about the industry-author relationships: speakers bureau member and consultant, which were not further defined in detail. Nevertheless, actual characteristics of speakers bureaus or consulting arrangements are not known, uniform, or found in disclosures required by journals. Ultimately, we expect that the interplay of perceptual influences and the imprecision in disclosures within our survey mirrors those faced by reviewers who are presented with actual conflict of interest disclosures on manuscripts under review. The reported reviewer perceptions, therefore, may be important influences on how reviewers assess submitted manuscripts.
Our findings may not generalize to other journals or specialties. Nevertheless, this particular reviewer population is similar to reviewers from other journals and specialties 
, including the ability of these reviewers to detect deliberate introduced flaws in a manuscript 
,  
We did not adjust our levels of statistical significance for multiple comparisons with the Bonferoni correction, which may be too conservative; readers should interpret our findings in light of multiple comparisons we made.
Based on our findings, we offer several practices that might help reviewers and readers better understand the degree and character of potential bias introduced by financial disclosures:
- Because there is neither evidence nor consensus to support a specified minimum monetary threshold below which bias does not exist, the monetary amount of all financial relationships should be reported with manuscript submissions. We recommend that all financial relationships should be disclosed because over one-half of our reviewers believed that any honorarium would most likely bias the author's judgment. Because the percentage of reviewers who held this belief increased as the specific amount of honorarium increased, we recommend that the exact monetary amount also be disclosed.
- Lead authors should be required to certify that they have had unrestricted access to all data and statistical analysis, and the right to publish in accordance with ICMJE recommendations. Several recent incidents support the need for adherence to this recommendation. These incidents exposed that important adverse events were not reported in publications, and that several authors received either incomplete data or only the final results tables with no access to the analysis undertaken or original data set , , .
- Because our results raise the concern that reviewers with personal ties to industry may be vulnerable to minimizing the possible bias associated with financial relationships with industry, all reviewers should disclose any financial ties to industry whether related to the article under review or not. In support of this recommendation, other studies suggest that individual physicians may be poor judges of whether a financial relationship with industry is relevant to the study at hand .
Finally, the majority of reviewers surveyed report a high level of skepticism regarding financial relationships between authors and industry without a clear or consistent translation of that skepticism into actions of manuscript assessment and recommendation. Organizations like ICMJE, COPE, and WAME have all developed increasingly specific and detailed guidelines on COI disclosures over the past 10 years, and exhorted all journals to do the same. Although these increasingly strict and comprehensive disclosure guidelines have been recommended, our results drive us to ask whether disclosure alone truly aids reviewers in identifying potential bias, accounting for the magnitude of its effect and translating that understanding into action. In addition to the greater detail about monetary amounts and actual activities associated with particular disclosures, we propose that research is needed to better identify the components of study design vulnerable to COI bias and the specific components of analysis and result reporting most likely to harbor that bias. Once these components are identified, journals could develop unique guidelines to aid reviewers in identifying bias more consistently, and in more accurately accounting for the effect of that bias in their manuscript assessment.